2898
Beyond matrix completion. Communications of the ACM
59 (11): 94-102.
Lewis, Gregory. 2011. Asymmetric information, adverse
selection and online disclosure: The case of eBay motors.
The American Economic Review 101 (4): 1535-1546.
Lieber, Ethan, and Chad Syverson. 2012. Online versus
offline competition. In Ox ford handbook of the digital
economy, ed. M. Peitz and J. Waldfogel, 189-223.
Oxford: Oxford University Press.
Luca, Michael. 2013. Reviews, reputation, and revenue: The
case of . HBS working knowledge.
Mayzlin, Dina, Yaniv Dover, and Judith Chevalier. 2014.
Promotional reviews: An empirical investigation of online
review manipulation. American Economic Review 104
(8): 2421-2455.
Nosko, Chris, and Steven Tadelis. 2015. The limits of
reputation in platform markets: An empirical analysis
and field experiment. Cambridge: National Bureau of
Economic Research.
Pallais, Amanda. 2014. Inefficient hiring in entry-level labor
markets. American Economic Review 104 (11): 3565-
3599.
Quan, Thomas W., and Kevin R. Williams. 2016. Product
variety, across-market demand heterogeneity, and the
value of online retail. New Haven: Connecticut Cowles
Foundation for Research in Economics, Yale University.
Rochet, Jean-Charles, and Jean Tirole. 2003. Platfomr
competition in two-sided markets. Journal of the
European Economic Association 1 (4): 990-1029.
Romanyuk, Gleb. 2017. Ignorance is strength: Improving the
performance of matching markets by limiting infor-
mation. Manuscript.
Sorkin, Andrew Ross. 2016. Why Uber keeps raising billions.
The New York Times.
Tadelis, Steven, and Florian Zettelmeyer. 2015. Information
disclosure as a matching mechanism: Theory and
evidence from a field experiment. The American Eco-
nomic Review 105 (2): 886-905.
Ursu, Raluca Mihaela. 2016. The power of rankings:
Quantifying the effect of rankings on online consumer
search and purchase decisions. Social Science Research
Network SSRN scholarly paper ID 2729325. Rochester.
Weyl, E. Glen. 2010. A price theory of multi-sided plat-
fomrs. American EconomicReview 100(4): 1642 1672.
Weyl, E. Glen, and Michal Fabinger. 2013. Pass-through as
an economic tool: Principles of incidence under imperfect
competition. Journal of Political Economy 121 (3): 528-
583.
White, Alexander, and E. Glen Weyl. 2016. Insulated
platform competition. Social Science Research Network
SSRN scholarly paper ID 1694317.
Zervas, Georgios, Davide Proserpio, and John Byers. 2015.
The rise of the sharing economy: estimating the impact of
Airbnb on the hotel industry. Social Science Research
Network SSRN scholarly paper ID 2366898. Rochester.
Zhang, Hongkai. 2017. Accelerated quality discovery through
sponsored search advertising in online marketplaces.
Manuscript.
Dimensions of Economic Quantities
Dimensions of Economic Quantities
P. H. Wicksteed
A unit is a concrete magnitude selected as a stan-
dard by reference to which other magnitudes of the
same kind may be compared. A derived unit is a
unit determined with reference to some other unit.
Thus the unit of area may be derived from the unit
of length by being defined as the area of the square,
erected on the unit of length. The unit of speed may
be derived from the unit of length and the unit of
time, by being defined as that speed at which the
unit of length is traversed in the unit of time. In
relation to the derived units of area and speed, the
units of length and time would then be fundamental
— ‘fundamental’ being a term correlative to
‘derived’.
The theory of dimensions is concerned with ‘the
laws according to which derived units vary when
fundamental units are changed’ (Everett). A
fundamental unit, together with the magnitudes of
like kind referred to it, is regarded as having one
dimension. Thus a length had the dimension L. The
unit of length enters twice into the unit of area, first
determining the base and then the altitude of the
unit rectangle, and therefore the dimensions of an
area are LL, usually written L2. If we alter the unit
of length, say from a foot to an inch (1:12) the unit
of area will be reduced in the same ratio twice
successively (1:144 in all). The variations of the
unit of area, therefore, are directly as the squares of
the variations in the unit of length. The units of
length and of time enter once each into the unit of
speed, but they do not enter on the same footing. If
the unit of time be the minute, and the unit of length
the foot, the unit of speed will be a foot per minute.
This unit will become smaller if we make the unit of
length smaller, since an inch per minute is a smaller
speed than a foot per minute; but it will become
larger if we make the unit of time smaller, a foot a
second being a greater speed than a foot a minute.
This is expressed by saying that the dimensions of
time T enters negatively into speed. The dimensions
of
Dimensions of Economic Quantities
speed, then, are expressed as LT-1. A unit into
which a dimension enters negatively is always a
unit of rate, and measures amount of x per unit of y,
—y being the quantity the dimension of which
enters negatively.
We have now examined simple cases of the
variations of derived units, but it is obvious that the
numerical values of concrete magnitudes vary
inversely as the units by reference to which they are
estimated. The smaller the unit the greater the
numerical value of any given magnitude. The
numerical value of a magnitude, therefore, will vary
inversely as the unit whose dimension enters into it
positively, and directly as the unit whose
dimensions enters into it negatively. Thus, let the
unit of speed (dimensions LT-1) be a foot per
minute, and let the numerical value of a certain
concrete speed be 10, i.e. let the speed be ten per
minute. Then change the unit of length to an inch
(1:12) and the unit of time to a second (1:60); the
derived unit will now be an inch per second, and its
relation to the former derived unit is obtained by
altering directly in the ratio of 1:12 (dividing by 12)
and inversely in the ratio of 1:60 (multiplying by
60), so that the new unit is five times as great as the
old one, an inch per second being five times as great
a speed as a foot per minute; but the numerical
value of the concrete speed we had to express must
be altered inversely as 1:12 and directly as 1:60,
and is now only 2 -
i. e. the speed is two inches per second - or
one-fifth of what it was before.
If we are measuring such a magnitude as feet of
vertical motion per foot of horizontal motion in the
path of a projectile, the dimensions will be LL-1
and will cancel each other. No change in the unit of
length, then, will in any way affect the numerical
value of this magnitude, and as no other dimension
enters into it at all, it may be said to have no
dimensions. Angular magnitudes, defined as ratios
between arcs and radii, trigonometrical functions,
and ratios generally are of this nature. They have no
selected units, and their numerical values are
absolute.
When the elements of the theory of dimensions
have been thoroughly grasped it will be easy to
apply it to economic questions; and it will be found
an invaluable check in the more intricate
2899
problems of co-ordination and analysis. Thus, if the
unit of value-in-use or utility be taken as
fundamental, and regarded as having the dimension
U, and if the commodity we are considering be
taken as having the dimension Q, then degree of
utility of the commodity, being the rate at which
satisfaction is secured per unit of commodity
consumed, will have dimensions UQ-1, and, will be
readily distinguished from rate of enjoyment,
accruing to the consumer, per unit of time, with
dimensions UT-1. Price, determined by marginal, or
final, degree of utility, will have dimensions UQ-1
or P; and hire, being price per unit of time, will
obviously have dimensions PT-1 or UQ-1T-1.
When the thing hired is money and is used
commercially, the utility derived from it is a
commodity of like nature with itself. The dimension
U then becomes Q, and the dimensions of interest
(as a rate) are QQ-1T-1 or T-l, which will be found
on reflection and experiment to be correct.
The theory of dimensions should be applied to
economics in close connection with the diagram-
matic method. But of course the connection
between dimensions, as now explained, and the
geometrical dimensions of the diagrams is purely
arbitrary. The physicist may, according to his con-
venience, represent the height of a projectile - a
magnitude of one dimension - by a line, or by an
area, and speed by a line of an inclination. So the
economist may represent a magnitude measured by
a complicated derived unit by a line, or a magnitude
measured by a fundamental unit by an area or a
solid; and if he keeps the theory of dimensions well
before him he may vary his methods indefinitely
without any danger of confusion. In all cases,
however, the dimensions of those quantities
represented by areas or solids will be compounded
of the dimensions of those represented by the lines
which determine them. Again, those who have any
acquaintance with the elements of the calculus will
see that if the equation of a curve be differentiated
to x then the area of the derived curve will have the
same dimensions as the ordinate ofthe fundamental
curve; the ordinate of the derived curve will have
the dimensions of the ordinates of the fundamental
curve positively, and those of its abscissae
negatively;
2900
Y some apparent errors and confusions which made
Dimensions of Economic Quantities, Fig. 1
and the abscissae of the two curves will have the
same dimensions. In other words, differentiation
introduces the dimensions of the variable to which
we differentiate negatively, and integration intro-
duces the dimensions of the variable to which we
integrate positively (Fig. ).
By way of illustration take a figure, on the
ordinate of which intensity of desire, or degree of
utility, is represented, while supply of commodity
per unit of time is measured on the abscissae. Now
imagine a third axis (of Z) perpendicular to the
page, along which time is measured. Such a figure
will enable us to represent all the quantities we have
to deal with in an ordinary problem of consumption.
Rate of supply is represented on axis ofX,
dimensions QT-1; degree of utility on axis of Y,
dimensions UQ-1; time on axis of Z, dimension T;
rate of enjoyment on areas parallel to plane of axes
of X and Y, dimensions UQ- 1QT-1 or UT-1; total
enjoyment on solid figure, dimensions UQ-1QT-1T,
or U; total supply on areas parallel to plane of axes
of X and Z, dimensions QT-1T, or Q, and in like
manner price, hire, total sum paid, etc., may be
read, and their dimensional relations seen at a
glance.
[The theory of dimensions was (according to
Jevons, Principles of Science, , p. 325) first clearly
stated by Joseph Fourier. He expounded it with
great lucidity in his Theorie Analytique de la
Chaleur, 1882, §§ 159-62. An excellent popular
statement of the theory, as it has since been
elaborated, will be found in the beginning of J. D.
Everett’s C.G.S. System of Units, 1891. Jevons was
the first to suggest the application of die theory to
economics (Theory of Political Economy, 1888, pp.
232-52), but he unfortunately fell into
Direct Taxes
the suggestion barren in his hands. A criticism of his
treatment of the subject and an independent
working-out of his suggestion, by the writer of the
present article will be found in the American Quar-
terly Journal of Economics for April 1889, pp. 297-
314.]
Bibliography
Everett, J.D. 1875. Illustrations of the centimetre-gramme-
second system of units, with tables of physical constants.
London.
Fourier, J. 1822. Theorie analytique de chaleur. Paris. Jevons,
W.S. 1871. The theoty of political economy. London:
Macmillan.
Jevons, W.S. 1887. Principles of science. London:
Macmillan.
Wicksteed, P.H. 1888. The alphabet of economic science.
London: Macmillan.
Wicksteed, P.H. 1889. On certain passages in Jevons’s Theoty
of Political Economy. Quarterly Journal of Economics 3:
293-314.
Wicksteed, P.H. 1895. In A symposium on value, ed.
J.H. Levy. London: Macmillan.
Wicksteed, P.H. 1910. The common sense of political
economy. London: Macmillan.
Direct Taxes
John Kay
The distinction between direct taxes and indirect
taxes traditionally rests on a view of the incidence
of the two kinds of tax. The incidence of a tax
identifies who suffers loss of income or welfare as a
result of the imposition of the tax. This may differ
from the location of the legal liability for payment
of the tax if the payer is able to shift part or all of
this liability to some other agent. The capacity to
shift the tax burden in this way depends on the
elasticities of demand and supply of the taxed factor
or commodity. Direct taxes are those for which the
legal liability and the incidence are identical:
indirect taxes are those where the tax is shifted,
most usually to final consumers.
Direct Taxes
Thus income taxes are generally regarded as
direct taxes and commodity taxes as indirect. This
supposes that factor supplies are completely
inelastic and commodity supplies perfectly elastic,
an empirical observation which may hold in
particular cases but which cannot be seen as a
universal truth. In reality, all taxes are shifted to
some extent and none completely. A more recent
reformulation of the direct/indirect distinction
(Atkinson and Stiglitz ) describes indirect taxes as
those differentiated by the nature of the transaction
and direct taxes are those differentiated by the
identity of the transactors; but this too breaks down
on closer examination and the classification is one
with no particular economic significance. The most
important direct taxes are progressive wealth,
income and expenditure taxes levied on individuals,
and tax imposed on the income of corporations.
Before the emergence of modem systems of
public finance, ad hoc wealth taxes were a primary
source of revenue. But this was possible only when
wealth mostly took the form of real property and
revenue requirements were relatively minor. Many
countries still have a tax on wealth, but there is
none in which it makes an important contribution to
revenue. Taxes are often levied on transfers of
capital, on death or sometimes when substantial
gifts are made. Such a tax may be donor or donee
based. An inheritance tax levied on the donor is a
progressive tax based on the total of gifts made by
the tax payer. An accession tax on the donee is one
in which the rate of tax is based on the cumulative
total of gifts received.
Although there is extensive academic discussion
of the potential of a direct tax on expenditure, no
major country has adopted one. Direct taxes are
primarily income taxes, on the incomes of
individuals and of corporations. We begin by
looking at the base of the tax and then consider the
criteria which should determine the rates at which
income - personal or corporate - should be taxed.
It seems trite to observe that in order to tax
income it is necessary to define it, but in fact the
taxing statutes of most states do not attempt to do
so. Income is exemplified rather than defined. For
economists, the classic definition of income is that
2901
of Hicks ( ) - ‘income is the maximum value
which a man can consume during a week and still
be as well off at the end of the week as he was at
the beginning’. By the same principle, corporate
income might be defined as the maximum which a
company can distribute, and still be as well off at
the end of the accounting period as at the begin-
ning. But these are not operational concepts for a
tax inspector. How is he to determine what a man
expects? And what is he to do if these expectations
are unreasonable?
Thus attention has instead been devoted to the
concept of ‘comprehensive income’, or Haig-
Simons income, so-called after its principal
advocates (Simons ). As Hicksian income looks
forward, so Haig-Simons looks back, and measures
not what a man could have expected to consume but
what he could in fact have consumed. If
expectations are always fulfilled, then the two
concepts are identical: but windfall gains, excluded
from the Hicksian concept of income, fall within
the Haig-Simons one. It follows that Haig-Simons
income requires that all accruing capital gains
should be included within the tax base and taxed as
income. In fact no country has gone as far as this;
some tax certain capital gains as income; almost all
tax most kinds of capital gain more lightly, if at all.
The application of either a Hicksian or a Haig-
Simons measure of real income implies indexation
of the tax base. This means not only that capital
gains should be adjusted for inflation, but that
investment income - paid or received - should be
adjusted also.
Such inflation adjustment should relate to indi-
vidual income, to capital gains, and to the income
of corporations. Inflation adjustment to the income
of individuals is very rare, although several
countries now provide for indexation in calculating
capital gains. Most attention to the effects of
inflation on the measurement of income has been
given in the corporate sector inflation. Accounting
profit becomes a misleading indicator of the returns
earned by a company under inflation because
depreciation is generally based on the historic cost
rather than the current cost of equivalent assets;
because the rise in the price of goods held in stock
(stock depreciation) is included in
2902
profits; and because interest paid or received is
expressed in nominal rather than real terms. All
countries with recent experience of high rates of
inflation have considered changes to accounting
standards to remove these distortions but agreement
on appropriate adjustments has proved elusive.
In the absence of accepted accounting principles,
tax systems have responded to inflation in ad hoc
ways. The inadequacy of historic cost depreciation
allowances has been partly compensated for by
acceleration of the rate at which such allowances
may be taken. Relief for the effect of inflation on
stock values has been given, either by accepting
accounting practices such as LIFO (last in, first out)
which automatically give relief at current prices, or
by particular measures of stock relief. Tax
authorities have been much more reluctant to make
allowance for the effect of inflation in eroding the
value both of the monetary assets of companies and
of their debts.
A tax system which was fully indexed in this
way would be neutral with respect to the rate of
inflation, but it would not equalize pre-and posttax
rates of return because the real return earned by the
company would continue to be subject to tax. Full
neutrality could be achieved by means of a cash
flow tax, which allows immediate deductibility of
all capital expenditure - either in stocks or on fixed
assets - but denies any relief for financing costs,
whether interest or otherwise. Such a tax was
proposed by the Meade Committee ( ) and it uses
as its base the flow of funds
from the real operations of the company to those
who finance it.
Once income has been defined, at what rate
should it be taxed? Differentiation between types of
personal income was a principal issue when income
tax was introduced in the nineteenth century. The
argument rested on the precariousness of income
from employment relative to property income, and
this, is was suggested, provided a reason for taxing
investment income more heavily. These arguments
read rather oddly in a twentieth-century context, in
which inflation and economic fluctuations have
generally made property income appear more
precarious than earnings, and this argument has
largely vanished
Direct Taxes
from discussion and its consequences from tax
schedules.
A tax schedule is progressive if the average rate
of tax increases with income. This does not require
that marginal tax rates are increasing and indeed a
linear tax schedule is progressive if its intercept is
positive. There is no unambiguous measure of
progressivity, and the same term is sometimes used
to cover both the extent to which the schedule
deviates from proportionality and the redistributive
effect of tax structure. It will be apparent that a tax
which departs substantially from proportionality but
generates little revenue will have less redistributive
effect than a more nearly proportional but heavier
tax.
Nineteenth-century utilitarian arguments
suggested alternative rate structures. The principle
of equal sacrifice, for example, demanded a
schedule which imposed equal utility losses on all
the taxpayers. This implied payments from those
with higher incomes, but not necessarily
proportionately larger payments, the outcome
depending on the elasticity of the marginal utility of
income. Utility maximization subject to a revenue
constraint requires equal marginal sacrifices, with
similar implications.
However, these analyses take no account of the
effects of tax schedules on labour supply. Like
indirect taxes, income taxes impose a deadweight
loss or excess burden in addition to the revenue
which they raise. The magnitude of these losses
depends on marginal tax rates and the wage elas-
ticity of labour supply. It follows that there is a
direct conflict between the progressivity of a tax
schedule - which implies high marginal rates of tax
- and its efficiency properties - which require low
marginal rates.
Mirrlees ( ) was the first to examine this
trade-off explicitly and although a substantial lit-
erature on optimal income tax structures has
developed since, relatively few results of general
application have emerged. There is some indication
that marginal tax rates should be lower at the
extremes of the distribution than in the middle of it.
The disincentive effects of high marginal tax rates
depend on the numbers of individuals in the
relevant range, whereas their redistributive function
depends on the number of individuals above
Direct Taxes
that range. As we move up the income distribution,
this redistributive effect steadily diminishes, while
the disincentive effect remains; and thus the balance
between the two factors changes in a direction
which points to lower marginal rates of tax. Similar
arguments can be applied at the lower end of the
distribution.
While the welfare effects of income taxation are
principally the product of marginal rates, the overall
effect on labour supply is determined by both
income and substitution effects, and is therefore
influenced by the average as well as the marginal
tax rate at any point in the distribution. For this
reason, while the efficiency costs of increasing
taxation are unambiguous the labour supply effect
may be positive or negative in sign. Labour supply
is presumably zero at tax rates of 100 per cent,
however, and if an interior maximum exists (which
is by no means certain) then there will be some rate
below this which yields maximum revenue. This
observation yields what has become known as the
Laffer curve.
The structural issues which influence redistri-
bution across the income distribution are concerned
with vertical equity in taxation. Horizontal equity
reflects its concern with the relative tax burdens at
the same point in the income distribution.
Horizontal equity implies that individuals in the
same circumstances should be treated similarly and
would exclude, for example, random taxation (even
though this might, under certain circumstances, be
efficient). However the principle of horizontal
equity has limited application because of the
difficulty of agreeing an objective definition of
‘similar circumstances’. The most important issues
of horizontal equity in practice concerns the tax
treatment of the family, an area of taxation in which
there is direct conflict between two conflicting
principles - the desire to respect the right of
individuals to individual treatment, which points to
an individual basis for taxation, and the desire to
relate liability to the whole of an individual’s
circumstances, which necessarily includes the
circumstances of those with whom he or she lives.
Most tax systems incorporate elements of both
individual and unit bases.
In fixing the rate of corporate income tax, it is
necessary to begin by asking why we tax
2903
corporate income at all. Although corporations have
distinct legal personalities, they have no economic
personality and ultimately generate no command
over resources other than those of the individuals
who work for them, manage them, buy their
products, or own their shares. It is these individuals
who pay corporation tax. The economic rationale
for corporate income taxes therefore requires
justification.
One such argument is that they are there: the
phenomenon of tax capitalization implies that if
particular assets, such as the equity of corporations,
are subject to discriminatory taxation then these
taxes will be reflected in the prices of the assets
concerned. To remove such a tax would effect no
current efficiency gain, and would confer windfall
gains on current shareholders; this is the
rationalization of the traditional maxim that ‘an old
tax is a good tax’. Corporation tax may also enable
countries to derive revenue from the assets of non-
residents; this is a powerful argument for such a tax
in many countries.
An important point is that in the absence of
corporate income tax, individuals would avoid the
personal income tax through incorporation. This
suggests that the income of corporations should be
attributed to its owners and taxed as their income.
Although the possibility of full integration of cor-
porate and personal income taxation has been
discussed, and was recommended for Canada by the
Carter Commission ( ), no country has yet
adopted it. The classical system of corporation tax
is one in which the income of corporations is taxed
at a flat rate entirely separate from the income of
shareholders. This is the system used in the United
States; most European countries, however, now
employ an imputation system in which the
shareholder receives some credit against his own
income tax bill on dividends for corporation tax
paid by the company from which he receives them.
This relieves the element of double taxation implicit
in the classical system, but still tends to tax income
accruing through corporations more heavily than
other kinds of income.
Corporation tax has usually been seen as a tax
on capital employed in the corporate sector. It
follows that this purpose is discriminated against
2904
relative to other uses of capital in the domestic
economy, such as agriculture or property. This is
the approach adopted in Harberger’s classic
(theoretical) analysis of the incidence of corporation
tax (Harberger ), which traced its effects on returns
to capital in different sectors of the economy, ft is
also implicit in most empirical studies of the impact
of corporation tax, such as those of Musgrave and
Krzyzaniak ( ), which
have considered the question of the extent to which
a tax on the capital employed by corporations can
be shifted forward into the prices of goods produced
by the corporate sector. Their work suggested that
the extent of such shifting might be substantial.
More recent analysis has challenged this
approach to the incidence of company taxation
(Stiglitz ). The argument is that corporation tax
cannot appropriately be represented as a tax on
capital employed. Most corporate taxes allow
extensive deductions for capital costs, such as
interest and depreciation. If capital costs are folly
deductible, then the corporate tax system is neutral.
Such neutrality can be achieved either if all
investment costs can be expensed, or if depreciation
allowances correspond to true economic
depreciation and financing costs are folly deduct-
ible, through tax relief on interest paid and impu-
tation for company dividends. If the tax regime
provides - as is common in many countries - both
for deductions for the costs of finance and for
accelerated depreciation, then the corporation tax
may actually act as a subsidy to corporate capital
rather than a tax. The post-tax rate of return may
exceed the pre-tax rate. Such a tax may still yield
revenue, since it will still fall on pure profits, i.e.
returns earned by the firm which are not directly
attributable to its capital employed.
Pure profits are generated by entrepreneurship, a
word which may describe the classic entrepreneurial
function of bringing different factors of production
together; the exploitation or establishment of
monopoly rents; or the generation of new means of
organization or invention. Thus the new view of
corporation tax sees it as a levy on those items,
combined with a rather arbitrary array of taxes and
subsidies to different types of investment. The rates
of these taxes and subsidies
Direct Taxes
depend on the degree to which a given activity can
be financed by debt rather than equity and the
relationship between true economic depreciation
and what is permitted for tax purposes.
Direct taxation can be adjusted sensitively to
bold social and economic objectives, and as modem
states have developed and their revenue
requirements have grown so reliance on them has
tended to increase. More recently, however,
dependence on personal income tax has been seen to
imply excessive rates. The result has been some
moves back towards broadly based indirect taxes,
particularly the value added tax, which has been
introduced throughout the European community and
in about thirty other states.
Similar pressures have been evident in the cor-
porate sector. Taxing corporate income is therefore
not the only means of taxing corporations and,
given the difficulties involved in identifying the
country within which income arises, measuring
income in a period of inflation, and taxing declining
real profitability, taxes on corporate income have
tended to diminish in importance. The average share
of total OECD tax receipts derived from corporation
tax fell from 9.2 per cent to 7.4 per cent between
1965 and 1983. At the same time, however, other
taxes on business, particularly payroll and social
security taxes, have tended to increase: implying an
overall shift in relative tax rates on capital and
labour as factors of production.
See Also
Bibliography
Atkinson, A.B.., and J.E. Stiglitz. 1980. Lectures on public
economics. New York: McGraw Hill.
Carter Commission. 1966. Report of the Royal Commission
on Taxation.
Harberger, A.C. 1962. The incidence of the corporation
income tax. Journal of Political Economy 70: 215-240.
Hicks, J.R. 1939. Value and capital. Oxford: Clarendon Press.
Meade, J.E. (chairman). 1978. The structure and reform of
direct taxation. London: Allen & Unwin.
Directly Unproductive Profit-Seeking (DUP) Activities
Mirrlees, J.A. 1971. An exploration in the theory of optimal
income taxation. Review of Economic Studies 3k: 175-
208.
Musgrave, R.A., and M. Krzyzaniak. 1964. The shifting of the
corporation income tax. Baltimore: Johns Hopkins Press.
Simons, H.C. 1938. Personal income taxation. Chicago:
Chicago University Press.
Stiglitz, J.E. 1976. The corporation tax. Journal of Public
Economics 5(3-4): 303-311.
Directly Unproductive Profit-Seeking
(DUP) Activities
Jagdish N. Bhagwati
Keywords
Chicago School; Directly unproductive profit-
seeking (DUP) activities; Endogeneous tariffs;
Immiserizing growth; Lobbying; Optimal tariffs;
Predation; Production subsidies; Public choice;
Regulation; Rent seeking; Revenue seeking;
Shadow pricing; Smuggling; Tariff seeking;
Tariffs; Transfer problem; Voluntary export
restrictions
JEL Classifications
F2
Directly unproductive profit-seeking (DUP) activ-
ities are defined (Bhagwati ) as ways of making a
profit (that is, income) by undertaking activities
which are directly (that is, immediately, in their
primary impact) unproductive, in the sense that they
produce pecuniary returns but do not produce goods
or services that enter a conventional utility function
or inputs into such goods and services.
Typical examples of such DUP (pronounced
appropriately as ‘dupe’) activities are (?) tariff-
seeking lobbying which is aimed at earning pecu-
niary income by changing the tariff and therefore
factor incomes; (ii) revenue-seeking lobbying which
seeks to divert government revenues towards
oneself as recipient; (iii) monopoly
2905
seeking lobbying whose objective is to create an
artificial monopoly that generates rents; and (iv)
tariff-evasion or smuggling which de facto reduces
or eliminates the tariff (or quota) and generates
returns by exploiting thereby the price differential
between the tariff-inclusive legal and the tariff-free
illegal imports.
While these are evidently profitable activities,
their output is zero. Hence, they are wasteful in
their primary impact, recalling Pareto’s distinction
between production and predation: they use real
resources to produce profits but no output.
DUP activities of one kind or another have been
analysed by several economic theorists, among
them (i) the public-choice school’s leading
practitioners, their major work having been brought
together in Buchanan et al. ( ),
(ii) Lindbeck ( ) who has worked on ‘endog
enous politicians’, and (iii) the Chicago ‘regulation’
school, led by Stigler, Peltzman, Posner and also
Becker ( ).
However, a central theoretical breakthrough has
come from the work of trade theorists who have
systematically incorporated the analysis of DUP
activities in the main corpus of general equilibrium
theory.
The early papers that defined this general-
equilibrium-theoretic approach, and which were set
in the context of the theory of trade and welfare,
were: Bhagwati and Hansen ( ) which
analysed the question of illegal trade (that is, tariff-
evasion), Krueger ( ) which analysed
the question of rent-seeking for rents associated
with import quotas specifically and quotas more
generally, and (iii) Bhagwati and Srinivasan ( )
who analysed the phenomenon of
revenue-seeking, the ‘price’ counterpart of
Krueger’s rent-seeking, where a tariff resulted in
revenues which were then sought by lobbies.
The synthesis and generalization of these and
other apparently unrelated contributions, showing
that they all related to diversion of resources to
zero-output activities, was provided in Bhagwati (
) where they were called DUP activities.
The following significant aspects of the theoretical
analysis of DUP activities are noteworthy.
First, they are generally related to policy inter-
ventions (but they need not be: plunder, for
2906
instance, pre-dates the organization of govern-
ments). In so far as policy interventions induce
DUP activities, they are analytically divided into
two appropriate categories (Bhagwati and
Srinivasan ):
Category I: Policy-triggered DUP activities.
One class consists of lobbying activities. Examples
include: rent-seeking analysis of the cost of
protection via import licences (Krueger ); revenue-
seeking analysis of the cost of tariffs (Bhagwati and
Srinivasan ), of shadow prices in cost-benefit
analysis (Foster ), of price versus quantity
interventions (Bhagwati and Srinivasan ), of
non-economic objectives
(Anam ), of rank-ordering of alternative distorting
policies such as tariffs, production and consumption
taxes (Bhagwati et al. ), of the optimal tariff
(Dinopoulos ), of the transfer problem (Bhagwati et
al. ), and of voluntary export restrictions relative to
import tariffs (Brecher and Bhagwati ).
Another class consists ofpolicy-evading activ-
ities. Examples include: analysis ofsmuggling
(Bhagwati and Hansen ),its implication for
optimal tariffs (Johnson ; Bhagwati and Srinivasan
), and alternative modelling by
Kemp ( ), Sheikh ( ), Pitt ( ) and
Martin and Panagariya ( ).
Categoiy IP. Policy-influencing DUP activities.
The other generic class of DUP activities is not
triggered by policies in place but is rather aimed at
influencing the formulation of the policy itself. The
most prominent DUP-theoretic contributions in this
area relate to the analysis of tariffseeking. Although
Brock and Magee ( , )
pioneered here, the general equilibrium analyses of
endogeneous tariffs began with Findlay and Wellisz
( ) and Feenstra and Bhagwati
( ), the two sets of authors modelling the
government and the lobbying activities in
contrasting ways. Notable among the later contri-
butions are Mayer ( ), who extends the anal
ysis formally to include factor income- distribution
and therewith voting behaviour, and Wellisz and
Wilson ( ). Magee ( ) has an
excellent review of many of these contributions.
The implication of endogenizing the tariff for
conventional measurement of the cost of
Directly Unproductive Profit-Seeking (DUP) Activities
protection has been analysed in Bhagwati ( )
and Tullock ( ).
The choice between alternative policy instru-
ments when modelling the response of lobbies and
governments to import competition has also been
extensively analysed. The issue was raised by
Bhagwati ( ) and analysed further by
Dinopoulos ( ) and Sapir ( ) in terms of
how different agents (for example, ‘capitalists’ and
‘labour’) would profit from different policy
responses such as increased immigration of cheap
labour and tariffs when import competition inten-
sified. It has subsequently been explored more fully
by Rodrik ( ), who compares tariffs
with production subsidies.
Second, Bhagwati ( ) has noted, general-
izing a result in Bhagwati and Srinivasan ( ),
that DUP activities, while defined to be those that
waste resources in their direct impact, cannot be
taken as ultimately wasteful, that is, immiserizing,
since they may be triggered by a suboptimal policy
intervention. For, in that event, throwing away or
wasting resources may be beneficial. The shadow
price of a productive factor in such ‘highly
distorted’ economies may be negative. This is the
obverse of the possibility of immiserizing growth
(Bhagwati ). Thus, Buchanan ( ), who has
addressed the issue
of DUP activities and defined them as activities that
(ultimately) cause waste, has been corrected in
Bhagwati ( ): the definition of DUP activ
ities cannot properly exclude the possibility that
DUP activities are ultimately beneficial rather than
wasteful. This central distinction between the direct
and the ultimate welfare impacts of DUP activities
is now universally accepted. DUP activities are
therefore defined now, as in Bhagwati ( ) and
subsequent contributions,
as wasteful only in the direct sense.
Third, Bhagwati et al. ( ) have raised yet
another fundamental issue concerning DUP
activities. Thus, where DUP activities belong to
Category II distinguished above, full endogeneity of
policy can follow. If so, the conventional rank-
ordering of policies is no longer possible. We have
the determinacyparadox: policy is chosen in the
solution to the full ‘political- economy’, DUP-
theoretic solution and cannot be
Directly Unproductive Profit-Seeking (DUP) Activities
varied at will. These authors have therefore
suggested that, where full endogeneity obtains, the
appropriate way to theorize about policy is to take
variations around the observed DUP- theoretic
equilibrium. Thus, traditional economic parameters
such as factor supply could be varied; similarly now
the DUP-activity parameters such as, say, the cost
of lobbying could be varied. The impact on actual
welfare resulting from such variations can then be a
proper focus of analysis, implying a wholly dif-
ferent way of looking at policy questions from that
which economists have employed to date.
Finally, DUP activities are related to Krueger’s (
) important category of rent-seeking activi
ties. The latter are a subset of the former, in so far as
they relate to lobbying for quota-determined
scarcity rents and are therefore part of DUP
activities of Category II distinguished above
(Bhagwati ).
See Also
►
► " Vcc
Bibliography
Anarn, M. 1982. Distortion-triggered lobbying and welfare: A
contribution to the theory of directly- unproductive profit-
seeking activities. Journal of International Economics 13
(August): 15-32.
Becker, G.S. 1983. A theory of competition among pressure
groups for political influence. Quarterly Journal of
Economics 93: 371-400.
Bhagwati, J. 1980. Lobbying and welfare. Journal of Public
Economics 14: 355-363.
Bhagwati, J. 1982a. Directly-unproductive profit-seeking
(DUP) activities. Journal of Political Economy 90: 988-
1002.
Bhagwati, J. 1982b. Shifting comparative advantage, pro-
tectionist demands, and policy response. In Import
competition and response, ed. J. Bhagwati. Chicago:
Chicago University Press.
Bhagwati, J. 1983. DUP activities and rent seeking. Kyklos
36: 634-637.
Bhagwati, J., and B. Hansen. 1973. Theoretical analysis of
smuggling. Quarterly Journal of Economics 87: 172-187.
Bhagwati, J., and T.N. Srinivasan. 1973. Smuggling and trade
policy. Journal of Public Economics 2: 377-389.
2907
Bhagwati, J., and T.N. Srinivasan. 1980. Revenue-seeking: A
generalization of the theory of tariffs. Journal of Political
Economy 88: 1069-1087.
Bhagwati, J., and T.N. Srinivasan. 1982. The welfare con-
sequences of directly-unproductive profit-seeking (DUP)
lobbying activities: Price versus quantity distortions.
Journal of International Economics 13: 33-44.
Bhagwati, J., R. Brecher, and T.N. Srinivasan. 1984. DUP
activities and economic theory. In Neoclassical political
economy: The analysis of rent-seeking and DUP
activities, ed. D. Colander. Cambridge, MA: Ballinger &
Co.
Bhagwati, J., R. Brecher, and T. Hatta. 1985. The generalized
theory of transfers and welfare: Exogenous (policy-
imposed) and endogenous (transfer-induced) distortions.
Quarterly Journal of Economics 100: 697-714.
Brecher, R. and Bhagwati, J. 1987. Voluntary export
restrictions and import restrictions: A welfare-theoretic
comparison. Essays in honour of W.M. Corden, H.
Kierzkowski. Oxford: Basil Blackwell.
Brock, W., and S. Magee. 1978. The economics of special
interest politics: The case of the tariff. American Eco-
nomic Review 68: 246-250.
Brock, W., and S. Magee. 1980. Tariff formation in a
democracy. In Current issues in International Commer-
cial Policy and Diplomacy, ed. J. Black and B. Hindley.
New York: Macmillan.
Buchanan, J. 1980. Rent seeking and profit seeking. In
Towards a general theory of the rent-seeking society, ed.
J. Buchanan, G. Tullock, and R. Tollison. College Station:
Texas A&M University Press.
Buchanan, J., G. Tullock, and R. Tollison, eds. 1980.
Towards a general theory of the rent-seeking society.
College Station: Texas A&M University Press.
Dinopoulos, E. 1983. Import competition, international factor
mobility and lobbying responses: The Schumpeterian
industry cases. Journal of International Economics 14:
395—410.
Dinopoulos, E. 1984. The optimal tariff with revenueseeking:
A contribution to the theory of DUP activities. In The
neoclassical political economy: The analysis of rent-
seeking and DUP activities, ed. D. Colander. Cambridge,
MA: Ballinger & Co.
Feenstra, R., and J. Bhagwati. 1982. Tariff seeking and the
efficient tariff. In Import competition and
response, ed. J. Bhagwati. Chicago: Chicago University
Press.
Findlay, R., and S. Wellisz. 1982. Endogenous tariffs, the
political economy of trade restrictions, and welfare. In
Import competition and response, ed. J. Bhagwati.
Chicago: Chicago University Press.
Foster, E. 1981. The treatment of rents in cost-benefit
analysis. American Economic Review 71: 171-178.
Johnson, H.G. 1974. Notes on the economic theory of smug-
gling. In Illegal transactions in International Trade,
Series in International Economics, ed. J. Bhagwati.
Amsterdam: North-Holland.
Kemp, M. 1976. Smuggling and optimal commercial policy.
Journal of Public Economics 5: 381-384.
2908
Krueger, A. 1974. The political economy of the rent- seeking
society. American Economic Review 64: 291-303.
Lindbeck, A. 1976. Stabilization policies in open economies
with endogenous politicians. Richard Ely Lecture.
American Economic Review 66: 1-19.
Magee, S. 1984. Endogenous tariff theory: A survey. In
Neoclassical political economy: The analysis of rent-
seeking and DUP activities, ed. D. Colander. Cambridge,
MA: Ballinger & Co.
Martin, L., and A. Panagariya. 1984. Smuggling, trade, and
price disparity: A crime-theoretic approach. Journal of
International Economics 17: 201-218.
Mayer, W. 1984. Endogenous tariff formation. American
Economic Review 74: 970-985.
Pitt, M. 1981. Smuggling and price disparity. Journal of
International Economics 11: 447-458.
Rodrik, D. 1986. Tariffs, subsidies and welfare with endog-
enous policy. Journal of International Economics 21:
285-299.
Sapir, A. 1983. Foreign competition, immigration and
structural adjustment. Journal of International Economics
14: 381-394.
Sheikh, M. 1974. Smuggling, production and welfare.
Journal of International Economics 4: 355-364.
Tullock, G. 1981. Lobbying and welfare: A comment.
Journal of Public Economics 16: 391-394.
Wellisz, S. and Wilson, J.D. 1984. Public sector inefficiency,
a general equilibrium analysis. Discussion Paper No. 254,
International Economics Research Center, Columbia
University.
Director, Aaron (1901-2004)
Steven G. Medema
Keywords
Antitrust; Cartels; Director, A.; Law and eco-
nomics; Oligopoly; Posner, R.; Predatory pric-
ing; Resale price maintenance; Tie-in sales
JEL Classifications
B31
Aaron Director’s enduring contribution to eco-
nomics came via his role in the development of the
Chicago law and economics tradition. Director was
bom in Charterisk (in present-day Ukraine) in
Director, Aaron (1901-2004)
1901 and emigrated to the United States with his
family in 1913. He received his undergraduate
degree from Yale University and his graduate
training at the University of Chicago. Although he
came to Chicago in 1927 to work with Paul Douglas
on labour economics, it was Frank Knight and Jacob
Viner who, via their price theory courses, had the
greatest influence on him. Director remained at
Chicago as a graduate student and part-time
instructor until 1934. The 1930s were a heady
period at Chicago, where the student body included
George Stigler, Paul Samuelson (who credits
Director’s teaching with stimulating his interest in
economics), and Milton Friedman - each of whom
helped to reshape economic thinking in the middle
third of the 20th century - as well as Rose Director
(Aaron’s sister and, eventually. Rose Friedman).
Aaron Director was very much part of this milieu.
He left the University of Chicago for the US
Treasury Department in 1934 and, save for an
aborted attempt to complete a dissertation on the
history of the Bank of England, remained in
Washington, DC, until 1946, when he returned to
the University of Chicago to take up a position in
the Law School, where he remained until his
retirement in 1966.
Director’s appointment in the Law School was a
result of the efforts of Henry Simons, the first
economist on the law faculty at Chicago, and Frie-
drich Hayek, whose Road to Serfdom was published
in the United States largely because of Director’s
intervention with the University of Chicago Press.
The plan, as laid out by Simons, was for Director to
head up the ‘Free Market Study’, a Volker
Fundfinanced project, housed in the Law School
and dedicated to undertaking ‘a study of a suitable
legal and institutional framework of an effective
competitive system’ (Coase , p. 246). However,
Simons committed suicide in the summer of 1946,
and Director was asked to take on Simons’s basic
Law School price theory course, ‘Economic
Analysis ofPublic Policy’. This provided Director
with an initial forum for bringing the perspective he
had learned from Knight and Viner into the Law
School’s teaching programme.
The transition from having an economist on the
Law School faculty to the establishment of a law
and economics tradition at Chicago began not
Director, Aaron (1901-2004)
long after this, when Edward Levi invited Director
to collaborate in the teaching of the antitrust course.
Levi would teach a traditional antitrust course for
four days each week; Director would then come in
on the fifth day and, using the tools of price theory,
show that the traditional legal approach could not
stand up to the rigours of economic analysis. The
basic pattern was very simple: Director would ask
whether the practice in question was, in general,
consistent with monopolistic profit maximization.
The answer was often negative, which meant that
there had to be some sort of legitimate rationale for
the supposedly anti-competitive practice in
question. What Director’s price theory showed was
that the ‘simple and obvious’ answers were often
wrong- headedly simplistic. This process had a
profound impact on students and colleagues alike.
Director’s antitrust students - a group that included
Robert H. Bork, Ward Bowman, Kenneth Dam,
Edmund Kitch, Wesley J. Liebeler, John S. McGee,
Henry Manne, and Bernard H. Siegan - have often
spoken of the ‘conversion’ they experienced in this
class, and even Levi himself became a partial
convert (see Kitch ; Director and Levi ). What was
perhaps Director’s most significant contribution on
the missionary front came after his retirement, when
he and Richard Posner spent time together at
Stanford in 1968 - Posner’s first year on the
Stanford Law School faculty. It was Director who
taught Posner to think like a Chicago economist,
introduced him to Stigler and Ronald Coase, and in
this and other ways was instrumental in Posner’s
move to the Chicago Law School after only one
year on the Stanford faculty. The rest, as they say, is
history.
Although Director’s published output was slight,
his influence extended well beyond the classroom.
His insights made their way into the antitrust
literature - and, eventually, antitrust policy -
through the writings of students and colleagues, as
Sam Peltzman ( ) has detailed.
Director’s primary legacies are in the analysis of
predatory pricing (via McGee ), resale price
maintenance (via Telser ), and tie-in sales
(see Director and Levi ; Bowman ;
Bumstein ), but his influence was also prominent in
Stigler’s view of oligopoly and antitrust
2909
policy, Posner’s ( ) perspective on oligopoly
and cartels, and Robert Bork’s influential articles on
antitmst (for example, Bork and Bowman ; Bork ).
These contributions coalesced in a distinctive
Chicago approach to antitmst analysis, an approach
that Herbert Hovenkamp ( p. 1020) says ‘has done
more for antitmst policy than any other coherent
economic theory since the New Deal’, and whose
influence is inescapable.
Director’s impact at the Law School went far
beyond antitmst: He was also the prime mover in
the early professionalization of law and economics.
Director formally established the nation’s first law
and economics programme, which maintained
visiting fellowships for law and economics scholars,
and, in 1958, founded the Journal of Law and
Economics. Within a few decades, Director’s efforts
at Chicago had been replicated in a set of thriving
and well-funded law and economics programmes at
major law schools around the country. One would
be hard pressed to name an individual in our
discipline who has had as much influence as
Director without a much more extensive
bibliography.
See Also
► mti-trust Enforcement
► Chicago School
► ,aw, Economic Analysis of
Selected Works
1931. (With P. Douglas.) The problem of unem-
ployment. New York: Macmillan.
1933. The economics of technocracy. Public Policy
Pamphlet No. 2. Chicago: University of Chicago
Press.
1940. Does inflation change the economic effects of
war? American Economic Review 35,351-361.
1951. (With E. H. Levi.) Law and the future: Trade
regulation. Northwestern Law Review 51,281-
296
1964. Parity and the economic marketplace. Jour-
nal of Law and Economics 7, 1-10.
2910
Bibliography
Bork. R.H. 1967. Antitrust and monopoly: The goals of
antitrust policy. American Economic Review 57: 242-253.
Bork. R.H.. and W.S. Bowman Jr. 1965. The crisis in antitrust
Columbia Law Review 65: 363-376.
Bowman. W.S. Jr. 1957. Tying arrangements and the leverage
problem. Yale Law Journal 67: 19-36.
Bumstein. ML. 1960. The economics of tie-in sales. Review
of Economics and Statistics 42: 68-73.
Coase. R.H. 1998. Director, Aaron. In The newpalgrave
dictionary of economics and the law. London: Macmillan.
Hovenkamp, H. 1986. Chicago and its alternatives. Duke Law
Journal 1986: 1014-1029.
Kitch, E.W. 1983. The fire of truth: a remembrance of Law
and Economics at Chicago. 1932-1970. Journal of Law
and Economics 26: 163-233.
McGee, J.S. 1958. Predatory price cutting: The standard oil
(N.l) case. Journal of Law and Economics 1: 137s-1369.
Peltzman, S. 2005. Aaron Director's influence on antitrust
policy. Journal of Law and Economics 48: 313-330.
Posner, R.A. 1969. Oligopoly and the antitrust laws: A
suggested approach. Stanford Law Review 21: 1562-1606.
Stigler, S.M. 2005. Aaron Director remembered. Journal of
Law and Economics 48: 307-311.
Telser, L.G. 1960. Why should manufacturers want fair trade?
Journal of Law and Economics 3: 86 105.
Discrete Choice Models
Takeshi Amemiya
These are those statistical models which specify the
probability distribution of discrete dependent vari-
ables as a function of independent variables and
unknown parameters. They are sometimes called
qualitative response models, and are relevant in
economics because the decision of an economic unit
frequently involves discrete choice: for example, the
decision regarding whether a person joins the labour
force or not, the decision as to the number of cars to
own, the choice of occupation, the choice of the
mode of transportation, etc.
Despite their relevance, however, it is only
recently (approximately in the last twenty years)
that economists have started using them
Discrete Choice Models
extensively. There seem to be three reasons for a
recent surge of interest in such models: (1) Econ-
omists have realized that econometric models using
only aggregate data cannot accurately explain
economic phenomena nor predict the fiiUire values
of economic variables well. (2) Large scale
disaggregated data on consumers and producers
have become available. (3) The rapid development
of computer technology has made possible
estimation of realistic models of this kind.
Note that when aggregated over many individ-
uals, discrete variables behave almost like continu-
ous variables and therefore can be subjected to
standard regression analysis. A discrete choice
model becomes necessary when we want to model
the behaviour of an individual economic unit.
As econometric applications of these models
have increased, we have also seen an increase of
theoretical papers which address the problem of
their specification and estimation. Biometricians
have in fact used such models longer than have
econometricians, using them, for example, to
analyse the effect of an insecticide or the effect of a
medical treatment. However, since the versions that
econometricians use arc generally more complex
than those used by biometricians, it has been
necessary for the former to develop new models and
new methods of statistical inference.
There are cases where a discrete decision of an
economic unit is closely interrelated with the deter-
mination of the value of a continuous variable. For
example, a decision to join the labour force neces-
sitates the decision of how many hours to work and
at what wage rate. A decision to buy a ear cannot be
separated from the decision of how much to spend
on a car. The joint determination of the values of
discrete variables and continuous variables belongs
to the topic of limited dependent variables.
Other closely related topics are Markov chain
models and duration (or survival) models. T hese
models introduce the time domain into discrete
choice models thereby making the models dynamic.
In Markov chain models time changes discretely,
whereas in duration models time moves
continuously.
Discrete Choice Models
Those who wish to study the subject in more and
detail than the present entry are referred to
Amemiya ( , ), Maddala ( ), and
McFadden ( ).
Univariate Binary Models
Model Specification
The simplest type of a discrete choice model is a
univariate binary model which specifies the binary
(1 or 0) outcome of a single dependent variable. Let
y, be the zth observation on the binary dependent
variable and x, the rth observation on the vector of
independent variables. Then a general univariate
binary model is defined by
P(yi = l)=F(x'j), i = 1,2, ...
(I)
where P stands for probability, F is a particular
distribution function, and jS is a vector of unknown
parameters. For example, the event yt = 1 may
signify that the zth individual buys a car and the
elements of the vector x, may include the income of
the rth individual and the price of the car the indi-
vidual must pay if he decides to buy a car.
Note that we have assumed the argument of F in
( ) to be a linear function of the independent
variables. As in the linear regression model, this
linearity assumption is more general than appears at
first, because x, need not be the original economic
variables like income and price, but instead could
contain various transformations of the original
variables. However, the model in which the function
F depends on a nonlinear function of the
independent variables and unknown parameters can
be handled with only a slight modification of the
subsequent analysis.
A vari ety of models arises as we choose dif-
ferent distribution functions for F. The most
commonly used functions are the standard normal
distribution function <P and the logistic distribution
function A. These functions are defined by
»X 1/2
<P(x) (2n) exp(—2_1Z2)df
J—
oo
2911
A(x) = (\+e-x)-1
When F = F, the model is called the probit
model, and when F = A, it is called the logit model.
The decision regarding which function to use
should be based both on theoretical considerations
and on how well a model fits the data. However, as
long as a researcher experiments with various
independent variables and with various ways in
which the independent variables appear in the
argument of F, the particular choice of F is not
cmcial.
Let us consider by way of an example how this
model arises as the result of an individual
,n, maximizing a utility function. Consider the decision
of a person regarding whether he drives a car to
work or travels by public transport. We suppose that
a level of utility is associated with each alternative
and the person is to choose the alternative for which
the utility is greater. Let Un and Ui0 be the zth
person’s utilities associated with driving a car and
travelling by public transport respectively. We
assume that they are linear functions of independent
variables with additive error terms as follows:
Un = x'nfii + fiiii
and
UiO = XfiPo + £ , 0 -
Here, the vector xtl may be thought of as consisting
of the time and the cost which would be incurred if
the zth person were to drive a car, plus his socio-
economic characteristics. The error term may be
regarded as the sum of all the unobserved
independent variables. Defining y, 1 if the zth
person travels by car and y, 0 if he travels
otherwise, we have
^ , = 1 ) =P(Un>Ui0)
= Fulfil -x'i0Po),
2912 Discrete Choice Models

where F is the distribution function of ei0 - an. Thus, Xi xt. Define n, = number of integers contained in 7t
a probit model will result from the normality of e;0 - and Pt - n, /yi. Then, by a Taylor
e,i. The normality may be justified on the ground of expansion, we have approximately
a central limit theorem.
If a probit model fits the data well, so will a logit
model because the logistic distribution function is
similar to the standard normal distribution function.
where F~ 1 denotes the inverse function of F. The
Estimation MIN x2 estimator fi is the weighted least squares
Let us consider the estimation of the parameter estimator applied to this last heteroscedastic
vector (3 in the model ( ). We shall first discuss the regression equation; that is,
maximum likelihood (ML) estimator and second,
the minimum chi-square (MIN /’) estimator. n -i
The likelihood function based on n independent P £ wxx TwtXtF 1
t t t
t=\
binary observations yy, y2, ..y„ is given by t=i

where

VV, = «,/2[L 1 (P,) ] / [P, ( l P,) .

i=1
The MIN x2 estimator has the same asymptotic
The ML estimator /> is obtained by maximizing distribution as the ML estimator. Its advantage over
In L. Under general conditions fi is consistent and the latter is computational simplicity, while its
asymptotically normal with the asymptotic variance- weakness is that it requires many observations for
covariance matrix given by each value of the independent variables. The
required number of observations increases with the
number of the independent variables. If an
independent variable takes many values it may be
f2(W) necessary to group the values into a small number of
yf=\E * W ) [ 1 -F(W)\ groups in order to define the MIN x 2 estimator. But
i=1 such a procedure will introduce a certain bias to the
estimator.
where/is the derivative of F.
Since an explicit formula for the ML estimator
cannot be obtained for this model, the calculation of
Multinomial Models
the estimator must be done by an iterative method.
The log likelihood function can be shown to be A multinomial model is a statistical model for
globally concave in the probit and logit models. In independent discrete variables, some of which take
these models, therefore, a standard iterative more than two values: Supposing that y,- takes in, +
algorithm such as the Newton-Raphson method will 1 integer values 0, 1,..., m the model is defined by
h
generally converge to the global maximum. specifying the N) (/«,■ probabilities:
The MIN x estimator, first proposed by Berkson
( ) for tiie logit model, works only if there are
many observations on y for each of the values taken P{yi = j) = Fij{x, P), i = 1,2, ..., n j=
by tiie vector x. Let us suppose that .r, takes T 1,2, ... ,m,-.
vector values x\, x2,..., xT and classify integers 1,
2,..., n into Tdisjoint sets ij, /2,..., // by the rule: i C /, Note that P fy, = 0) need not be specified
if because the sum of the m, + 1 probabilities is
Discrete Choice Models 2913

one for each i. It is important to let m depend on i F, the last equation leads to an ordered model
because the number of alternatives available to defined by
different individuals may differ.
Defining | {nij + 1) binary variables p< F a F a X
Ji = ./ ) = ( j+1 - W) - ( j - ',P) ■

As in the binary case, the choice of <P and A

yij = i if yi=j for F is most frequently used.
= o if yt ^j, i = 1,2, . ,.,n j =0,1, An ordered model is attractive because of its
simplicity. However, in many economic applica-
the likelihood function of the model can be written tions it may be an oversimplification to assume that
as the outcome of a multinomial variable can be
n mi yij
completely determined by the outcome of a simple
l
rniw continuous variable. For example, for owning cars
!=i;=o it is probably more realistic to assume that the rth
person owns j cars if > U,/; for all k / j, where Ujj is
Note that this reduces to the L equation of
the utility that accrues to the /th person if he owns j
Section “ ” if m, = 1 cars. In this case m continuous variables Ujj — Ujj
for all i. + i,y = 0, 1,..., m, determine the outcome of the
The ML estimator of /> is consistent and discrete variable.
asymptotically normal with its asymptotic variance- A multinomial model which is not an ordered
covariance matrix given by model is called an unordered model. The models
discussed in the next parts of this section are all
unordered.

V$ = ' d2iogU\ Multinomial Logit Model

dpdp'\ ’ A multinomial logit model is described below by
defining the probabilities of the rth individual who
which will be equal to Vfi equation in Section “ faces three alternatives 7 = 0, 1, and 2. A gener-
” in the binary alization to the case of more alternatives can be
case. The MIN %2 estimator can be also defined for easily inferred. The three probabilities are given by
the multinomial model, although the definition will
not be given here.
P{yi = 2) = D lexp(x'2/i)
Ordered Models
Piy. = 1) = D 'exp(x'|/i),
An ordered multinomial model arises when there is
an unobserved continuous random variable y* P(yt = 0) = D - \
which determines the outcome ofy, by the rule where D 1 I exp + exp(x'2/i).
McFadden ( ) showed how a multinomial
logit model can be derived from the maximization
yt = j if and only if x, < y* < ocj+u j =0,1 xo = of stochastic utilities. Suppose that the rth individ-
-oo, am+i = oo. ual’s utility Uj associated with the j th alternative is
the sum of the nonstochastic part /r y and the
Such a rule may be appropriate, for example, if yi = stochastic part i= and that the individual chooses
the alternative for which the utility is a maximum.
j signifies the event that the rth individual owns j
Suppose further that e;0, e,i and e,2 are independent
cars and y* refers to a measure of the intensity of
and identically distributed according to the
the rth individual’s desire to own cars. If the
distribution function exp[ exp( s)] - called the
distribution function of by y] x'fi is
2914
type I extreme value distribution. T h e n we
can show
P(yt = 2 ) = P{Ua > UiU Ui2 > Ui0)
= exp(pn) /[exp(^0) + exp(pn)
+ e x p (pi2)],
and similarly for P (y, = 1) and P (y, = 0). Thus, the
model defined by three equations above follows
from putting pa — pi0 = xJi2fi and pn pi0
= *'nP-
The multinomial logit model has been exten-
sively used in economic applications, such as the
choice of modes of transportation, the choice of
occupations, and the choice of types of appliances.
The likelihood function of the model can be shown
to be globally concave; consequently, the ML
estimator can be computed with relative ease.
A major limitation of the multinomial logit
model lies in its independence assumption. Consider
the choice of transportation modes and suppose first
that the alternatives consist of car, bus, and train.
Then the assumption of independent utilities may be
reasonable. Next, to use McFadden’s famous
example, suppose instead that the choice is among a
car, a red bus, and a blue bus. Then it is clearly
unreasonable to assume that the utilities associated
with the red bus and the blue bus are independent. In
the next subsection we shall consider a multinomial
model which corrects this deficiency.
Nested Logit Model
We continue the last example. Let Uj = p, +
/ 0, I and 2, be the utilities associated with car,
red bus, and blue blus, respectively. (The subscript i
is suppressed to simplify notation. ) Following
McFadden ( ), suppose that e0 is
distributed according to the type I extreme value
distribution and independent of £i and e2 and that the Multivariate Models
joint distribution of and e2 is given by
P(o, ei) = exp — [exp^/U'e/)
+ [exp(-p~1£2)]P,0 < p < 1.
Discrete Choice Models
This distribution is called Gumbel s type B
bivariate extreme value distribution. The corre-
lation coefficient is 1 — p2, and if/) = I (the case of
independence), ::2) becomes the product of two type
I extreme value distributions.
Under these assumptions it can be shown that
P(y = 0) = exp(p0) /{exp(p0) + exp^-'pi) +
exp(p~V2)f}
and
P(y= ib^o)
= exp(p Vi)/[exp(p Vi) +exp(p V2)].
The other probabilities can be deduced from the
above. Note that the last equation shows that the
choice between red bus and blue bus is made
according to a binary logit model, while the pre-
vious equation shows that the choice between car
and noncar is also like a logit model except that a
certain weighted average of exp(/q)and cxp(/r 2) is
involved.
Multinomial Probit Model
A multinomial probit model is derived from the
assumption that the utilities Ui0, Un,..., Uimi are
multivariate normal for every i. Its advantage is that
general assumptions about the correlations among
the utilities are allowed. Its major disadvantage is
that the calculation of the choice probability
requires the evaluation of multiple integrals of joint
normal densities, which is feasible only for a small
number of alternatives.
A multivariate discrete choice model specifies the
joint probability distribution of two or more discrete
dependent variables. For example, the joint
distribution of two binary variables jq and v 2 each
of which takes values 1 or 0 is determined by the
four probabilities Pjk = P (jq = j, y2 = k), j, k = 0,1.
(Of course, the sum of the probabilities must be
equal to 1. )
Discrete Choice Models
A multivariate model is a special case of a
multinomial model. For example, the model of two
binary variables mentioned in the preceding
paragraph may be regarded as a multinomial model
for a single discrete variable which takes four
values with probabilities Pn, P10, P01, and Poo-
Therefore, all the results given in section “ ”
apply to multivariate
models as well. In this section we shall discuss three
types of models which specifically take into account
the multivariate feature of the model.
Log-Linear Model
A log-linear model refers to a particular parame-
terization of a multivariate discrete choice model. In
the previous bivariate binary model, the log-linear
parameterization of the four pro- babilitites is given
as follows:
Pn = £>-1exp(ai + a2 + an),
Pio = D_1exp(ai),
Poi = £»_1exp(a2),
and
Poo=D-\
(III)
where D = 1 + exp(ai) + exp(a2)
+exp(ai + a2 + “12)
There is a one-to-one correspondence between
any three probabilities and the three a parameters of
the log-linear model; thus, the two
parameterizations are equivalent. An advantage of
the log-linear parameterization lies in its feature that
7,2 = 0 if and only if Vj and >’2 are independent.
Equations ( ) may be represented by the fol-
lowing single equation:
P(y\, y2) exp^yj + “2^2 + ai^iTa)-
Each equation of ( ) is obtained by inserting
values 1 or 0 into y\ and y2 in this equation. This
formulation can be generalized to a log-linear
2915
model of more than two binary variables. The case
of three variables is given below:
exp(aiy] + a2y2 + ^3 +
+ + a23.y2.y3
+ 0Cl23A’lA2>'3)-
The first three terms in the exponential function
are called the main effects. Terms involving the
product of two variables are called second- order
interaction terms, the product of three variables
third-order interaction terms, and so on.
Note that the last equation has seven parameters,
which can be put into one-to-one correspondence
with the seven probabilities that completely
determine the distribution ofyi, y 2, and y3- Such a
model, without any constraint among the param-
eters, is called a saturated model. Researchers often
use a constrained log-linear model, called an
unsaturated model, which is obtained by setting
some of the higher-order interaction terms to zero;
e.g. Goodman ( ). See also Nerlove and
Press ( ) for an example of a log-linear model
in which some of the a parameters are specified to
be functions of independent variables and unknown
parameters.
Multivariate Nested Logit Model
The multivariate nested logit model is a special case
of the nested logit model discussed in section “ ”,
which is useful
whenever a set of alternatives can be classified into
classes each of which contains similar alternatives.
It is useful in a multivariate situation because the
alternatives can be naturally classified according to
the outcome of one or more of the variables.
For example, in the bivariate binary case, the
four alternatives can be classified according to
whether y\ 1 or 0. Let Ujk be the utility associated
with the choiceyi = j andy2 = Kj< k = 0, 1 and
assume as before that Ujk = njk + e.jk, where /ris are
nonstochastic and e’s are random. As a slight
generalization of the Gumbel distribution in section
“ ” assume
that
2916
F
(fi/1, fi/o) = aj exp { - [exp (~Pj 1 £jl)
+ exp^-p71£y2^]«}j j — i,o
and that (en, £10) are independent of (£0i, £oo>- Then
the resulting multivariate nested logit model is
characterized by the following probabilities:
P{y\ = 1) = «i [exp(prVn) + exp(prVio)]Pl
[exp(pfVn) +exp(pfV10)]pl
+ a0[exp(po Voi) + expfpoVoo)]''0},
P{yi = Ibi = !) = exp(pfVn)/
[exp(prVn) +exp(p]-V10)],
p
{yi = Ibi = 0) = exp(po Voi)/
[exp(p0Vtoi) +exp(po Voo)]-
We may further specify pjk = W-
Multivariate Probit Model
This model is conceptually different from the
models of the preceding two sections in that here
the marginal probabilities are specified first and the
joint probabilities are then defined in a certain
natural way.
As an example of a bivariate binary probit
model, let us suppose 1), / 1, and
2, and vj is unobservable and its value determines
the value of the observable binary variable y, bythe
rale
yj = i if yj > o
= 0 otherwise.
This rale determines the marginal probabilities
P(yi = 1) = <P(fij),j= 1 and 2.
Thus, the model will be complete when we specify
the joint probability P (y\ b J;2 = 1). A natural way characterize P (ijx) are not related to the
to specify it would be to assume
Discrete Choice Models
that yj and y\ are jointly normal with a correlation
coefficient p and define
P(y = hy2 = i) = P(y\ > Puy*2 > Pi)-
Usually, a researcher will further specify p \ =
x\ fj and p2 = x2p an(i estimate the unknown
parameters [1 and p; see Morimune ( ) for an
econometric example of this model.
A bivariate logit model may be defined simi-
larly. But, unlike the probit case, there is no natural
choice among many bivariate logistic distributions
with the same marginal univariate logistic
distributions.
Choice-Based Sampling
In models () or ( ), the independent variables x, were
treated as known constants. This is equivalent to
considering the conditional distribution of y\ given
x,. This practice was valid because it was implicitly
assumed that y, and x, were generated according to
either random sampling or exogenous sampling.
Under random sampling, y and x are sampled
according to their true joint distribution P (y\x) /(x).
Thus the likelihood function denoted ZR, is given by
L
R = n^Cy/k/)/(■*/)■
i=\
Under exogenous sampling, a researcher sam-
ples x according to a certain distribution g (x),
which may not be equal to the true distribution /(x)
of x in the total population, and then samples y
according to its true conditional probability P (jjx).
Thus the likelihood function, denoted LB, is given by
n
L
E = n^Cy/ki) s{xi)-
i= 1
In either case, as long as the parameters that
Discrete Choice Models 2917

parameters that characterize / (x) or g (x), the attain a higher efficiency of estimation by sampling
maximization of LR or LE is equivalent to the bus riders at a bus depot to augment the data
maximization of gathered by random sampling.
An interesting problem in choice-based sam-
pling is how to determine H (y) to maximize the
l
=’ efficiency of estimation. Although there is no clear-
11
cut solution to this problem in general, it is
which is equivalent to the L of Section “ expected that if Q (j) is small for some j then the
Under choice-based sampling, a researcher value of// (/) which is larger than Q (j) will yield a
samples y according to fixed proportions H (y), and more efficient estimator than the value of H (j)
then, given y, samples x according to the which is equal to Q (j). Note that if in the formula
conditional density /fx[y). By the formula of con- for Lc, H(j) = Q (j) for every j, then Lc is reduced to
ditional density, Lr.

My) =P(y\x)f(x)/Q(y),
Distribution-Free Methods
where Q (v) = Ex P (y|x), and Ex denotes the
expectation taken with respect to the random vector Consider the univariate binary model (). There, we
x. Thus, the likelihood function under choice- based assumed that the function F( ) is completely
sampling, denoted Lc, is specified and known. Recently, Manski ( )
and Cosslett ( ) have shown how to estimate
n
L
c=n I jl consistently (subject to a certain normalization)
i= 1 without specifying F(-).
Manski’s estimator is based on the idea that as
Unlike random sampling or exogenous sam-
long as F satisfies the condition /TO) = 0.5, one can
pling, choice-based sampling requires new analysis
predict y\ to be 1 or 0 depending on whether x-f) is
because the maximization of Lc is not equivalent to
positive or negative. His estimator of fj is chosen so
the maximization of L on account of the fact that Q
as to maximize the number of correct predictions. If
(y) depends on the same parameters that
we define the characteristic function 7 of the event
characterize P (y|x).
E by
In particular, it means that the standard ML
estimator which maximizes L is not even consistent
%{E) = 1 if £ occurs
under choice-based sampling. The reader should
consult Amemiya ( ) or Manski and = 0 otherwise,
McFadden ( ) for the properties of the choice-
based sampling ML estimator which maximizes Lc the number of correct predictions can be mathe-
in various situations. matically expressed as
Choice-based sampling is useful when only a
small number of people sampled according to
S
random sampling are likely to choose a particular (P) = [yixttP-0) + -y^mfi < °)]-
i=1
alternative. For example, in a transportation study,
random sampling of individual households in a
Manski calls this the score function - and hence
community with a small proportion of bus riders
his estimator the maximum score estimator. The
may produce an extremely small number of bus
estimator has been shown to be consistent, but its
riders. In such a case a researcher may be able to
asymptotic distribution is unknown.
Cosslett proposed maximizing the likelihood
function L in Section “ ” with respect to
2918 Discriminating Monopoly

both [1 and F, and called his estimator the gener- Morimune, K. 1979. Comparisons of normal and logistic
models in the bivariate dichotomous analysis.
alized ML estimator. For a given value of /J, the
Econometrica 47: 957-976.
value of F which maximizes that L is a step Nerlove, M., and S.J. Press. 1973. Univariate and multi-
function, and Cosslett showed a simple method of variate log-linear and logistic models, R-1306-EDA/
determining it. Finding the optimal value of (i, NIH. Santa Monica: Rand Corporation.
however, is the computationally difficult part. Like
the maximum score estimator, the generalized ML
estimator of [> is consistent but its asymptotic
distribution is unknown. Discriminating Monopoly

John M. Hartwick
See Also

► Censored Data Models

► Labour Supply of Women
► Limited Dependent Variables
► Logit, Probit and Tobit
► Selection Bias and Self-Selection
Bibliography
Amemiya, T. 1981. Qualitative response models: A survey.
Journal of Economic Literature 19: 1483-1536.
Amemiya, T. 1985. Advanced econometrics. Cambridge,
MA: Harvard University Press.
Berkson, J. 1944. Application of the logistic function to
bioassay. Journal of the American Statistical Association
39: 357-365.
Cosslett, S.R. 1983. Distribution-free maximum likelihood
estimator of the binary choice model. Econometrica 51:
765-782.
Goodman, L.A. 1972. A modified multiple regression
approach to the analysis of dichotomous variables.
American Sociological Review 37: 28-46.
Maddala, G.S. 1983. Limited-dependent and qualitative
variables in econometrics. Cambridge: Cambridge
University Press.
Manski, C.F. 1975. The maximum score estimation of the
stochastic utility model of choice. Journal of Econo-
metrics 3: 205-228.
Manski, C.F., and D. McFadden (eds.). 1981. Structural
analysis of discrete data with econometric applications.
Cambridge, MA: MIT Press.
McFadden, D. 1974. Conditional logit analysis of qualitative
choice behavior. In Frontiers in econometrics, ed. P.
Zarembka, 105-142. New York: Academic Press.
McFadden, D. 1977. Qualitative methods for analyzing travel
behavior of individuals: Some recent developments.
Cowles Foundation Discussion Paper No. 474.
McFadden, D. 1984. Econometric analysis of qualitative
response models. In Handbook of econometrics, vol. 2,
ed. Z. Griliches and M.D. Intriligator, 1385-1457.
Amsterdam: North-Holland.
In The Wealth of Nations, Adam Smith refers to two
instances of price discrimination. In Book V,
Chapter I, Part III, he ruminates on the problem of
finding the best set of levies for toll roads and
commends the practice of charging for luxurious
carriages more than for working men’s wagons even
though the vehicles are of the same weight. He
suggests that the rich can subsidize the poor by this
tariff scheme. In Book IV, Chapter V, he notes that
some groups of producers have sold their produce
abroad at lower prices than at home. He views this
as cross-subsidization and deplores the high prices
which he sees as resulting in the domestic market
Smith’s first problem, how to set tolls, has occupied
economists to this day, although the solution was
laid out in principle by Dupuit ( )
and with considerable precision by Edgeworth (
): let each user’s levy in excess of his or her
marginal cost of usage (which may be zero on a toll
bridge) be proportional to his or her intensity of
preference as expressed by his or her elasticity of
demand. Edgeworth in fact worked out details of
two sorts of price discrimination - that practised by
a private profit-maximizing monopolist and that
practised by a ‘state monopoly’ interested in raising
Z dollars of profit from the users of the monopoly
while at the same time reducing welfare as little as
possible. The solution to this state monopoly or
public utility pricing problem we refer to today as
Ramsey pricing (Ramsey , who attributes the idea
for his paper to Pigou).
Smith’s second instance of price discrimination
can most usefully be viewed as the case of a
monopolist selling at distinct prices in two
Discriminating Monopoly
separate markets, domestic and foreign, with dis-
tinct demand curves. Barone ( , pp. 291-2)
analysed it from this perspective diagrammati- cally
and Yntema ( ) filled in the algebraic
details.
Pigou ( ) presented his synthesis of results
and introduced the terms first, second and third
degree price discrimination - degrees referring to
the fineness with which separate prices can be
assigned to separate units demanded of the
monopolist. He graphically worked out the two
market case with linear demands in his Appendix
III, pointing out that given two markets, only one
might be served under uniform pricing whereas
both might be served under price discimination. In
1904 he had in fact independently of Dupuit
analysed what we now call perfect price discrim-
ination, or the situation in which each unit produced
by a monopolist fetches a different price, each of
which being bounded above by the buyer’s
willingness to pay. In her synthesis, J. Robinson (
, p. 205) asked whether the
monopolist would produce more under third degree
price discrimination relative to his output under a
uniform price.
The toll-setting problem turns on the fact that if
users were charged commensurate with the wear
and tear they cause (marginal cost pricing), insuf-
ficient revenue would generally be raised to cover
the cost of building on obviously desirable road,
bridge, railroad, telephone network etc. For exam-
ple, the wear and tear caused by the marginal bridge
user is approximately zero and thus no revenue
would be raised by charging according to costs of
usage. Dupuit realized that an individual’s
willingness to pay for a trip could far exceed his or
her incremental cost of usage and suggested
collecting revenue on the basis of each individual’s
maximum willingness to pay. The revenue so
collected ‘would not have the slightest relation to
the costs of production’ (p. 271) but would reflect
the total utility in dollars per day’s use of the
project. This is price discrimination: each user pays
generally a different price for the same service.
To sharpen his exposition Dupuit turned to
shipping tons of ore across a bridge. At high prices
obviously fewer tons will be shipped, since buyers
2919
of the ore will be obliged to absorb the charges and
will demand less at high prices inclusive of delivery
charges. Dupuit discussed the hypothetical case of
each additional ton crossing the bridge ‘paying’ a
slightly lower toll evaluated at the maximum
willingness to pay for the ton in question. This is the
case of perfect price discrimination and a variant is
practised in the form of block pricing. Firms
occasionally sell the first say 1000 bricks at $3 each,
the second thousand at $2.50 each, the third
thousand at $2.25 and so on. Robinson reflected on
the issue of the monopolist brick seller selecting the
break points (1000 bricks, 2000 bricks etc.)
simultaneously with price per brick in order to
maximize profits. One can see that perfect price
discrimination is a procedure for transferring
consumer surplus (the area under an individual’s
demand curve up to the quantity consumed less the
amount actually paid) to the seller of the product.
Price discrimination is practised by a monopolist
because it permits profits to rise above what they
would be if a single or uniform price were charged.
To see this suppose that in two separate markets the
monopolist were practising price discrimination and
maximizing aggregate profit. If he were now
obliged to sell in both markets at a single uniform
price his optimand can never rise since the single
price represents a new constraint on his pursuit of
maximum profit. Pursuing this case in more detail,
let Q\(p) be the demand curve for gadgets by
citizens abroad (or for return rail car trips for wheat
shippers on a line) and Q2(p) the demand curve for
gadgets from local people (or for return rail car trips
for potash producers located in the wheat farming
area). Then the monopolist’s profit under price
discrimination is 71 = PiQiiPi) + PiQiiPi) - C(Qi +
Q2) where C( ) is total cost, increasing and convex in
Q = Qi + Qi, and pt is the price in market i with
quantity sold Qi(p,). Profits attain a maximum when
Cq = pi[ 1 + (1/ £:;)], where
Cn=clC/clQ and et = ^ • — < -1 d Pi Qi
is the elasticity of demand in market i. This profit-
maximizing condition is referred to as the
2920
Robinson-Yntema condition and was first set out by
Edgeworth ( ). The left-hand side is mar
ginal cost and the right-hand side is the marginal
revenue in market i. For <?i|§ 6i\,Pi^ Pi- For n > 2,
the analysis is the same. (Edgeworth made Cq a
constant at c, defined his elasticity as
dQ (Pi ~ c)
d
O i -c) Q
and arrived at his ‘equal elasticity condition’ for
profit-maximizing monopoly price discrimination.)
A monopolist forced to sell at a single price
(presumably because the product can be readily
resold) will maximize profit when Cq = p{l +
[l/(ffli ei + m2 e2)]}, where to,- = QJQ.
Edgeworth investigated when deviations in pi
and p2 from a uniform p would increase welfare
(consumer surplus), while Robinson argued that
price discrimination would raise Q from the level
corresponding to a uniform profit-maximizing price
if the more elastic demand curve is concave and the
less elastic demand curve is convex.
The basic first order condition for monopoly
price discrimination can be written as
AQi _ AQ2 _
Qi Qi ~ '
where
[Pi -C Q ] dQi
A Qi
d Pi '
This illuminating formula indicates that each
output would to a first approximation rise propor-
tionately if there were no monopoly and no price
discrimination, and it can orient one’s intuition in
viewing Robinson’s result on concavity and con-
vexity of demand schedules. Schmalensee (
and Varian ( ) have pointed out that a neces
sary condition for total net consumer surplus to rise
as the monopolist switches from a uniform price to
profit-maximizing price discrimination is that there
be a rise in total output delivered.
The public utility or state monopoly problem in
price discrimination is to raise Z dollars of profit
Discriminating Monopoly
by charging diverse prices to distinct customers
while reducing welfare least. These profits might be
assigned to cover the fixed costs of a public facility.
Let Blip) and B2(p) be the areas under each demand
curve for price p. The state monopoly pricing
problem is to maximize W = Bxipx) + B2(p2) - C(Qi
+ Q2) subject to Z =PiQi + P2Q2 - C(Qi + Qi). The
first order condition can be expressed as
A(3i _ AQ2 ,
Qx~ Qi ’
where 2 is a function of the level Z of profit sought
and the AQ’s were defined above. This problem is
an instance of Ramsey ( ) optimal excise tax
analysis and has been put into a general equilibrium
context in Hartwick ( ). Contemporary
price discrimination schemes (e.g. Oi ) have
incorporated income elasticities of demand as well
as price elasticities and arrived at two-part tariffs
involving a ‘membership’ fee and a user’s fee for
service.
In closing, we note that high prices in peak times
and low in off-peak times are not forms of
monopoly price discrimination, since such peakload
prices vary with changes in the marginal costs of
production. Under price discrimination, prices
deviate from marginal costs of production only in
accord with variations across buyers in their
‘intensities’ of demand. Marginal cost remains the
same for each buyer.
See Also
► Basing Point System
► i fziz'i z:j
► Price Discrimination
) ► Public Utility Pricing
References
Barone, E. 1921. Les syndicats (cartels et trusts). Revue de
Metaphysique et de Morale 28(2): 279-309.
Dupuit, J. 1844. On the measurement of the utility of public
works. Reproduced in Readings in welfare
Discrimination
economics, ed. K.J. Arrow, T. Scitovsky, and American
Economic Association. Homewood: Irwin, 1969.
Edgeworth, F.Y. 1910. Applications of probabilities to
economics. Economic Journal 20: 284-304, 441-465.
Reprinted in F.Y. Edgeworth, Papers relating to political
economy, vol. II. New York: Burt Franklin, 1925.
Hartwick, J.M. 1978. Optimal price discrimination. Journal
of Public Economics 9(1): 83-89.
Oi, W.Y. 1971. A Disneyland dilemma: Two-part tariffs for a
Mickey Mouse monopoly. Quarterly Journal of Eco-
nomics 85(1): 77-96.
Pigou, A.C. 1904. Monopoly and consumers’ surplus.
Economic Journal 14: 388-394.
Pigou, A.C. 1920. The economics of welfare. London:
Macmillan.
Ramsey, F. 1927. A contribution to the theory of taxation.
Economic Journal 37: 47-61.
Robinson, J. 1933. The economics of imperfect competition.
London: Macmillan.
Schmalensee, R. 1981. Output and welfare implications of
monopolistic third-degree price discrimination. American
Economic Review 71(1): 242-247.
Smith, A. 1776. The wealth of nations. Ed. E. Cannan,
reprinted, New York: Modem Library, 1937.
Varian, H. 1985. Price discrimination and social welfare.
American Economic Review 75(4): 870-875.
Yntema, T. 1928. The influence of dumping on monopoly
price. Journal of Political Economy 36(6): 686-698.
Discrimination
Peter Mueser
Discrimination may be said to occur in a market
where individuals face terms of trade that are
determined by personal characteristics which do not
appear directly relevant to the transaction. Most
concern has centred on differential treatment by
race or ethnic group, and by sex. The primary focus
has been on the labour market and housing market,
with research motivated, in large part, by
controversy over the role of government in
maintaining or eliminating observed differentials.
The first extensive literature on the economics
of discrimination dates to the equal pay controversy
in Britain beginning before the turn of the century,
focusing on the lower wages of women. Although
interest in the economics of pay differentials by sex
abated in the two decades following
2921
World War II, many aspects of the more recent
theory appeared in this literature. The modem
development of systematic models of economic
discrimination began with the publication of Gary
Becker’s The Economics of Discrimination ( ).
With the passage of laws prohibiting dis
crimination in the US, Britain and other countries in
the 1960s and 1970s, research in the area has again
grown.
Market Discrimination and Personal
Preferences
Becker’s ( ) treatment took market discrimi
nation to be the result of personal tastes of partic-
ipants, providing a simple, closed model with a
variety of testable implications. Earnings differ-
entials, and discrimination in housing and other
markets stem from the attempts of owners, workers
and customers to avoid contact or interaction with
certain groups.
Consider first the influence of employer pref-
erences. Rather than maximizing profits, employers
maximize a utility that incorporates the personal
characteristics of employees. If employers prefer to
hire workers from group A rather than group B and
are willing to sacrifice profits to do so, they may be
said to have discriminatory tastes. Where such
employers dominate the market, the relative wages
of group B workers must adjust downward if any
are to be hired, and the resulting difference equals
the pecuniary value of the employed preference.
Where there are variations in taste among
employers, relative wages are determined by the
shape of the taste distribution, and the proportion of
A and B workers to be hired.
Employees may also be taken to have discrim-
inatory preferences over the group membership of
their co-workers. If discriminating workers and
members of the group they shun are perfect sub-
stitutes in production, employers have an incentive
to provide separate facilities for groups, but no
wage differential between groups will occur. In
order for such preferences to cause wage differen-
tials, the technology of production must preclude
complete separation. For example, if it is
2922
necessary for supervisors to interact with assembly
line workers, and supervisors prefer one group of
workers to another, an employer who has no taste
for discrimination and faces a competitive labour
market will hire members of both groups only if
their wages differ correspondingly.
Customers’ tastes may also influence market
wages in the absence of employer preferences. The
extension of this approach to a variety of markets is
clear. Housing discrimination would occur if
owners required a premium in order to sell or rent to
individuals in certain groups. However, in markets
for goods, in contrast to services, no appreciable
price differential could survive unless restrictions
on resale were binding.
In any of these cases, some market participants
may have a taste for discrimination without market
differentials occurring if there are a sufficient
number of non-discriminating participants to
interact with the disliked group. One obvious source
of non-discriminating participants is fellow group
members. If groups are equally represented among
employers, various kinds of workers, and
customers, complete segregation can occur without
economic loss to any group. Preferences against
trading with those outside one’s own group can only
affect terms of trade where groups have different
resoiuces or skills. The formal model is, however,
silent on the source of such differences.
Competition and Discriminatory
Preferences
It is widely argued that, in the long ran, competition
in markets for output and capital wi 11 drive out
discriminating employers. Since discrimination by
race and sex has long existed, this result has
frequently been taken as grounds for rejecting the
model (Arrow ). In fact, this conclusion follows
only from a particular version of the model, in
which the taste for discrimination imposes a direct
utility loss on the employer for each employee hired
from the disliked group. Since non-discriminating
employers suffer no such utility loss, under free
competition, where they can expand production or
buy out
Discrimination
discriminators, they will take over the market. This
need not be the case. A polar example has been
termed nepotism, in which preferences for hiring
one group act much like a net subsidy for the
employer. In this case, those with the strongest
discriminatory preferences ultimately dominate the
market.
In general, discriminating employers earn lower
money profits than those who do not discriminate,
but this does not imply that under competition they
will be driven from the market. Foregone profits
must be recognized as consumption expenditures,
and, so long as employer resources are sufficient to
permit any consumption, there is no inconsistency
between perfect competition and the existence of
stable, long-mn wage differentials stemming from
employer preferences.
While market discrimination may survive
competition, restrictions on competition will often
result in more severe discrimination. In a
competitive market, any market differential trans-
lates into a pecuniary cost that the discriminating
employer must pay. But where prices do not
equalize supply and demand, the employer’s cost of
discrimination may decline.
An effective minimum wage or, during times of
economic decline, a wage that is downwardly rigid,
allows the employer to hire both more productive
workers and those most preferred without paying a
higher wage. In contrast, if wages do not adjust
immediately in an economic upturn, the cost of
discrimination will increase.
Where a union successfully bargains for wages
above the competitive level, discriminatory hiring,
on the basis of either employers’ or union
members’ preferences may take place at lower cost.
In fact, discrimination against blacks by unions in
the US was explicit and widespread until the 1930s,
but by the 1960s union representation for blacks
and whites was nearly identical. Black
representation has been greatest in industrial
unions, where unionization often hinges on the
ability to organize both black and white workers,
and proportionally lowest in craft unions, where
unions frequently exercise power by limiting
membership. Despite the historically high level of
union discrimination, it appears to
Discrimination
have contributed relatively little to observed black-
white wage differentials in the US.
The cost of discrimination to firms may decline
where there are restrictions on profit maximization.
Those who manage non-profit organizations,
regulated monopolies or government bureaucracies
will devote more resources to improving their own
working conditions; unless faced with direct
constraints, they will be more likely to exercise
personal preferences in the kinds of workers they
hire (Alchian and Kessel ).
Finally, in markets with search costs, discrim-
ination may occur even if there are sufficient non-
discriminating participants to trade with members
of the disliked group, since the appropriate
matching cannot occur.
Discrimination as Exploitation
It is frequently asserted that discrimination is
engaged in because it is profitable. In general, there
is some level of discrimination by members of any
group that will improve the terms of trade so as to
increase their money incomes. Discrimination by
white employers, under some conditions, may
increase the incomes of whites by increasing both
employer profits and the earnings of white workers.
Similarly, tastes that restrict blacks and women to
certain kinds of jobs may increase money income
both for employers and white male employees.
If discriminators’ preferences are taken seri-
ously, however, the impact of discriminatory pref-
erences on money income is irrelevant. Although
Becker’s original treatment calculated such welfare
effects, like any such comparison it required an
arbitrary normalization to compare individuals with
differing preferences. Changes in money income
due to discrimination can be taken to represent
group welfare if discriminatory behaviour does not
reflect actual personal preference. However,
individual incentives in a competitive market can
no longer explain discrimination, since those who
discriminate least receive the greatest gains. It is
necessary that some process exist by which the
group enforces its will on
2923
individual choices. Exploitation must have its roots
in a social or political process.
Historically, there is no question that the enact-
ment of discriminatory laws and provision of
unequal public services has often represented the
exploitation of groups with little political power.
J.S. Mill ( ) argued that limitations on
women’s legal rights and the restrictions they faced
in entering certain occupations, were part of a
policy to provide men both with higher earnings in
the labour market and greater authority over their
wives at home. The history of governmental action
regarding blacks in America since the Civil War is
replete with examples of policies designed to
benefit whites with political power at the expense of
blacks.
Despite the government’s often central role in
furthering dominant group interests, there are
clearly other channels by which groups exercise
influence. An ethnic group is generally bound
together by an ideology that dictates members’
actions in a wide variety of contexts. Although
some members may internalize such ideology,
compliance is enforced by systems of social norms
and sanctions within the group. The process by
which such systems develop is not well understood,
although it has been shown that discriminatory
norms may be self-enforcing once established
(Akerlof ). Nonetheless, it is clear empirically that
economic relations among groups, and their
relations within the power structure, are critical in
determining group ideology and, in turn, individual
actions.
For example, it is a recurrent observation that
severe ethnic or racial antagonism often can be
traced to the point at which groups first find them-
selves competing in the labour market. Some
writers have argued that all discrimination by race
or ethnic group can be traced to such a dynamic, in
which groups mobilize political and economic
resources to further their material interests. The
goal of such action is seen to be the exclusion of the
competing group from the labour market or, failing
this, the creation of a caste system providing the
dominant group with preferential treatment
(Bonacich ).
As a rule, it is among lower income groups that
racism appears most virulent and associated
2924
violence most common. In part, this reflects the fact
that racism is a source of power to those groups
whose alternatives are limited. In some measure,
social norms, personal animosity and collective
violence substitute for political power and state
action.
In contrast to the assumption of the preference-
based model, the treatment here implies that dis-
criminatory preferences cannot be taken as exog-
enous. Tastes, or apparent tastes, may develop to
further group interests. This is not to say that
individual actions are ever completely determined
by group interests, even where these are unambig-
uous. Within the most tightly structured groups, for
example where ethnic identity is strong, dis-
criminatory collusion against outsiders relies
heavily on the availability of explicit policing
mechanisms. Where individual behaviours are
difficult to observe, and the benefits of violating
collusive rules are great, discrimination will be less
successful.
It must be stressed that many of the conclusions
of the taste-based model may apply even where
groups’ interests play a critical role in shaping
individuals’ actions. For example, the model tells
us that if white workers who compete with black
workers merely refuse to work with them, white
workers obtain no net gain in income. It is only
through the adoption of discriminatory practices by
employers that white workers realize gains.
Statistical Discrimination
Participants in a market have an incentive to con-
sider personal characteristics if these provide
information that is relevant to the exchange but
costly to obtain by other means. Statistical dis-
crimination occurs where an ascribed characteristic
serves this function. The widely accepted use of sex
in markets for various kinds of insurance is an
obvious example. Markets for credit and rental
housing have similar structures, as does the market
for labour. Initial screening is particularly critical in
hiring for entry level positions in firms with internal
labour markets, where a firm often undertakes
extensive worker training, and implicit
Discrimination
contracts limit the employer’s ability to adjust
wages in accord with realized productivity.
Some labour market analysts have attempted to
limit the term statistical discrimination to contexts
in which an employer distinguishes groups that do
not differ in average productivity (Aigner and
Cain ). For example, where an employer favoured
men over women because women were more likely
to quit after receiving firm-specific training, this
would not be labelled discrimination. In contrast,
statistical discrimination would be said to occur if
an employer screened by race in jobs where
expected ability was critical because he was unable
to judge the abilities of blacks. Such a distinction
becomes muddied when it is recognized that
matching the worker to the job is part of the
productive process.
In certain respects, observed patterns of
employment for women and blacks are consistent
with statistical discrimination. Both are seriously
under-represented in jobs offering extended pro-
motion ladders, and, historically, firms often
explicitly reserved for white males the training that
prepared an employee for promotion.
Since such persistent statistical discrimination
results from the efficient use of information, the
basis of wage differentials would seem to rest on
pre-market influences, not market dynamics.
However, it is possible that group differences are
themselves the result of employer expectations.
Assume employers believe that members of a
particular group have lower levels of the skills
necessary for success in screened jobs. In so far as
performance is ultimately rewarded for those placed
in screened positions, members of this group,
because they are less likely to be hired into such
positions, will have reduced incentives to invest in
relevant skills. Any one employer who hired
members of that group into these positions would
find workers to be less productive, so beliefs would
be confirmed.
In addition to a number of technical conditions
(Arrow ), in order for such a ‘self-fulfilling
prophecy’ to be stable, the actions of a single firm
must not alter individual incentives. If it were
possible for a firm to contract with individuals to fill
positions prior to the point when they acquire such
skills, individuals who entered contracts,
Discrimination 2925

whatever their group membership, would face the of discrimination. However, such a division is also
same incentives to obtain skills, and the vicious consistent with joint optimization by husband and
circle would be broken. The acquisition of such wife: if the bearing and rearing of chi Idren are even
skills must therefore occur well in advance of the weakly complementary, it is efficient for the family
point that individual workers and firms can easily to have the women specialize in both these non-
enter into agreements. Differences in socialization market tasks. Sex-typed socialization would then
by sex, or cultural differences by race or ethnic merely reflect preparation for anticipated roles.
group, if in response to disparate treatment in the That perfect equality would occur in the absence
labour market, could reflect this kind of vicious of all labour market discrimination seems unlikely.
circle. Nonetheless, unless there are strong sanctions,
employers have an incentive to practise statistical
discrimination, magnifying whatever sex
Explaining Market Differentials differences would occur in its absence.
The labour market disadvantages suffered by
Earnings for women have been appreciably below
many ethnic and racial groups is similarly open to
those of men in almost all societies, past and
interpretation. Thomas Sowell ( ) has
present. Differences in levels of market participa-
argued that cultural differences between arriving
tion and other observable personal attributes explain
immigrant groups and blacks in the US are more
only a portion of the differential. Historically, some
important in explaining their economic progress
of the difference may be identified with
than the levels of discrimination they faced. While
governmental or institutional discrimination.
it is clear that cultural factors are critically
Nevertheless, the enactment of laws in many
important, the degree to which these or other pre-
countries prohibiting discrimination in the 1960s
market differences explain observed earnings
and 1970s have had little effect on the overall
differentials is unclear. The theory implies that
distribution of wages by sex. There is no obvious
discrimination will be most common and most
way to identify the impact of discrimination in
damaging against groups with low levels of
explaining observed differences. Any unobserved
resources, those who would be disadvantaged in its
direct market discrimination may induce differences
absence.
in labour market participation and measurable pre-
For blacks in the US, slavery and subsequent
market factors, yet any unmeasured differences
governmental discrimination induced shortfalls in
between men and women that are not due to
human resources that would have limited black
discriminatory treatment, may also contribute to the
achievement under the best of conditions. None-
wage differential.
theless, up through the 1960s, measured pre-market
Women have historically performed the bulk of
differences explained only a modest portion of
household work and child care, participating in the
observed earnings differentials. Although
labour market less continuously and less intensively
unobserved premarket differences may have played
than have men. Given this division of labour within
a role, given the pervasiveness of explicit market
the family, women who expect to marry have less
discrimination, it seems likely that discrimination
incentive to develop skills requiring continuous
further depressed the black position. To what degree
labour market participation. Non-discriminating
labour market differences that persist despite the
employers may simply pay women less because
prohibition of discrimination since 1965 - most
they have not invested in those skills that are most
notably in rates of unemployment - are due to
valuable in the market.
unmeasured pre-market differences, possibly
How the earnings gap is viewed must depend
associated with statistical discrimination, or to other
partly on the source of the family’s division of
market discrimination, is an open question.
labour. If it results from market discrimination, or
social norms constructed to benefit dominant males,
it may be analysed in terms of the models
2926
See Also
► qua y
► lender
► iumar >: ta
► Inequality Between Persons
► Inequality Between the Sexes
► Uoo v A:: 1/ c :
► I-
► Voir v Wages
Bibliography
Akerlof, G. 1976. The economics of caste and of the rat race
and other woeful tales. Quarterly Journal of Economics
90(4): 599-617.
Aigner, D., and G. Cain. 1977. Statistical theoies of dis-
crimination in labor markets. Industrial and Labor
Relations Review 30(2): 175-187.
Alchian, A., and R. Kessel. 1962. Competition, monopoly,
and the pursuit of pecuniary gain. In Aspects of labor
economics, NBER Special Conference Series. Princeton:
Princeton University Press.
Arrow, K. 1972. Models of job discrimination. In Racial
discrimination in economic life, ed. A. Pascal. Lexington:
D.C. Heath.
Becker, G. 1957. The economics of discrimination, 2nd ed.
Chicago: University of Chicago Press, 1971.
Bonacich, E. 1972. A theory of ethnic antagonism: The split
labor market. American Sociological Review 37: 547-559.
Mill, J.S. 1869. The subjection of women. New York: Stokes,
1911.
Sowell, T. 1981. Ethnic America: A history. New York: Basic
Books.
Disequilibrium Analysis
Jean-Pascal Benassy
A convenient way to define ‘disequilibrium’ is of
course as the contrary of ‘equilibrium’. Unfortu-
nately this leaves us with no unique definition as the
word equilibrium itself has been used in the
economic literature with at least two principal
meanings. The first one refers to market equilib-
rium, i.e. the equality of supply and demand on
markets. This is the meaning we shall retain in
Disequilibrium Analysis
this entry, and therefore the disequilibrium analysis
we shall be concerned with here is the study of
nonclearing markets, also called non-Walrasian
analysis by reference to the most elaborate model of
market clearing, the Walrasian model.
The second meaning of equilibrium is somewhat
more general. A typical definition is given by
Machlup ( ) as ‘. . . a constellation of
selected interrelated variables, so adjusted to one
another that no inherent tendency to change prevails
in the model which they constitute’. Dealing with
disequilibrium in this second meaning would be a
quite formidable (and actually extremely imprecise)
task, which is why we want to limit ourselves in this
entry to disequilibrium analysis in the first sense.
We should note that the entry RATIONED
EQUILIBRIA presents concepts of equilibria in the
second, but not in the first sense of the word, i.e.
more specifically equilibria without market clearing
or non-Walrasian equilibria.
The Essence of the Theory
Disequilibrium analysis is best appraised by ref-
erence to the standard equilibrium market clearing
paradigm, corresponding to the notions of
Marshallian or Walrasian equilibrium. There all
private agents receive a price signal and assume that
they will be able to exchange whatever they want at
that price. They express demands and supplies,
sometimes called ‘notional’, which are functions of
this price signal. An equilibrium price system is a
set of prices for which demand and supply match on
all markets. Transactions are equal to the demands
and supplies at the equilibrium price system.
Two characteristics deserve to be stressed: all
private agents receive price signals and make
rational quantity decisions with respect to them. But
no agent makes any use of the quantity signals sent
to the market. Also no agent actually sets prices, the
determination of which is left to the ‘invisible hand’
or to the implicit Walrasian auctioneer. This logical
hole of the theory was pointed out by Arrow ( )
when he noted
Disequilibrium Analysis
there was ‘. . . a logical gap in the usual formula-
tions of the theory of the perfectly competitive
economy, namely, that there is no place for a
rational decision with respect to prices as there is
with respect to quantities’, and more specifically
‘each individual participant in the economy is
supposed to take prices as given and determine his
choices as to purchases and sales accordingly; there
is no one left over whose job is to make a decision
on price’.
Disequilibrium analysis takes this strong logical
objection quite seriously, and its purpose is to build
a consistent theory of the functioning of
decentralized economies when market clearing is
not axiomatically assumed. The consequences of
abandoning the market clearing assumption are
actually quite far-reaching: (i) The transactions
cannot be all equal to demands and supplies
expressed on markets. Rationing will be experi-
enced and quantity signals will be formed in addi-
tion to price signals (ii) Demand and supply theory
must be substantially modified to take into account
these quantity signals. One thus obtains a theory of
effective demand, as opposed to notional demand
which only takes price signals into account (iii)
Price theory must also be amended in a way that
integrates the possibility of non-clearing markets,
the presence of quantity signals, and makes agents
themselves responsible for price making, (iv)
Finally expectations, which in market clearing
models are concerned with price signals only, must
now include quantity signals expectations as well.
History
Though roots may be found earlier, an
uncontestable grandfather of disequilibrium anal-
ysis in the sense we use here is of course Keynes (
). He rightfully perceived that one of his
main contributions in the General Theory was the
introduction of quantity adjustments, and more
specifically income adjustments, in the economic
process, whereas the then dominant ‘classical’
economists focused on price adjustments only. As
Keynes ( ) wrote ‘As I have said
above, the initial novelty lies in my maintaining
2927
that it is not the rate of interest, but the level of
incomes which ensures equality between savings
and investment’.
Unfortunately for many decades things did not
go much further: macroeconomists added the level
of income in their equations, thereby allowing for
unemployment. But concentration on the
‘equilibrium’ of the goods and money markets,
exemplified by the dominant IS-LM model,
obscured the ‘disequilibrium’ nature of the model.
As for microeconomics, it was basically unaffected
by the Keynesian revolution, and cor- relatively a
growing gap developed between microeconomics
and macroeconomics.
A few isolated contributions in the post-war
period made some steps toward modem disequi-
librium theories. Samuelson ( ), Tobin and
Houthakker ( ) studied the theory of demand
under conditions of rationing. Bent Hansen ( )
introduced the ideas of active demand,
close in spirit to that of effective demand, and of
quasi-equilibrium where persistent disequilibrium
created steady inflation. Patinkin ( , ch. 13)
considered the situation where the firms might not
be able to sell all their ‘notional’ output. Hahn and
Negishi ( ) studied non-tatonnement pro
cesses where trade could take place before a general
equilibrium price system was reached. Hicks ( )
discussed the ‘fixprice’ method as opposed
to the flexprice method.
A main impetus came from the stimulating
works of Clower ( ) and Leijonhufvud
( ). Both were concerned with the microeco
nomic foundations of Keynesian theory. Clower
showed that the Keynesian consumption function
made no sense unless reinterpreted as the response
of a rational consumer to a disequilibrium on the
labour markets. He introduced the ‘dual-decision’
hypothesis, a precursor of modem effective demand
theory, showing how the consumption function
could have two different functional forms,
depending on whether the consumer was rationed
on the labour market or not. Leijonhufvud ( )
insisted on the importance of short-run
quantity adjustments to explain the establishment of
an equilibrium with involuntary unemployment.
These contributions were followed by the mac-
roeconomic model of Barro and Grossman (
2928 Disequilibrium Analysis

), integrating the ‘Glower’ consumption func- d* < d s* <s

tion and the ‘Patinkin’ employment function in the
first ‘disequilibrium’ macroeconomic model. which implies that:
Then the main development was that of micro- d* = s* < m i n (d, s)
economic concepts of non-Walrasian equilibrium
proposed notably by Benassy ( , , , Actually in this simple example, there is not reason
), Dreze ( ) and Younes ( ). These, why these two agents would exchange less than the
which generalize the notion of Walrasian general minimum of demand and supply, as they would be
equilibrium to non-market clearing situations, gave both frustrated in their desires of exchange. This
solid microeconomic foundations to the field. The simple ‘efficiency’ assumption leads us to take the
main concepts are reviewed in the entry transaction as:
RATIONED EQUILIBRIA.
From then on, the field has developed quite d* = s* = m i n id, , v )
rapidly, notably in the direction of macroeconomic
applications and econometrics, as we shall outline the well-known ‘rule of the minimum’.
below. We shall now review quickly the main Now at the same time as transactions take place,
elements of disequilibrium analysis. Longer quantity signals are set across the market: faced
developments can be found notably in Benassy ( with the supply ,v, and under voluntary exchange,
). the demander knows that he will not be able to
purchase more than S. Symmetrically the supplier
knows that he cannot sell more than d. Each agents
Non-Clearing Markets and Quantity
thus receives from the other a quantity signal,
Signals
respectively denoted as d and ,v, which tells him the
A most important element of the theory is obvi- maximum quantity he can respectively buy and sell.
ously to show how transactions can occur in a In this example:
market in disequilibrium, and how quantity signals
are generated in the decentralized trading process. d=ss=d
To make things clear and intuitive, we shall start
with the simple case of two agents, one demander and the transactions can thus be expressed as:
and one supplier in a market which does not
necessarily clear. They meet and express, d* = min(d, d)
s* = m i n (s, s)
respectively, an effective demand d and supply s
(note that we do not use notional demands and
Let us move now to the general case (which is
supplies which are fully irrelevant in this context).
explored more formally in the entry on RATIONED
We shall now indicate how transactions and quan-
EQUILIBRIA). Agents, indexed by i, exchange
tity signals are formed in this example. Transac-
goods indexed by h. On each market h a rationing
tions will be denoted d* and s* respectively and
scheme transforms inconsistent demands and
they must of course satisfy d* s*.
supplies, denoted dn% and .?,■/,, into consistent
The first principle we shall use is that of vol-
transactions, denotedand d*h and s*h, which balance
untary exchange, i.e. that no agent can be forced to
identically (i.e. total purchases always equal total
trade more than he wants on a market. This con-
sales). At the same time, and continuing to assume
dition is quite natural and actually verified on most
voluntary exchange, each agent receives a quantity
markets, except maybe for some labour markets
signal, respectively t/,/, or s,i, for demanders and
which are regulated by more complex contractual
suppliers, which tells him
arrangements. It is written in this example:
Disequilibrium Analysis 2929

the maximum quantity he can buy or sell, and the There is thus an equality between the signal the
rationing scheme is equivalently written: agent sends to the market (demand or supply) and
the transaction he will obtained from it.
d*ih = m i n (dih, dih) In disequilibrium analysis there is of course a
s*h = m i n (sih, sih) difference between the signals sent (effective
demands and supplies) and their consequences (the
where d,/, and Sih are functions of the demands and transactions actually realized). Effective demands
supplies of the other agents on the market. These and supplies expressed by an agent in the various
quantity signals may result from the signals sent to markets are the signals which maximize his
each other by agents in decentralized pairwise expected utility of the resulting transactions,
meetings (as in the above two agents example) or knowing that these transactions are related to the
result from a more centralized process (as in a demands and supplies by equalities of the type seen
uniform rationing scheme). above, i.e.:
We thus see that on a market we may have
unrationed demanders or suppliers, or rationed ones. d*h = min{dih, dih)
The rationing scheme is called efficient if there are s*h min (slh, sih)
not both rationed demanders and rationed suppliers
in the same market. An efficient rationing scheme The results of such expected utility maximization
implies the well-known ‘rale of the minimum’, programmes may be quite complex, depending for
according to which aggregate transactions equal the example on whether quantity constraints are
minimum of supply and demand. Such an expected deterministically or stochastically, or
assumption, which was very natural for our example whether agents act or not as price markers, as we
with two agents, may not be valid if one considers a shall see in the next section. In the case of deter-
macroeconomic market, as not all demanders and ministic constraints, there exists a simple and
suppliers meet pairwise. In particular it is well workable definition of effective demand, which
known that the property of market efficiency may generalizes Clower’s original ‘dual decision’
be lost in the process of aggregating submarkets, method: effective demand (or supply) on one par-
whereas voluntary exchange remains. Note, ticular market is the trade which maximizes the
however, that the concepts that follow do not agent’s criterion subject to the constraints encoun-
require that property of market efficiency. tered or expected on the other markets. This def-
Now it is clear that the quantity signals received inition thus naturally integrates the well-known
by the agents hould have an effect on demand, ‘spillover effects’, which show how disequilibrium
supply and price formation. This is what we shall in one market affects demands and supplies in the
explore now. other markets.
We shall immediately give an illustrative
example of this definition, due to Patinkin ( ) and
Effective Demand and Supply Barro and Grossman ( ), that of
the employment function of the firm. Consider a
Demands and supplies are signals that agents send firm with a production function y = m exhibiting
to the ‘market’ (i.e. to the other agents) in order to diminishing returns, and faced with a price p on the
obtain the best transactions according to their output market and a wage w on the labour market.
criterion. The traditional ‘notional’ or Walrasian The traditional ‘notional’ labour demand results
demands and supplies are constructed under the from maximization of profit py - wl subject to the
assumption (which is actually verified ex-post in a production constraint y = F([), which yields
Walrasian equilibrium) that each agent can buy and immediately the usual Walrasian labour demand
sell as much as he wants on each market. l<1 \w/p). Assume now that the firm faces a constraint
y on its sales
2930 Disequilibrium Analysis

of output (i.e. a total demand y). According to the a particular organization of the pricing process
above definition the effective demand for labour / is where agents on one side of the market (usually the
the solution in l of the following programme: suppliers) quote prices and agents on the other side
act as price takers. Other modes of pricing
Maximize py — wl s.t. (bargaining, contracting) are currently studied, but
y = F(l) have not yet been integrated in this line of work. As
y<y we shall see this model of price making is quite
reminiscent of the imperfect competition line:
the solution of which is:
Chamberlain ( ), Robinson ( ), Tri-
l
/ = min|.F' {w/p), , 1
F ~ (y)j fifin ( ), Bushaw and Clower ( ), Arrow
( ), Negishi ( ).
Consider thus, to fix ideas, the case where
We see that the effective demand for labour may
sellers set the prices (things would be quite sym-
have two forms: the Walrasian demand just seen
metrical if demanders were setting the prices), and
above if the sales constraint is not binding, or, if
in order to have only one price per market, let us
this constraint is binding, a more ‘Keynesian’ form
characterize a market by the nature of the good sold
equal to the quantity of labour just necessary to
and its seller (we thus consider two goods sold by
produce the output demand. We see immediately on
different sellers as different goods, a fairly usual
this example that effective demand may have
assumption in microeconomic theory since these
various functional forms, which intuitively explains
goods differ at least by location, quality, etc...). On
why disequilibrium models often have multiple
each ‘market’ so defined we thus have one seller,
regimes (see for example the three goods-three
the price maker, facing several buyers. As we saw
regimes model in the entry FIX-PRICE MODELS.
above, for a given price this seller faces a quantity
In the case of stochastic demand, the programme
constraint s, actually equal to the demand of the
yielding the effective demand for labour becomes
other agents on that market. But the price level is
evidently more complex. One obtains some results
now a decision variable for the seller, and this
quite reminiscent of the inventories literature as
quantity constraint (the others’ total demand) can be
developed for example by Arrow et al. ( ) or
modified by changing the price: for example in
Bellman ( ). See Benassy
general the seller who wants to sell more knows
( ) for the link between these two lines of work.
that, others things being equal, he should lower the
price. The relation between the maximum quantity
Price Making he expects to sell and the price set by the price
maker is called the expected demand curve. If
We shall now address the problem of price making
demand is forecasted deterministically, this
by decentralized agents, and we shall see that there
expected demand curve will be denoted as:
too quantity signals play a prominent role. It is
actually quite intuitive that quantity signals must be
S(p,6)
a fundamental part of the competitive process in a
truly decentralized economy. Indeed, it is the where 6 is a vector of parameters depending on the
inability to sell as much as they want that leads exact functional form of that curve (for example
suppliers to propose, or to accept from other agents, elasticity and a position factor for isoelastic curves).
a lower price, and conversely it is the inability to If demand is forecast stochastically, the expected
buy as much as they want that leads demanders to demand curve will have the form of a probability
propose, or accept, a higher price. Various modes of distribution on s (i.e. total demand) conditional on
price making integrating these aspects can be the price.
envisioned. We shall deal here with For a given expected demand curve, the price
maker chooses the price which will maximize
profits, given the relation between price and
Disequilibrium Analysis 2931

maximum sales. For example, continuing to con- expectations into the microeconomic setting was
sider a firm with production function F(l), the made in Benassy ( , , ). Macroeco
programme yielding the optimum price is the nomic applications of the corresponding concepts
following in the case of a deterministic expected can be found in Hildenbrand and Hildenbrand (
demand curve: ), Muellbauer and Portes ( ), Benassy
( , ), Neary and Stiglitz ( ).
Maximize py — wl s.t.
y = m
y < Sip, 9) Scope and Uses of Disequilibrium Analysis
/</
We have briefly outlined the basic elements or
where 7 is the constraint the firm possibly faces on
building blocks of disequilibrium analysis. We saw
the labour market, where it is a ‘wage taker’. Note
that it generalizes the traditional theories of
that, according to our definition above, the effective
demand, supply and price formation to cases where,
demand for labour of this same firm would be given
in the absence of an auctioneer, markets do not
by the above programme, from which the last
automatically clear. This theory is thus a quite
constraint would be deleted.
general one, and the scope of its applications very
Both the price and quantity decisions of price
broad. Up to now there have been in the literature
makers depend on the parameters 9. Of course it
three particularly active areas of development: (1)
would require quite heroic assumptions on the
The construction of various concepts of equilibria
computational ability and information available to
with rationing, or non-Walrasian equilibria. These
price setters to assume that they know the ‘true’
concepts, which generalize the traditional notion of
demand fimction (i.e. the ‘true’ functional form
Walrasian equilibrium to the cases where not all
with the ‘true’ parameters). But the theory
markets clear, show how mixed price-quantity
developed here gives a natural way of learning
adjustments can bring about a new type of
about the demand curve. Indeed, each realization p,
equilbrium in the short run. (2) The development of
s in a period is a point on the ‘true’ demand curve in
numerous macroeconomic applications, which
that period (Bushaw and Clower ). Using the
basically use the above concepts in the framework
sequence of these observations, plus any extra
of aggregated macromodels, and derive policy
information available (including for example the
implications, for example to fight involuntary
price of its competitors), the price maker can use
unemployment. (3 ) Finally new econometric
statistical techniques to yield an estimation of the
methods have been developed to deal with such
demand curve. Whether this learning would lead to
models, as traditional methods were more suited to
the ‘true’ demand curve is still an unresolved
the study of equilibrium markets.
problem.
Microeconomic concepts of non-Walrasi;in
equilibria are reviewed in the entry RATIONED
EQUILIBRIA. We shall now very briefly outline
the macroeconomic and econometric developments.
Expectations
Of course, the modifications we outlined Macroeconomic Applications
concerning the signal structure affect not only the
current period, but the future periods as well, and as Many contributions in the field started from a
compared to traditional ‘competitive’ analysis, dis- reconsideration of Keynesian models, and it is
equilibrium analysis introduces expected quantity therefore no surprise that many macroeconomic
signals in addition to expected price signals. Such
an introduction allows for example to rationalize the
traditional Keynesian accelerator (Grossman ). The
introduction of such quantity
2932 Disequilibrium Analysis

applications have been made. The early model of Let us consider the very simplest case, that of a
Barro and Grossman ( ) has been followed by single market with a rigid price. The most basic
a huge macroeconomic literature, notably aimed at system to estimate is then:
policy analysis and the study of involuntary
unemployment. A very valuable feature of dis- Xd = adZd + Ed
equilibrium macromodels is that, like the micro- Xs = asZs + t:s
economic models, they endogenously generate X = mm(Xd, Xs)
multiple regimes in which various policy tools may
have quite different impacts. These models are thus where X1 is quantity demanded, Zd is the set of
a particularly useful tool for synthesizing hitherto variables affecting demand, ad is the vector of
disjoint macroeconomic theories. One finds a corresponding parameters and s:d is a demand
number of macroeconomic applications in books by disturbance term (and symmetrically on the supply
Barro and Grossman ( ), Benassy side). The market is assumed for the moment to
( , ), Cuddington et al. ( ), function efficiently so that transaction X is the
Malinvaud ( ), Negishi ( ). We may note minimum of demand and supply Xd and Xs. The
that the same methods can also be used to study the problem in estimating such a model, as compared
problems of centrally planned economies (Portes ). with an equilibrium model, where by assumption
A few lessons can be drawn from these macro-
disequilibrium models. The first is that, even though X = Xd = X s
these models were at the very beginning aimed at
bridging the gap with Keynesian analysis, they is that only A is observed, not X1 or A®.
proved to be of more general relevance, and were Techniques for dealing with these problems are
able to generate non-Keynesian results as well as reviewed in Quandt ( ). Of course this is the
the traditional Keynesian results. Secondly, and simplest
more generally, whether or not a policy tool is possible model, and numerous extensions are now
efficient may depend very much on the ‘regime’ the considered: (1) The prices may be flexible, either
economy is in. A famous example is the Barro and within the period of estimation, or between
Grossman fixprice macroeconomic model, with its successive periods. The price equation must then be
‘Classical unemployment’ and ‘Keynesian estimated simultaneously with the demand-supply
unemployment’ regimes. Finally, it appears that the system. (2) Since some applications are made on
results of these models are quite sensitive to both macroeconomic markets, the ‘minimum’ condition
the price formation mechanism on each market, as may not be satisfied, and is replaced by an explicit
well as on the expectations formation mechanisms procedure of aggregation of submarkets. (3) Finally
on both price and quantities (cf. for example, models with several markets in disequilibrium have
Benassy ( ), which been estimated, notably at the macroeconomic level.
experiments with various hypotheses).
This quite naturally leads to the need of further
Concluding Remarks
theoretical work, and to the necessity of empirically
testing these models, an issue to which we now The development of disequilibrium analysis has
turn. clearly led to an enlargement and synthesis of both
traditional microeconomics and macroeconomics.
Usual microeconomic theory in the market
Disequilibrium Econometrics clearing tradition has been generalized in a number
of directions: the study of the functioning of non-
In order to estimate microeconomic or macroeco- clearing markets and the formation of quantity
nomic disequilibrium models a whole new econo- signals, a theory of demand and supply
metric technology has developed in recent years.
Disequilibrium Analysis
responding to these quantity signals as well as to
price signals, the integration of quantity expecta-
tions into microeconomic theory. This line of
analysis further includes a theory of price making
by agents internal to the system which also brid-
ges the gap with the traditional theories of imper-
fect competition.
As for the corresponding macroeconomic
models, they turn out to be a very useful syn-
thetical tool, as they cover all possible disequilib-
rium configurations. They are more general than
either traditional Keynesian macromodels, which
considered only excess supply states, or than ‘new
classical’ macromodels which postulate market
clearing at all times. They are of course the natural
tool to study problems such as involuntary
unemployment.
Still richer developments lie ahead with further
developments in the theories of price and wage
formation in markets without an auctioneer. The
methodology outlined here will permit us to
derive the micro and macro consequences, as
well as the consequences in terms of economic
policy prescriptions. Much is also to be expected
of the development of the associated econometric
methods, which should allow us to choose the
most relevant hypotheses, and to characterize spe-
cific historical episodes.
See Also
► liqui i
► Ratic : ::
► empo
► Jncertaint
Bibliography
Arrow, K.J. 1959. Towards a theory of price adjustment. In
The allocation of economic resources, ed.
M. Abramowitz. Stanford: Stanford University Press.
Arrow, K.J., S. Karlin, and H. Scarf. 1958. Studies in the
mathematical theory of inventory and production.
Stanford: Stanford University Press.
Barro, R.J., and H.I. Grossman. 1971. A general disequi-
librium model of income and employment American
Economic Review 61: 82-93.
Barro, R.J., and H.I. Grossman. 1976. Money, employment
and inflation. Cambridge: Cambridge University Press.
2933
Bellman, R. 1957. Dynamic programming. Princeton:
Princeton University Press.
Benassy, J.P. 1975. Neo-Keynesian disequilibrium theory in a
monetary economy. Review of Economic Studies 42: 502-
523.
Benassy, J.P. 1976. The disequilibrium approach to
monopolistic price setting and general monopolistic
equilibrium. Review of Economic Studies 43: 69-81.
Benassy, J.P. 1977a. A Neo-Keynesian model of price and
quantity determination in disequilibrium. In Equilibrium
and disequilibrium in economic theory, ed. G.
Schwodiauer. Boston: D. Reidel Publishing Company.
Benassy, J.P. 1977b. On quantity signals and the foundations
of effective demand theory. Scandinavian Journal of
Economics 79: 147-168.
Benassy, J.P. 1982. The economics of market disequilibrium.
New York: Academic Press.
Benassy, J.P. 1986. Macroeconomics: An introduction to the
Non-Walrasian approach. New York: Academic Press.
Bushaw, D.W., and R. Clower. 1957. Introduction to math-
ematical economics. Homewood: Richard D. Irwin.
Chamberlin, E.H. 1933. The theory of monopolistic
competition. Cambridge, MA: Harvard University Press.
Clower, R.W. 1965. The Keynesian counterrevolution: A
theoretical appraisal. In The theory of interest rates, ed.
F.H. Hahn and F.P.R. Brechling. London: Macmillan.
Cuddington, J.T., P.O. Johansson, andK.G. Lofgren. 1984.
Disequilibrium macroeconomics in open economies.
Oxford: Basil Blackwell.
Dreze, J.H. 1975. Existence of an exchange equilibrium under
price rigidities. International Economic Review 16: 301-
320.
Grossman, ELI. 1972. A choice-theoretic model of an income
investment accelerator. American Economic Review 62:
630-641.
Hahn, F.H., and T. Negishi. 1962. A theorem on non
tatonnement stability. Econometrica 30: 463-469.
Hansen, B. 1951. A study in the theory of inflation. London:
Allen & Unwin.
Hicks, J.R. 1965. Capital and growth. London: Oxford
University Press.
Hildenbrand, K., and W. Hildenbrand. 1978. On Keynesian
equilibria with unemployment and quantity rationing.
Journal of Economic Theory 18: 255-277.
Keynes, J.M. 1936. The general theory of money, interest and
employment. New York: Harcourt Brace.
Keynes, J.M. 1937. Alternative theories of the rate of interest.
Economic Journal 47: 241-252.
Leijonhufvud, A. 1968. On Keynesian economics and the
economics of Keynes. Oxford: Oxford University Press.
Machlup, F. 1958. Equilibrium and disequilibrium: Mis-
placed concreteness and disguised politics. Economic
Journal 68: 1-24.
Malinvaud, E. 1977. The theory of unemployment
reconsidered. Oxford: Basil Blackwell.
2934
Muellbauer, J., and R. Portes. 1978. Macroeconomic models
with quantity rationing. Economic Journal 88:
788-821.
Neary, J.R, and J.E. Stiglitz. 1983. Towards a reconstruction
of Keynesian economics: Expectations and constrained
equilibria. Quarterly Journal of Economics
98(Supplement): 199-228.
Negishi, T. 1961. Monopolistic competition and general
equilibrium. Review of Economic Studies 28: 196-201.
Negishi, T. 1979. Microeconomic foundations of Keynesian
macroeconomics. Amsterdam: North-Holland. Patinkin, D.
1956. Money, interest and prices. New York: Row, Peterson
& Co.; 2nd ed. New York: Harper & Row, 1965.
Portes, R. 1981. Macroeconomic equilibrium and disequi-
librium in centrally planned economies. Economic
Inquiry 19: 559-578.
Quandt, R.E. 1982. Econometric disequilibrium models.
Econometric Review 1: 1-63.
Robinson, J. 1933. The economics of imperfect competition.
London: Macmillan.
Samuelson, PA. 1947. Foundations of economic analysis.
Cambridge, MA: Harvard University Press.
Tobin, J., andH.S. Houthakker. 1950. The effects of rationing
on demand elasticities. Review of Economic Studies 18:
140-153.
Triffin, R. 1940. Monopolistic competition and general
equilibrium theoiy. Cambridge, MA: Harvard University
Press.
Younes, Y. 1975. On the role of money in the process of
exchange and the existence of a non-Walrasian equilib-
rium. Review of Economic Studies 42: 489-501.
Disguised Unemployment
Amit Bhaduri
Marx set out the notion that a ‘reserve army’ of
unemployed labour is more or less continuously
maintained in the course of capitalistic develop-
ment. In the initial phases, this reserve army may be
created through the destruction of the pre-
capitalistic modes of production while, in later
phases, a systematic bias in favour of labour-
displacing innovations could serve the same
purpose. This entails a broad vision of capitalistic
development under extremely elastic supply
conditions for labour where the actual level of wage
employment is usually demand-determined. This
means that the supply of labour tends to
Disguised Unemployment
adjust to its demand through various routes such as,
higher participation rate (e.g. as more married
women join the labour force or the average
schooling period is shortened), interregional and
international migration of labour etc., all this taking
place against the background of continuous induced
innovations. Under these circumstances, it is not
very usefi.il to think of a ‘natural’ rate of growth,
set by the growth of labour force and of labour
productivity, as the maximum feasible growth rate
of a capitalist economy (Marglin pp. 103-8).
The elastic nature of the labour supply and its
adjustability to the level of demand entail the
existence of open or disguised unemployment as an
untapped reservoir of labour in the normal course of
capitalist development. However, such disguised
unemployment although real is a somewhat
amorphous phenomenon in an advanced capitalist
economy for two distinct reasons. First, under
normal circumstances, many potential entrants (e.g.
married women, late school-leavers, young people
on the farm) may not actually even try to enter the
labour market unless demand is seen to be high with
all sorts of job vacancies exceeding their
corresponding numbers in registered
unemployment. Second, the economic nationalism
in the richer countries often takes the form of
strictly regulating the migration of ‘guest-workers’
so that open unemployment in the (potentially)
labour-exporting countries, rather than disguised
unemployment in the advanced capitalist countries,
becomes the normal pattern. And yet, prolonged
stagnation in economic conditions in an advanced
capitalist country may make this phenomenon of
disguised unemployment more visible, as the
redundant workers either seek various forms of self-
employment with virtually no invested capital or try
to sell their labour services directly as porters, odd-
jobmen, domestic servants, farm-hands etc.
(Robinson , pp. 157-8). Their earnings in these
peripheral jobs would then become the ‘reservation
price’ of this marginalized labour force. When
unemployment dole and social security set a higher
reservation price, some of this unemployment may
come out in the open instead of being disguised. In
this sense, it is probable that the
Disguised Unemployment
growth of the welfare state may openly register as
unemployed some who would have been otherwise
unemployed in a disguised fashion earlier. And, the
reverse could happen if the social security measures
are cut by the government.
The existence of such disguised unemployment
on a significant scale is usually accommodated by a
secondary or informal labour market mostly in the
service sector. This is much more easily visible in
the phenomenon of massive migration to urban
centres from rural areas in many developing
countries. While all such migrants from rural areas
aspire to limited job opportunities in organized
industries located in urban areas, only a small
fraction among them are actually able to find proper
jobs at any given point of time. The rest spend their
time, waiting in search of appropriate jobs. In the
meantime, they somehow manage to disguise their
unemployment either by self-employing themselves
with tiny amounts of invested capital (e.g. polishing
shoes, cleaning cars etc. ) or by selling their labour
services directly in odd jobs or even, simply taking
recourse to the support of the elaborate kinship
system in more traditional societies e.g. by living
off better-placed relatives and migrant workers from
their home areas. Thus, the phenomenon of
disguised unemployment in the urban areas of many
developing countries becomes closely linked with
the massive migration from rural areas during the
course of industrialization.
A distinguishing feature of disguised unem-
ployment in such an informal sector is the irregular
and often long hours of work per day. This is
evident enough in the case of most self-employed
persons in the informal sector; but even those who
are employed on a wage-labour basis usually have
highly flexible wage contracts in many respects
(e.g. domestic servants, odd-jobmen etc.). Partly the
explanation lies in the lower unionization of this
sector. However, a deeper explanation lies in the
fact that most self-employed persons as well as
workers paid at the piece-rate have to work
extended hours per day simply to make a livelihood.
But this also could have a limited advantage for
some of them insofar as the entire family can
participate in the work (e.g. traditional carpet
making, weaving and other types of artisan work
2935
are often done by many members of the family
working together). In this context, we have to make
a sharp distinction between labour-service and the
labourer providing such service: the same amount
of labour service (say, 18 hours per day) may be
spread out over several family members working as
labourers (say, three). In some cases, each family
member (labourer) may on an average have a lighter
work load (of only six hours) per day compared to
an average worker in the organized industry. This
brings us to a somewhat different analytical
dimension of disguised unemployment: some
persons may be unemployed in a disguised manner
not only in the sense of having a very low earning
rate i.e. income-wise unemployment but also in the
sense of relatively light work-intensity per day, i.e.
time-disposition-wise unemployment. And, unless
one believes in the neoclassical proposition that
income necessarily reflects the marginal product,
one would have to devise, a third (and separate)
criterion of disguised unemployment in terms of
abnormally low productivity of labour. However,
given the structure of reward in a capitalist
economy, one needs to be careful in applying these
concepts. Thus, an ‘important person’ belonging to
the board of directors of several large corporations,
may be making a well-above- average income by
attending only a couple of board meetings per
month. Such a person may very well be considered
to be disguised unemployed by the time-disposition
criterion and even perhaps by the labour-
productivity criterion although, he cannot, by any
means, be considered unemployed, disguised or not,
by the income criterion! Also recall in this context
that ‘unproductive labour’ was a common category
used in the classical tradition of political economy
and, all those engaged in unproductive labour (e.g.
‘priests, prostitutes and professors’ according to a
picturesque phrase employed by Rosa Luxemburg)
could be considered to be disguised unemployed by
the productivity criterion.
In the normal organization of factory work
under the capitalist system, the threefold distinction
between income-wise, time-wise and productivity-
wide disguised unemployment may not be
particularly relevant. Thus, an unemployed
2936
industrial worker is both income- and time-wise
unemployed and of course, he does not have much a
chance to be productive either. However, such a
distinction can be highly relevant in the context of
traditional, family-based agriculture, especially for
characterizing such phenomena as nual poverty or
the existence of surplus labour. Consider for
example a typical nual woman in the poorest strata:
in addition to all her other work inside and outside
the house, she may have to spend long hoius
collecting wood for fuel and carrying water home
from a distance. Although she has exceptionally
hard and long working hoius every day and must be
considered time - and disposition- wise fully
employed and certainly productive in every normal
sense of the term, in keeping her family going under
most difficult circumstances, in all probability she
would not be classified as ‘gainfully employed’ by
the income criterion. Indeed her case is the opposite
of that oiu ‘important person’ who has a high
income by attending a couple of board meetings
every month. It is to be noted that the worst kind of
nual poverty is often concentrated among people
who are hilly employed by the time-disposition
criterion, but may be described as disguised
unemployed by the income criterion, because of
their miserably low earning rate per hour of work.
After all, this is what the phrase ‘eking out a living’
usually means.
There can hardly be any serious doubt that in the
backward agriculture of many populous countries
(e.g. in South Asia), a high proportion of the
population engaged in cultivation have extremely
low income and, in this sense suffers from disguised
unemployment by the income-criterion.
Nevertheless, it is far more problematic to identify
what such disguised unemployment by the income
criterion implies in terms of either the time-disposal
or the productivity criterion. If one were to believe
in the ideologically potent neoclassical slogan that
all ‘factors of production’ including labour always
tend to get paid according to their marginal product
even in pre-capitalist, backward agriculture, then
that proportion of population with extremely low
income could be said to be rather unproductively
engaged in agriculture. Their low income would be
the ‘evidence’ of
Disguised Unemployment
their low productivity which in turn would imply a
corresponding level of disguised unemployment in
agriculture. But this would involve implicit
theorizing based on the dubious assumption that
income (earning) is always positively associated
with productivity, even in traditional agriculture.
Such implicit theorizing apart (a sophisticated
example of which is the so-called ‘efficiency wage’
hypothesis e.g. Bliss and Stem ) the important
question remains as to whether there is any
meaningful sense in which one can argue about the
existence of significant surplus labour and
disguised unemployment in backward agriculture,
judged by the productivity criterion. This would
imply that some surplus labour can be withdrawn
from agriculture without adversely affecting the
level of agricultural output. Or, in more textbookish
jargon, ‘at the margin’ labour contributes nothing to
output so that, the marginal product of labour is
zero in such agriculture. Put in such general terms,
the formulation is too fuzzy to be useful. For
instance, if by ‘margin’, one means the intensive
margin of higher labour input per unit of land, then
considerable empirical evidence exists, at least in
India, to suggest that the smaller-sized land
holdings usually do use family labour more
intensively, both in current agricultural operations
and in direct investment of labour for improving
land quality. As a result, the total output, taking all
crops together over the year, tends to be higher per
unit of land on smaller holdings (Bharadwaj , chs.
2, 3 and 7 provide an excellent account). This
tendency towards an inverse relation between farm
size and productivity per acre in traditional
agriculture would tend to cast doubt on the simple-
minded proposition that the ‘marginal’ product of
labour is zero, especially if the notion of intensive
margin is used.
Without going into such finer points of inten-
sive and extensive margin, Schultz ( , ch. 4)
proposed the ‘epidemic test’: the 1918-19 influenza
epidemic in India killed 6.2% of the 1918
population and 8.3% of the working population in
agriculture (the latter according to Schultz’s esti-
mate). Schultz found that, although the weather
conditions were roughly similar in 1916-17 and in
1919-20, in the latter year agricultural output
Disintermediation
was lower by about 3.8%, providing circumstantial
evidence that withdrawal of labour from agriculture
did affect output level. However, apart from many
statistical and conceptual problems (e.g. the relation
between acrage change in the sense of extensive
margin and output change which is a resultant of
both extensive and intensive margin in his macro-
level statistical investigation), this ‘epidemic test’
must be deemed to be over-simplistic despite its
apparent ingenuity. At best, it showed that a
random x% withdrawal of labour from cultivation
did affect the acrage and/or output level. But it does
in no way establish the impossibility of selectively
withdrawing x% labour through suitable
reorganization of agricultural production at the
family and regional level (e.g. Sen ). And yet, most
of the important initial proponents of the ‘surplus
labour’ doctrine had in mind such selective (but not
random) withdrawal of labour that may be induced
by industrialization and expansion in urban
employment opportunities (Nurkse ; Lewis ). And,
once it is recognized that such withdrawal of labour
from agriculture can be accompanied by
reorganization of labour in the family farm through
adjusting the hours of work of the family members
staying back on the farm or through higher
availability of land per cultivating family, it seems
plausible to argue analytically (e.g. Takagi ) as well
as empirically that, labour can usually be released
from agriculture without adversely affecting the
level of agricultural output. Indeed, post-
revolutionary experiences of agrarian reorganization
in China and Vietnam demonstrated the possibility
of using surplus labour to improve the quality of
land through better drainage and irrigation without
significant drop in short-run agricultural output,
despite all the serious problems of lack of adequate
incentive to private production in cooperative and
collective agriculture.
See Also
► Harris-Todaro Hypothesis
► Labour Surplus Economies
► Robinson, Joan Violet (1903-1983)
2937
References
Bharadwaj, K. 1974. Production conditions in Indian agri-
culture (A study based on farm management surveys).
Occasional Paper 33, Department of Applied Economics.
Cambridge: Cambridge University Press.
Bliss, C., and N. Stem. 1978. Productivity, wages and
nutrition. Parts I and II. Journal of Development Eco-
nomics 5(4): 331-398.
Lewis, W.A. 1954. Economic development with unlimited
supplies of labour. Manchester School of Economic and
Social Studies 22: 139-191.
Marglin, S.A. 1984. Growth, distribution, and prices.
Cambridge, MA: Harvard University Press.
Nurkse, R. 1953. Problems of capital formation in under-
developed economies. Oxford: Clarendon Press. Robinson, J.
1956. The accumulation of capital London: Macmillan.
Schultz, T.W. 1964. Transforming traditional agriculture.
New Haven: Yale University Press.
Sen, A.K. 1967. Surplus labour in India: A critique of
Schultz’s statistical test. Economic Journal 77: 154-160.
Takagi, Y. 1978. Surplus labour and disguised unemployment
Oxford Economic Papers 30(3): 447-457.
Disintermediation
Charles Goodhart
‘Intermediation’ generally refers to the interposition
of a financial institution in the process of
transferring funds between ultimate savers and
ultimate borrowers. The forms of services that such
financial intermediaries provide, the characteristics
of their liabilities and assets, and the rationale for
their existence is described elsewhere. For this
purpose, we only need to assume that a certain
pattern of financial intermediation is given, say by
actual historical development, or is theoretically
optimal.
Disintermediation is then said to occur when
some intervention, usually by government agencies
for the purpose of controlling, or regulating, the
growth of financial intermediaries, lessens their
advantages in the provision of financial services,
and drives financial transfers and business into
other channels. In some cases the transfers of funds
that otherwise would have gone through the
2938 Displacement in the Middle East: Where the Past is Prologue?
books of financial intermediaries now pass directly
from saver to borrower. An example of this is to be
found when onerous reserve requirements on banks
leads them to raise the margin (the spread) between
deposit and lending rates, in order to maintain their
profitability, so much that the more credit-worthy
borrowers are induced to raise short-term funds
directly from savers, for example, in the
commercial paper market. Another, more recent,
example arises when stringent capital adequacy
requirements lead banks to provide funds to
borrowers in a form that can be packaged into
securities of a kind that can be on-sold to ultimate
savers, rather than kept on the books of the banks
involved, and thereby need larger capital backing.
Disintermediation not only refers to those
instances where financial flows are constrained by
intervention to pass more directly from saver to
borrower (than in an unconstrained context), but
also where such flows pass through different, and
generally less efficient, channels than would
otherwise be the case. This latter is just as common
in practice. For example, where constraints and
regulations are imposed on some sub-set of
domestic financial institutions, substitute services of
a similar kind will become provided by ‘fringe’
financial institutions that are not so constrained.
More generally, in the absence of exchange control,
constraints and burdens on the provision of
domestic financial services will encourage financial
institutions to provide these same services abroad,
notably in the international Euro-markets. Indeed,
the development of the Euro-markets provides a
case study of the power of disintermediation out of
more rigorously controlled domestic financial
markets into an international milieu not subject to
such controls. The likelihood of such
disintermediation imposes a limit on the authorities’
ability to impose controls and regulations on
financial intermediaries. If such controls are to be
effective, they presumably force financial inter-
mediaries to behave in a way that they would not
voluntarily do, and hence represent a burden on
them. There will then be an incentive for the
controlled financial intermediary to seek to escape
such a burden, for example through disintermedi-
ation. This represents a perennial problem for the
monetary authorities. Logically, it might seem to
lead to a tendency for the authorities to be forced to
extremes, either to prevent disintermediation
altogether by extending the ambit of controls to all
forms and kinds of financial intermediation, or
alternatively to allow complete laissez-faire within
the financial system, despite the dangers of
financial instability that might ensue. In practice,
however, the authorities try to seek a compromise
in the form of regulations sufficiently well-designed
to maintain monetary control and financial stability,
without being sufficiently burdensome to cause
large-scale disintermediation. This is not, however,
an easy exercise and requires continuous
adjustment by the authorities as the financial
system evolves.
See Also
► financial Intermediaries
► Monetary Policy
Displacement in the Middle East: Where
the Past is Prologue?
Dawn Chatty and Maira Seeley
Abstract
The Middle East is now the major refugee-
producing region of the world, as well as the
major hosting region, with nearly 63% of the
world’s refugees (UNHCR statistics ). These
major population flows and changes going back
to the middle of the 19th century have had
significant impacts on the development of the
region throughout the 20th century and into the
21st.
Keywords
Community; Development; Diaspora; Dis-
placement; Homeland; Identity; Immigration;
Iraq; Jordan; Lebanon; Middle East;
Displacement in the Middle East: Where the Past is Prologue? 2939

Migration; Nationalism; Refugee; Regeneration; (Al-Rasheed ; Khalidi ; Lemer et al.

Ottoman; Palestine; Syria; Turkey ). These refugees, exiles and ‘exchangees’
found new homes and created new communities
without much attention or assistance from the new
JEL Classifications international order. They established themselves in
J15; J61; N34 new soil, but managed their memories so as not to
lay down new roots, but rather to strengthen the
commonality and trust in their immediate social
network. They were creating moral and economic
Displacement at the End of the Ottoman communities with social capital that oiled internal
Empire social cohesion (Chatty ).
These forced migrant movements also had pro-
The Middle East has been the focus of centuries, if found impacts on the economic development ofthe
not millennia, of movements of people. For much of region in the early 20th century. Armenians fleeing
the past 500 years the largely involuntary move- Turkish persecution began to arrive in the Arab
ment of people was supported by a system of world in the early 1900s, moving into Palestine and
government which encouraged and tolerated vari- Transjordan in the early 1910s (Al-Khatib
ations among people, drawing out differences ). While economic data from this early period
between neighbours and encouraging the formation is scarce, these Armenian refugees contributed
of unique identities based on cultural, linguistic or expertise in trades such as shoemaking, machine
religious grounds. In this heartland of the Ottoman mechanics and still photography, as well as finan-
Empire, belonging was not based on physical cial expertise, supporting the development of local
birthplace alone, but specifically included the social economies (Becker and El-Said ). Circassian and
community of origin (Humphreys ; Kedourie ). Chechen refugees arriving in Palestine, Syria and
Immigrants - forced and voluntary - were readily Jordan during the same period expanded agri-
accepted into the fabric of this multicultural empire cultural production, and much of Jordan’s develop-
and institutionalised mechanisms were set up by the ing market economy was initially structured along
state to assist in their integration (Chatty ). The ethnic lines, with displaced ethnic groups playing a
Ottoman Empire upon which such identities were critical role (Becker and El-Said ). Refugees from
based came to an end with the First World War. all three ethnic groups also played crucial roles in
Amid the nibble at the end of the Great War was Jordan’s early urban development, particularly in
a startling range of social movement. This included and around Amman.
social groups on the Russian-Ottoman border lands, The transnational links created by forced
such as the Armenian, Circassian and other migration in the Middle East also contributed to a
Northern Caucasus peoples (Barkey and Von Hagen continued re-imagination of geographic and social
; Brubaker ). Other dispossessions had their origins space in response to population movements. In this
in the lines drawn on maps by the great Western region the mass movement of people over the past
powers (e.g. the Sykes-Picot agreement) to create 150 years makes the attempt to regard the area as a
new nation states (Bocco et al. ; Gelvin ; Helms ; set of homelands or cultural regions bewildering, to
Morris ; Wilkinson ). These include the say the least. The Assyrian Arabs, once largely
Palestinians, the Kurds, the pastoral Bedouin and a found in pre-and post-colonial Iraq, have
variety of ‘stateless peoples’. In some cases, such as reappeared in Chicago; the Circassians have centred
those of the Yazidis, the Assyrians and some their diasporic headquarters in New Jersey; and
Armenian groups, migration was closely linked to Iraqi refugees and exiles have found new
regional efforts to create a pan-Arab, socialist or community nodes in London and other major
Islamic society western cities. Remittance links between these new
communities and the Middle East also contributed
significantly to the economic
2940 Displacement in the Middle East: Where the Past is Prologue?

development of the region; charitable organisations European nation states, does not translate as easi ly
formed by, for example, the Iraqi diaspora in the to the contemporary states which make up the
USA, including Chaldeans, previously provided modem Middle East. In the new states which
emerged after the demise of the Ottoman Empire,
critical financial support for Iraqis seeking refuge in
Syria and Jordan, as well as education and medical the violent displacement of people, often through
aid (Blayney ). The ‘here’ and the ‘there’ have compulsory exchange, was generally accompanied
become blurred in such translocal or diasporic by a variety of state and international assistance,
situations and the cultural certainty of the ‘centre’ which included the granting of citizenship and
becomes equally unclear. Thus the experience of housing aid, the provision of land, and sometimes
displacement was not restricted to those who have financial packages as well as employment. Thus,
moved to the periphery, but also affected those in for example, Asia Minor Greeks were taken and
the core. given space to live by the Greek state. The League
It is clear that nationalism played an important of Nations’ Refugee Settlement Commission (the
role in the politics of ‘place-making’ out of territo-effective predecessor of the United Nations High
rial spaces. Thus, the creation of natural links Commission for Refugees - UNHCR), financed
between places and people lay largely with the such resettlement by high-interest international
dominating cultural group which controlled the loans, assisted with land allocations and agricultural
state. However, contestation or opposition to these start-up packages. Between 1923 and 1930, it set up
natural links was common among the dispossessed, some 2000 villages which were created at the Greek
as evidenced by the emergence of ethnic ‘counter- state’s direction in the newly conquered zones from
nations’ such as the Circassians, die Palestinians, which Muslims had been forced to leave
die Kurds and the Armenians. Palestinians, for ‘voluntarily’ (Hirschon ; Loizos ).
example, expressed a deeply felt relationship to die Most of the dispossessed, uprooted and
‘villages of origin’ and the ‘land’ in general. deported, who straggled to build new lives and re-
In many of the states of the region, Syria being create communities in the early 1920s, however,
just one case in point, the sense of national unity were not provided with much national or
was created through the struggle for independence international assistance. They were often left to
(Brandell and Rabo ). Beginning in 1920, with the their own devices to survive and reconstruct their
awarding of the League of Nations mandate to the networks and communities. Not having interna-
French administration, the territory was divided into tional support was balanced, however, by being in
a number of states. Through common cause and the midst of supportive social environments made
hostility, the population of the territories rebelled up of discrete communities of people who had
and continue to fight the French policy of ‘dividing migrated into the region decades earlier and who
and ruling’ Syria as six separate statelets. After shared common beliefs about their identities based
more than a decade of insurrection and conflict the on ideas of religion and, also, ethnicity (Barth ;
French government agreed to reunite the territory Eriksen ). Thus when the Muslims from Crete
administratively into a single state. The exceptions arrived on the Levantine and Asia Minor coasts,
were the areas that had been attached to Mount they expected to be resettled on abandoned
Lebanon, to create the new state of Greater Lebanon properties. On their arrival, however, they often
and the Sanjak of Alexandretta. found that the formerly Greek Orthodox-owned
lands and houses which should have been available
to them had been appropriated by local people or
Displacement and the Creation of New government officials (Loizos , p. 245). But their
States more widely flung social networks along both
coasts meant that they were able to tap into a
The close link between culture, national identity and supportive environment made up of similarly
territory, which has been so characteristic of discrete
Displacement in the Middle East: Where the Past is Prologue?
communities who had arrived and settled in these
territories a century earlier.
They ‘healed’ each other and built new com-
munities based on trust, exchange and mutuality.
They, too, consciously retained a separate identity
from the rest of their surroundings and thus actively
sought to mark themselves out as an unassimilated
minority. The Cretans Muslims in Turkey re-created
their past by retaining certain selected key elements
of their culture while other parts diminished in
importance (cf. Hirschon )•
The Muslim Circassians and related peoples, the
Armenians and Eastern Christian peoples, the
Palestinians and the Kurds represented a significant
range of the ethno-religious communities that were
dispossessed, uprooted and eventually, largely
through their own efforts, re-established in the Arab
Middle East. These four communities are elaborated
in Chatty ( ), bringing the
voice of the forced migration to the fore. This work
uniquely extends our understanding of displacement
and dispossession in the modem Middle East
beyond the Palestinian case, which has rightly
dominated contemporary scholarship. The work
pulls in the nearly 3 million ‘others’ (mainly
Muslim and Christian forced migrants of the 19th
and 20th centuries) and sets the Circassians and
Chechnyans, the Armenians and other Christian
groups as well as the Kurds on a level playing field
with Palestinian refugees.
These four cases give us a deeper and more
complex understanding of the meanings of home
and homeland, myth and myth-making, community
regeneration, economic development and resilience,
as well as the local rej ection of diaspora and
transnationalism to reaffirm the process of social
and economic integration without cultural
assimilation in new physical spaces (Chatty ). It
addresses the unique roles that each of these forced
migrant communities have played in the modem
development of, for example, commercial
monopolies in particular trades (jewellery, tailoring,
textiles and photography), as well as in the security
services (Circassian and Armenians in Syrian
gendarmerie and Circassians and Chechnyans in the
Royal Household protocol and security services of
Jordan).
2941
The contributions of these displaced ethno-
religious communities to the development of their
host countries have been significant. As noted
above, the early 20th century influxes of
Circassians, Chechens and Armenians were
instrumental in economic and urban development in
the Levant, especially in previously isolated
Transjordan. The displacement of Palestinians in
1948 and after also impacted the region in several
critical ways. The majority of Palestinians seeking
refuge during and after the 1948 war settled in
Lebanon, Syria, Transjordan, Egypt and Iraq, and
the impact of Palestinians on the development of
Amman, Jordan’s capital, provides insight into the
effects of Palestinian displacement. By 1950,
500,000 Palestinian refugees were in Jordanian-
controlled territory, and additional refugees fled to
Amman in the 1950s (Hanania ). These mass
arrivals prompted urban development in the capital
and the construction of new roads and municipal
infrastructure. Palestinian refugees also contributed
significantly to the formation of a new middle class
of professionals and skilled workers in Amman
(Hanania ). Many refugees brought higher levels of
education and technical skills that fuelled economic
and social development in the capital, and the
arrival of wealthy Palestinian families drove
demand for imported and manufactured goods, as
well as investment (Hanania ). Additional refugees
arrived during and after the 1967 war, contributing
further to Amman’s expansion. Remittances from
Palestinians employed in the Gulf in later decades
were also an important factor in the development of
Jordan’s public and private sectors, as a source of
start-up funding (Chatelard ). After the 1990 Gulf
War, most of the educated professional Palestinians
who had helped to build the economy of Kuwait in
the 1960s, 70s and 80s were expelled and ‘returned’
to Jordan (many had grown up in Kuwait and had
never visited Jordan before). They invested their
savings largely into property, fuelling a construction
spurt that witnessed an exponential period of
growth in Amman in the decade that followed.
However, the influx of refugees (many of whom
had little education or formal technical training)
also led to a sharp rise in unemployment
2942 Displacement in the Middle East: Where the Past is Prologue?
in Jordan, with the majority of Palestinians living in
refugee camps and reliant on humanitarian aid, as
well as heavy pressure on Amman’s existing water
resources (Hanania ). It is also critical to note that
the development impact of Palestinian refugees
differed in other host countries, such as Lebanon
(Chaaban et al. ).
21st Century Displacement Within
the Region
The 21st century has seen a new wave of dispos-
session in the Middle East, so large and so sudden
as to threaten the economic and political stability of
many countries in the region. Lebanon, Iraq, Syria
and Jordan have seen waves of dispossessed and
displaced enter their countries over the past five
years, which have dwarfed the refugee loads
experienced in Europe in previous decades.
Critics of the 2003 Iraq war have tended to
focus on the cost in money and lives rather than on
the catastrophic consequences for Iraq. One
consequence in particular deserves more attention
than it has received: the plight of Iraq’s 4 million
refugees, most of whom have remained in the
region. Crucially, Iraqis’ recent refuge in the
neighbouring countries of Syria, Jordan and Leb-
anon rapidly became a protracted crisis, notwith-
standing the tolerance of their hosts. Unwilling to
return and largely unable to emigrate further west
or north, Iraq’s refugees in the Middle East remain
in a perilous situation.
In the aftermath of the invasion of Iraq in March
2003, the western powers prepared for 1 million
Iraqi ‘refugees’ to flee their country. Camps were
duly set up to receive those who might try to escape
the conflict. However, six months after the fall of
the Iraqi regime, few Iraqis actually had fled their
country. The international aid regime had
miscalculated the Iraqi people’s response to the
invasion; the empty emergency camps were
dismantled and pre-positioned food and equipment
were removed.
Three years later, in 2006, the West was caught
off-guard as hundreds of thousands of Iraqis fled
their homes to escape the deadly sectarian violence
which had escalated with the al-Askari
mosque bombing in Samarra in the February of that
year. That single event became the iconic image of
sectarian violence and the ‘unmixing’ of people
which followed. Nearly 4 million Iraqis fled their
homes in 2006 and 2007, with 1-1.5 million
crossing national borders into Syria and Jordan. The
UNHCR and affiliated NGOs raced to set up
reception centres and to provide emergency aid. In
both Syria and Jordan, Iraqis were not regarded
legally as refugees by the host governments,
partially because neither country was a signatory to
the 1951 United Nations Convention relating to the
Status of Refugees.
Many of the Iraqis seeking asylum were from
the educated, professional middle class. The extent
of the extraordinary ‘brain drain’ from Iraq during
this period has been well documented (Sassoon ). A
number managed to escape with savings, which
helped to ease their transition in exile. Migrations
during previous decades meant that some Iraqi
social networks were already in place in the host
countries. The residual cultural memory of the
‘millet’ system of the Ottoman Empire, which gave
minority/religious communities a limited amount of
power to regulate their own affairs, meant that Iraqi
arrivals in these cities were generally tolerated, if
not actively comforted. Also, memory of the Pan-
Arab aspirations in the region meant that Iraqis
were seen as temporary guests and ‘Arab brothers’.
The Iraqi displacement crisis has reached a
critical stage. International interest in Iraq is
declining. Yet the lack of security, continuing civil
conflict and economic uncertainty makes it unlikely
that a mass Iraqi return will occur. More likely,
Iraqi refugees will remain in neighbouring states
under increasingly difficult circumstances. As their
savings diminish and their circular movements into
and out of Iraq to make money become more
precarious, it is likely that irregular and long-
distance migrations will occur in larger numbers.
Iraqi displacement has had a number of effects
on host countries’ development. As noted above,
movements in and out of Iraq have both enabled
some of the displaced to continue to transfer income
generated within Iraq to neighbouring
Displacement in the Middle East: Where the Past is Prologue?
countries, while family members are based in Syria,
Lebanon and Jordan (Crisp et al. ). Many Iraqi
refugees in these countries have come from middle
class origins, with some bringing considerable
investment to their countries of refuge, fuelling
economic development. Most of the 25,000 Iraqis
who have gained residence rights in Jordan since
the beginning of the Iraqi conflict in 2003 have
done so through investment (Chatelard ). In Jordan,
a majority of displaced Iraqis surveyed by UNHCR
in 2009 were professionals and 35% held a
university degree (Crisp et al. ).
However, very few Iraqis in the three host
countries are able to access employment, and many
suffer from high levels of need. As of 2008, 42% of
Iraqis in Jordan depended on remittances from Iraq
for survival, while 24% of Iraqis in Syria depended
on remittances from abroad (Harper ). In Syria, the
arrival of refugees caused a dramatic increase in the
price of basic necessities and rents, as well as strain
on public services in education and healthcare,
lowering the quality of services provided (Al-
Miqdad ).
The speed with which Syria disintegrated into
violent armed conflict after 2011 shocked the
world; it has also left the humanitarian aid regime in
turmoil as agencies struggled to react effectively to
the massive displacement crisis. By 2015 there were
more than 5 million Syrians seeking refuge in
neighbouring countries, as well as another 6 million
internally displaced in the country. The
international aid regime has attempted to provide
assistance to refugees who register with the UN if
they are deemed needy. Perhaps only 5-10% of
these millions have been able to access food
vouchers and basic survival kits. For the most part,
refugees in the neighbouring countries are not
permitted to work, making their reliance on the UN
particularly significant.
Each country bordering on Syria has responded
differently to this complex emergency: Turkey
rushed to set up its own refugee camps for the most
vulnerable groups, but generally permitted self-
settlement; Lebanon refused to allow the
international humanitarian aid regime to set up
formal refugee camps; and Jordan facilitated the
2943
creation of a UN refugee camp near its border with
Syria. Turkey, Lebanon and Jordan have all granted
refugees “guest” status under domestic legislation.
And although Turkey has signed the 1951
Convention, it has reserved its interpretation of the
Convention to apply only to Europeans seeking
refuge/asylum in Turkey. UN estimates are that
over 80% of the Syrian refugee flow across
international borders is self-settling in cities, towns
and villages where they have social networks.
Despite a general rejection of encampment among
those fleeing, still some 15-20% of the Syrian
refugee population is in camps.
Each of these states has established a variety of
temporary measures to deal with the crisis, which
has reached proportions far outstripping the dis-
placement crisis at the end of the Second World
War. With refugees from Syria now estimated at
nearly 3 million in Turkey and 1.5 million in
Lebanon and officially nearly 700,000 in Jordan, as
well as nearly 1 million now in Europe, the crisis
threatens not only the security of the hosting nations
in the region but also the unity of the European
Union. Syrians who have sought refuge in the
neighboring states are largely not permitted to work
(although Jordan has recently sought to increase the
accessibility of work permits for Syrian refugees).
Many are unable to access adequate education, food
and healthcare for their families.
The Syrian refugee crisis has also had wide-
ranging effects on the socioeconomic development
of neighbouring countries in the region, who have
received the vast majority of refugees from the
conflict. In Lebanon, increasing demand for public
services has both strained public finances and
lowered the quality of services provided (World
Bank ). The Lebanese Ministry of Health and Social
Affairs reported a 40% increase in use of its health
and social programmes (World Bank ). This strain
on social safety nets and services is ultimately
pushing tens of thousands of Lebanese below the
poverty line and may be increasing the youth
unemployment rate in a country already
experiencing significant poverty (World Bank ). It
is important to note, however, that high
unemployment has long been present in the Middle
East and that current unemployment may not have
been caused directly
2944 Displacement in the Middle East: Where the Past is Prologue?

people dispossessed of their property as a result of

by locals’ competition with Syrian refugees (IRC )•
Overall, competition from Syrian refugees hasthe upheaval leading to and including the end of
driven down wages in the informal sector and empire and ensuing neo-colonial enterprises
negatively affected working conditions in host endorsed by the League of Nations. The Middle
countries (IRC ). In Jordan, overall unemploymentEast has provided comfort and relief both on an
has risen from 14.2% to 22.1% since the beginningindividual basis and also for social groups. Perhaps
as a residual trait of the tolerance which the
of the refugee crisis, with more significant increases
among youth and less educated sectors of the Ottoman empire had enshrined in its millet system
towards multi-ethnic and plural society, the states to
population, and there is some evidence that Syrians
may have replaced Jordanians in specific sectors,emerge from the Arab Ottoman provinces all
such as construction, wholesale and retail (Stavetolerated, if not actively, the development of these
and Hillesund ). Child labour is more prevalent minority cultures.
among refugee families, whose children have lower Only in the mid-20th century did a different
school enrolment rates than Jordanian children: instrument for managing and ordering the displaced
and disposed emerged - the refugee camp. Here, a
while 95% of Jordanian children are still in school
system of control and standardised routine emerged
at 17, less than 40% of Syrian children are still in
school at age of 15 (Stave and Hillesund ). As inas the principal tool for managing large numbers of
displaced and refugee populations around the world.
Lebanon, the refugee influx in Jordan has strained
In the Middle East, the United Nations Relief and
social and public services, particularly in education,
Works Agency, established in 1949, was set up to
deteriorating the quality of services provided
deal with nearly 1 million Palestinians displaced by
(REACH )•
the 1947-48 War. Here, the basics of life, food,
The presence of Syrian refugees has, however,
shelter, healthcare and primary education were pro-
brought benefits for host countries that may have a
vided by the Agency, but the interstitial nature of
positive impact on their long-term development. In
the lives of the individual refugees was not
Jordan, direct investment by Syrians has stimulated
addressed (Brand ; Farsoun and Zacharia ; Peteet ;
industry and created employment opportunities for
Rosenfeld ).
both Jordanians and Syrians (IRC ). In Turkey, 26%
For the earlier wave of involuntary migrants of
of newly registered businesses in 2014 were either
the Middle East, return to the homelands of origin
Syrian-owned or possessed Syrian capital (Del
was a hope, a nostalgic dream or a unifying myth.
Carpio and Wagner ). The increase in humanitarian
Those early Muslim refugees of the 19th and early
aid flows to host countries has also boosted local
20th century knew they could not go back. They
economies: each US$1 spent on humanitarian
had to create their homelands in new spaces. None
assistance had a multiplier value of 1.6 for the
of the populations exchanged after the 1923 Treaty
Lebanese economy (UNHCR and UNDP ). In
of Lausanne had any ambiguity about their
addition, each US$1 of cash assistance spent by
condition. The hminality might have been physical,
Syrian refugees in Lebanon generated US$2.13 of
but there was no question of their future. They had
Lebanon’s GDP (IRC ). Lebanon, Jordan and
to create a new community, both imagined and
Turkey have all proven relatively economically
moral, in which new ties or kinship and trade could
resilient during the refugee crisis (IRC ).
emerge. The Kurds, perhaps more than any other
group, held out for a return and alternated between a
realistic hope and a nostalgic dream. Their
Conclusion
homeland remains divided between four modem
Over the past 150 years the Middle East has pro- states.
vided refuge and asylum to numerous groups of Many Palestinian refugees live within a hundred
miles of their original villages and urban
neighbourhoods. Some can even see the lights of
Displacement in the Middle East: Where the Past is Prologue?
their hometowns and settlements at night. Some
Armenians have travelled back to visit the
homeland - both in Turkey and in the Republic of
Armenia. So, too, have the Circassians and other
Caucasians. A few Kurds, recent migrants to Syria,
have managed to smuggle themselves across the
border, sometimes on the backs of Peshmergas
fighters, to visit their mountainous places of birth.
The effort to reverse the misfortune of displacement
and dispossession and to em-place themselves has
become a strategy for survival and its success is a
measure of the resilience of the forced migrant as
exhibited by the new communities established by
Circassians, Armenians, Palestinians and Kurds in
the Arab Middle East.
How successful forced migrants are in re-
creating and re-placing themselves depends on the
nature of the displacement and dispossession itself.
The way people experience movement to a new
place and the extent to which this is a shocking and
disruptive experience is determined by the
conditions under which they move and whether they
can extend their notions of territorial attachment to
new areas not necessarily adjacent to each other.
Thus the Cretan Muslims were able to re-create
their identity in several new locations outside of
Crete, on the northern coast of Lebanon and Syria
as well as on an island off the coast of Izmir in
Turkey. For most forced migrants, however, the
move is generally conducted in more traumatic
conditions. The task of re-creating a place, a home
or a neighbourhood, of ‘producing a locality’, is
dominated by the effort to re-establish some
continuity with the past places of origin. This work
of continuity maintenance and management of
memory is clearly articulated in the writings of
Hirschon ( ), Parkin ( ),
Malkki ( ), Loizos ( ) and Chatty ( ).
The nature of post-Ottoman Arab society - as
separate from its politics - has been such that it has
tolerated and acknowledged multiple layers of
belonging in the straggle to make new places in the
world. Although not physically displaced, the
peoples of the Arab provinces of the post- Ottoman
Empire have spent most of the 20th century creating
new identities, and em-placing themselves in a new
social order. This process of
2945
re-placing and re-creation has had a variety of
impacts on the development of Middle Eastern
nations absorbing waves of refugees, from the first
arrivals of Circassians, Chechens and Armenians in
the early 20th century, to the current Syrian refugee
crisis. Despite the challenges of integrating large
and often destitute populations into still-developing
regional economies, the contributions of early
forced migrants to the economic and urban
development of Jordan, Lebanon and Syria, as well
as the contributions of professionally skilled
Palestinians to their host countries’ development,
both in the region as well as in the Arabian Gulf,
demonstrate the strong link between forced
migration and both local and regional development.
More recent arrivals of Iraqi and Syrian refugees
have had complex impacts on the development of
host countries, many of which are still emerging.
Despite the significant strains on public finances
and services that these refugee influxes have
created, as well as social tensions, there are
indications that the presence of Iraqi and, in
particular, Syrian refugees may positively impact
their host countries’ longterm development.
Ethnic minority communities in the Middle East
have found ways to economically, physically and
socially integrate themselves in their new
surroundings, but at the same time resist the natural
phenomena of assimilation over the long term.
Patronage and real as well as ‘Active’ kinship
networks are powerful positive forces; so too are the
religious and charitable associations which these
groups set up to help those less fortunate in their
communities.
Whether the current wave of dispossessed from
Syria can weather the storms of dislocation with
similar support and equanimity to that of their
forefathers remains to be seen. Much will depend
upon the way in which international humanitarian
emergency assistance can unfold and develop into
conceited measures to educate and assist the
displaced in finding sustainable livelihoods. An
educated population has agency and will contribute
to the development of its host state. A current
refugee population with no access to education or
employment remains vulnerable and passive and a
drain on the national economy. It remains to be
2946 Displacement in the Middle East: Where the Past is Prologue?

seen whether lessons from the late Ottoman reforms Chaaban, J., H. Ghattas, R. Habib, S. Hanafi, N. Sahyoun, N.
will be learned regarding the integration of refugees Salti, K. Seyfert, and N. Naamani. 2010. Socioeconomic
survey of Palestinian refugees in Lebanon. Beirut:
and other forced migrants, recognising their American University of Beirut and the United Nations
potential contributions to the long-term Rehef and Works Agency for Palestinian Refugees in the
development of host countries. Near East (UNRWA).
Chatelard, G. 2010. Jordan: A refugee haven. Migration
Pohcy Institute. Available at:
Accessed
See Also 19 July 2016.
Chatty, D. 2010. Dispossession and displacement in the
modem Middle East. Cambridge: Cambridge University
Press.
Crisp, J., J. Janz, J. Riera, and S. Samy. 2009. Surviving in
the city: A review of UNHCR s operation for Iraqi
refugees in urban areas of Jordan, Lebanon and Syria.
Bibliography Geneva: Policy Development and Evaluation Service,
UNHCR . Available at: http://
Al-Khatib, M.A. 2000. Language shift among the Armenians . Accessed 19 July 2016.
of Jordan. International Journal of the Sociology of Del Carpio, X.V., and M.C. Wagner. 2015. The impact of
Language 152: 153-177. Syrian refugees on the Turkish labor market. Pohcy
Al-Miqdad, F. 2007. Iraqi refugees in Syria. Forced Migra- Research working paper no. WPS 7402. World Bank
tion Review Special Issue (June 2007): 19-20. Available Group, Washington, DC. Available at:
at: me:: ;c ”::. M :: /: :: g: ;sc7s:: 315 22 5 :. . 7/
FMRdownloads/en/syria/dahi.pdf. Accessed 20 July . Accessed
2016. 20 July 2016.
Al-Rasheed, M. 1994. The myth of return: Iraqi Arab and Eriksen, T.H. 1993. Ethnicity’ and nationalism: Anthropo-
Assyrian refugees in London. Journal of Refugee Studies logical perspectives. London: Pluto.
1(2 J>). 199 219. Farsoun, S.K., and C.E. Zacharia. 1997. Palestine and the
Barkey, K., and M. Von Hagen, ed. 1997. After Empire: Palestinians. Westview: Boulder.
Multiethnic societies and nation-building: The Soviet Gelvin, J.L. 1998. Divided Loyalties: Nationalism and mass
Union and the Russian, Ottoman, and Habsburg politics in Syria at the close of empire. Berkeley:
Empires. Boulder: Westview. University of California Press.
Barth, F., ed. 1969. Ethnic groups and boundaries: The Hanania, M.D. 2014. The economic impact of the Palestinian
social organization of culture difference. Oslo: Scandi- Refugee Crisis on the development of Amman, 1947-
navian University Press. 1958. British Journal of Middle Eastern Studies 41(4):
Becker, K., and H. El-Said. 2013. Management and inter- 461-482.
national business issues in Jordan. London: Routledge. Harper, A. 2008. Iraq’s refugees: Ignored and unwanted.
Blayney, C. 2011. Iraqi American diasporic philanthropic International Review of the Red Cross 90: 869 (March):
remittances to Iraqi refugees in Jordan: past projects 169-190. Available at:
and potential for future partnerships. American Uni- . Accessed 20 July 2016.
versity in Cairo, John D. Gerhart Center for Philan- Helms, C. 1981. The cohesion of Saudi Arabia. London:
thropy and Civic Engagement. Available at: Croom Helm.
>26/42 . Accessed 19 July Hirschon, R. 1998. Heirs of the Greek Catastrophe: The
2016. social life of Asia minor refugees in Piraeus. Oxford:
Bocco, R., R. Jaubert, and F. Metral, ed. 1993. Steppes Berghahn.
d’Arabies, Etats, Pasteurs, Agriculteurs et Hirschon, R. 2001. Surpassing Nostalgia: Personhood and
Commerqants: le Devenir des Zones Seches. Paris: the experience of displacement (Colson Lecture). Oxford:
Presses Universitaires de France. unpublished.
Brand, L.A. 1988. Palestinians in the Arab World: Institution Humphreys, R.S. 1999. Between memory and desire: The
Building and the Search for State. New York: Columbia middle east in a troubled age. Egypt: University of
University Press. California Press.
Brandell, I., and A. Rabo. 2003. Nations and nationalism: International Rescue Committee (IRC). 2016. Economic
Dangers and virtues of transgressing disciplines. Ori- impacts of Syrian refugees: Existing research review and
entalia Suecana LI-LII: 35-46. key takeaways. IRC Policy Brief No. 1 (January 2016).
Brubaker, R. 1995. Aftermaths of Empire and the unmixing Available at:
of peoples: Historical and comparative perspectives.
Ethnic and Racial Studies 18(2): 189-218. . Accessed 20 July 2016.
Dispute Resolution 2947

Kedourie, E. 1984. Minorities and majorities in the Middle

East. European Journal of Sociology 25(2): 276-282. Dispute Resolution
Khalidi, R. 1997. Palestinian identity: The construction of
modern national consciousness. New York: Columbia
University Press.
Amy Farmer and Paul Pecorino
Lemer, D., L.W. Pevsner, and D. Riesman. 1958. The passing
of traditional society: Modernizing the middle east. In
Free Press. New York: Collier-Macmillan.
Loizos, P. 1999. Ottoman half-lives: Long term perspectives
on particular forced migrations. Journal of Refugee Abstract
Studies 12(3): 237-263. The high cost of disputes creates an incentive for
Malkki, L.H. 1995. Purity and exile: Violence, memoty, and
parties to disputes to settle. In civil litigation and
national cosmology among Hutu refugees in Tanzania.
Chicago: University of Chicago Press. arbitration, settlement failure may arise from
Morris, B. 1987. The birth of the Palestinian refugee prob- asymmetric information or optimism. Devices to
lem, 1947-1949. Cambridge: Cambridge University induce settlement include voluntary disclosure
Press. and mandatory discovery. The effects of these
Parkin, D. 1999. Mementoes as transitional objects in human
displacement. Journal of Material Culture 4(3): 303-320. are considered, as are the English rule (whereby
Peteet, J.M. 1991. Gender in crisis: Women and the Pales- the loser at trial pays the reasonable legal costs
tinian resistance movement. New York: Columbia Uni- of the winner), the use of contingency fees, and
versity Press. the operation of conventional arbitration and
REACH. 2014. Evaluating the effect of the Syrian refugee
crisis on stability and resilience in Jordanian host com- final offer arbitration. Researchers continue to
munities: Preliminary impact assessment. REACH. propose new arbitration mechanisms in the hope
Available at: of improving the dispute resolution process.
. Accessed 20 July 2016
Rosenfeld, M. 2004. Confronting the Occupation: Work,
Education and Political Activism of Palestinian Families
in a Refugee Camp. Stanford: Stanford University Press.
Sassoon, J. 2009. The Iraqi refugees: The new crisis in the Keywords
middle east. London: I.B.Tauris.
Stave, S. E. and Hillesund, S. 2015. Impact of Syrian refugees Asymmetric information; Bargaining; Signal-
on the Jordanian labour market. International Labour ling; Screening; Arbitration; Final offer arbi-
Organization (ILO). Available at: tration; Contingency fees; English rule; Fee-
shifting; Optimism; Self-serving bias
Accessed
20 July 2016.
UNHCR. 2014. UNHCR statistical yearbook. 14th ed.
Geneva: UNHCR. JEL Classification
United Nations High Commissioner for Refugees (UNHCR) C7
and United Nations Development Program (UNDP).
2015. Impact of humanitarian aid on the Lebanese
economy. UNHCR and UNDP. Available at:
Disputes may arise in a variety of settings,
including labour negotiations, civil disputes and
. Accessed 20 July 2016. family conflict. If individuals fail to reach an
Wilkinson, J. 1983. Traditional concepts of territory in South agreement, there exist a variety of mechanisms for
East Arabia. The Geographical Journal 149(3): 201-315.
World Bank. 2013. Lebanon: Economic and social impact resolving the dispute. These include civil litigation,
assessment of the Syrian conflict. Washington, DC: arbitration and, in labour relations, the strike.
World Bank. Available at: Resolving disputes in these ways is costly, thereby
creating a contract zone within which both parties
strictly prefer to settle. Given the high cost of
disputes, considerable research has been devoted to
understanding why settlement sometimes fails to
occur and how different mechanisms affect the
20 July 2016. dispute
2948 Dispute Resolution

rate. Here we focus on dispute resolution in the screening offer is more likely to be optimal for the
context of civil litigation and arbitration. defendant when there is a high prevalence of weak
plaintiffs, when court costs are low, and when the
difference in expected trial awards for the two
Why Settlement Fails plaintiff types is large.
If the informed plaintiff is allowed to make the
The dominant rational choice explanation for set-
offer, this is called the signalling game. While these
tlement failure is asymmetric information. An
games generally have multiple equilibria, the D1
alternative explanation, not consistent with the
refinement (Cho and Kreps ) has been employed to
assumption of full rationality, is optimism. If agents
focus on a separating equilibrium in which the weak
have symmetric information and beliefs about the
plaintiff submits a low demand to the defendant,
expected outcome of a dispute, theory suggests a
while the strong plaintiff submits a high demand.
settlement will occur. However, if one party has
Under Dl, it is assumed that an out of equilibrium
private information about the expected outcome of
offer is made by the plaintiff willing to make that
the dispute, settlement failure can occur. Similarly,
offer for the largest set of acceptance probabilities.
if one or both parties to the dispute are subject to
In equilibrium, the high demand must be rejected
optimism, a contract zone may fail to exist.
with a sufficiently high probability so as to
There are two basic models in the asymmetric
discourage the weak plaintiff from also making this
information literature, which make different
demand. These rejections lead to a positive
assumptions about the structure of information.
probability of trial with the strong plaintiff.
When the uninformed party makes the offer, we
While we used a two-type model to motivate the
have a screening model which was developed by
discussion, the Bebchuk and Reinganum and Wilde
Bebchuk ( ). When the informed party makes
models employ a continuum of types whose
the offer we have a signalling model developed by
distribution is known by the uninformed party.
Reinganum and Wilde ( ). Both models’ pre
These models have been extended in numerous
dictions are consistent with the existence of costly
ways by allowing for two-sided information
disputes in equilibrium.
asymmetries (Schweizer ; Daughety and Reinganum
To explore the intuition behind these models,
) and multiple offers (Spier ) among other
consider a civil dispute in which the failure of
extensions. While the effects of policy variables
negotiations would result in a trial. Suppose a
(such as cost shifting at trial) are often sensitive to
plaintiff known to be harmed has private informa-
the modelling details, the prediction that
tion concerning the damages she has incurred and
asymmetric information can result in costly disputes
that this information would be revealed at trial.
is quite robust. Excellent surveys of the literature
Further, suppose the plaintiff is one oftwo types: a
are provided by Spier ( )
weak type with a low expected payoff at trial or a
and Daughety ( ).
strong type with a high expected payoff. A risk-
The empirical studies by McConnell ( ),
neutral plaintiff will accept a settlement offer if and
Conlin ( ) and Osborne ( ) support the
only if it equals or exceeds her expected net payoff
model of asymmetric information.
from trial. The defendant knows the probability that
The optimism or self-serving bias explanation
he is facing a weak or strong plaintiff but not the
for settlement failure relies on bargainers who have
plaintiffs exact type. In a screening model, the
potentially inaccurate beliefs about the expected
uninformed defendant makes an offer to the
outcome at trial. For example, the plaintiff’s belief
plaintiff. He will choose between a low (screening)
about the probability she will prevail at trial may
offer that only weak plaintiffs would accept and a
exceed the defendant’s belief about this same
high (pooling) offer that both types would accept. If
probability. If these differences in beliefs are not
he makes the low offer, then a strong plaintiff
based on differences in information, then we are in
would proceed to trial. The
the realm of the optimism
Dispute Resolution
model. Versions of this model have been employed
by Landes ( ), Posner ( ),
Shavell ( ), and Priest and Klein ( ). Opti-
mism violates rationality, but Bar-Gill ( ) the dispute rate.
finds that cautious optimism can allow the opti-
mistic party to obtain a larger portion of the joint
surplus from settlement. As a result, cautious
optimism can persist in an evolutionary setting.
Babcock and Loewenstein ( ) survey an
experimental literature documenting the existence
of a self-serving bias which leads players in the role
of a plaintiff to expect a greater payoff at trial than
the defendant, even though both are exposed to the
same set of facts. When players are exposed to the
facts of the case before being assigned then- role as
plaintiff or defendant, the self-serving bias tends to
disappear.
Waldfogel ( ) and Farmer et al. ( ) find
empirical evidence that is consistent with the opti-
mism model. Note that the optimism and asym-
metric information explanations are not mutually
exclusive. It is possible that each factor is respon-
sible for some proportion of observed disputes.
Mandatory Discovery and Voluntary
Disclosure
If asymmetric information causes disputes, it is
logical to ask whether voluntary disclosures and
mandatory discovery can eliminate these
asymmetries. In a screening model where credible
disclosure is costless, Shavell ( ) shows
that plaintiffs with strong cases will reveal enough
information to ensure that all cases settle. Plaintiffs
who do not reveal their information (those with
weak cases) receive a pooling offer that all accept.
However, the work of Sobel ( ) shows that
this result is not robust to the
introduction of positive costs of disclosure. He also
shows that a costless (to the plaintiff) discovery
procedure will lead to greater settlement. Farmer
and Pecorino ( ) consider costly dis
covery and disclosure in both the signalling and the
screening games. Costly disclosures may be made
in the signalling game but not the screening game,
while costly discovery may be invoked in the
screening game but not the signalling game.
2949
If the cost of these procedures is not too high, the
combination of the two will lead to a great deal of
information transmission and a large reduction in
Why then do disputes persist? Perhaps, as
Shavell ( ) suggests, private information has
strategic value if withheld until trial. Hay ( )
develops a model in which an initial informational
asymmetry on the merits of the case is resolved by
mandatory discovery, but by the time this occurs a
new asymmetry - namely, the extent of attorney
preparation - has emerged. This second asymmetry
leads to trials in the equilibrium of the model. Hay
notes that the extent of attorney preparation is not
subject to discovery. This is also true of pref-
erences. Farmer and Pecorino ( ) show that
asymmetric information on risk preferences can
lead to trial, and that this information is neither
subject to discovery nor easy to credibly transmit.
As a result, this type of asymmetry may tend to
persist in the face of mandatory discovery and
opportunities for voluntary disclosure.
Other Institutional Features
There is a voluminous literature which examines
how a variety of institutional features affect settle-
ment in civil litigation. What follows is a much
abbreviated discussion of a large and complex
literature. One difficulty in addressing this question
is that even a single institution is likely to have
multiple effects on the litigation process. Thus, a
single institution may have conflicting effects on
the dispute rate and may also have important influ-
ences on other aspects of the litigation process.
A classic example of this difficulty is reflected
in the analysis of the English rule under which the
loser at trial pays the reasonable legal costs of the
winner. If the probability of a finding for the
plaintiff at trial is private information, then fee
shifting at trial reduces settlement rates by, in effect,
spreading out the distribution of player types
(Bebchuk ). If players are optimistic in their
assessments of the probability that the plaintiff will
prevail, then fee shifting will aggravate this
optimism and reduce the probability that a contract
zone will exist (Shavell ).
2950
This prediction - that fee shifting will increase
the probability of trial - is made with expenditure at
trial held constant. It is well established that the fee
shifting at trial will increase expenditure
(Braeutigam et al. ). If the expenditure effect is
strong enough, it can result in fewer (but more
costly) disputes (Hause ). The English rule also
affects the mix of cases which are filed. It
discourages cases where there are large stakes but a
low probability of success, and encourages low
stakes cases with a high probability of success
(Shavell ).
Many of the theoretical predictions on fee
shifting at trial appear to be borne out in the data
(see Hughes and Snyder ).
Under a contingency fee, the plaintiff’s lawyer
receives a percentage of the judgment at trial if the
plaintiff wins the case and nothing if she loses. The
effects of contingency fees on the litigation process
are very complex and wide ranging (see Rubinfeld
and Scotchmer , for a survey). However, one effect
of contingency fees on settlement is clear: if the
attorney controls the settlement decision, he will
have an excessive incentive to settle the case
relative to the interests of his client. The reason is
that the attorney bears most of the costs of a trial
but is paid only a fraction of the award. On the
other hand, if the client controls the case, she may
have an excessive incentive to reject a settlement
offer and bring the case to trial. (This is particularly
true if the contingency percentage is not lower for
cases which settle early. )
When a single defendant faces multiple plain-
tiffs in sequence, some interesting issues regarding
settlement arise. Spier ( , ) and
Daughety and Reinganum ( ) have analysed
the use of most favoiued nation (MFN) clauses in
the context of repeat litigation. Suppose a plaintiff
settles early under MFN. If another plaintiff later
settles for more, the early settlement is adjusted
upward. An MFN clause can be a mechanism
whereby the defendant commits to not raising his
offer to plaintiffs who settle later in the process.
While there is some ambiguity of the effects of
MFN on settlement rates and the overall dispute
costs (see especially Spier ), the general thrust of
these papers suggests that MFN clauses
Dispute Resolution
are efficiency enhancing in the sense that they will
reduce the expected dispute costs associated with
litigation.
Arbitration
Under conventional arbitration (CA), the arbitrator
is free to impose her preferred settlement on the
bargaining parties. Under final offer arbitration
(FOA), each party to the dispute submits an offer to
the arbitrator who must pick one of the submitted
offers. While there is some evidence that submitted
offers affect the outcome in CA (Farber and
Bazerman ), for the purpose of the following
discussion we assume that they do not. From a
modelling standpoint, this makes CA look exactly
like a simple version of civil litigation. Under FOA,
the submitted offers clearly affect the outcome, a
feature which has important implications for dispute
resolution.
Consider the two-type version of the screening
model where the plaintiff can have a strong or a
weak case. In CA, the defendant will either make an
offer that only a weak plaintiff will accept or a
pooling offer that both types will accept. If all nego-
tiation takes place prior to the submission of offers
to the arbitrator, then under FOA it is possible that
the defendant will make an offer that neither type
will accept, resulting in a 100% dispute rate
(Farmer and Pecorino ). This can occur because the
sequentially rational offer submitted to the arbitrator
influences the acceptable settlement prior to
arbitration. The lack of early settlement allows the
defendant to commit to an offer which is optimal
against the entire distribution of plaintiff types.
Farmer and Pecorino ( ) also show (in contrast
to CA) that
costless voluntary disclosure never takes place
when FOA is the dispute resolution mechanism.
The reason is that information has strategic value in
this game. Both of the impediments to settlement
discussed above disappear if bargaining is permitted
after offers are submitted to the arbitrator but prior
to the arbitration hearing.
While not totally conclusive on this point, the
results of Farmer and Pecorino ( ) also sug
gest that allowing for bargaining after offers are
submitted to the arbitrator can increase settlement
Dispute Resolution
for reasons different from those discussed above.
Because a submitted offer affects the outcome of
arbitration, it will tend to reflect private informa-
tion. This may in turn promote settlement. Taken
together, the results on FOA suggest that the effects
of this institution on settlement are sensitive to
whether or not bargaining occurs in the face of
offers submitted to the arbitrator. In major league
baseball, a prominent use of FOA, a good deal of
bargaining and settlement occurs after offers have
been submitted to the arbitrator.
FOA was proposed by Stephens ( ) and has
since become an important alternative to CA.
Researchers continue to propose new arbitration
mechanisms in the hope of improving the dispute
resolution process. Combined arbitration (Brams
and Merrill ) is a mixture of FOA and CA. Other
proposed mechanisms include tri-offer arbitration
(Ashenfelter et al. ) and amended final offer
arbitration (Zeng ).
See Also
► ipistemio Game Theory: An Over. ie v
► Epistemic Game Theory: Incomplete
Information
Bibliography
Ashenfelter, O., J. Currie, H. Farber, andM. Spiegel. 1992.
An experimental comparison of dispute rates in alter-
native arbitration systems. Econometrica 60: 1407-1433.
Babcock, L., and G. Loewenstein. 1997. Explaining
bargaining impasse: The role of self-serving biases. Journal
of Economic Perspectives 11: 109-126. Bar-Gill, O. 2002.
The success and survival of cautious optimism: Legal rule,
and endogenous perceptions in pre-trial settlement
negotiations. Public law working paper No. 35. Cambridge,
MA: Harvard Law School. Bebchuk, L. 1984. Litigation and
settlement under imperfect information. RAND Journal of
Economics 15: 404-415.
Braeutigam, R., B. Owen, and J. Panzar. 1984. An economic
analysis of alternative fee shifting systems. Law and
Contemporary Problems 47: 173-185.
Brams, S., and S. Merrill. 1986. Binding versus final-offer
arbitration: A combination is best Management Science
32: 1346-1355.
Cho, I., and D. Kreps. 1987. Signaling games and stable
equilibria. Quarterly Journal of Economics 102: 179-222.
2951
Conlin, M. 1999. Empirical test of a separating equilibrium in
National Football League contract negotiations. RAND
Journal of Economics 30: 289-304.
Daughety, A. 1999. Settlement. In Encyclopedia of law and
economics, vol. 5, ed. B. Bouckaert and G. de Geest.
Cheltenham: Edward Elgar.
Daughety, A., and J. Reinganum. 1994. Settlement nego-
tiations with two-sided asymmetric information: Model
duality, information distribution, and efficiency. Inter-
national Review of Law and Economics 14: 283-298.
Daughety, A., and J. Reinganum. 2004. Exploiting future
settlements: A signaling model of most-favored-nation
clauses in settlement bargaining. RAND Journal of
Economics 35: 467-485.
Farber, H., and M. Bazerman. 1986. The general basis of
arbitrator behavior: An empirical analysis of conventional
and final offer arbitration. Econometrica 54: 819-854.
Farmer, A., and P. Pecorino. 1994. Pretrial negotiations with
asymmetric information on risk preferences. International
Review of Law and Economics 14: 273-281.
Farmer, A., and P. Pecorino. 1998. Bargaining with infor-
mative offers: An analysis of final offer arbitration.
Journal of Legal Studies 27: 415-432.
Farmer, A., and P. Pecorino. 2003. Bargaining with voluntary
transmission of private information: Does the use of final
offer arbitration impede settlement? Journal of Law,
Economics, and Organisation 19: 64-82.
Farmer, A., and P. Pecorino. 2005. Civil litigation with
mandatory discovery and voluntary transmission of pri-
vate information. Journal of Legal Studies 34: 137-159.
Farmer, A., P. Pecorino, and V. Stango. 2004. The causes of
bargaining failure: Evidence from major league baseball.
Journal of Law and Economics 47: 543-568.
Hause, J. 1989. Indemnity, settlement, and litigation, or I’ll be
suing you Journal of Legal Studies 18: 157-179.
Hay, B. 1995. Effort, information, settlement, trial. Journal of
Legal Studies 24: 29-62.
Hughes, J.W., and E.A. Snyder. 1998. Allocation of litigation
costs: American and English rules. In The new Palgrave
dictionary of economics and the law, vol.
1, ed. P. Newman. London: Macmillan.
Landes, W. 1971. An economic analysis of the courts.
Journal of Law and Economics 14: 61-107.
McConnell, S. 1989. Strikes, wages and private information.
American Economic Review 79: 801-815.
Osborne, E. 1999. Who should be worried about asymmetric
information in litigation? International Review of Law
and Economics 19: 399—409.
Posner, R. 1973. An economic approach to legal procedure
and judicial administration. Journal of Legal Studies 2:
399-458.
Priest, G., and B. Klein. 1984. The selection of disputes for
arbitration. Journal of Legal Studies 13: 215-243.
Reinganum, J., and L. Wilde. 1986. Settlement, litigation, and
the allocation of litigation costs. RAND Journal of
Economics 17: 557-566.
Rubinfeld, D., andS. Scotchmen 1998. Contingent fees. In
The new Palgrave dictionary of economics and the law,
vol. 1, ed. P. Newman. London: Macmillan.
2952 Distortions

consumer. Also, the failure to achieve aggregate

Schweizer, U. 1989. Litigation and settlement under two-
sided incomplete information. Review of Economic production efficiency, in the sense of not producing
Studies 56: 163-178.
on the boundary of the set of production
Shavell, S. 1982. Suit, settlement, and trial: A theoretical
possibilities given available resources and tech-
analysis under alternative methods for the allocation of
legal costs. Journal of Legal Studies 11: 55-82. nology, is deemed a distortion.
Shavell, S. 1989. Sharing ofinformation prior to settlement or
Given non-interdependent consumer preferences
litigation. RAND Journal of Economics 20: 183-195.
the above equalities and production efficiency are
Sobel, J. 1989. An analysis of discovery rules. Law and
Contemporary Problems 52: 133-159. indeed necessary conditions (leaving aside comer
optima) for a feasible allocation to be Pareto
Spier, K. 1992. The dynamics of pretrial negotiation. Review
of Economic Studies 59: 93-108. Optimal. Any combination of production,
Spier, K. 1998. Settlement of litigation. In The new Pal-
consumption, and foreign trade vectors such that
grave dictionary of economics and the law, vol. 3, ed. P.
Newman. London: Macmillan. there is no positive excess demand for any
commodity is feasible. Pareto Optimality ensures
Spier, K. 2003a. ‘Tied to the mast’: Most-favored-nation
clauses in settlement contracts. Journal of Legal Studies
that no other feasible allocation can make at least
32: 91-120.
one consumer better off without making some other
Spier, K. 2003b. The use of ‘most-favored-nation’ clauses in
consumer worse off. These conditions are also
settlement of litigation. Rand Journal of Economics 34:
78-95. sufficient if it is assumed that the aggregate
Stephens, C. 1966. Is compulsory arbitration compatible with
production set and individual preferences are
bargaining? Industrial Relations 5(1): 38-52.
convex. Further, if each consumer has a positive
Waldfogel, J. 1998. Reconciling asymmetric information and
endowment of every commodity, any such Pareto
divergent expectations theories of litigation. Journal of
Law and Economics A\ \ 451—476. Optimal allocation can be sustained as a
Zeng, D. 2003. An amendment to final offer arbitration.
competitive equilibrium provided redistribution of
Mathematical Social Sciences 46(1): 9-19.
endowments among consumers or of their incomes
through lump sum transfers in feasible. This is in
essence the second fundamental theorem of
neoclassical welfare economics.
Bhagwati, and others analysing distortionary
Distortions taxation (Atkinson and Stiglitz ), factor market
distortions (Magee ) etc., all seem to take
T. N. Srinivasan (implicitly) the second fundamental theorem as the
point of departure. This suggests the following
definition: a distortion exists in an economy in
which lump sum transfers or their equivalents are
feasible redistribution instruments, if some Pareto
The voluminous literature on distortions, including
Optimal allocations from the point of view of
a masterly survey by Jagdish Bhagwati ( ),
consumers of that economy cannot be characterized
contains no formal definition of the term
as competitive equilibria. This is consistent with the
distortion. The analysis often proceeds in terms of
argument of Arrow ( ) that
specific examples. Bhagwati analyses distortions in
the context of foreign trade policies and the welfare the best developed part of the theory (of
externalities) related only to a single problem: the
of home consumers. He characterizes distortions as statement of a set of conditions, as weak as possible,
departures from the equality of the marginal rate of which insure that a competitive equilibrium exists
transformation of one commodity into some other and is Pareto-efficient. Then the denial of any of
through foreign trade (the so-called foreign rate of these hypotheses is presumably a sufficient
condition for considering resort to nonmarket
transformation) with transformation through channels of resource allocation - usually thought of
domestic production (the domestic rate of as government expenditures, taxes, and subsidies.
transformation) and with the marginal rate of
substitution in the consumption of the same pair of
commodities by each
Distortions
A distortion can clearly arise in situations when
some of the premises of the theorem are violated.
Obviously, if lump sum income transfers are
infeasible, even if all the other premises are satis-
fied some Pareto Optima cannot be characterized as
competitive equilibria. Hence any redistribution
achieved through other instruments may be Pareto
dominated by an equilibrium achievable with lump
sum transfers. Violations of other premises will in
general call for the use of additional policy
instruments besides lump sum transfers. If
externalities or increasing returns in production lead
to non-convexity of the aggregate production set, in
general, a set of Pigouvian taxes and subsidies are
the needed additional instruments. If price-taking
producers and consumers do not perceive a
country’s market power in its foreign trade then
taxes on foreign trade are needed.
The preceding discussion implicitly views the
distortions as structural features of the economy or
in Bhagwati’s terminology, as endogenous and the
policies as ‘first best’ responses that assure Pareto
Optimality. However, if the same policy instru-
ments are used in the absence of distortions or at
levels that are not ‘first best optimal’ in their
presence, the resulting equilibrium will be Pareto
dominated by another achievable equilibrium by
refraining from their use in the former case and by
using them at first best optimal levels in the latter.
Bhagwati characterizes such inappropriate use of
policy as autonomous policy imposed distortion.
Policymakers may have other objectives besides
consumer welfare. Johnson ( ) termed
such social concerns non-economic objectives. The
literature that followed (Bhagwati and Srinivasan )
addressed two policy questions. The first derives the
‘first best’ policy that achieves a given non-
economic objective, with the least cost in terms of
consumer welfare. One feasible policy for achieving
the non-economic objective has a higher cost than
another, if the equilibrium associated with the
former is Pareto dominated by an equilibrium
achievable using the latter and making lump sum
transfers between consumers as needed. The second
question is the ranking of alternative feasible
policies starting from the ‘first best’ to ‘second
best’, ‘third best’ etc. in terms of their cost in
achieving the non-economic objective.
2953
Governments may wish to raise the output (for
instance, for reasons of national defence) of some
industry above its level in a laissez-faire competi-
tive equilibrium through policy intervention. A
production subsidy (or its equivalent) to that
industry is the appropriate first best policy for
achieving the production (non-economic) objective.
Such a subsidy would be non-optimal, and hence a
distortion, in the absence of the objective. Bhagwati
characterizes such interventions as policy imposed
instrumental distortions, the word instrumental sig-
nifying that the policy is an instrument for achiev-
ing non-economic objectives. Trade tariffs and
quotas, consumption taxes and subsidies, wage sub-
sidy and similar factor use taxes or subsidies turn
out to be the first best policy instruments for achiev-
ing suitably specified non-economic objectives.
Each such policy could also be a feasible policy for
achieving non-economic objectives for which it is
not the first best
The impacts, measured in alternative ways, of
particular policies or processes in the presence of a
distortion rather than the optimal policy response to
it have been analysed. An example is the impact of
an import tariff in an economy with a distortion in
the labour market in the form of a minimum wage
above its market clearing level. The literature on
immiserizing growth (Bhagwati ) and its offshoots
analyse the impact of the processes factor
accumulation, technical change, external capital
inflow, etc. on an economy with a tariff distortion.
This diverse literature establishes two important
common propositions. First, given an existing
distortion, the impact of policies other than the first
best or of processes could be in a direction opposite
to that they would have taken had there been no
distortion or had the distortion been addressed with
a first best policy. For example, the accumulation of
capital which would have increased consumer
welfare had there been an optimal tariff could be
welfare-worsening in its absence. The shadow price
of a factor to be used in social cost benefit analysis
is a small open economy with a distortionary tariff
can be negative. Thus the withdrawal of that factor
from its existing employment for use in a project
instead of adding to the project’s cost increases its
social value! (Srinivasan and Bhagwati ). An
2954
implication of this is that some of the production
activities in an economy are subtracting rather than
adding value at shadow prices, while obviously
their value added at market prices is positive.
The second proposition shows that policies
other than the first best, even if distortionary, can
increase welfare. Thus given an existing distortion,
introduction of another can improve welfare. In the
so called Harris-Todaro ( ) economy a
distortionary minimum wage is enforced in urban
manufacturing activity. Rural workers migrate to
urban areas as long as their expected wage (taking
into account the probability of being unemployed)
exceeds the rural wage. If the first best policy of a
wage subsidy to both sectors is not feasible, an
output subsidy to agriculture can improve welfare
compared to the laissez-faire equilibrium. The
second proposition is an illustration of the general
theorem of the second best: ‘if there is introduced
into a general equilibrium system a constraint
which prevents the attainment of one of the Paretian
conditions, the other Paretian conditions, though
still attainable, are in general, not desirable’ (Lipsey
and Lancaster ). The constraint of this theorem is
the equivalent of a distortion and violating other
attainable Paretian conditions is equivalent to
introducing other distortions. The theory of the
second best is rigorously analysed by Guesnerie (
, ).
The literature on rent seeking (Krueger ) and
directly unproductive profitseeking (DUP) activities
(Bhagwati ) has highlighted another aspect of
distortions. A distortion by raising the demand price
of a commodity above its relevant supply price
creates a rent that may trigger a competition for
acquiring it. For example, an import quota (tariff)
by raising the domestic price above the import price
could trigger a competition for quota rents (tariff
revenues), thereby diverting resources from
production. However, such a diversion takes place
in the context of an existing distortion (an
inappropriate quota or tariff) and as such,
paradoxically, it can improve consumer welfare if it
succeeds in reducing the welfare loss associated
with the distortion more than the welfare loss it
creates in reducing resources available for
production. It has also been shown that the
Distortions
welfare ranking of policies that achieve a given non-
economic objective can be reversed once seeking
activities triggered by such policies are taken into
account.
To sum up, a distortion by definition creates a
welfare loss; first best optimal policies could often
be devised to offset this loss; if for some reason,
first best policies are infeasible, other welfare-
improving policies may exist and sometimes, they
can be ranked as ‘second best’, ‘third best’ etc.;
however, such policies, can have effects in
directions opposite to those they would have had in
the absence of a distortion; distortions have
implications for social cost-benefit analysis; finally
distortions can trigger rent-seeking activities.
See Also
►
Bibliography
Arrow, K.J. 1964/1970. Political and economic evaluation of
social effects and externalities. In Analysis of public
output, ed. J. Margolis. New York: National Bureau for
Economic Research.
Atkinson, A.B.., and J.E. Stiglitz. 1980. Lectures on public
economics. New York: McGraw Hill.
Bhagwati, J.N. 1968. Distortions and immiserizing growth: A
generalization. Review of Economic Studies 35: 481^485.
Bhagwati, J.N. 1971. The generalized theory of distortions
and welfare. Chapter 12 in Trade, balance of payments
and growth: Papers in international economics in honor
of Charles P. Kindleberger, ed. J.N. Bhagwati, R.W.
Jones, R. Mundell, and J. Vanek. Amsterdam: North-
Holland.
Bhagwati, J.N. 1982. Directly unproductive profit Seeking
(DUP) activities. Journal of Political Economy 90: 988-
1002.
Bhagwati, J.N., and T.N. Srinivasan. 1969. Optimal inter-
vention to achieve non-economic objectives. Review of
Economic Studies 36: 27-38.
Guesnerie, R. 1979. General statements on second best Pareto
optimality. Journal of Mathematical Economics 6: 169-
194.
Distributed Lags
Guesnerie, R. 1980. Second-best policy rules in Boiteux
tradition. Journal of Public Economics 13: 51-58.
Harris, J.R., andM.P. Todaro. 1970. Migration, unemploy-
ment and development: A two sector analysis. American
Economic Review 60: 126-142.
Johnson, H.G. 1965. Optimal trade intervention in the
presence of domestic distortions. Chapter 11 in Trade,
growth and the balance of payments, ed. R.E. Caves,
H.G. Johnson, and P.B. Kenen. Amsterdam: North-
Holland.
Krueger, A. 1974. The political economy of the rent- seeking
society. American Economic Review 64: 291-303.
Lipsey, R.G., and K. Lancaster. 1956. The general theory of
the second-best Review of Economic Studies 24: 11-32.
Magee, S.P. 1976. International trade and distortions in
factor markets. New York: Marcell Dekker.
Srinivasan, T.N., and J.N. Bhagwati. 1978. Shadow prices for
project selection in the presence of distortions: Effective
rates of protection and domestic resource costs. Journal
of Political Economy 86: 97-116.
Distributed Lags
Philip Hans Franses
Abstract
This article reviews various aspects of distrib-
uted lag models. Specific attention is paid to the
interpretation of model parameters.
Keywords
Almon lags; ARMA models; Cointegration;
Distributed lags; Error correction models;
Koyck transformation; Maximum likelihood;
Multicollinearity; Nonlinear least squares;
Ordinary least squares; Vector autoregressions
JEL Classifications
C22
Distributed lag models correlate a single dependent
variable with its own lags and with current and
lagged values of one or more explanatory variables.
Examples concern the current and dynamic
correlations between output and
2955
investment and between sales and advertising.
Distributed lag models typically assume that the
explanatory variable is exogenous. (In case of
doubt, one usually resorts to vector autoregressive
models where two or more variables can be
endogenous; see Sims .)
This article highlights a few aspects of distrib-
uted lag models. The two main aspects are repre-
sentation and interpretation. Useful extended
surveys appear in Dhrymes ( ), Griliches
( ) and Hendry et al. ( ).
Representation
Consider a dependent variable y, and, for ease of
notation, a single explanatory variable xt. Indicator t
runs from 1 to n and it can concern seconds, hours,
days or even years. A general (autoregressive)
distributed lag model is given by
y, = p + + - + u-Py,-P + Po*t
+ PiXt-i + ... + + £/, (1)
where p and m can take any positive integer value,
and where it is usually assumed that s:t is an
uncorrelated variable with mean zero and variance
a2. (Part of the literature assumes the label distrib-
uted lags model for the case where p 0 and m = DC.
Below we will see that such a model is often
approximated by a model as in ( ).)
As the model contains the lagged dependent
variables, it is called an autoregressive distributed
lag model with orders p and m, in short ADL(p, in).
The model allows for delayed effects of xt, as /)
„ can be 0, and it also allows for time gaps in these
effects when some j> parameters are zero and
others are not.
Reducing the Number of Parameters
Basically, given fixed and finite values ofp and m,
the parameters in ( ) can be consistently estimated
with ordinary least squares (OLS). (Typically one
uses information criteria as those of Akaike or
Schwarz to choose the relevant values of p and in in
practice. ) In practice, p and m can be large, and in
theory even as large as oo. This can be
2956
inconvenient, for two reasons. First, the variables
yt and xt each can be strongly autocorrelated, and
then the regression in ( ) suffers from multi-
collinearity. Second, with many parameters in a
model there might be many values to evaluate and
interpret.
To reduce the number of parameters and to
facilitate interpretation, one can impose restric-
tions. Early suggestions are the Almon and Shiller
lag structures, where the parameters are made
functions of i, i = 0, 1, 2,... , m (see Almon
, and Shiller ), and the so-called Koyck
transformation (see Koyck ).
Almon and Shiller Transformations
Consider the version of ( ) with p = 0 and m =
in and set ji at 0 for convenience, that is, consider
yt = Poxt + P\xt-1 + ••• + Pmxt-»1 + Et- (2)
Almon ( ) proposes to reduce the number
of parameters by assuming the approximation
j8,- = a0 + cM + a2i2 + ... + a qiq.
with q > in. This makes the sequence of /S, param-
eters a polynomial and hence a smooth function
without possibly implausible spikes.
Working out the Almon lags, one can derive
that the structure implies that
f t + i - 2 f t + /?,._!=)>,
where y, is a function of a, values. Shiller (
considers this as too restrictive and he proposes to
assume that
/Vt-2 +
for i = 1, 2, ... , m — 1.
Koyck T ransformation
The Koyck model can be interpreted as a model
which includes adaptive expectations. Suppose
that
y, = « + fix* + £ , ,
Distributed Lags
where x* denotes the expected value of x„ an
expectation formed at t 1. When the adaptive
expectations schedule is assumed, like
x* = Xx*_j + (1 - X)xh (7)
with again |2| < 1, then substituting ( ) into ( )
gives
y, = a ( l - X) + lyt_1 + / ? ( 1 - X)x, + e,
- A e, _ j . (8)
The short-run effect of x, on y, is /i(l — X),
while the long-run effect is = fX as could be
expected given ( ).
Consider the case were m equals oo, and where
all a parameters are set to zero. When it is further
assumed that [lj = /i0A x 1, with |2| < 1 for j is 1,2,...,
then ( ) becomes
y, P + Poxt + Po^t-i + l\)^xt 2 + . . .
+ e,. ( 9)
Subtracting Xyt.\ from this expression gives
(3)
yt = { 1 - + Xy,_x + [)0x, + e, - Xe,
-1, (10)
which is again ( ). This Koyck transformation leads
to a rather simple model with a moving average
(MA) error term. The appropriate estimation
method is maximum likelihood, as it is
(4) described in, for example, Hamilton ( ,
p. 132) for general ARMA models. Note that the
)
parameter X appears in the autoregressive part and
in the MA part.
Restructuring the Model
(5)
An alternative way to reduce the number of
parameters, also in order to facilitate interpretation,
is to restructure the model.
To overcome multicollinearity, one can rewrite
model ( ) in the so-called error correction format.
This format combines levels and differences of
levels, which is convenient as these are usually
much less correlated than the levels themselves,
(6)
Distributed Lags 2957

and hence multicollinearity will be much less of a Interpretation

problem. An additional feature of the error cor-
rection format is that it provides an immediate We now turn to the interpretation of distributed lag
look at key parameters such as the total effect, models.
the current effect, and the speed at which the
Long-Run and Short-Run Effects
total effect is accomplished.
The error correction model in ( ) provides
With A, denoted as the /-th order differencing
immediate estimates of current and dynamic effects.
filter, that is, Ajyt = y, — yt.j, an error correction
(Fok etal. , show that when the series yt and xt have
representation for ( ) reads as
a unit root and are cointegrated, as defined by Engle
and Granger , one should speak of the long-run
Ai y, = p
effect, while when the series are stationary there is a
TOLoPi total or cumulative effect. For the latter, see also
•S'- ) 1 y t ~i - ■X,-\ Hendry et al. .) The current effect is /i 0 and the long-
m
run or total effect is
+ P Q A\X I — Ay_lJVf_l
i=
2P 1_
— “/Az-iy,-! + Si, J2j=i
(13)
j=°-
( 11)
Note that the long-run effect can be larger or
where lagged levels are suitably combined into smaller than the short-run effect, depending on the
differenced variables such that at each lag a higher- values of the parameters. The parameters in the
order differenced variable appears. This error correction model, when written as
representation even further reduces chances of
having multicollinearity. Note that the model can
Aiy, = n + p[y t -1 - ywi] +
also be written in terms of lagged levels and first m
differences only, that is as
i=2
P
A i y, = p -X^A-iLi-i +fif. 04)
i=2
&-1 y__^Lr ,
St—1 1 --
i - E l i <i=i can be estimated using non-linear least squares.
\j=i
This method provides direct estimates of the long-
^0AIAV-^7;AIAV_ run effect y and its associated standard error.
i=\

+ ^2 SjAi-iyi + £/
J=I Duration Interval
As well as the long-run and short-run effects, one
( 12)
may also be interested in the speed with which the
With the use of ( ), all but two parameters (that effect of xt decays over time. To be able to compute
is, ai and [> ]) can be directly estimated by using this so-called duration interval, one needs explicit
;
OLS, while 71 and [>, straightforwardly follow expressions for for all values of k running from 1
from applying OLS to ( ). Note that model ( ) to, potentially, oo. Given the expression in ( ), these
can also be written such that the levels (now at 1 expressions are easily derived as
1) enter at t — 2 or, say, t — p.
2958 Distributed Lags

dyt there are at least two further issues that one needs to
Po
dx, address, that is, next to selecting p and m and a
dy , dy, useful transformation. The first concerns the sta-
+1 = fii + *ITT
tistical analysis of the model. For example, if yt and
dy, dx, dxt
—
% »#±i+ x, are not stationary, one needs to rely on
+2
UXf OXf cointegration techniques that involve non-standard
dx,
d
asymptotic theory. The theory that is most relevant
dy, yt+(k-j) here is formulated in Boswijk ( ). Also, in the
+k dx.
dx, case of the Koyck model,
one faces the so-called Davies ( ) problem.
where it should be noted that txk 0 for k > p, and Under the null hypothesis that /i0 = 0, the model
that /?/ 0 for k > m. Hence, the final form of a collapses to y, = st and hence A is not identified
distributed lag model (see Harvey ), is then.
The second issue concerns aggregation over
00 time. It may be that y, and x, are not available at the
y, = ^2 ^iX,-‘+ error’ (15) same sampling frequency. For example, television
i=0
commercials last for 30 seconds and recur each
hour, say, while sales data are available only at the
where
weekly level. Tellis and Franses ( ) have
dy,+i
Si dx. (1 6) a few recent results, but more work is needed.

With these S,-, one can derive all kinds of sum- Bibliography
mary effects (like mean and median, or half lives of
Almon, S. 1965. The distributed lag between capital appro-
shocks) of x, on y,.
When Sj decays monotonically, it is useful to Boswijk, H. 1995. Efficient Inference on cointegration
define the decay factor by
dy, 9yl+k
dx, dx,
Pk =
dx.
This can be computed only for discrete values of
k as there are only discrete time intervals. This
decay factor is a function ofthe model parameters.
Through interpolation, one can decide on the time k
it takes for the decay factor to be equal to some
value ofp, which typically is equal to 0.95 or0.90.
This estimated time k is then called the p per cent
duration interval. This measure is frequently used in
advertising research (see Clarke ; Leeflang et al. ;
Franses and Vroomen ).
Final Issues
Distributed lag models continue to be a standard
empirical approach. When the models are applied,
priations and expenditures. Econometrica 33: 178-196.
parameters in structural error-correction models. Journal
of Econometrics 69: 133-158.
Clarke, D. 1976. Econometric measurement ofthe duration of
advertising effect on sales. Journal of Marketing
Research 8: 345-357.
Davies, R. 1987. Hypothesis testing when a nuisance
parameter is present only under the alternative.
Biometrika 64: 247-254.
Dhrymes, P. 1971. Distributed lags: Problems of estimation
and formulation. San Francisco: Holden-Day.
Engle, R., and C. Granger. 1987. Cointegration and error
correction: Representation, estimation, and testing.
Econometrica 55: 251-276.
Fok, D., C. Horvath, R. Paap, and P. Franses. 2006. A
hierarchical Bayes error correction model to explain
dynamic effects of price changes. Journal of Marketing
Research 43: 443-461.
Franses, P., and B. Vroomen. 2006. Estimating confidence
bounds for advertising effect duration intervals. Journal
of Advertising 35(Summer): 33-37.
Griliches, Z. 1967. Distributed lags: A survey.
Econometrica 35: 16-49.
Hamilton, J. 1994. Time series analysis. Princeton: Princeton
University Press.
Hansen, B. 1996. Inference when a nuisance parameter is not
identified under the null hypothesis. Econometrica 64:
413-430.
Distribution Theories: Keynesian
Harvey, A. 1990. The econometric analysis of time series.
London: Philip Allan.
Hendry, D., A. Pagan, and J. Sargan. 1984. Dynamic
specification. In Handbook of econometrics, ed. Z.
Griliches and M. Intriligator, Vol. 2. Amsterdam: North-
Holland.
Koyck, L. 1954. Distributed lags and investment analysis.
Amsterdam: North-Holland.
Leeflang, P, D. Wittink, M. Wedel, and P. Naert. 2000.
Building models for marketing decisions. Boston:
Kluwer.
Shiller, R. 1973. A distributed lag estimator derived from
smoothness priors. Econometrica 41: 775 788.
Sims, C. 1980. Macroeconomics andreality. Econometrica 48:
1-48.
Tellis, G., and P. Franses. 2006. The optimal data interval for
econometric models of advertising. Marketing Science
23: 217-229.
Distribution Theories: Keynesian
Mauro Baranzini
As Kaldor has pointed out, Keynes was never
interested in the problem of distribution of income
as such; the determination of its level was his main
concern: ‘One may nevertheless christen a partic-
ular theory of distribution as ‘Keynesian’ if it can
be shown to be an application of the specifically
Keynesian apparatus of thought’ (Kaldor p. 94).
Since the middle Fifties a large number of neo-or
post-Keynesian models of economic growth and
income distribution have appeared, originating
mainly in the University of Cambridge. Post-
Keynesian distribution theory now occupies an
undisputed place in most macro- economic
textbooks. These models have been labelled as
‘post-Keynesian’ since savings passively adjust to
the externally given full- employment investment,
via redistribution of income between wages and
profits and/or among social classes. This contrasts
with the pre-Keynesian or neo-classical framework,
where investment is governed by saving, and where
the production function and marginal productivity
theory play a crucial role in determining income
distribution. The ‘post-Keynesian’ model
2959
also differs from the static Keynesian scheme,
where changes in the level, rather than in the
distribution, of income ensure equality between
saving and investment.
The Role of Social Classes
In order to study income distribution in classical,
post-Keynesian and neo-Ricardian theories it is
relevant to define what kind of relationships exist
between property and income earners. Property
rights are fundamental determinants of distribution
if the production process requires some form of
cooperation from individuals having the power of
‘withdrawing’ certain essential inputs. As in most
classical theories, social classes remain crucial for
post-Keynesian theories, and their distinctive
feature is given by their saving and consumption
behaviour. In more sophisticated theories (Pasinetti)
the assumption of ‘separate appropriation’ of each
production factor is no longer in the foreground,
and workers’ income is made up by wages and
profits on accumulated savings. In a general sense
in post-Keynesian theories different rates of saving
are associated with different economic or social
classes. And the distribution of income among
classes will be such as to yield an overall saving
equal to the desired level of full-employment
investment. We shall start by considering the origin
of Keynesian income distribution theories.
The Hharrod-Domar Dilemma
When in the late Thirties and in the Forties the first
macro-economic models of economic growth were
developed, the theory of income distribution was
caught in an impasse, represented by the well-
known Harrod-Domar equilibrium condition ^ =
g(K/Y), where s is the aggregate saving ratio, g the
natural rate of growth (which can include ‘labour
saving’ technical progress), and K/Y the
capital/output ratio. If these three variables are all
given, then it is unlikely that the Harrod-Domar
condition can be satisfied. Hence, in order to have a
model in which the
2960
possibility of steady growth is assured, it is nec-
essary to relax one or another of the assumptions.
The equality between s' and g(K!Y) can be obtained
by: (a) flexibility in KJY (also referred to as the
technology assumption); (b) flexibility in s (saving
assumption); (c) flexibility in g (labour- market
and/or labour-supply assumption).
The above three cases can, of course, be com-
bined in various ways as, for instance, in Samuel-
son and Modigliani’s ( ) model, where (a) and
(b) apply simultaneously.
Two Different Ways of Answering
the Harrod-Domar Dilemma
Solution (a) above was adopted by the neoclassical
or marginalist school:
Instead of there being fixed coefficients in produc-
tion there may exist a production function offering a
continuum of alternative techniques, each involving
different capital- labour ratios; ... The consequence is
that the capital-output ratio v is adjustable, instead of
being fixed, and this provides a way in which s/v and
n may be brought into equality (Hahn and
Matthews , p. 785).
The second answer to the Harrod-Domar
dilemma, that is, the assumption of a flexible
aggregate saving ratio, was primarily adopted by
the (neo-) Keynesian or Cambridge School. Of
course there are many ways in which one can give
flexibility to s ; but the one which has played the
major role is the hypothesis of a two-class society
(namely workers and capitalists, or consumers and
entrepreneurs), each with a different (constant)
propensity to save. In this way there always exists a
distribution of income between the two classes
which produces precisely that saving ratio that will
equal the value g(K/Y), so satisfying the Harrod-
Domar equilibrium condition. The validity of this
approach is reinforced by the fact that the
assumption of a uniform aggregate saving ratio
ignores all possible differences in saving (and
consumption) behaviour between, for instance,
different classes of income receivers, or categories
of income or even different sectors of the economy.
Moreover the problem of aggregating savings might
give rise to particular and unknown difficulties, so
that it may be safer to
Distribution Theories: Keynesian
consider it in a disaggregate way, as the neo-
Keynesian model does. Thirdly, this assumption
also receives empirical support from the observed
high rates of saving out of corporate profits and
lower rates out of labour income. Considering a
full-employment long-run equilibrium growth
model with a capitalis’ class (whose income is
derived entirely from capital) and a workers’ class
(whose income is derived from wages and
accumulated savings), both with constant
propensities to save, the Cambridge economists
were in a position to (1) provide a solution to the
Harrod-Domar dilemma (by specifying an
aggregate saving ratio ,v which equals g(K/Y),
where g and K/Y are both exogenously given);
(2) determine the long-run equilibrium value of the
rate of profits, the distribution of income between
profits and wages, and the distribution of disposable
income between the two classes;
(3) allow the existence of an income residual,
namely the wages, consistent with the assumption
of a relationship between the savings of that class of
individuals (the capitalists) who are in the position
to control the process of production and the patterns
of capital accumulation; and (4) give some insights
into the process of accumulation of capital by
specifying the equilibrium capital shares of the two
classes. This range of results is obtained within a
fairly simple framework and on the basis of
relatively few assumptions, much less ‘hybrid,
opposite and extreme’ than those of the marginalist
model.
Kaldor's Theory of Distribution
Kaldor’s distribution theory plays a fundamental
role in the Cambridge or post-Keynesian theories of
income distribution. His original analysis appeared
first in the Review of Economic Studies, 1956 and,
in a slightly different form in Essays on Value and
Distribution in . Kaldor considers a one-sector
growing economy in which there are two classes:
one whose income is derived entirely from capital
(the capitalists, who are not wage- earners) and a
second one which derives its income uniquely from
wages (the workers). At each of these two groups he
attaches a fixed
Distribution Theories: Keynesian
propensity to save, ,vc and sw respectively, higher
for the capitalists and lower for the workers.
Kaldor’s model, as well as all other neo- or post-
Keynesian models, is based on the assumption of
long-run, full-employment equilibrium.
Assuming that national income (V) is divided
into wages (W) and profits (P) and a situation of
steady growth, where all variables grow at the same
rate g and where all ratios among macro- economic
variables remain constant, Kaldor derives explicit
formulae for the overall rate of profits and share of
profits in national income. Additionally, by making
the ‘classical’ assumption that sw (the propensity to
save of the workers) is zero, he obtains the
following two simple relationships: P/K = g/sc
andP/Y = gK/sc. The first solution shows that the
equilibrium rate of profits depends only on the
exogenously given rate of growth (g) and on the
constant propensity to save of the capitalists’ class.
The second solution shows that the long-run share
of profits in national income is determined by the
rate of growth, the capital/output ratio (K) and the
propensity to save of the capitalists (all exogenously
given).
Pasinetti's Theorem
As we have seen, Kaldor’s saving function con-
siders, essentially, two types of income. What
happens if we assume that the saving propensities
differ, not according to classes of income, but
according to classes of individuals (a more realistic
assumption, referring to the weak definition of
social class discussed above)? It is at this point that
the basic contribution of Pasinetti may be brought
in, where he assumes that saving propensities differ
by class (assumed inter-generationally stable),
rather than by type of income. His contribution, in
his own words
has come from the discovery of a fundamental
relation (passed unnoticed in the whole of previous
economic literature) which links profits to savings
through the ownership of the capital stock. This
relation simply follows from the institutional prin-
ciple that profits are distributed in proportion to the
ownership of capital and that the ownership of
capital derives from accumulated savings (Pasinetti ,
p. 127)
2961
In this way the workers’ class is allowed to own
a share of the total capital stock, from which it
derives an interest income. By solving the model
Pasinetti obtains an explicit value for the rate of
profits and share of profits in national income; in
particular the former turns out to be P/K = g/sc,
which has been defined as Pasinetti’s Theorem, or
‘Cambridge equation’ (as a matter of fact it should
be defined as the ‘New Cambridge equation’, since
the original one had been found by Kaldor).
Pasinetti’s analytical results are similar to those
obtained by Kaldor, there is, however, a fundamen-
tal difference, since Pasinetti’s solutions have been
obtained without making any assumption whatso-
ever on the propensity to save of the workers, which
may assume positive values indeed. These results
are undoubtedly of importance and establish a direct
and simple relationship between the rate of profits
and the rate of growth, through the interaction only
of the capitalists’ propensity to save. More
precisely, the value of the rate of profits shows that
on the long-run equilibrium growth path, the
propensity to save of the workers, through
influencing the distribution of income between
capitalists and workers, does not influence the dis-
tribution of income between profits and wages.
Implications of the 'New Cambridge
Equation'
The first thing that can be stressed is that the rate of
profits and the share of profits in national income
both vary inversely with sc. Hence, all other things
being equal, the less the capitalists save and the
greater is their return on capital (but with a smaller
share of the capital stock). Exactly the opposite is
true for the workers: the more they spend, the less
they will receive for their future consumption
(through a reduced share of the capital stock).
Secondly, as Pasinetti himself points out in the
original exposition and more recently (Pasinetti ,
Ch. VI) the irrelevance of workers’ propensity to
save gives the neo-Keynesian growth model much
more generality than it appears at first sight. It is not
necessary to make any hypothesis whatever on the
aggregate saving behaviour
2962
of the workers for the simple reason that both the
rate of profits and the distribution of income are
determined independently of the propensity to save
of the workers. Therefore the workers could be
divided into any number of sub-categories we
wanted. Again, the particular saving behaviour of
any sub-category of workers would influence the
distribution of income among the various sub-
classes of workers and, of course, between the
workers and capitalists, but the distribution of
income between wages and profits would not be
affected at all, given the constancy of the cap-
italists’ propensity to save.
Third and finally, the ‘New Cambridge equa-
tion’ shows and uncovers, for the first time in
modem economic theories, the ‘absolute strategic
importance’ of the saving behaviour of just one
group of individuals (the capitalists) for the deter-
mination of the most vital relationships of the
model. On the other hand the saving behaviour of
the other class (or sub-classes) has nearly no power
at all: they can save as much as they want, and of
course receive an interest on it, but they will not
influence the distribution of income between profits
and wages. Moreover the share of wages in national
income is a residual, once the share of profits (a
function of the capital/output ratio, the rate of
growth and the propensity to save of the capitalists )
has been determined. The concept of residual of the
classical economists is to be found again: but while
for Ricardo the residual was represented by profits,
in post-Keynesian models wages are a residual,
once profits have been determined.
The Marginalists' Reply to
the 'Cambridge Equation'
The results obtained by the neo-Keynesian econ-
omists did of course attract the attention of the
neoclassical economists, who defined the Cam-
bridge equation of income distribution as a ‘para-
dox’. Their reaction was not surprising, since the
Cambridge equation makes the whole ‘well-
behaved’ production function framework irrelevant.
With the aim of defending the theory of marginal
productivity of capital, Meade (
Distribution Theories: Keynesian
) and Samuelson and Modigliani ( ) set
out to find a condition for which Kaldor- Pasinetti’s
Theorem would be prevented from operating, by
arguing that when the propensity to save of the
workers is exactly equal to the propensity to save of
the capitalists times the profits share, then the
capitalists cannot in equilibrium survive in the
system and their propensity to save cannot
determine the rate of profits. In such a situation all
equilibrium savings of the system would be
provided by the workers only, and the two-class
system would become a singleclass model where
the marginalist scheme could be applied again to
determine income distribution. But, as the ensuing
debate has shown, such a situation is very unlikely
to happen in the real world and, more importantly, it
does represent a ‘knife-edge’ solution since in
equilibrium it applies only when ,v„ = s. To devise
one ‘knife- edge’ in order to answer another one
(the Harrod-Domar’s) may not represent the best
counter-argument.
Other Criticisms of the 'Cambridge
Equation'
The most common criticisms of post-Keynesian
income distribution models (cf., for instance, Bliss ,
Ch. 6; Samuelson and Modigliani ) seem to
concentrate on: (a) the assumption of the equality,
in the long-mn, between the rate of profits earned
by the capitalists and the rate of interest earned by
the workers on their accumulated savings; (b) the
constancy of the propensity to save of the two
classes, exogenously given and hence independent
of other variables as, for instance, the rate of
interest or the rate of population growth; and (c) the
assumption and identification of individuals who
retain their class identity forever, that is, of classes
which are intergenerationally stable.
Let us consider these points in some detail. In
the late Sixties and early Seventies several authors
have suggested that if one were to assume a dif-
ferentiated rate of return for workers’ and capital-
ists’ savings, the Cambridge equation would no
longer apply. As a matter of fact, as Pasinetti
Distribution Theories: Keynesian
( , pp. 13 9-41) has formally proved, this may
not be true: the hypothesis of a differentiated
interest rate comes to reinforce his analysis, since
‘A rate of interest lower than the rate of profit has
the same effect of a higher propensity to save of the
capitalists, as it redistributes income in favour of the
class that owns the physical capital stock.’ The
second criticism was put forward by economists
who thought that the introduction of the life-cycle
hypothesis on savings into the two-class model
(where individuals would make their saving plans
on the basis of the level of the interest rate and of
other life-cycle parameters) would make the
equilibrium interest rate a function of all parameters
of the model. The assumption of the life-cycle
hypothesis is of course not strictly compatible with
the neo-Keynesian framework, where investment is
independent of savings; nonetheless it has been
shown that even in the context of a two-class life-
cycle model, as long as there exists a class of ‘pure’
capitalists, the equilibrium interest rate is a function
of the behavioural parameters of the capitalists
only. The third main criticism concerns the
assumption of intergenerationally stable classes;
one would expect that the relaxation of this
assumption would invalidate the relevance of the
Cambridge theorem. But it is not really so, as few
authors seem to conclude.
Vaughan ( ), for instance, in his analysis
obtains a third solution for the interest rate, which
approaches Pasinetti’s solution when the net
transference of individuals between classes is low,
as it may be the case over the very long run which
constitutes the framework of these models.
Conclusions
Summing up we may say that post-Keynesian
theories place themselves half-way between clas-
sical and marginalist theories of income distribu-
tion, since on the one hand they reject the strong
version of the social-class theory of distribution
postulated by classical economists, where each
income share meets a strong ‘claim’ associated with
the property of an essential input (labour,
2963
capital or land). Instead post-Keynesian theories put
forward a much more flexible concept of social
class, characterized by a given saving and
consumption behaviour (for Pasinetti the workers
may even be divided into sub-classes). But on the
other hand post-Keynesian theories differentiate
themselves from the models of competitive eco-
nomics where individuals react only with respect to
the markets on which they have little effect. Post-
Keynesian theories do moreover allow for elements
of monopoly power, and retain the concepts of
residual income and circularity of the production
process contemplated by classical economists. It
may well be that their extension to include certain
elements of the life-cycle theory of saving and
consumption behaviour will give them some micro-
foundations.
See Also
References
Baranzini, M. 1975. The Pasinetti and the anti-Pasinetti
theorems: A reconciliation. Oxford Economic Papers 27:
470-473.
Baranzini, M., and R. Scazzieri. 1986. Knowledge in eco-
nomics: A framework. In Foundations of economics:
Structures of inquiry and economic theoiy, ed. M.
Baranzini, and R. Scazzieri. Oxford/ New York: Basil
Blackwell.
Bliss, C.J. 1975. Capital theory and the distribution of
income. Amsterdam: North-Holland.
Hahn, F.H., and R.C.O. Matthews. 1964. The theory of
economic growth: A survey. Economic Journal 74: 779-
902.
Harcourt, G.C. 1972. Some Cambridge controversies in the
theory of capital. Cambridge: Cambridge University
Press.
Kaldor, N. 1956. Alternative theories of distribution. Review
of Economic Studies 23(2): 83-100.
Kaldor, N. 1960. Essays on value and distribution. London:
Duckworth.
Meade, J.E. 1963. The rate of profit in a growing economy.
Economic Journal 73: 665-674.
Meade, J.E. 1966. The outcome of the Pasinetti process: A
note. Economic Journal 76: 161-165.
Pasinetti, L.L. 1962. The rate of profit and income distri-
bution in relation to the rate of economic growth. Review
of Economic Studies 29: 267-279.
2964 Distribution Theories: Marxian

production relations are ‘essentially coincident’, he

Pasinetti, L.L. 1974. Growth and income distribution, essays
in economic theory. Cambridge: Cambridge Universityargued, since ‘both share the same historically
Press.
transitory character’. (Marx , pp. 883, 878).
Pasinetti, L.L. 1981. Structural change and economic growth.
A theoretical essay on the dynamics of the wealth of And yet, despite these reasonably self-evident
nations. Cambridge: Cambridge University Press. theoretical connections, the analysis of distribution
Robinson, J. 1962. Essays in the theory of economic growth.
has remained substantially underdeveloped in the
London: Macmillan.
Samuelson, P.A., and F. Modigliani. 1966. The Pasinetti
historical evolution of Marxian economics. While
such classic issues as crisis theory, the
paradox in neoclassical and more general models. Review
of Economic Studies 33: 269-302. transformation problem and the usefulness of the
Vaughan, R.N. 1979. Class behaviour and the distribution of
labour theory of value have been intensively and
wealth. Review of Economic Studies 46: 447-465.
vigorously reviewed, the determination of distri-
bution patterns over time and cross-sectionally has
been elided in synthetic treatments of Marxian
analytics and largely ignored in more focused
scholarly investigations.
Distribution Theories: Marxian More recent developments in Marxian econom-
ics, fortunately, have finally begun to overcome this
David M. Gordon traditional reticence. This essay provides a brief
review of traditional attention - or, more accurately,
^attention - to the problem of distribution and then
surveys some promising recent cultivations of this
It is hard to imagine a more important topic within historically fallow terrain.
Marxian economics than the distribution of income
and the means of production among the principal
Terms of Analysis
classes in capitalist economies. For example: (1)
The share of profits (or, inversely, the share of Before beginning that review, however, it will be
wages) constitutes one important component of the useful to clarify the defining boundaries of this
rate of profit. (2) The rate of profit operates as a topic.
fundamental determinant of the pace of investment It is probably most usefiil to begin with the role
and, therefore, of accumulation. (3) The rate of of distribution in the determination of profitability,
accumulation serves as a kind of life-force that central fulcrum of economic behaviour. A
invigorating capitalist economies over time - familiar accounting identity reminds us that the rate
regulating their growth and development, and the of profit of the individual firm, r, can be expressed
wealth of their participants. (4 ) Distribution, as the product of the share of profits in firm value-
production and accumulation are thus added, sn the ratio of output to utilized capital stock,
fundamentally interconnected, forming the yu, and the ratio of uti lized to owned capital stock,
foundation of lives and livelihoods in capitalist K , or
societies.
In this respect, indeed, Marx himself regarded r=sr-yu-k*,
‘distribution relations’ as part of the core of the
capitalist economy. Criticizing those who ventured (1)
an ‘initial, but still handicapped, criticism of
where
bourgeois economy’ by seeking to distinguish
between the level of priority of production and r = n / A T 0 ; ® r = n / y ; yu = YjKu\
distribution, Marx affirmed that both production k*=Ku/K0;
relations and distribution relations are part of the (2)
‘material foundations and social forms’ of any
given historical epoch. Distribution relations and and n is firm profits, Kq is the value of the firm’s
owned capital stock, Tis firm value-added, and Ka
Distribution Theories: Marxian 2965

is the portion of the owned capital stock which is agents, beyond our starting groups of capitalists
currently utilized. In the aggregate, abstracting from and workers, which may seem relevant or
variation among firms for such purposes, the same necessary for our analyses.
accounting identity applies. 3. A share of revenue need not necessarily translate
In this accounting identity, distribution relations into an exactly equivalent share of real income,
primarily affect the level of and changes in the since the prices confronting workers and
share of profits in firm revenue. Factors affecting capitalists may not exactly parallel each other
the rate of capital accumulation and the productivity overtime. The relative purchasing power of their
of the means of production primarily affect yu- revenues received, and therefore the distribution
Secular trends in the robustness of aggregate of income, may consequently vary as a result of
demand and its fluctuations over the business cycle changes in the relative prices of capital goods
have their most direct impact on k*. and wage goods as well. It is conceivably useful,
At this first level of approximation, then, anal- therefore, to decompose the profit share in Eq.
ysis of distribution relations among the principal into two terms, one involving a ratio of‘real’
classes of a capitalist economy can begin with a profits to real income and the other a ratio of
focus on the determinants of sT. Such analyses capital-goods prices to an index of (weighted)
would immediately concern themselves with the output prices. (See Weisskopf , for useful
wage share, sw, as well, since sw = (1 ,vr). elaboration of this kind of decomposition.)
This is, of course, only a first level of approx-
imation. At a second level of investigation, we must A final consideration seems critical for defining
deal with three further refinements of focus. the scope of our analysis. It is taken for granted
within the Marxian tradition that a given class’s
1. Accounting Eq. is formulated in revenue terms, share of revenues is conditioned, at the most basic
not in value terms, so it does not yet encompass level, by the extent of its power over the means of
the Marxian concern with the value-theoretic production. And yet, over time, a given class’s
determinations of economic relations. But this relative control of the means of production will be
additional consideration requires simply that we responsive to systematic changes in its share of
add an analysis of the rate of exploitation (or the revenues. It is not at all inappropriate, therefore, to
rate of surplus value), s, to the definition of our treat the class distribution of revenues and the class
task, since conventional Marxian value analytics distribution of control over the means of production
establish a straightforward transformation as interdependent and mutually-determining over
between the profit share and the rate of the long term. We may therefore define our task
exploitation. In one simple formulation, for most broadly, in this respect, as the analysis of the
example, the rate of exploitation is equal to the determination of class (and group) shares of
ratio of profits (n) to wages (W) weighted by the revenue (and therefore of income) and of the class
capital-labour ratio (&!>, or s = & L • (II W) . distribution of relative control over the means of
(See Marglin production.
, pp. 57-60 and 191-192, for a useful elaboration of Marx was himself clear on the importance of
these relations of equivalence.) defining the analysis of distribution in both of these
2. The first level of approximation, represented by two senses. ‘It may be said ...’ he wrote at the end of
Eqs. and , also allows for the existence of only Volume III of Capital ( , p. 879),
two classes in capitalist economies, abstracting ‘that capital itself ... already presupposes a dis-
from all other relevant economic groupings or tribution: the expropriation of the labourer from the
subsidiary classes. At a second level of conditions of labour [and] the concentration of these
approximation, therefore, we must also consider conditions in the hands of a minority of
the existence of and determination of the shares individuals ...’ This underlying dimension of dis-
of any other categories of economic tribution ‘differs altogether’, he continued, ‘from
2966 Distribution Theories: Marxian

what is understood by distribution relations ... [as] hour and therefore, given the wage, the profit share
the various titles to that portion of the product as a residual. The behavioural hypothesis that
which goes into individual consumption’. This does capitalists save all profits combines with the
not in any way suggest, he insisted, that distribution determination of consumption by customary wage
in this former sense does not involve ‘distribution levels to create the conditions for a feasible and
relations’ or should somehow remain peripheral to warranted steady-state combination of profit rates
our analysis: and growth rates. Marglin concludes ( , p. 62):
The aforementioned distribution relations, on the ‘In contrast with the inherited neoclassical
contrary, are the basis of special social functions approach, in which resource allocation determines
performed within the production relations by certain
of their agents . . . . They imbue the conditions of
income distribution, causality here runs from
production themselves and their representatives with [exogenously-determined] distribution to growth.’
a specific social quality. They determine the entire There is, of course, nothing intrinsically wrong
character and the entire movement of production. with these assumptions about directions of cau-
sality. Treating distribution as exogenous to the
internal operations of the capitalist economy has
Traditional Analysis
simply meant that Marxian economists have tended
Inherited approaches to the problem of distribution to elide the factors determining distribution, setting
them aside as consequences of ‘historical and moral
are most easily viewed through three somewhat
elements’ and the ‘technical’ conditions of
separable lenses: the growth-theoretic perspective,
production.
crisis-theoretic hypotheses of a rising profit share,
and antipodal crisis theories based on a falling profit
share.
Hypotheses of a Rising Profit Share

Distribution has played a somewhat more explicit

Long-Term Trajectories role in analyses of tendencies toward economic
crisis. One group of theories has built upon
Marxian economics has not always found it
hypotheses of a secular tendency toward an
congenial to reflect upon the long-term growth
increasing profit share.
paths of capitalist economies, since such perspec-
Perhaps the first systematic example of this
tives are tainted in some minds by associations with
hypothesis emerges in Lenin’s account of imperi-
concepts like ‘stability’ and ‘equilibrium’. It is
alism and monopoly capitalism ( ). In its
nonetheless possible to extract from traditional
essence, Lenin’s argument begins with the rela-
Marxian analyses a clear approach to the logic of
tively simple hypothesis of increasing oligopoly
determination of ‘steady-state’ tendencies -
and therefore, ‘since monopoly prices are
provided this exercise is understood, in Marglin’s
established’ (p. 241), of relatively reduced com-
words ( ,
petitive pressures. With the help of financial oli-
p. 52), ‘as a subset of Marxian theory and not as an
garchies, corporations are able to achieve a
attempt to represent the whole’.
continuously rising profit share and therefore to
It seems reasonably clear, in that context, that
amass ‘an enormous “surplus of capital’” (p 212).
distribution relations are exogenously given to the
With this surplus of capital, capitalists are prompted
traditional model, determined outside the set of
to export capital overseas and, eventually, to reduce
basic interactions which jointly establish ‘equilib-
efforts at technical improvements. Over time, ‘the
rium’ rates of growth and rates of profit. Historical
tendency to stagnation and decay, which is
conditions, not directly subject to internal economic
characteristic of monopoly, continues to operate ...’
analysis, establish a ‘customary’ wage. Existing
(p. 241; emphasis in the original).
levels of productiveness, also exogenous to the
The model begins, therefore, with a strong
system, determine the level of output per
hypothesis about distribution - presuming a
Distribution Theories: Marxian
strong initial tendency under monopoly capitalism
towards a rising profit share. And yet, the
conditions which would be necessary to derive this
as a prevailing long-term tendency are unexplored.
There is no real analysis of wages, although
prevailing assumptions about competitive labour
markets are implicitly incorporated into the model.
There is equal taciturnity about the initial
determination of real productivity, even though the
rate of growth of real productivity must exceed the
growth of real wages for the initial condition of a
rising profit share and an ultimate ‘surplus of
capital’ to hold. And, despite the international
orientation of the analysis, there is no real
incorporation of a model of international pricing
and exchange which would support the hypothesis
of rising profit shares in all the advanced countries.
These elisions are subsequently reproduced in
most 20th-century analyses of underconsumption
and monopoly capital. The models begin with a
premise of growing capitalist power, most fre-
quently from increasing monopoly control over
product markets. This power leads to a rising
‘surplus’ and therefore to a rising profit share. From
that set of initial premises, the problems of effective
demand and urgent efforts to absorb the surplus
follow naturally (Bleaney ; Baran and Sweezy ). As
with Lenin, however, there is remarkably little
attention to the conditions which permit this initial
increase in the profit share. What about wages? Or
labour productivity? Or conditions of international
pricing? There is, in general, the simple
presumption that conditions have evolved in a such
a way as to permit consistent increases in the profit
share, but little reflection on the relations which
make those conditions possible. Baran and Sweezy
admit some of this inattention, particularly to the
social relations which would allow real productivity
growth to outstrip real wage growth ( , pp. 8-9):
We do not claim that directing attention to the
generation and absorption of surplus gives a com-
plete picture of this or any other society. And we are
particularly conscious of the fact that this approach,
as we have used it, has resulted in almost total
neglect of a subject which occupies a central place in
Marx’s study of capitalism: the labour process.
2967
Hypotheses of a Falling Profit Share
For completeness, it is useful to consider the alter-
native hypothesis of a falling profit share, although
attention to this possibility has only emerged within
Marxian analysis more recently, primarily in the
post-World War II era.
This hypothesis has relatively simple analytic
foundations. For whatever reasons, working-class
power may increase sufficiently to allow wages to
rise more rapidly than labour productivity and
therefore to result in a persistent increase in the
wage share of revenues.
The hypothesis follows most naturally in a
cyclical context and bears close connections to
Marx’s own analysis of cyclical dynamics in
Chapter XXV of Vol. I of Capital ( ). In the
short run, rapid expansion may lead to tight labour
markets, increasing workers’ bargaining power and
resulting in a rising wage share. (Boddy and Crotty
( ) provide a useful development of
this cyclical model in relatively traditional terms.)
The hypothesis needs further grounding in order
to serve as the basis for a theory of economic crisis,
however. The forces which lead to tight labour
markets in short-term expansions could plausibly
result in comparably slack labour markets during
short-term contractions and therefore to a recovery
of the profit share. In order properly to ground a
theory of secular crisis upon this hypothesis of a
falling profit share - and therefore fully to develop a
‘profit squeeze’ theory of economic crisis - one
must show why cyclical contractions do not restore
the profit share and, other things equal, the rate of
profit. This requires analyses of conditions which
permit rising worker power - even in the age of
oligopolistic competition - from one business cycle
to the next. Until the mid-1970s, Glyn and Sutcliffe
( ) were the
principal Marxian economists to have formally
developed such an analysis, and in their case
primarily for the sole case of England.
Even in their case, however, the analytic
requirements for the secular version of the ‘profit
squeeze’ theory of crisis are not fully developed.
What are the explicit conditions of labour market
competition which explain particular patterns of
wage growth? Under what conditions in the
2968
organization of production and the promotion of
technical change would real productivity growth
fail to keep pace with real wage growth? What are
the conditions of international economic linkages
which would or would not support tendencies
towards a falling profit share? A further problem
involves the closeness of the relationship between
profits and surplus value; Shaikh ( ) reviews
some of the problems with casual assumptions
about this connection.
Kalecki and Mandel as Connecting Writers
We can find in the work of Michal Kalecki and
Ernest Mandel some early instances of the kinds of
concerns which have fuelled more recent
explorations.
Particularly in his later essays, Kalecki identifies
but does not yet develop some of the lines of
inquiry which would be necessary for a more
advanced analysis of distribution. In ‘Class Struggle
and Distribution of National Income’ ( ),
Kalecki refines the analysis of the relationship
between wages and the profit share, noting that
analyses of the conditions of product market com-
petition are necessary ‘to arrive at any reasonable
conclusion on the impact of bargaining for wages
on the distribution of income’ (p. 159); that trade
union power is likely, ceteris paribus, to reduce the
level of the mark-up; and that, in general,
class struggle as reflected in trade-union bargaining
may affect the distribution of national income but in
a much more sophisticated fashion than expressed by
the crude doctrine: when wages are raised, profits fall
pro tanto. (p. 163)
In ‘Trend and Business Cycle’ ( ), Kalecki
develops what he regards as a more satisfactory
analysis of the relationship between short-and
longer-term determinants of investment and there-
fore, a fortiori, the conditions which are likely to
affect movements in the profit share over time.
Both of these analyses are entirely preliminary,
however, since they constitute more of a pro-
gramme for further work than a report on completed
analyses. In particular, Kalecki notes that
Distribution Theories: Marxian
most of his analysis hangs on a handful of coeffi-
cients which he takes as given for his purposes,
including the level of labour productivity, the share
of gross profits flowing into capitalist consumption,
capitalists’ propensities to invest, and the rate of
embodied technical progress. ‘To my mind’, he
concluded, ‘future inquiry ... should be directed ...
towards treating ... the coefficients used in our
equations ... as slowly changing variables rooted in
past development of the system’ (p. 183). The real
problem, in short, is not to assume the central
parameters of the determination of profits and
investment but rather to derive them from
determinant structural and historical analysis.
Mandel serves as a transitional figure in a differ-
ent way. Although much of Mandel’s analysis is
hard to pin down precisely, he has nonetheless
helped highlight the importance of an integration
between formal Marxian analytics and structural/
historical analysis. In Late Capitalism ( ), in
particular, he suggests the rich possibilities for anal-
ysis of the particular conditions which might or
might not give rise to variations in the rate of
surplus value. There is much to learn, he urged (p.
183):
Late capitalism is a great school for the proletariat,
teaching it to concern itself not only with the imme-
diate apportionment of newly created value between
wages and profits, but with all questions of
economic policy and development, and particularly
with all questions revolving on the organization of
labour, the process of production and the exercise of
political power.
Recent Explorations
As this review is being written, a rich range of
Marxian work on distribution in advanced capitalist
societies has recently been completed or is currently
under way. Since much of it is still in progress and
unpublished, full references are difficult and
probably inappropriate for an enduring
encyclopedia. This final section will therefore
concentrate on a synthetic review of the kinds of
explorations which have recently been undertaken
and the promising possibilities which have begun to
emerge.
Distribution Theories: Marxian
Changing Power Relations
One central problem in traditional Marxian analysis,
which the examples of Kalecki and Mandel as
connecting figures help to highlight, was the
reluctance to forge determinate linkages between
formal analytic categories, on one side, and the
structure of and changes in power relations, on the
other. Many appear to have felt either that these two
loci of investigation operated at different levels of
logical abstraction or that power relations, with all
the social complexity of phenomena like the class
struggle, could not be rendered analytically or
studied empirically in any kind of formal or
rigorous fashion. One is left with analyses, to quote
Harris ( , p. 166), which
remain ‘essentially ad hoc and tentative’.
Recent work has begun to overcome these
hesitations. It has pursued careful and analytically
determinate investigations of the relationship
between power relations and, among other vari-
ables, the profit share. Attention has been focused
primarily on three different dimensions of power
relations: capital-labour relations, global linkages,
and contests over state policy and practice.
Capital-Labour Relations
It has been recognized since Marx that class
straggle over wages could conceivably affect
distribution. But the formal linkage of conditions of
class straggle to the determination of wage and
profit shares has been hampered by the impression
that levels and rates of change of productivity are
determined orthogonally - by technical conditions
and the pace of investment - and therefore that the
two kinds of concerns could not somehow be
combined into a single, inclusive, determinate
analysis of changes in the profit share itself.
This problem appears to have been overcome. In
recent work, particularly by Weisskopf, Bowles and
Gordon (1983), a ‘social model of productivity
growth’ has formally linked factors affecting
capital-labour relations with the more traditional
analyses. Several hypotheses about
2969
factors affecting the level of labour intensity in
production have been both elaborated mathe-
matically and tested empirically. This ‘social
model’ appears to provide a robust explanation of
variations in rates of productivity in the United
States in the decades following World War II.
One crucial insight in that work is also begin-
ning to invigorate Marxian wage analysis. Tradi-
tional perspectives on wage determination, building
upon the ‘reserve army’ effect, focused on the
relationship between wage bargaining and the threat
of unemployment. As capitalist societies have
developed, however, the threat of unemployment
has been tempered by the availability of various
components of what is typically called the ‘social
wage’ - such as unemployment insurance and
income maintenance expenditures. This has
prompted the development of a more inclusive
measure of the threat to workers of job dismissal: an
index of ‘the cost of job loss’. It calculates the
expected income loss resulting from job termina-
tion, usually calculated as a percentage of the
expected annual income if still employed, and
incorporates estimates of the average wage in
employment, expected unemployment duration,
available income-replacing benefits, and available
non-income-replacing benefits (which workers
receive whether employed or not). (For provisional
definition and measurement, see Weisskopf et al. ).
Building upon these insights, it is likely that we will
soon see much more fully developed and
sophisticated analyses both of the determinants of
wage growth and of the relationship between wage
growth and labour demand.
Taken together, these new hypotheses about
wage change and productivity growth themselves
combine to provide the possibility of much more
advanced hypotheses about determinants of changes
in the profit share. Given that it is formally true that
the rate of change of the real profit share is equal to
the rate of change of real productivity minus the
rate of change of real wages, analytic
determinations of changes in the class distribution
of revenues can now properly reflect both ‘social’
and ‘technical’ determinations.
2970
Global Power
As noted above, another elision in traditional
Marxian analyses of distribution has involved
international connections. Traditional analyses have
either assumed perfect competition, an awkward
first approximation, or have tended, following
models of monopoly capitalism, to assume a
constant or rising price mark-up. But in an open
economy, neither assumption seems useful, even as
a first approximation, because of the likelihood of
secular changes in a given economy’s relations with
other suppliers and buyers in global markets. And
these changes are quite likely to affect the
distribution of revenues, since they are bound to
affect either relative input prices or the mark-up and
through either path potentially to influence the real
profit share.
Analyses of international linkages have lagged
behind studies of capital-labour relations, but some
promising initial exploration are under way. Two
principal avenues of approach seem to be emerging.
One seeks explicitly to model the effects of changes
in the level and variability of the terms of trade on
domestic productivity and profitability. The other
aims at understanding and eventually modelling the
effects of changing conditions of international
power and, in particular, the effects of the
internationalization of capital and growing
multinational corporate leverage over domestic
labour. (Bluestone and Harrison ( ) provide a
useful early account of some of
these latter effects for the US.) This kind of work is
still in its early stages but seems increasingly
essential in a more and more interdependent
economy.
State Policy and Practice
The state can obviously have important effects on
the private distribution of income among classes,
both through tax policies and through the effects of
expenditures on the costs of production and the
relative bargaining power of the respective classes.
Work on these connections has not yet moved
beyond its early stages. Gough ( )
Distribution Theories: Marxian
reviews the paths of likely effect on both the tax and
expenditure side. Bowles and Gintis ( )
provide one provisional study of the effects of state
policies on the profit share in the United States. And
some of the studies of capital-labour relations
discussed above are beginning to shed important
light on the effects of ‘social wage’ expenditures on
private-sector wage and productivity determination.
Combined Effects
These three dimensions of power relations need not
be quarantined in separate cells of analytic-
isolation. It is possible to derive an inclusive model
of their combined effects which retains a focus on
the power relationships incumbent in each. Bowles
et al. ( ) provide one such
model of the determination of the profit rate; it
includes factors affecting labour intensity, relative
international power, and relationships with the state.
Applied econometrically, the model appears to
provide the most robust account available of
variations in the rate of profit in the US in the
postwar era. Although the study focuses on the rate
of profit as a dependent variable, its approach could
also permit more focused analysis of the profit share
as a potentially separable component of
profitability.
Comparative Analysis
It seems equally important, finally, to advance our
understanding of the factors which explain cross-
sectional variations in the levels and time patterns
of the class distribution of revenues and income.
This task must inevitably come rather late in the
game, since it largely presupposes the availability of
existing models of distribution which work for at
least one country or groups of countries on their
own terms. At the time of writing, some promising
initial studies of cross-national variations in the
determination of profit rates and shares are just
under way. The best existing review of the political
economic history upon which such studies
Distribution Theories: Marxian 2971

must build is the excellent comparative analysis analyses of labour segmentation have emerged in
provided by Armstrong et al. ( ). studies of various countries, and it is not at all clear
that a single uniform model of labour segmentation
in advanced capitalist formations can or should
One, Two ... Many Classes? emerge. These studies nonetheless suggest the
promise and importance of studying (a) the effects
One final analytic task remains. Almost all recent
of different structures of production and labour on
studies of distribution have accepted the traditional
the opportunities and realized incomes of individual
preoccupation with a two-class model of capitalist
members of the working class; and (b) the potential
economies - focusing almost exclusively on the
impact of systematically structured divisions within
single pair of opposing magnitudes, the profit share
the working class on the wage share of the class as a
and the wage share. It is important at least to
whole. Gordon et al. ( ) provide
consider the possibility that a more variegated
one important analysis of segmentation for the
categorization of individuals would be fruitful, even
United States; Wilkinson ( ) offers one useful
for traditional Marxian problematics. What about
early compilation of comparative studies; while
managers? The petty bourgeoisie? Financiers?
Bowles and Gintis ( ) provide a formal analytic
Different strata of the working class?
integration of segmentation analysis within the
Empirical analyses aimed in this direction have
value-theoretic context of more traditional Marxian
lagged in large part because of continuing uncer-
theory.
tainty and conflict over the appropriate definition of
These two approaches are potentially comple-
group boundaries and their inter-relationships. Two
mentary, not conflicting, since the former concen-
main approaches appear to have emerged as the
trates largely on the group distribution of realized
principal lines of inquiry within the Marxian
surplus value while the latter primarily explores the
perspective.
group distribution of variable capital. They have not
One approach seeks to derive a more complex
yet been properly vetted, compared, and integrated,
mapping of primary and ‘intermediate’ or ‘sub-
however, so we still await a complete and
sumed’ classes from the method and essential
satisfactory theoretical and empirical account of the
categories of traditional Marxian analysis. Sharp
distribution of revenues among all the relevant
debates nearly overwhelmed these efforts in the
categories of individuals in capitalist economies.
mid-to late-1970s, but it is conceivable that a
relatively widespread agreement on the terms of
analysis may be emerging in the mid-to late- 1980s.
See Also
Almost all of these analyses presuppose the
usefulness of a single category of ‘productive ► Marxian Value Analysis
workers’ and seek to distinguish, as carefully as ► Surplus Approach to Value and Distribution
possible, among various groups of intermediate ►value
agents and non-productive workers whose incomes
largely draw upon realized surplus value. Wright (
Bibliography
) offers one useful early
review of the possibilities and problems in this Armstrong, P., A. Glyn, and J. Harrison. 1984. Capitalism
approach, while Resnick and Wolff ( ) present since World War II. London: Fontana.
an interesting recent treatment. Baran, P.A., andP.M. Sweezy. 1966. Monopoly capitalism.
New York: Monthly Review Press.
A second approach, usually encompassed under Bleaney, M. 1976. Underconsumption theories. New York:
the general heading of ‘segmentation theory’, has International Publishers.
paid primary attention to the importance of various Bluestone, B., and B. Harrison. 1982. The deindustrialization
divisions within the working class. Different of America. New York: Basic Books.
2972
Boddy, R., and J. Crotty. 1975. Class conflict and macro
policy: The political business cycle. Review of Radical
Political Economics.
Bowles, S., and H. Gintis. 1977. The Marxian theory of value
and heterogeneous labour: A critique and reformulation.
Cambridge Journal of Economics 1(2): 173-192.
Bowles, S., and H. Gintis. 1982. The crisis of liberal
democratic capitalism: The case of the U.S. Politics and
Society.
Bowles, S., D.M. Gordon, and T.E. Weisskopf. 1986. Power
and profits: The social structure of accumulation and the
profitability of the postwar U.S. economy. Review of
Radical Political Economics.
Glyn, A., and B. Sutcliffe. 1972. British capitalism, workers
and the profits squeeze. Harmondsworth: Penguin.
Gordon, D.M., R. Edwards, and M. Reich. 1982. Segmented
work, divided workers. New York: Cambridge University
Press.
Gough, I. 1979. The political economy of the welfare state.
London: Macmillan.
Harris, D.J. 1978. Capital accumulation and income dis-
tribution. Stanford: Stanford University Press.
Kalecki, M. 1968. Trend and business cycle. In Selected
essays on the dynamics of the capitalist economy, 1933-
1970, ed. M. Kalecki. Cambridge: Cambridge University
Press, 1971.
Kalecki, M. 1971. Class struggle and distribution of national
income. In Selected essasys, ed. M. Kalecki. op. cit.
Lenin, V.I. 1917. Imperialism, the highest stage of capitalism.
In Selected works, one-volume ed. New York:
International Publishers, 1971.
Mandel, E. 1972. Late capitalism, English ed. Trans. Joris De
Bres, London: New Left Books, 1975.
Marglin, S. 1984. Growth, distribution, and prices.
Cambridge, MA: Harvard University Press.
Marx, K. 1867. Capital, vol. I. New York: International
Publishers, 1967.
Marx, K. 1894. Capital, vol. HI. New York: International
Publishers, 1967.
Resnick, S.A., and R.D. Wolff. 1985. A Marxian
reconceptualization of income and its distribution. In
Rethinking Marxism, ed. S.A. Resnick and R.D. Wolff.
Brooklyn: Autonomedia.
Shaikh, A. 1978. An introduction to the history of crisis
theories. In U.S. capitalism in crisis. New York: Union
for Radical Political Economics.
Weisskopf, T.E. 1979. Marxian crisis theory and the rate of
profit in the postwar U.S. economy. Cambridge Journal
of Economics 3(4): 341-378.
Weisskopf, T.E., S. Bowles, and D.M. Gordon. 1983. Hearts
and minds: A social model of U.S. productivity growth.
Brookings Papers on Economic Activity, No. 2.
Wilkinson, F. (ed.). 1981. The dynamics of labour market
segmentation. London: Academic Press.
Wright, E.O. 1978. Class, crisis, and the state. London: New
Left Books.
Distribution Theories: Neoclassical
Distribution Theories: Neoclassical
Christopher Bliss
Whenever a theory becomes involved in contro-
versy the question of what constitutes that theory
itself becomes a contentious issue, and the neo-
classical theory of distribution is no exception to
that general rale. Some have seen marginal pro-
ductivity as an essential feature of neoclassical
theory. Others have regarded the aggregation of
capital or an aggregate production function (even a
function of the Cobb-Douglas form) as essential.
Neoclassical distribution theory is viewed as
general equilibrium theory by many but Friedman
has defended the ‘Marshallian’ or partial equilib-
rium approach.
The truth is that any body of ideas widely
maintained for a long time inevitably develops and
transforms itself, absorbs some ideas, discards
others, and fathers traditions and sub-traditions. As
the neoclassical theory of distribution has been the
predominant view in the leading countries for the
development of economics for over 100 years, it is
not surprising that it conformed to this pattern and
expressed itself in diverse even contradictory
voices. Many, whether or not they like neoclassical
theory, hold that one voice represents the true
message, but neoclassical theory, like Christian
doctrine, may stand on certain fundamentals but is
not and could not be monolithic.
It is important to distinguish between ‘neoclas-
sical theory’ on the one hand and the history of the
development of that theory on the other. Both are
valid subjects for study but a scientific assessment
of the theory should address itself to the best
modem statements. This principle has not always
been respected, particularly in the heat of contro-
versy, and some maintain that the theory went
wrong from the start, and that if one could only go
back to where the vital mistakes were made
everything would become clearer. (For an extensive
development and discussion of this line of
argument, see Baranzini and Scazzieri .)
Distribution Theories: Neoclassical
However, the development of economic theory is
not like a complicated calculation in which every
step is supported by every earlier step. As with any
other discipline, the logical standing of a theory and
the history of the development of that theory are
distinct entities.
By way of illustration of the last point, consider
the way in which the theory developed in its early
stages. The ‘neoclassical’ movement, whose leading
members may be taken to include Bohm- Bawerk,
Edgeworth, Gossen, Jevons, Marshall, Menger,
Walras, Wieser and Wicksell, did not begin with a
theory of distribution but quite neglected that side
of the economic problem. By focusing on marginal
utility and the demand for given resources in a
barter economy, the neoclassical economists were
able to develop a powerful and flexible method, the
marginal principle, so impressive that it has often
been taken to define their approach. The so-called
‘psychic’ notion of marginal utility represented the
refinement, no more, of the old idea of ‘value in
use’. However with its help the neoclassical
eventually succeeded in clarifying, as Smith,
Ricardo and Mill had all failed to clarify, how value
in use, value in exchange and cost of production
could coexist. Only the Austrians with their concept
of ‘imputation’ hung on to the idea that utilities
were in some sense primary and other values
derived.
Put in unashamedly modem terms, the central
neoclassical idea is that the pricing of goods and the
pricing of factors of production are governed by
common principles, mainly the forces of supply and
demand generated by agents who maximize their
objectives. From the perspective ofthe history of the
development of the theory the definition is
anachronistic. Economics did not develop and refine
the notion of a factor of production or the concept
of maximizing an objective and later arrive at the
neoclassical theory of distribution. Rather the two
processes took place in tandem. Despite the lip
service to classical ideas paid by some members of
the neoclassical school, notably Marshall, neoclas-
sical is a misnomer. The neoclassicals were not
revivalists of classical economic ideas, an Oxford
Movement of classical political economy. They
were revolutionaries.
2973
The Distribution of Income
The theories with which we are concerned are
designed to explain the levels of payment to the
various factors of production - rents, wage rates,
and rates of profit - and by extension the shares of
the various factors in the total product. That is to
say that they are concerned with the functional
distribution of income.
We shall not discuss the distribution of personal
or household incomes, sometimes called the size
distribution. The size distribution of household
incomes takes the form of a function relating the
level of income and the number of units receiving
that income. It is true that given the distribution of
the ownership of factors among units, strictly the
quantities supplied to the various markets, and
given also the rates at which those factors are
remunerated, the size distribution may be derived.
However, except in the short run, the
interrelationship between the functional and size
distributions is more complicated. This is mainly so
because the quantities of factors which may be
accumulated by individual units, land and capital,
and even the quantity of labour, respond to rates of
return to the various factors. Pasinetti ( ) pre
sents a model which unusually takes this inter-
relationship into account. For a discussion of the
Pasinetti model and some of the criticisms which it
has attracted, see Marglin ( , pp. 324-8). On
the distribution of personal income and wealth, see
Atkinson ( ).
Factors of Production
It is not surprising that the concept of a factor of
production plays a leading role in neoclassical
theory because it lends itself to the view that the
inputs used in production stand to each other in a
relation of symmetry, governed by common
principles. This is not to say that no differences
between the conditions applying to factors are
admitted. The symmetry is most marked in the
treatment of the demand for factors, while on the
supply side important differences are recognized.
2974
The membership of the trinity of land, labour
and capital, which have always been taken to be
factors of production, goes back to the classical
writers, and an additional factor called
‘entrepreneurship’ is widely recognized by
neoclassical and classical writers alike. The
development of the theory along formal lines has
tended until recently to suppress the role of the
entrepreneur and to make the firm into a rather
lifeless object. However lately the increasing
employment of economic theory in industrial
economics has given rise to some richer treatments
of the firm.
The employment of the concept of a factor of
production has been criticized. It has been argued
that labour in particular does not submit itself to the
laws of supply and demand like any other input.
The introduction of distinctive features of the
various factors and their markets tends to
undermine the simple symmetry of pure theory'.
Some have detected apologetics in the designation
of capital as a factor of production. On this see the
discussion of ideology below.
Marginal Productivity
and the Determination of Factor Prices
Do marginal productivities determine factor
rewards? This apparently straightforward question
conceals conceptual complications and, depending
on the context to which the question is applied,
either ‘no’ or ‘yes’ may be defended as reasonable
answers. Robertson ( ) argued that
the wage rate ‘measures’ the marginal productivity
of labour. The reference is to the demand curve for
labour, which is the schedule of the marginal
productivities of various quantities of labour.
Robertson was reminding his readers that the wage
rate in a competitive market is determined by the
intersection of the demand curve and the supply
curve - both blades of the scissors cut the paper. If
marginal productivities are values determined by
the equilibrium solution as much as are wages and
prices, talk of one determining the other is
misplaced. The same point applies when the
marginal product of capital and the return to capital
are under consideration.
Distribution Theories: Neoclassical
In certain contexts however it is reasonable to
see marginal productivity as the determinant and the
payment to the factor as determined. Consider the
claim that managers of large enterprises are paid
very large salaries because the marginal value
productivity of a good manager amounts to a great
deal of money. Supposing this argument correct, the
high marginal productivity is a general feature
which does not depend upon solving out the whole
equilibrium. Contrast this with the case of a micro
unit, say a farm, facing a given wage rate for labour
and able to vary the quantity of labour employed.
For that exercise the wage rate is given and the
marginal product is determined by it.
A Simple Neoclassical Model
In this section we examine a static model. Growth
and capital will be considered below. The idea is to
construct a model in which factor prices will drive
everything else, including goods prices through cost
functions. This requires special assumptions but
makes for a model which can be easily presented
and which suffices to illustrate some points about
the neoclassical model of distribution. For a much
more thorough review of neoclassical models, see
Ferguson ( ).
We assume factors and goods to be distinct and
that factors are not directly consumed. Let there be
F factors available in given quantities, and G goods
producible from those factors, F and G need not be
equal and there may be more goods than factors, or
less or the same number. The production function
for the rth good is:
S=ftf) ( * = 1, . . . , G ) ; (1)
where v1 is the output of the r th good and x' is a
vector of factor inputs to the production of the rth
good. /() is a concave constant returns production
function. The cost function shows the unit cost of
producing good r given factor prices. Factor prices
are a vector w and the unit cost of the i th good is
C(w), where C(w) is the solution to the programme:
min wxl; (2)
x‘
Distribution Theories: Neoclassical
subject to:
/V) >1- (3)
We denote the prices of goods by vector c(w),
where the i th element of c(w) is C'(vv). There
are //households. Let the h th household own factors
^, in which case its income will be w - x11. All the
household’s income is assumed spent on goods and
the vector of goods demanded by household h is
denoted zh and is given by the h th household’s
demand function:
^ = 2^ [ c ( w ) , w - ^ ] (h = l ,. . . ,H ) .
(4)
Now note that factor prices w imply demands
for factors as may be seen by the following line of
reasoning. Given w, we have household incomes w
■ xh and goods prices c(w). Hence we have total
demands for goods:
^zh[c(vv),M--xh] = z. (5)
h undefined. In an extreme case there is only one
The amount of factor j used in the production of
good i is the partial derivative of C'(u) with respect
to uJ, denoted cj. The matrix of these coefficients,
denoted C, depends on w only. Hence demand for
factors is C • z, supply is Vv\ and we have shown
that excess demands for factors are a function of
factor prices.
To prove the existence of factor prices such that
factor demands and supplies are equal (strictly such
that there is excess demand for no factor), one has
to establish the continuity of the relationship
between factorprices and excess demands for
factors, and then employ a fixed point theorem (see
Arrow and Hahn, , ch. 5).
We note some salient features of this model.
First, prices of factors are determined by the supply
and demand for those factors although demands for
factors are derived demands depending on their
employment to produce goods. Secondly, both the
technology of production and tastes influence the
solution for factor prices. Thirdly, factorprices
measure the marginal products of factors, a property
which is ensured by
2975
the process of cost minimization. However there is
clearly no sense in which marginal products are
prior to prices.
More and Less General Models
The model of the previous section is designed to
illustrate the manner in which the determination of
the distribution of income may be viewed as the
outcome of a general equilibrium of supplies and
demands for factors of production. The model is
less general than the standard general equilibrium
model. It exhibits, for example, constant returns to
scale production functions, no joint production and
no direct consumption of factor services. Also,
goods are not used as inputs to the production of
goods. The introduction of those features would
undermine the model’s neatness without
introducing fundamentally new principles.
More striking results are produced when the
model is made still more specialised. The factor
input coefficients may be treated as constants
independent of w. In this fixed coefficient case the
marginal product of a factor in producing a good is
factor, usually labour, with the result that relative
goods prices are independent of demand. (For a
discussion of this non-substitution result and its
extension to an economy which uses fixed capital,
see Bliss, , ch. 11.) Models of this kind typically
introduce the use of goods as intermediate inputs to
the production of goods. However so long as there
is no genuine joint production (the term genuine
joint production is used to distinguish the
production of final demands jointly from the
notional joint production that arises when fixed
capital goods are treated as one of the products of
the productive process. ), the inputs used to produce
final output may all be reduced to the quantities of
the factor incorporated in them.
The model of Sraffa ( ), sometimes known
as the neo-Ricardian model, will be seen to be a
version of this model, but including an elegant
extension to fixed capital goods. Hahn ( )
has argued against the claim that the neo-Ricardian
approach leads to new insights by
2976 Distribution Theories: Neoclassical

pointing out that the model is a special case of the force. Let y be the total stock of goods available
general equilibrium model. next period for consumption and as inputs to next
period’s production. The production function
corresponding to a unit labour input is:
The Problem with Capital
F{y,x) = 0.
The introduction of capital into the theory of dis-
tribution raises two issues which should be distin- (6)
guished, even though they are not entirely
In steady state growth with a per capita con-
unrelated. One is the aggregation of capital, the
sumption of acQ, y will be acQ + (1 + y )x. Hence:
other is the nature of the supply of capital in the
long run. F[ OLC 0 + (1 + y)x,x] = 0 (7)
Although many expositions of the theory have
been expressed in terms of an aggregate called Given a particular per capita consumption ac 0,
capital, and there have even been attempts to (7) may be satisfied by various values of x, but only
formally underwrite this approach, it is now gen- one of these will be the efficient and equilibrium
erally recognized that there is no rigorous method of value. To see this let V(x]) be the maximum value of
aggregating a heterogeneous collection of capital jS such that jSc0 is a sustainable per capita
goods. (The most famous attack on the use of consumption starting with a capital stock x 1. If x1 is
aggregate capital is Robinson, -4); see also the steady state composition of the capital stock for
Champemowne, -4; Harcourt, and consumption [lc0 then x2=x1 must solve:
Marglin, , eh. 12.) In this respect capital stands on a
par with other types of input, labour for example. max a (8)
Highly aggregated models should therefore be seen
as simple devices for illustrating how a type of subject to:
model functions and not as descriptions of the
world. Unfortunately, some writers who emphasize F[(l + A)x2 + ac0,x'] > 0; (9)
the problems of aggregating capital are quite
cavalier when it comes to discussing the and
aggregation of labour or output. Formally however
there is tittle difference between the cases. V(x2) > pc 0 . (10)
With many distinct capital goods, demands for
inputs are demands for the services of particular Let the Lagrange multipliers attaching to the
capital goods. However the supply of capital in the constraints (9) and (10) be respectively n and r| and
long run is the supply of saving, which may let F,(i = y, x) denote the vector of partial
translate itself as required into particular capital derivatives of F with respect to the output and input
services. Hence a long-run neoclassical theory of vectors. The necessary conditions for a solution to
distribution depends on a model of long-run saving, (8)—(10) are:
a point which deserves emphasis.
1 + ncD ■ Fy = 0;
We show how the solution for the quantities of
capital goods and equilibrium prices may be (11)
obtained in a simple constant returns growth model.
Let there be N goods, and let the quantities of them and
which make up the capital stock used by one unit of
labour be represented by the elements of a vector x. /tfy( 1 +1) + rjVx = 0. (12)
Let consumption be proportional to a vector c Q, and
y the rate of growth of the labour Equation ( ) states that the marginal rates of
substitution between outputs of the various goods
shall be equal to the marginal rates of substitution
Distribution, Ethics of
between those same goods as inputs to the longterm
provision of future consumption; compare Dorfman
et al. ( , eh. 12). This condition
reduces the degrees of freedom enjoyed by the
steady state capital stock to one. That last degree of
freedom depends on the level of steady state Arrow, K.J., and F.H. Hahn. 1971. General Competitive
consumption the determination of which requires a
saving condition.
Theory and Ideology
According to its Marxist critics, neoclassical distri-
bution theory is irredeemably apologetic in charac-
ter, and it is indeed the case that some economists in
die past saw the theory, and in particular the
concept of marginal productivity, as throwing a
relatively favourable light on capitalism. When the
justification for tiie earnings of capital owning
rentiers was being questioned, the notion that
capital earns no more than its ‘contribution’ to
production was not unwelcome in the salons. The
idea that the rich are rewarded according to the
marginal productivity of their ‘waiting’ sounded
better still.
It can need a positive effort to see that all this is
strictly irrelevant to the scientific standing of neo-
classical theory. No one supposes that Newton’s
mechanics should be dismissed because its author
saw in it the justification of a hierarchical organi-
zation of social life. A play on the overtones of
words such as ‘earning’ or ‘waiting’ to justify the
distribution of income should be similarly
disregarded. Of course the neoclassical theory of
distribution can be used to analyse the effects of
policy, including policies to redistribute income. In
a perfect world the conclusions which emerged
from such investigations would be independent of
the political stance of the investigator. We do not
live in a perfect world, but the fact that the scientific J. B. Clark
ideal is never fully attainable should not lead us to
conclude that economics can know nothing but self-
serving apologetics.
See Also
► Adding-up Problem
► Clark, John Bates (1847-1938)
2977
References
Analysis. Amsterdam: North-Holland.
Atkinson, A.B.. 1975. The economics of inequality. Oxford:
Clarendon Press.
Baranzini, M., andR. Scazzieri. 1986. The foundations of
economic knowledge. Oxford: Basil Blackwell.
Bliss, C.J. 1975. Capital theory and the distribution of
income. Amsterdam: North-Holland.
Champemowne, D.G. 1953. 4The production function and the
theory of capital: A comment. Review of Economic
Studies 21(2): 112-135.
Dorfman, R., P.A. Samuelson, and R.M. Solow. 1958. Linear
programming and economic analysis. New York:
McGraw-Hill.
Ferguson, C.E. 1969. The neoclassical theory of production
and distribution. Cambridge: Cambridge University
Press.
Hahn, F.H. 1982. The neo-Ricardians. Cambridge Journal of
Economics 6(4): 353-374.
Harcourt, G.C. 1972. Some Cambridge controversies in the
theory of capital. Cambridge: Cambridge University
Press.
Marglin, S.A. 1984. Growth distribution and prices.
Cambridge, MA: Harvard University Press.
Pasinetti, L.L. 1962. Rate of profit and income distribution in
relation to the rate of economic growth. Review of
Economic Studies 29: 267-279.
Robertson, D.H. 1931. Wage grumbles. In Economic
fragments, ed. D.H. Robertson. London: P.S. King &
Son.
Robinson, J.V. 1953. 4The production function and the theory
of capital. Review of Economic Studies 21 (2): 81-106.
Srafifa, P. 1960. The production ofcommodities by means of
commodities. Cambridge: Cambridge University Press.
Distribution, Ethics of
The primary fact of economics is the production of
wealth. The division of the product among those
who create it is secondary in logical order and, in a
sense, in importance. Yet the most important subject
of thought connected with social
2978
economy is distribution. If the term be used broadly
enough it designates all of the economic process
that presents moral problems for solution. On the
settlement of the ethical questions concerning the
division of the social income depends not only the
peace of society but the fhiitfulness of industry. It is
a striking fact that Ricardo, whose smdies carried
economic science forward in the direction of the
truth concerning distribution, but stopped short of
that goal, and so strengthened the hands of social
agitators, realized the paramount importance of the
subject on which his thought was chiefly
concentrated: ‘To determine the laws which
regulate this distribution’, he says in his preface, 'is
the principal problem in political economy.'
Scientific errors concerning the law of distri-
bution react more harmfully on production than
do errors of doctrine concerning production itself.
Among self-asserting people, industry loses fruit-
fulness whenever the belief is widely diffused that
products are shared according to an unjust principle.
If it were a general conviction that social evolution
is in the direction of iniquity - that distribution
already robs the workers and will rob them more
hereafter - no force could prevent a violent
overturning of the social order.
Industry has its fruits and its sacrifices; it creates
useful things at the cost of working and waiting.
Where production is carried on in a collective way,
both the products and the burdens of the process
have to shared by different classes of men
according to some principle. The apportionment
that has to be made is not only of products, which
represent positive values, but of sacrifices, which
may be treated as negative values of a ‘subjective’
kind. While the term distribution, as currently used,
designates only the apportionment of the positive
values, or products, it is capable of being used in a
more complete sense, and made to include the
apportionment of the negative ones also. It would
then include all of economic science that involves
moral problems.
Both parts of this twofold distributive process
must in any ease be studied if the ethical questions
connected with industry are to be solved. There is
no independent standard of justice in the distribu-
tion of products only. What a man ought to get out
Distribution, Ethics of
of the collective income of mankind depends on
how much he or some one who represents him has
sacrificed in helping to create it. The apportionment
of the positive values referred to is inseparably
connected with that of the negative values. Political
economy must tell us how both products and
burdens are actually shared, and ethics must tell us
how both of them ought to be shared, if the existing
plan of social industry is to be morally tested.
Political economy has not as yet furnished a
theory of the actual distribution of positive values,
or products of industry, that has met with general
acceptance. It has scarcely attempted to furnish a
theory of the distribution of the negative values.
Ethical science has not furnished a clear standard of
justice in the double apportionment.
lively producer experiences in his own person
the double effect of industry; he is first burdened
and then rewarded. The net effect of the two
influences on the man’s well-being may be termed
the subjective resultant of production. A complete
science of distribution must study the economic
resultants in the case of different classes of men.
How is a labourer on the whole affected by indus-
try? What is the measure of the net benefit that
comes to him from this source? How is a capitalist
affected? How do the net effects compare with each
other? What tendencies are at work to change the
two, both absolutely and relatively? These are
economic questions; while the ethical question is
what the resultants in the two cases ought to be.
The personal resultant of industry is always a
positive quantity. Work yields a net gain; the fruits
of it are worth more than they cost. For the most
hardly-used classes an industrial life is, by eco-
nomic tests, more than worth living. The hours of
labour in a day are increasingly burdensome as the
period of work is prolonged. A man might labour
three hours a day with little weariness and no
injury. The eighth hour is wearying, and the tenth is
more so. There comes a time at which work
naturally stops, if the man is free, because working
longer would cost more in the way of pain than it
would secure in the way of pleasure. Final or
marginal labour is that which just pays for the
weariness that it costs. The gain that comes through
labour offsets the burden that it entails at
Distribution, Ethics of
the point in the working day at which the burden is
greatest. The less onerous labour of the earlier hours
affords a net personal gain. If the man is paid by the
hour he earns a part of his wages very easily.
Intramarginal labour, as we may term it, affords a
net subjective gain, what some would call
producers’ rent.
Though the wages of all hoius may be equal by
money standards, they are of unequal utility to the
man who gets them. His first earnings are spent on
necessities, later ones on comforts, and final or
marginal ones on things that figure in his estimate
as luxuries. The last hoiu of his labour may ensure
to him only the least important thing that he gets at
all. It is the minimum benefit seemed by an hour’s
labour that offsets the maximum sacrifice caused by
it. There is therefore a second net gain coming to
the worker in the spending of his money. As the
sixpence or dime that is spent for a luxury benefits
the man enough to offset the weariness of final or
most fatiguing labour, those that are spent for food,
clothing, etc., afford an additional benefit. The man
en ri ches himself whenever he buys a loaf of bread.
In general the sacrifices and the benefits of
production just offset each other at the point at
which the sacrifices are the greatest and the gains
are the least. Everywhere except at the margin the
gains are greater and the sacrifices are less.
Again the positive resultant of industry is
increased by social organization. Anarchy, even if it
were peaceful, would increase sacrifices and
diminish rewards. Whatever might be true of a
sparsely settled world, a crowded world is depen-
dent on the multiplying of productive power that
combination brings. All classes are debtors to
society. No serious case can be made against the
existing social order on the ground that it lessens
the gain that labour naturally brings.
The indictments brought against the social order
are based on the comparative treatment that society
accords to men of different classes. Are the benefits
conferred on different ones what they ought to be
relatively? Does society proceed capriciously in the
allotment of rewards and sacrifices? Do some
classes fail to get the proportionate benefit that is
properly theirs? Are social tendencies in the
direction of equity or away from it? These are the
ethical questions to be
2979
solved by a comparision of the ideally just distri-
bution with the actual one.
Of the ideals of distribution that have been
advanced none has been crude enough to provide
for the apportionment of the products of industry
and take no account of the burdens. A rule of equal
rewards for unequal sacrifices would have no moral
support. Ethical studies in this field really have as
their object the attainment of a rule for adjusting
what we have termed the personal resultants of
industry, or a rule that, if followed in practice,
would make the net effect of industry on the welfare
of different classes equitable. Communistic theories
make equality nearly synonymous with equity; but
the thing that is to be equalized is seldom mere
property or income. If the principle of equality be
carried into refinements, so as to bring to one level
the net benefits that society confers on all its
members, the rule approaches, though it is still far
from reaching, the ultimate moral ideal of
distribution.
The better socialistic ideals are refinements of
the rale of equality. In applying the rale to indi-
viduals, inheritance is the first disturbing influence
encountered. The law of inheritance is based on a
certain solidarity of families. Where it is in force the
sacrifices of a parent may accrue to the benefit of a
child. What we have termed the resultant of
industry in the case of the heir to an estate is not to
be measured by adding together positive values,
represented by the enjoyments that the property
brings, with negative values, represented by the
inheritor’s own sacrifices. If he be considered apart
from his family the values in the case are nearly all
positive. A crude levelling of individuals’ net gains
accruing from industry demands the abolition not
only of inheritance, but of gifts from parents to
children. Where it is advocated it is in the interest of
purely individualistic equality.
The handing over of all capital to the state
sweeps away even more completely inequalities of
wealth in permanent possession. In theory it might
avoid the evil connected with the abolition of
inheritance, that, namely of reducing the capital that
is necessary if wages are to be sustained at a high
rate; since it is conceivable that the state itself might
accumulate capital with needed rapidity.
2980
This measure also would, in effect, disregard the
solidarity of families and tend to put men on a
footing of individualistic equality.
Economic difficulties do not need to be con-
sidered in the shaping of a moral ideal. The vesting
of all capital in the state would save the student of
applied ethics one serious difficulty, that, namely,
of determining whether the sacrifice of abstinence is
unduly rewarded as compared with that of labour,
or, in other words, whether interest is too high as
compared with wages. A socialistic state has its
moral duty simplified, since it has only to reward
different kinds of labour equitably.
A scheme that is too crude to have much support
makes the wages and the working hours equal for
all. Estimate the wages in money or its equivalent,
gauge labour by time only, and bring both to an
equality in the case of the whole adult population.
Even the rewards are not thus in reality' equalized,
and the sacrifices are very unequal. In real rewards
unmarried men would be favoured and large
families would suffer. The real sacrifices incurred
would vary according to the nature of the work
performed.
An improvement on this scheme provides a
stipend for each dependent member of a family, and
tries to equalize sacrifices by so reducing the
number of hours of labour per day in occupations
that are disagreeable or hurtful, as to bring all
employments to a certain uniformity of
burdensomeness. In the case of very disagreeable
work the hours would be reduced to a minimum,
while in occupations that are less and less repellent
they would be shortened proportionately less.
Production would of course suffer by this arrange-
ment, and the ideal that the plan of division presents
is that of small but equal pay, with easy work, for
all.
Another scheme does not content itself with
equalizing what we have termed the personal
resultants of industry, but aims to level inequalities
of condition that lie at the back of industry itself.
Society should do more for the lame and the blind
than for those who have all faculties in possession,
in order that the ultimate condition of all may be
made as nearly equal as is possible. Here is the
levelling policy in perhaps its most
Distribution, Ethics of
ambitious mood. It is not the treatment of men by
society that is to be equalized, but the treatment of
them both by nature and society'. The industrial
organism is to deal with its members unequally in
order that it may somewhat neutralize the partiality
of nature.
A rule of division that is often regarded as
ethically lower than either of those above specified
is that of compensation according to acUial
production. Give to a man the wealth that he
creates, neither more nor less. Every one owns what
he brings into existence; let not society wrest or
filch from him any part of it. Let it keep itself clear
from robbery and fraud.
If workers lived side by side in peaceful anar-
chy, with no division of labour and no exchanges,
each man would get what he created. He would get
little, but he would get all that would be his own.
Introduce now a social union that multiplies prod-
ucts ten-fold but increases some men’s returns only
five-fold, and you seem to benefit these men and to
rob them at the same time. If in organized industry
some of the product that is distinctly attributable to
labour itself finds its way into the hands of men
who do not create it, the labourer suffers a wrong,
even though the share that he still keeps may be
larger by reason of the fact of his connection with
the men who rob him. Such is the conception of
industrial society that exists in many minds. The
socialistic indictment against society is that it
filches from workers a part of their share of the
extra product of industry due to organization. Does
society, under natural law, take from labour a prod-
uct that is distinctly attributable to it? This is one of
the most important questions in economics. A
successful analysis of social production answers it.
What needs to be known is what part of the
composite result of industry is distinctly due to
labour itself. In a land peopled by isolated producers
and managing to live in peace, each man would get
his own; does exchange vitiate this result? If so,
organization proceeds here on an unusual principle;
since the complications of society as a rule disguise
essential facts of primitive industry, but do not
annual them. The presumption is that the man who
got his own when he worked alone gets it when he
trades with his neighbour on terms of genuine
freedom, and that a true analysis
Distribution, Law Of
of social relations will show the fact. If so, society
tends actually to conform to the rule ‘to every man
the product that is distinctly attributable to the
sacrifices that he or others in his interest have
made’. There is common honesty in the distribution
that takes place under natural law.
The literature of the subject of economic ethics
is not as scanty as it is one-sided. The basis of the
socialistic movement is ethical, and much of the
literature is designed to prove that society is orga-
nized on a plan that systematically wrongs workers
in the apportionment of the social income. A
defence would naturally aim to show that the law of
distribution is not itself iniquitous, however many
particular cases of injustice might arise under it. A
weak point in the defence is the lack of a clear
demonstration of the complete nature of the actual
law of distribution, a lack that, as is hoped, may
soon be supplied. In the meanwhile statistics are
appealed to on both sides to prove, on the one hand,
that the actual apportionment of wealth is departing
more and more from the ideal standard, and, on the
other, that it is tending towards it.
R e p r i n t e d f r o m Palgrave s Dictionary
of Political Economy.
Distribution, Law Of
J. B. Clark
The most important share of the income of society
is the one falling to labour. The so-called ‘wage
fund’ theory accounted for the rate at which
labourers are paid on the ground that wages come
from a fund of capital devoted to the fund and that
the rate per man depends on the size of the fund and
the number of the claimants. The discovery of the
fact that wages come from the product of industry,
and not from capital, has made a new theory
necessary, and has opened the way to the discovery
of a general law of distribution.
The parties in the division of the general product
of industry are - (1) those who contribute to
2981
production the element labour; (2) those who con-
tribute instruments, or wealth in productive forms;
and (3) those who bring labour and productive
wealth into co-ordination by hiring both of these
agents, and receiving and selling their products. The
labour furnished includes the work of management,
as well as other kinds of industrial effort; and the
productive wealth, as the term is here used, includes
land as well as other instruments. The co-ordinating
function is, in this enumeration, kept distinct from
the other two; the man who performs it is not to be
treated in this connection as a labourer or as a
capitalist, but as the employer of both labour and
capital.
The shares to be accounted for are thus wages,
interest, and pure profit, and these shares will
include the rent of land and the wages of superin-
tendence. The generic varieties of gain come from
putting forth productive effort of some kind, from
furnishing productive wealth in some form, and
from bringing the effort and the wealth into
coordination.
The scientific law of distribution determines
what reward shall attach to the performing of one of
these functions. It does not gauge the income of a
particular man, since a man nearly always performs
more than one function. A capitalist usually works,
a labourer usually has capital, and an entrepreneur,
or coordinator of labour and capital, almost
invariably owns some productive wealth, and does
some directive work. A scientific study aims to
discover what determines the gain that attaches to
the working, to the saving, and to the coordinating.
As a man is a composite functionary, it tells us how
much he naturally gets in each of his various
capacities.
The nature of the distributive process. Social
production is a synthesis of distinguishable ele-
ments. Distribution is an analysis; and it reverses
the synthetic operation step by step. In organized
production one worker does not complete a product
from the beginning; if he applies his energy to crude
nature and begins the making of something that the
wants of society require, he passes the product in an
incomplete state to a successor. This man in turn
advances the article nearer to completion and hands
it over to a third man. The product, when ready for
final use, has passed through the
Subproduct Resulting from:
1. Elementary utility: Joint result of Capital and
wool Labour.
Joint result of C' and L'.
2. Place utility: transporting
3. Form-utility: Joint result of C" and L".
manufacturing
4. Form-utility: tailoring Joint result of C'" and L'".
hands of a series of workers each of whom has put
his touch on it and passed it to his successor.
The process may be represented by Table . The
garment, when completed, is an aggregate of
distinct utilities, and we use the term sub-product to
denote the quality imparted to it by each specific
group of producers. The sharing of the value that a
coat represents among the groups that have
performed the specific operations of production is
an analytical operation, that follows, in a reverse
direction, the steps of the productive synthesis.
The first sub-product in the series is wool. It
embodies an ‘elementary utility’, or one that results
from calling a raw material into existence. The
merchant’s sub-product is only the special utility
imparted to the wool by conveying it to his
warehouse, assorting it, and dividing it into
quantities convenient for purchasers. It is mainly a
‘place utility’, which is the service-rendering
quality that a thing acquires by being taken to the
place where it can be used; though in a complete
statement it would be necessary to recognize a
‘form-utility’ due to assorting and dividing. The
manufacturer’s sub-product is not the cloth, but the
‘form-utility’ imparted to the wool by transmuting it
into cloth. The tailor’s sub-product is the further
‘form-utility’ imparted to the cloth by making a coat
of it. Each specific utility is created by the joint
action of labour and capital; and each of these
agents must have its share of the value embodied in
its sub-product.
In order that the action of labour and capital
within the sub-groups may be a joint-action at all, it
is necessary that a certain coordinating act be done.
Some one must hire labour of the right kind, borrow
capital and invest it in the proper forms,
and cause the two to cooperate. This is the work of
the entrepreneur, in an unusually limited sense of
the term. This functionary, in his capacity as entre-
preneur, is not a capitalist and not a labourer,
however frequently it may happen that the man who
performs the coordinating function may perform
others as well. The coordinator, as such, is not a
business manager or superintendent. The
performing of this function does not require salaried
labour; indeed, after the process is begun, it scarcely
requires effort at all. Bargaining operations first
divide the total product of industry among the
general groups of which society as a whole is
composed. How much wealth shall come to the
entire group of workers, capitalists, and
entrepreneurs who are engaged in the creating of the
finished products, woollen garments? That depends
on the price for which the garments sell. A myriad
of finished products from other groups in the world
at large must come, by way of exchange, to minister
to the wants of the men in this one group; and the
quantity and quality of those products is fixed by
the sale of the clothing. This sale, and others like it,
perform the first and most generic dividing act that
takes place in the process of distribution. It
determines the total income of those who contribute
to the production of clothing.
What fixes the part of the income of this general
group that goes to each of the sub-groups that
compose it? Bargains again. Each group must buy
the utilities made by those that come earlier in the
series, and sell them, with the addition of its own
utility, to the group that succeeds it. The
manufacturing group buys wool and sells cloth; and
what it receives, less what it pays, constitutes the
reward of the manufacturing operation. As the first
division of the income of society resolves it into
rewards of general producing groups, the first
subdivision resolves the portion falling to one
general group into shares for the sub-groups that
constitute it.
A fijrther division is to be effected: it is that of
the shares falling to labourers, to capitalists, and to
entrepreneurs in each sub-group. Here is the test
operation of distribution; in this smallest of fields is
created and divided the wealth that rewards each
class in industrial society.
Distribution, Law Of
The productive operation from the fruit of which
labour and capital get their pay is intra- groupal; it
goes on within the specific industry in which a
particular force of men and their quota of capital are
engaged. The value that rewards woollen weavers
and spinners and the men who furnish them capital
is created wholly within the mill, and the sum that is
divided between these classes is a sum on which no
others have any claim. Yet the fact that labour and
capital both migrate freely from group to group, so
that workers from any group are able to share in the
special gains that may come to the earners in any
other, creates a certain solidarity of labour on the
one hand, and capital on the other. Give to the wool
spinners an advance of wages, and movements of
labour will in the end distribute the gain among the
whole working class. On the other hand, change the
cardinal relations of labour and capital as a whole,
and you change them in the end within every sub-
group. Labour is in reality trans- group al, and
capital is the same. Each is a productive agent, the
field of which extends directly across the sub-
groups of the diagram. It is the relation of all capital
to all labour that determines wages and interest. The
law of wages is nothing if not general, and the same
is true of the correlative law of interest.
It is a familiar fact that interest and wages tend
toward uniformity in different occupations. Men of
different productive powers may earn different
rewards, even within a single trade; and the labour
of management regularly receives more than work
of the ordinary kinds. Men differ in the amount of
working force that they possess, but men of like
power tend to receive uniform wages throughout the
series of industrial groups. If wages are high in the
woollen mill the young men and women who are
about to enter the field seek out this part of it, and
by their competition reduce the wages there
prevalent to the rate that prevails elsewhere. Interest
tends to a similar uniformity; under free competition
it tends to keep the same rate in all industries.
With interest has often been vaguely grouped
what we have termed pure profit itself; the gross
gains loosely attributed to capital tend toward
equality. It is, however, in a special way that the
2983
element that we have distinguished as pure profit
tends toward equality in different industries.
Wherever it comes into existence it sets at work
forces that tend to sweep it again out of existence.
In away this gain is self-annihilating. The uniform
rate toward which pure profit tends - though it never
reaches it in all groups at once - is a zero rate. Here
indeed, we reach controverted ground, and can
claim only to present one theory, not a view that has
universal support; but the evidence in favour of the
correctness of the view is simple and conclusive.
Competition tends to annihilate pure profit. The
existence in one sub-group of a gain that is in
excess both of interest on all the productive wealth
that is there used, and of pay for all labour, is an
inducement to the entrepreneurs of the group to hire
in the market both capital and labour, and secure the
pure profit that their joint industry creates. Let
woollen mills pay wages, including salaries, and a
double interest on the capital that they use, and the
mills will speedily enlarge their capacity. The
increase in the product will then reduce the price of
it, and ultimately bring the enlargement to an end.
Under natural law the sub-groups are in stable
equilibrium when, aside from insurance and taxes,
each earns wages on all labour, including the labour
of management, interest on all capital employed,
and nothing more. On this point the testimony of
experience confirms the conclusions of theory.
The equilibrium is never in practice perfect.
Causes that cannot here be analysed in any fulness
cause the element pure profit to continually
reappear. Inventions, as applied in particular
industries, give to one and another of the sub-
groups a gain that is in excess of that which
perfectly stable conditions would afford. The
occupation of new land creates, in a local way, a
pure profit for the earlier comers. Continually
appearing in particular parts of the field, and slowly
disappearing by reason of competition - such is this
element of the social income. If we watch a single
sub-group we find the profit at intervals appearing
and disappearing; if we watch the industrial field as
a whole we find it everywhere present, though not
long at the same points. Pure profit depends on a
relation between industrial groups. What the
manufacturer pays to
2984
the earlier groups in the series above represented,
and what he receives from the tailoring group,
determine this part of his gain. The actual position
of the entrepreneur himself, in the diagram that
describes the sub-groups, is on the line that sepa-
rates his own industry from the following one. He is
a purchaser of everything that is produced on the
left of that line. In the buying of materials he
purchases the products of the earlier sub-groups,
and in the paying of wages and interest he virtually
buys the sub-product created in the group to which
he himself belongs. The entrepreneur of the woollen
mill buys wool, and so pays for the sub-products
created by wool growers and merchants; and he
buys the form-utility created in the woollen mill
itself by making bargains with workmen and
capitalists, giving them fixed sums, and inducing
them to relinquish their claims on the cloth. As the
place of a particular workman and of a particular
amount of capital is, in the diagram, intra-groupal,
so that of a particular entrepreneur is inter-groupal.
Workers and capitalists get their pay from results
seemed wholly within their own industries, while
entrepreneurs get theirs from the fruits of mercantile
transactions between earlier groups and later ones.
Pure profit does not depend on the relation between
capital and labour. Moreover, where this profit
exists it is local, it depends on the relations between
adjacent groups.
We have shown that there is no law of wages
that is merely local. There is no force that gauges
the pay of wool-spinning independently of the
wages paid in other employments. There is a level
toward which all wages tend. There is likewise a
level toward which interest in every group tends.
What is the law that fixes these levels? What is the
general law of wages and interest? Here again wc
arc on ground that is actively contested, and wc
therefore only indicate the namre of a certain theory
without claiming for it a position of general
acceptance, and without arguing any points in
controversy.
In presenting it we may utilize a Ricardian
formula for determining the rent of land. If we
apply to a fixed area of land an increasing amount
of labour, we get returns that diminish per capita.
The first man set working on 100 acres creates a
certain amount of wealth as the result of the
Distribution, Law Of
tillage. Adding a second man does not double the
crop. Adding a third docs not increase by a half the
product due to the former two. Each man, as he
comes into the field, adds less to the total output of
the industry than did any of his predecessors.
This hypothesis makes the men enter the field in
a certain order of time, and the one who is the final
man is so in a literal sense - he is the last to arrive.
Actually putting the men into the field one at a time
is not necessaiy in order to reveal the principle that
governs the final productivity of labour. Let the frill
complement of men occupy the field at once, and
there will still be what may be treated as the final
increment of labour. Take any man away from the
force that tills the field, and the remaining men will
gain in per capita productivity by reason of his
absence. The departure of one man out of a force
numbering twenty does not reduce the crop by a
twentieth, since the nineteen men remaining work at
better advantage by reason of the withdrawal of
one. The final productivity of labour is gauged by
what would be lost if one man out of the force were
to stop working. We may, by way of illustration,
acUially set the men working one at a time, and find
what the last comer creates; or wc may set them all
working at once and see what would be lost by the
departure of one. The conclusion is the same in
either case: the final unit of labour is the least
productive.
If, now, land were the only form of productive
wealth that figured in the case, wages would equal
the amount created by this final or twentieth man.
That would gauge the amount that the employer
would lose through the departure of any one man in
the force. It would determine what he could afford
to pay to any one. Each man tends to get what he is
separately worth.
What would be true in the case of labour applied
to land, and using no other capital worth
considering, is actually true of labour applied to a
fixed amount of general capital, or to a fixed
quanUim of wealth in all productive forms, includ-
ing both land and other instruments. For the field of
limited extent in the Ricardian illustration sub-
stimte a fixed value, expressible in pounds or
dollars, and invested in such appliances of every
Distributive Justice
kind as working the needs of the working com-
munity require. If there are a hundred men in the
force, the departure of one of them will not reduce
the product by 1 per cent. His departure will add
somewhat to the productivity of the remaining
workers. After he is gone the capital will adapt itself
in form to the needs of the ninety-nine, and it will
be in a slight degree more ample in quantity per
man. Wages are gauged, as in the former case, by
the final productivity of labour. What on the whole
is lost by the departure of one man fixes the
importance to employers of every man. If each man
gets what employers would lose by his absence, he
gets that he is effectively worth.
This principle in a reversed application fixes the
rate of interest. It is the productivity of the final
increment of capital, as employed by a fixed labour
force, that gauges the pay of each increment. Let
there be 100 men using 100 units of capital. Take,
now, one unit of capital away, and you will not
reduce the product by 1 per cent. The 99 units of
capital will have gained in productivity per unit in
consequence of the departure of the hundredth. The
loss inflicted on the entrepreneur by the withdrawal
of the one unit of capital gauges the importance of
any single unit. Each unit of capital gets as its
compensation what would be lost if one unit of
capital were withdrawn. This diminution of the total
product due to the departure of the final unit of
capital gauges the importance to the entrepreneur of
each separate unit. It determines what he will pay
for the use of each one. Interest is therefore gauged
by the final productivity of capital. Each pound or
dollar tends, under natural law, to secure for its
owner what, in production, it is separately worth.
Bibliography
Caimes, J.E. 1874. Some leading principles of political
economy. London: Macmillan.
Clark, J.B. 1886. The philosophy of wealth. Boston: Ginn &
Co.
Clark, J.B. 1888. Capital and its earnings. Baltimore:
American Economic Association.
Clark, J.B., and F.H. Gidding. 1888. The modern distributive
process. Boston: Ginn & Co.
George, H. 1879. Progress and poverty. New York:
Appleton.
2985
Longe, F.D. 1886. A refutation of the wage-fund theory of
modern political economy. London: Longmans, Green.
Thornton, W.T. 1869. On labour. London: Macmillan.
von Bohm-Bawerk, E. 1884-9. Kapital und Kapitabins.
Innsbruck: Wagner.
Walker, F.A. 1876. The wages question: A treatise on wages
and the wages class. New York: H. Holt & Co.
Walker, F.A. 1883. Political economv. New York: H. Holt &
Co.
Wieser, F.F.B. 1889. Der Naturtiche Werth. Vienna: Holder.
Distributive Justice
Edmund S. Phelps
Social justice is justice in all of the relationships
occurring in society: the treatment of criminals,
children and the elderly, domestic animals, rival
countries, and so forth. Distributive justice is a
narrower concept for which another name is eco-
nomic justice. It is justice in the economic rela-
tionships within society: collaboration in
production, trade in consumer goods, and the pro-
vision of collective goods. There is typically room
for mutual gain from such exchange, especially
voluntary exchange, and distributive justice is
justice in the arrangements affecting the distribution
(and thus generally the total production) of those
individual gains among the participants in view of
their respective efforts, opportunity costs, and
contributions.
In earlier times the discussion of distributive
justice tended to focus upon the obligations of the
individual toward those with whom he or she had
exchanges. So an employer was expected to be just
or not to be unjust, and the problem was to
demarcate employer injustice. With the rise of
governments capable of redistribution and the
spread of economic liberalism, the focus shifted to
the distributional obligations of the central gov-
ernment. Let enterprises and households pursue their
self interests while the government attends to
distribution (within the limits of its just powers).
Distributive justice is largely about redistributive
taxation and subsidies. The latter may take many
2986
forms such as public expenditures for schooling and
vocational training (beyond the point justified only
by the Pareto principle from the status quo ante) as
well as cash subsidies for the employment of labour
or low-wage labour (whether paid to employer or
employee).
Note that the so-called negative income tax,
whatever the claims for or against it as a tool of
social justice, does not appear to be an instrument
for distributive justice unless restricted somehow to
those participating (more than some threshold
amount?) in the economy (and thus in the gener-
ation of the gains to be (re)distributed). In any case,
it will not be discussed here, although some
propositions about subsidies apply also to the
negative tax.
The suggestion that distributive justice might (at
least in principle) require subsidies, not merely tax
concessions or tax forgiveness for the working poor,
tends to raise the eyebrows of some and accounts
for the fact that distributive justice raises the
hackles of a few. As long as the Iroquois and the
Sioux have no contact, there are no gains to be
distributed and distributive justice does not apply; if
they are let free to engage in hi lateral inter-tribal
exchanges, however, the payment of a subsidy to
pull up the wage of the lowest earners, who are
Sioux, say, would come partly or wholly at the
expense of the Iroquois. Now some commentators
object to the notion that the Sioux, whose
exchanges with the Iroquois are entirely voluntary
and all of whom have benefited (or could have), we
may suppose, might deserve an additional payment
from the Iroquois, perhaps through some supra-
tribal authority. Ayn Rand ( ),
for example, argues that it is one thing to require of
a poor person a fare for riding a bus with empty
seats that the other riders can finance out of the
benefits they receive from the bus - she has no
qualms about such a free ride - and another thing for
the poor person to tax the other riders. But she has
got the economics wrong in the application of her
(actually rather Rawlsian) ethical premise. Up to a
point, a subsidy to the poorest-earning group (the
Sioux in the above example) would have the others
(the Iroquois) still with a net gain - a gain after the
tax needed to pay the subsidy. This is because of
diminishing returns: When the group
Distributive Justice
of Sioux workers is added to the fixed pool of
Iroquois’ labour and land, the extra product added
by the first arrivals - and, more generally, the
average of the extra products added by the suc-
cession of Sioux workers - is larger than the extra
product resulting from the last of these workers,
which is the ‘marginal product’ of Sioux labour; the
Iroquois could afford a subsidy equal to the excess
of the average extra product over the marginal
product. Correctly applied, then, the Randian
objection is to a gain-erasing or, at any rate, a gain-
reversing subsidy, not to any subsidy whatsoever.
Another objection to the concept of distributive
justice and to the admissibility of subsidies argues
that if these notions were sound it would make
sense, by analogy, to apply them to marriage
allocation, to the matching of husbands and wives;
since we never hear of such applications the ideas
are presumably unsound. Of course, it would strike
us as novel and foreign to see a proposal for a tax on
marriage with Iroquois men and a subsidy to
marriage to Sioux men on the ground that the
former were apparently more attractive to women
(from either tribe) and the resulting inequality of
benefits unjust and demanding correction. But the
reasons might be other than the supposed
unacceptability of the ideas of distributive justice.
Maybe the impracti- cality of deciding on the taxes
and subsidies stands in the way. Perhaps a marriage
subsidy would be demeaning while employment
subsidies would not, being graduated or even a flat
amount per hour. Yet the key observation may be
that, although there is economic exchange here and
although racial discrimination or racial prejudices
could cause real injustices, the Sioux and Iroquois
men in this example are not cooperating for mutual
gain and so no problem about the just division of
such gains can arise; they are competing, or
contesting, for partners, not forming partnerships
with one another. Thus distributive justice cannot
apply here.
The terms offered to the working poor, as
already implied, is the locus classicus to which
notions of distributive justice have been applied.
However, two other arenas in which issues of
justice are being fought out should be
Distributive Justice
mentioned. One of these is the problem of inter-
generational justice. It was first addressed in a
celebrated paper in by Frank Ramsey, who adopted
as the criterion of optimality the standard associated
with utilitarianism - the sum of utilities over time.
This conception of inter- generational justice
encountered difficulties when in the 1960s it was
applied to optimum saving of a society in which the
population is to grow without bound, although that
odd demographic case may have put utilitarianism
to an unfair (and absurd) test. In John Rawls
struggled with the problem of intergenerational
justice in a famously problematic section of his,
only to conclude that ‘. . . the difference principle
[i.e., Rawls’s maximin or, more accurately, leximin
principle] does not apply to the savings problem.
There is no way for later generations to improve the
situation of the least fortunate first generation.’ This
seems to say that intergenerational justice, if there is
such a thing, is not a problem of distributive justice,
since there is no cooperation for mutual gain among
generations, not even between adjacent ones in the
chain. But the premise that the current generation
cannot be helped by succeeding generations
appears, on the face of it, to be a slip in Rawls’s
economics. In a closed economy, we can help future
generations by providing them with more capital -
even in an open economy enjoying perfect capital
mobility, we can provide them with social overhead
capital that the world capital market would not
provide (or not so cheaply) - and, if overlapping
with us, they can help us by meeting consumption
claims we make through our issue of public debt
and pension entitlements. Thus distributive justice
does apply here, with a precision fit. What Rawls
may be interpreted to mean is that if, being the least
fortunate owing to heaven-sent technological
discoveries over the future, the present generation
were permitted to invest nothing (not even gross of
depreciation!) - rather as we can imagine the poorest
in the static problem to begin by sullenly asking for
equality - the future generations could not bribe the
present one to do something in their mutual interest
- unlike the static problem in which the rich can
explain the benefits of trickle-down. But in fact the
next generation can bribe the present one with
2987
some old-age consumption in return for some
investment. It may be conjectured that a maximin-
optimal growth path would still exist in a model
along the lines of the Phelps-Riley model
notwithstanding the introduction of technological
progress.
The other arena in which we find a debate over
distributive justice is the international trade field.
When a giant nation trades with a small number of
pygmy countries, not large enough even in the
aggregate to influence relative prices in the giant
state, the latter receive all the gains from trade and
the former gets nothing and loses nothing; this is
exactly the Rawlsian maximin solution if perchance
the pygmy countries are poorer (in some suitably
defined way) than the giant. But if these tiny
countries ‘spoil the market’, worsening their terms
of trade in the course of exporting to and importing
from the giant, because they are not of negligible
size at least in the aggregate, then the Rawlsian
solution is not obtained by the free market. The
recent North-south problem of which the ‘Southern’
countries complain can be understood as the
tendency of the ‘Northern’ countries that are
already the richest countries, such as the North
American and European countries, to retain the gain
from trade resulting from the aforementioned
change in the terms of trade caused by the
‘Southern’ countries through their trade with the
‘Northern’ ones. The ‘Southern’ countries believe
justice to require that the ‘Northern’ countries
arrange to give back that gain through some
appropriate international transfer mechanism.
There are able and serious philosophers who
would be happy to see distributive justice left to the
economists. In fact, the history of philosophy has
been seen as a process of divesting itself of a sub-
field as soon as it could thrive independently.
Likewise, there are economists who would leave the
subject to philosophers. But, whichever group
receives the lion’s share of the contract to work on
it, it seems that the economics (as well as philos-
ophy) of the problems being studied is an essential
element of the subject. In this sense and for this
reason, the necessary cross listing notwithstanding,
distributive justice is an important field under
economics.
2988
See Also
► 3 i.v.v:.:..-v:.\:
► 3;;:.;':. 7
► goods; Proposal power; Targeted public
► 3/:voi33o::.
►
Bibliography
Phelps, E.S., and J.G. Riley. 1978. Rawlsian growth:
Dynamic programming of capital and wealth for
intergeneration ‘maximin’ justice. Review of Economic
Studies 45(1): 103-120.
Ramsey, F.P. 1928. A mathematical theory of saving. In
Economic justice, ed. E.S. Phelps. Hamrondsworth:
Penguin.
Rand, A. 1973. Government financing in a free society. In
Economic justice, ed. E.S. Phelps. Hamrondsworth:
Penguin.
Rawls, J. 1970. A theory of justice. Oxford/Cambridge, MA:
Oxford University Press/Harvard University Press.
Distributive Politics and Targeted Public
Spending
Brian G. Knight
Abstract
This article analyses common pool problems
associated with the provision of local public
goods by central legislatures. In response to
incentives associated with common pool
problems, legislators act to maximize spending
for their home jiuisdiction but to restrain
spending elsewhere due to the associated tax
costs. The resolution of this conflict between
jiuisdictions depends in the United States upon
the distribution of political power across
Congressional delegations. Incumbents are
rewarded for delivering federal spending to their
jiuisdiction through increased voter support.
Distributive Politics and Targeted Public Spending
Keywords
Common pool problems; Distributive politics;
Earmarked projects; Lobbying; Local public
spending
JEL Classifications
H5
While conventional models of political economy,
such as the median voter model, focus on the
provision of national public goods, most federal
spending programmes, such as the US interstate
highway system, are more aptly characterized as
local in nature. While in the United States the
benefits of federal spending are concentrated in
specific geographic units, such as states, counties,
and Congressional districts, the associated tax costs
are, by contrast, geographically dispersed. This
common pool featiue of federal spending -
concentrated spending but dispersed financing -
leads to a geographic tug-of-war in which
jiuisdictions attempt to increase own-jurisdiction
spending but to reduce spending elsewhere due to
the associated tax costs. This conflict between
jiuisdictions is reflected most intensely in the
budget process within the US Congress, whose
members are locally elected and thus naturally
respond to these common pool incentives.
In this article, I first summarize evidence
suggesting that Congressional representatives are
responsive to the common pool incentives associ-
ated with concentrated spending but dispersed
costs. Having established the empirical saliency of
this common pool problem in Congress, I then
summarize the literature examining how this con-
flict is resolved. In particular, I analyse the effects
of Congressional delegation characteristics, such as
size, ideology, seniority, and committee assign-
ments, on the geographic allocation of federal
funds. Finally, I review evidence on the effects of
the geographic distribution of federal funds on
electoral outcomes.
Distributive Politics and Targeted Public Spending
As described in Knight ( ), common pool
problems underpin several theoretical models of the
legislative process, such as the universalism model
of Weingast, Shepsle, and Johnsen (1981) and the
legislative bargaining model of Baron and Ferejohn
( ). Whether or not Con
gressional delegations respond to these incentives in
practice, however, is primarily an empirical
question. It may be the case, for example, that
political parties, or related Congressional organi-
zations, serve as collective mechanisms through
which legislators internalize the tax costs in other
jurisdictions associated with own-jurisdiction
spending. One of the first papers to directly measure
the responsiveness of representatives to common
pool problems is by DelRossi and Inman ( ),
who examine the geographic distribution
of water projects authorized by the Water Resoiuces
Development Act of 1986. In particular, the authors
compare the size of project requests before and after
changes in local matching requirements, which
significantly increased the fraction of project costs
financed by local governments. As hypothesized,
districts experiencing larger increases in matching
rates requested significantly less funding for water
projects. In a similar vein, Knight ( ) examines
Congressional voting in 1998 over whether to
finance a set of transportation projects, which were
earmarked for specific Congressional districts and
were funded primarily via federal gasoline taxes. As
predicted, support for funding was concentrated in
those districts receiving more in funding and also in
those districts with lower gasoline tax burdens.
How is this geographic battle between jurisdic-
tions resolved? Which states and Congressional
districts win and why? Regarding the mere size of
delegations, an important feature of the US Con-
gress is its bicameral structure in which each state
has an equal number of delegates in the Senate but
in which seats are apportioned between states
according to population in the House of Repre-
sentatives. This equality of delegation sizes in the
US Senate provides small states with power dis-
proportionate to their population; Senators from
2989
California, the largest state, currently have over 60
times as many constituents as do senators from
Wyoming, the smallest state. In attempting to
measiue the magnitude of this small-state bias,
Atlas et al. ( ) and Lee ( ) find that small
states receive significantly more per capita in
aggregate federal spending than do large states.
While this finding is certainly provocative, it is
difficult to distinguish between the role of Senate
representation and other factors, such as population
density, that make small states inherently different
from larger states. In attempting to address this
issue of unobserved differences between small and
large states, Knight ( )
demonstrates that small states receive considerably
more per-capita funding in projects earmarked in
Senate bills; in House bills, by contrast, small and
large states receive similar project spending on a
per-capita basis. Knight ( )
also identifies two theoretical channels underlying
this small-state bias in the US Senate. Relative to
their population, small states are disproportionately
represented on key committees (the proposal power
channel) but are also cheaper coalition partners (the
vote cost channel) given that they pay a smaller
share of federal taxes. Interestingly, both channels
are shown to be empirically important and, taken
together, explain over 90 per cent of the measiued
small-state bias. In a related study of the size of
delegations, Falk ( ) studies
discontinuities in the apportionment of seats in the
US House arising from both timing (re-
apportionment occurs once every ten years) and
rounding issues (delegation sizes must be integers).
Using this variation in delegation sizes, he finds that
increases in seats per capita lead to statistically
significant increases in federal spending per capita.
Delegations of similar sizes, however, may
differ significantly in their composition. Key dif-
ferences between delegations in the degree of
political power include majority party affiliation,
seniority, and representation on key committees.
Regarding majority party affiliation, Levitt and
Snyder ( ) find that the Democratic Party
used its majority control of Congress to channel
2990
federal fluids into Congressional districts with a
high percentage of Democratic voters during the
period 1984-90. However, they find no evidence
that, conditional on the percentage of Democratic
voters, districts represented by Democrats received
higher federal spending. Levitt and Poterba ( )
report that states with very senior
Democratic representatives experienced more rapid
economic growth than did other states. However,
they find no relationship between the partisan
affiliation of delegations and the geography of
federal spending, a key hypothesized channel of the
measured differences in economic growth.
Regarding the role of Congressional committees,
Knight ( ) finds that Congressional
districts represented on key committees received
substantially more funding in projects earmarked in
transportation bills authorized in 1991 and 1998. He
interprets this result as evidence of the importance
of proposal power associated with the committee’s
ability to set the legislative agenda. De Figueiredo
and Silverman ( ) examine
interactions between committee representation and
lobbying in an empirical examination of earmarked
projects for universities. In particular, they find a
strong correlation between lobbying outlays by
universities and the receipt of federal funding; this
link between lobbying and spending, however, is
found to be much stronger for those universities
located in districts that are represented on key
appropriations committees.
We have focused throughout this survey on the
determinants of the geographic distribution of
federal funds. Politicians, however, have an
incentive to put forth the effort to secure project
funding only if they perceive that the associated
political gains are sufficiently high. While clearly
important, measurement of the effects of federal
spending on incumbent vote shares is plagued with
endogeneity problems. For example, incumbents
facing the strongest opposition have the strongest
incentives to put forth effort in securing funds.
Thus, there may be a downward bias in ordinary
least squares (OLS) estimates of the effect of
federal spending on incumbent vote shares. As an
instrument for district-specific federal spending,
Levitt and Snyder ( ) use fed
eral spending outside of the district but within the
Distributive Politics and Targeted Public Spending
state. The idea is that other actors, such as Senators
or governors, also play a role in the geographic
distribution of federal funds. Using this exogenous
variation in federal spending, they conclude that an
additional $100 per capita in spending translates
into an additional two percentage points in
incumbent vote shares.
We conclude that common pool problems
associated with concentrated project benefits but
dispersed costs are reflected not only in the behav-
iour of Congressional delegations but also in the
resulting distribution of federal funds. Who wins
and who loses in this geographic battle is deter-
mined in part by state size and the political power
of delegations. Consistent with these results, evi-
dence suggests that incumbent re-election prospects
are significantly enhanced by increases in federal
spending.
See Also
► atergovemmental Grants
► .ocal Publ ance
► 'olitical ;itiitions, Economic Approaches to
► Public Choice
Bibliography
Atlas, C.M., T.W. Gilligan, R.J. Hendershott, and M.A.
Zupan. 1995. Slicing the federal government net spend-
ing pie: Who wins, who loses, and why. American
Economic Review 85: 624-629.
Baron, D.P., and J.A. Ferejohn. 1989. Bargaining in legisla-
tures. American Political Science Review 83: 1881-1207.
de Figueiredo, J.M. and B.S. Silverman. 2002. Academic
earmarks and the returns to lobbying. Working Paper No.
9064. Cambridge, MA: NBER.
Falk, J. 2006. The effects of Congressional district size and
representative’s tenure on the allocation of federal funds.
Working paper, University of California, Berkeley.
DelRossi, A.F., and R.P. Inman. 1999. Changing the price of
pork: The impact of local cost sharing on legislators’
demands for distributive public goods. Journal of Public
Economics 71: 247-273.
Knight, B.G. 2004a. Legislative representation, bargaining
power, and the distribution of federal funds: Evidence
from the U.S. Senate. Working Paper No. 10385.
Cambridge, MA: NBER.
Diversification of Activities
Knight, B.G. 2004b. Parochial interests and the centralized
provision of local public goods: Evidence from con-
gressional voting on transportation projects. Journal of
Public Economics 88: 845-866.
Knight, B.G. 2005. Estimating the value of proposal power.
American Economic Review 95: 1639-1652.
Knight, B.G. 2006. Common tax pool problems in federal
systems. In Democratic constitutional design and public
policy analysis and evidence, ed. R.D. Congleton and B.
Swedenborg. Cambridge, MA: MIT Press.
Lee, F.E. 1998. Representation and public policy: The
consequences of Senate apportionment for the geographic
distribution of federal funds. Journal of Politics 60: 34-
62.
Levitt, S.D., and J.M. Poterba. 1999. Congressional dis-
tributive politics and state economic performance. Public
Choice 99: 185-216.
Levitt, S.D., and J.M. Snyder. 1995. Political parties and the
distribution of federal outlays. American Journal of
Political Science 39: 958-980.
Levitt, S.D., and J.M. Snyder. 1997. The impact of federal
spending on house election outcomes. Journal of Political
Economy 105: 30-53.
Weingast, B.R., K.A. Shepsle, and C. Johnsen. 1991. The
political economy of benefits and costs: A neoclassical
approach to distributive politics. Journal of Political
Economy 89: 642-664.
Diversification of Activities
A. Cosh
Diversification is the process by which the modem
corporation extends its activities beyond the
products and markets in which it currently operates.
It is a major determinant of the structure of modem
industrial economies and has important implications
for competition and efficiency. Robinson ( , p.
114) defines diversification as ‘the
lateral expansion of firms neither in the direction of
their existing main products, as with horizontal
integration, nor in the direction of supplies and
outlets, as with vertical integration, but in the
direction of other different, but often broadly sim-
ilar, activities’. The extent of diversification can be
measured in a number of ways, but is hampered by
the difficulty of precisely defining the boundaries
between different products, markets and industries.
It is not a simple task to assess the
2991
degree to which a firm spreads its operations over
different activities. The more narrowly defined are
these activities the greater will be the apparent
degree of diversification. These problems are not
unique to the measurement of diversification and
similar difficulties arise in the measurement of
concentration in industry. Indeed the process of
diversification itself has played a major part in
blurring the distinction between industries and in
creating these measurement problems. However, it
is clear that diversification must involve the firm in
producing new products which are sufficiently
different from its existing products to involve the
firm in new production or distribution activities.
Diversification may therefore involve only a small
change of direction, or a dramatic switch into an
entirely new line of business. In the literature the
former is referred to as related, or narrow spectrum
diversification and the latter as unrelated, or broad
spectrum diversification.
One possible measurement of the extent of
diversification involves identifying the number of
industries, or products in which the firm is involved.
The other main approach is to measure the
proportion of the firm’s activity in its core business
in comparison with the proportions in its diversified
activities. This measure has been refined in a
number of ways to take account of the number and
importance of these diversified activities (e.g. Berry
; Jacquemin and Berry ; Utton ).
The process of diversification is not a new
phenomenon, but the principal empirical studies
(Gort ; Rumelt ; Berry ; Utton
) have demonstrated a marked increase in the
degree of diversification over the past few decades.
The studies suggest that diversification tends to be
narrow spectrum diversification into similar
industries. However, both Gort and Rumelt were
able to discern some shift towards broad spectrum
diversification. The intensity of narrow spectrum
diversification was found to be industry related, but
the extent of broad spectrum diversification was
independent of the primary industry from which
diversification was occurring. Rumelt was also able
to identify a growth in importance of acquisitive
conglomerates and
2992
there can be little doubt that their importance has
grown further since his study. There was general
agreement that firms tended to diversify into
industries characterized by high research and
development intensity and rapid technological
change. The industries also tended to be faster
growing, but showed no significant differences in
terms of profits variability, or the degree of
concentration, than industries less popular with
diversifying firms. The industries from which
higher levels of diversification occurred were not
slower growing than other industries, but did tend to
be characterized by a higher degree of seller
dominance. Such industries might give less scope
for firm growth by capturing market share. The
more rapid diversifiers tended to be larger firms
with higher proportions of scientific and technical
employees and this is consistent with the impor-
tance of techno logical industries as diversification
choices which was noted above. Finally firms with
above average rates of diversification tended to
have above average rates in subsequent periods.
This may be related to the organizational changes
associated with diversification which have been
identified by Chandler ( ) and others
(e.g. Williamson ; Channon ). This issue is explored
further below.
The growth of firms and the role of diversifi-
cation in this growth process were elucidated in the
pioneering work of Penrose ( ). Penrose
identified three explanations for diversification:
first, as a response to specific opportunities; second,
as a response to specific threats; and third, as a
general strategy for growth. The opportunity to
diversify arises naturally as a byproduct of the
existing activities of the firm. A key area is the
research and development activities of the firm.
Such activities develop the firm’s knowledge of its
technology which is unlikely to be product specific.
Furthermore whether research is carried out only to
improve the firm’s existing products, or the develop
new products, it is likely to provide new
opportunities for diversification. The knowledge of
the markets for its existing products and their
channels of distribution provide the firm with other
opportunities for diversification. Another
opportunity for diversification arises from retained
earnings from existing activities. The
Diversification of Activities
finding that these earnings are invested in diversi-
fication rather than, for example, paying dividends,
is probably associated with the growth orientation
of management and the tax position of shareholders.
Thus the normal operations of the firm create both
new opportunities for expansion and the availability
of unused productive resources to meet these
opportunities. The second explanation offered by
Penrose concerns the exposure declines in demand
for their products. Diversification is a means of
spreading risk through reducing the firm’s
dependence on a few products. The reduction in
perceived risk may also reduce the cost of capital to
the firm. Diversification may also occur in response
to diversification by a competitor. This type of
competitive strategy raises the question of the
implications of diversification for competition and
this issue is examined below. Finally diversification
may occur as part of a general policy for growth.
This part of Penrose’s work has been taken further
by Marris ( ). Marris gives diversification a cen
tral role in his model of the growth of firms. The
management of firms have a strong motivation to
seek growth since it confers on them improved
status, salary and security. But growth within their
existing markets will eventually be limited by the
growth of demand for these products and diversi-
fication is the means by which this demand con-
straint may be overcome. It has been argued above
that a certain degree of diversification will be both
natural and beneficial as opportunities are exploited.
However, Marris argues that management will be
prepared to press growth, and hence diversification,
beyond the level which is optimal for shareholders.
The drawback of too rapid a rate of diversification
is a higher failure rate of new products due to a lack
of managerial, financial, development and
marketing resources. This may be a less reasonable
proposition when the possibility of growth through
merger is recognized. However before considering
this it is worth looking at the changing structure of
firms which has evolved with diversification.
The development of the M-form, divisionalized
company was identified by Chandler ( ) to
be a response to the growth and, more
particularly, diversification of the modem
Diversification of Activities
corporation. Subsequent research (e.g. Channon ;
Rumelt ; Williamson , ) has
reinforced their inter-connection to such a degree
that it is necessary to interpret the consequences of
diversification within the context of the
divisionalized company structure. In this structure
responsibility for profitability is restored to divi-
sional managers whose performance can be
assessed. Top management is freed from day-to-
day operational decisions and can concentrate on
the allocation of funds between the divisions and
other aspects of strategy. The divisional structure
significantly reduces the organizational constraints
of diversified growth, particularly growth by
acquisition. The acquisition of new divisions, or
sub-divisions, by takeover can be achieved quickly
and with minimum disruption. Diversification
through merger is often seen as less risky since it
involves the acquisition of the physical assets,
existing products and channels of distribution
required and brings with it management and
employees who are experienced in this area of
activity. Furthermore, entry is achieved without
initially having to compete for a market share. On
the other hand if the motive for diversificati on i s to
utilize spare resources within the firm, or to exploit
some technological development, then
diversification by internal growth may be preferred.
It appears that the importance of diversification
mergers has increased in recent decades, partly as a
response to the increase in strength of competition
policy.
The US merger laws have evolved into a potent
deterrent against sizeable horizontal and vertical
mergers. It is doubtful, however, whether they have
had much impact on the overall level of merger
activity which has continued at high levels (Scherer ,
p. 588).
A substantial controversy surrounds the question
of what impact diversification has on competition
and efficiency. At first sight the creation of large,
non-specialized firms would be expected to reduce
both, but there are counter arguments. The evidence
does not suggest that diversification raises market
concentration. Indeed, broad spec- frum
diversification may be a force for reducing
concentration in individual markets. Large firms
diversifying are able to overcome many barriers to
2993
entry and may promote competition by their entry.
Diversification may be the only means by which
firms may grow large enough to reap pecuniary
economies of scale, without becoming too dominant
in a single market. It is also argued that the diversity
of products, as well as large size, brings a greater
potential benefit from research. Therefore large,
diversified firms may be more likely to engage in
intensive research and development, to the benefit
of the whole economy. The associated introduction
of the M-form organization is argued to lead to
improved internal efficiency of the firm as divisions
strive to meet profit targets and compete for funds.
It is also argued that the internalizing of the capital
market within the large, diversified firm can lead to
improved allocative efficiency. This is created by
top management, who hold better information than
investors, allocating fluids to their most profitable
use. On the other hand there are several arguments
which suggest that the growth of the diversified
firm has the potential to create reduced competition
and efficiency. It was noted earlier that there has
been a high proportion of narrow spectrum
diversification.
At least one possible interpretation of this finding is
that the diversification that has led to relatively rapid
rates of corporate growth (or has accompanied it) has
not in general been to markets where the entering
firm is a new and potentially competitive force.
Rather, that ‘diversification’ has been to markets that
are related to - and potentially if not actively
competitive with - those in which the entering firm
will frequently share what ever market power
already exists. This kind of diversification is only
one small step removed from the consolidation of
market power through horizontal acquisition (Berry ,
pp. 74-5).
Furthermore, the internalizing of capital markets
has led to the removal of information and decision-
making from the investor and led to a concentration
of economic power. ‘This means that the
diversified, divisionalized firm is increasingly
becoming the arbiter of intersectional shifts in
funds’ (Rumelt , p. 155). Another focus of concern
has been the potential for predatory pricing
behaviour in which the diversified firm uses cross-
subsidization between divisions to eliminate, or
discipline, more specialized rivals and
2994
so achieve higher long-run profits. A further pos-
sibility is reciprocal purchasing agreements when a
firm is significant both as a seller to and buyer from
another firm. It is argued that such practices are
more likely to be found amongst large, diversified
firms, but there is little evidence for the widespread
existence of either predatory pricing, or reciprocal
purchasing behaviour. Finally there is the spheres of
influence hypothesis which recognizes the
pervasive influence of large, diversified firms in
almost all markets. Conglomerates might recognize
that aggressive behaviour against another
conglomerate in one market would have adverse
consequences in other markets. It is possible that a
symmetiy of market power might emerges which
would blunt competition. The answer to many of
the empirical issues concerning diversification are
as yet unresolved. This is in part due to a lack of
sufficient research, but also in part due to the fact
that the process of diversification is continuing.
When, and if, a more stable period emerges the
uncompetitive consequences outlines above may
become more apparent.
Bibliography
Berry, C.H. 1975. Corporate growth and diversification.
Princeton: Princeton University Press.
Chandler, A.D. 1963. Strategy and structure: Chapters in the
history of the industrial enterprise. Cambridge, MA: MIT
Press.
Channon, D.F. 1973. The strategy and structure of British
enterprise. London: Macmillan.
Gort, M. 1962. Diversification and integration in American
industiy. Princeton: Princeton University Press. Jacquenrin,
A., and C.H. Berry. 1979. Entropy measure of diversification
and corporate growth. Journal of Industrial Economics 27(4):
359 369.
Marris, R.L. 1964. The economic theory of managerial
capitalism. London: Macmillan.
Penrose, E.T. 1959. The theory of the growth of the /inn.
Oxford: Basil Blackwell.
Robinson, E.A.G. 1958. The structure of competitive
industry. Rev. ed. Cambridge: Cambridge University
Press.
Runrelt, R.P. 1974. Strategy, structure and economic per-
formance. Cambridge, MA: Harvard University Press.
Scherer, F.M. 198(1. Industrial market structure and eco-
nomic performance. Chicago: Rand McNally.
Utton, M.A. 1979. Diversification and competition, the
National Institute of Economic and Social Research
Divided Populations and Stochastic Models
Occasional Paper No. 31. London: Cambridge University
Press.
Williamson, O.E. 1970. Corporate control and business
behavior: An enquiry into the e f f e c t s of organisation
form on enterprise behavior. Englewood Cliffs: Prentice-
Hall.
Williamson, O.E. 1975. Markets and hierarchies: Analysis
and anti-trust implications, a study in the economics of
internal organization. New York: Free Press.
Divided Populations and Stochastic
Models
D. G. Champernowne
Introduction
The title of this entiy requires some explanation.
1 use the term ‘stochastic models’ to distinguish
those theoretical models which include one or more
stochastic variables from 'determinist models’
which do not. I shall confine attention to some
stochastic models which are obtained by introducing
into a determinist model a single stochastic variable
(which can be multivariate, but will in illustrative
examples be univariate). I shall use the term
‘generating system’ to mean a determinist model in
which from an initial state of the system an
unending sequence of successive states of the
system can be exactly predicted by means of a set of
rules such as lagged equations. It is convenient to
distinguish generating systems from stochastic
models rather than extend the former class to
include some or all of the latter. The important
feature of stochastic models is that they can make
allowance for wide margins of uncertainty and
ignorance.
By a ‘divided population’ I shall generally mean
a frequency distribution most of which is closely
clustered around two or possibly more peaks, but
fairly empty elsewhere: an extreme case would be
that where the peaks were completely separated by
an unoccupied stretch.
Divided Populations and Stochastic Models 2995

However, the term ‘divided population’ can occa- growth-rates. But it is also possible to frame fairly
sionally be extended to refer to a society which is simple rules which lead to oscillations which persist
divided into groups with contrasted living condi- at a constant amplitude. Often these will be smooth
tions, prospects and aims. and sinusoidal, but there is another possibility which
The term ‘crisis’ refers to an unstable situation is the one relevant to crises, where there are periodic
where a small disturbance could tip the scales jumps from one smooth steady path (which we
between the prospects of two widely different might call boom) to another smooth steady path
eventual outcomes. In a determinist model the (which we might call slump) alternately to and fro
representation of such a crisis would be a point of indefinitely.
unstable equilibrium, and in a stochastic model A convenient tool for the representation of such
based on that determinist model, one would still systems when there are sufficiently few equations
regard the point of instability as indicating crises involved is the phase diagram. If we are dealing
facing that part of the population found close to it, with difference equations of the kind just described,
by ‘crises’ meaning here the crises that chance the axes ofthe diagram could measure the values of
might play a predominant part in determining their one important variable along the horizontal axis as
future prospects. independent variable and the change of that same
As a preliminary to the main discussion, it will variable over the next time- unit along the vertical
be helpful to consider some standard tools for use axis as dependent variable. In such diagrams the
with determinist models involving divided curve relating the change of the variable as a
populations and crises. function of the value itself will reveal points of
equilibrium by its intersections with the horizontal
axis: however, where the curve cuts the axis from
Some Standard Methods for the Study below on the left, the equilibrium will evidently be
of Unstable Situations unstable, and we shall call such equilibrium points
crisis points. Figure is a phase diagram applicable to
A standard method of constructing a model of the the determinist model described below in section “
response of an economic system to the passage of
time, or to possible changes ofpolicy or of outside In Fig. the horizontal axis is cut by the graph
influences, is to set up a generating system giving a ITJKLSB in three points J, K and L denoting equi-
set of initial conditions containing the present and librium levels of the index of prosperity, but the
recent values of a set of economic variables and point K is a ‘crisis point’ indicating an unstable
policy parameters, together with a set of rules for equilibrium value. The arrows following the paths
calculating the set of the same variables one time- starting from A and from near K illustrate how,
unit later and repeating this operation successively given the level of the index in an initial period, the
for any required number of time-units. Such rules chart may be used to predict its values in later
would normally take the form of a number of periods assuming the rules of the model to be
equations giving the values of each variable as obeyed. For example, to follow the changes from
functions of the values of other variables, mainly at the initial value of 1000 at A on the horizontal axis,
earlier dates, taking account of the present values measure horizontally the same (negative) distance
assumed for any policy parameters. We may AA\ as the vertical distance of the graph from A.
confine attention to very simple examples of such Having marked A i, for the value after one unit of
models. time, repeat the operation from A,, to mark in A2
It is quite usual to find in simple models that and so continue as illustrated in Fig. . It is apparent
given the initial information, the application ofthe that the series of such values must converge to the
system of rules with fixed policy parameters will value marked by S, and similarly that starting from
generate a sequence of sets of values of the vari- C, near K on the right, when the distances concerned
ables which tend to a long-run equilibrium set, apart will now be positive, (to the
possibly from one or more constant common
2996
right or upwards), it should again be clear that the
series obtained must converge to the same stable
equilibriiun value S. Finally, starting from any point
to the left of K, the usual procedure must result in a
series arriving at the other stable equilibrium point
7 : (this is so because the gradient of the line ITJ is
— 1 in this example, which entails that as soon as
the path hits ITJ it leads directly to T). This
illustrates why the equilibrium at points, such as K,
where the axis is cut by the curve from below on the
left are unstable, while the points such as Tand S
mark stable equilibrium values.
Fig. will be used again in section “
” to illustrate
the numerical example there, which involves
equations ( ) to ( ): Table in that section provides
the first few values of the sequences that would be
obtained by applying the rules, starting from values
1000, 100 and 70 respectively.
An early example of a determinist economic
model involving oscillation between two points of
stable equilibrium across a gap containing a crisis
point of unstable equilibrium, due to the interven-
tions of a disturbing force moving the phase- curve,
was the model of the trade cycle published
Divided Populations and Stochastic Models, Fig. 1 Phase diagram with crisis point K
Divided Populations and Stochastic Models
by Kaldor in the March 1940 issue of the Economic
Journal. This contained a diagram closely related to
a phase diagram of the elementary type shown here
in Fig. , and which relied on the property that the
curve itself moved upwards or downwards,
depending on whether the currently relevant point
representing equilibrium was on the right or left of
the diagram.
Figure is a transposition of Kaldor's diagram
into a phase diagram of the type outlined above.
Three positions of the curve are shown marked 0, +
and *. Initially the relevant point of intersection is B
a stable prosperous stable equilibrium point: K and
S mark the currently irrelevant crisis and slump
equilibrium points on this curve. During the boom
the curve moves down to the position + + + at
which B and K meet at the point K+ of tangency
and the curve loses contact with the horizontal
equilibrium axis so that the relevant equlibrium
shifts rapidly to D*, the slump stable equilibrium,
and now the curve moves upwards past position 0 to
position +, at which S and K coalesce at the new
point K* of tangency, and again the curve loses
contact with the equilibrium line, so that now the
relevant
Change of
prosperity index
 Divided Populations and Stochastic Models 2997
Generation number 1 2 3 4 5 6 7 8 9 10
Divided Populations and Stochastic Models, Table 1 Three lines of bequests over ten generations
Level of bequest
Line 1 70.0 61.3 45.1 24.0 24.0 24.0 24.0 24.0 24.0 24.0
Line 2 100 117 150 210 324 443 518 565 593 611
Line 3 1000 865 780 728 695 675 662 654 649 646

Divided Populations
and Stochastic Models,
Fig. 2 Phase diagram:
three curve positions
-----end of
stump
-----Norm boom
or slump
----- end of
boom

-20 0 20
Activity level X

stable equilibrium moves rapidly to the right to B*,
the boom stable equilibrium. Then the curve moves
downwards again through the position 000 and the
story is repeated again and again with alternative
stable equilibria B and S in boom and in slump.
Since World War II a number of models have
been based on such non-linear differential or dif-
ference equations to produce fairly regular
switching between temporarily stable situations of
slump and boom. An early and particularly neat
example was provided by R. M. Goodwin in a paper
delivered in June 1955 to a meeting of the
International Economic Society in Oxford. An
account of that and other such early models will be
found in chapter 8 of Mathematical Economics by
R. G. D. Allen (London: Macmillan; New York: St
Martin's Press, 1956).
The model in section “
” with its crisis point has
much in common with those of Kaldor and
Goodwin and later writers, but in section “ Rules for
a CtochastR 'U see) ? Z::ves s ” we shall develop it
in a different direction by introducing stochastic
disturbances so that it
may be used for modelling the development of
bimodal distributions. The point which that model is
intended to illustrate is that a few simple ingredients
which may underlie a number of complex situations
in which bimodal frequency distributions may alone
be sufficient to produce bimodality, without any of
the many further influences which may also be
possible explanations of it. It is quite plain that such
a model is not in fact a complete explanation, but it
may be helpful as illustrating a method of taking a
first step in a variety of investigations of situations
where divided populations are observed.
The simple ingredients alluded to above are as
follows:
(T) A set of largely unidentifiable and unexpected
disturbances to each member value of the
population whose distribution is being gener-
ated. This may well increase the dispersion.
(2) A set of influences encouraging the growth of
large member values and the declines of small
member values.
(3) Opposing these influences, 1 and 2: specific
measures taken to discourage further growth
2998 Divided Populations and Stochastic Models

of very large member values and reverse the

fall of very small ones. [B, - T A X . ( B, - W E X ) - ( C - E ) / R ]
+{C — E)/R
These three ingredients will often be sufficient (2)
to produce a bimodal distribution. We have not Both ( ) and ( ) operate if B, = WEX.
included in (3) the many influences that there may The meanings of the symbols T, E, C, R, TAX
be operating to diminish or reverse the effects of and WEX are as follows: T = length of generation in
ingredient (2) at intermediate levels: where this years: we take T 25 for examples, E = level of
omitted set of influences is strong, a unimodal earnings per annum: we take E= 10 for examples, C
distribution is likely to be found. = consumption expenditure per annum: C = 12 for
examples, R = interest rate for dividends per annum:
R = 2.5% for examples, P = level up to which
Rules for a Model Generating Lines
bequests less than it are subsidized, P = 50 for
of Bequests
examples, TAX = rate of tax of bequests starting at
exemption level WEX for tax on bequests: TAX =
In this section we consider a population consisting
2/3 or 3/5; WEX = 250 or 400, in examples.
of family lines within which bequests are passed
The four rules which lead to the equations ( )
down generation after generation according to a
and ( ) are:
mechanical system of rules governing inheritance,
earnings, consumption, taxation, subsidies and Rule 1. So long as any of a bequest remains it
dividends, and which lead all family lines eventu- constitutes a fund attracting interest at the rate R
ally to ruin or to considerable wealth. per annum and provides a source from which
The same rules apply to each and every line of the excess expenditure (C E) can be maintained.
bequests, which differ only in the level of the initial Rule 2. If the whole of a bequest gets used up before
bequest. We may denote the level of the initial the end of a generation, consumption is cut from C
bequest in a representative line as B0 and the level to E, (debt is ruled out) and in this case the bequest
of the bequest in that line t generations later as Bt. (before subsidy) must be zero. Rule 3. Every
We shall set out in the next paragraph a set of rules bequest consists of the accumulated fund at
which entail that the following bequest, Bt+1 in the retirement (before tax or subsidy), which fund may
line is always obtainable from one or two linear be zero.
equations from the current bequest Bt. For reference Rule 4. The tax or subsidy on the bequest Bt is
these equations ( ), ( ), are set out below and applied at the moment of payment to the heir, so
followed by an explanation of the notation and by a that the heir receives W, out of B, < P where
description of the rules governing the accumulation Wt = Bt— TAX. (B, - WEX) i f 5 , > WEX,
of wealth for bequests and implying these Wt = P i f Bt < P a n d W, = Bt
equations. otherwise.
W h e n B, < WEX,
If we denote the value of the fund after u years
Bt+i = eRT.[Bt - (C - E) / R ] + { C - E ) / R i f by F(u), the derivation of equations ( ) and ( )
follows directly from the rules by solving the
Bt e x c e e d s P b u t o t h e r w i s e differential equation dF/Fu = R ■ F(u) C + E by
Bl+i = e R T . [ P- ( C - E) / ( R ) } standard methods to obtain F(T) given F(0) = W,
+ (C — E ) / R which may be found by Rule 4.

(1) The equations ( ), ( ) and a knowledge of the

values of the parameters T, E, C, R etc. and of the
or if this is <0, B t+1 0. initial bequest B0 of any line now enable us to
W h e n B t > WEX,
Divided Populations and Stochastic Models
derive the whole line of bequests 5 0, 51; B2, . . . as
far as we wish and to find the limiting value in
long-run equilibrium, by repeated application of the
relevant equations.
The operation of the model can be illustrated by
a phase diagram if we select values for the
parameters.
Putting
R = 2.5%,T = 2 5, C = 12 , E = 10 ,
TAX = 2/3, WEX = 250, P = 50
we obtain, when Bt < 250,
5,+1 — B, = 0.868246(5, - 80), if B,
> 40 (4)
but if Bt < 40;
5,+1 -B, = 23.95 -5,. (5)
But when 5t > 250,
5,+1 - 5, = -0.3772515, + 241.91465 (6)
and can derive the phase diagram, Fig. in which the
curve cuts the horizontal axis in the three
equilibrium points at T, K and S at which 5, = 23.95,
80 and 641.26. The values of the turning points J
and L are those of P and WEX, 50 and 250 (see
section “
”)•
Table covers ten generations and gives the
values of the bequests in three lines starting at 70,
100 and 1000.
Statistical Methods for Studying
Distributions with Many Peaks
Twin-peaked distributions often arise in situations
where there are three equilibrium points of which
two are stable, but the third is unstable, and lies
between them. In such unstable situations the fact
that an initial distribution contains individuals on
both sides of the unstable crisis point will ensure
that the population will eventually be divided into
2999
two groups at or close to the two stable equilibrium
points.
The same mathematical device that underlies the
crisis models generating alternative progressions to
the two stable equilibrium points, or in some
models regular switching from one to the other
across an unstable one, may be adapted to represent
situations which produce a frequency distribution
consisting of two peaked distributions each centred
on stable equilibrium points on either side of an
unstable one. The adaptation may consist of the
introduction of rules for moving the curve that
indicates the equilibrium points, as in Kaldor's
models, or by the introduction of rules disturbing
the point indicating the current state of affairs off
that curve: that is the line we shall investigate.
The whole frequency distribution may either
continue strictly positive across the neighbourhood
of the unstable point between the two peaks, or be
split into two entirely separate distributions, with
the unstable point left in the gap between. In the
class of models which will be discussed in section “
” of the entry, the
split version can only emerge if the rules governing
the stochastic disturbances to the movements of the
points (representing the individual values whose
frequency distribution is being generated) do not
enable individuals to arrive at or cross the unstable
equilibrium point. This is a very stringent condition,
but it represents an intermediate case between the
stochastic model generating the unbroken two-
peaked equilibrium distribution and the cruder
determinist models generating a long-term
equilibrium with all individual points concentrated
at the two stable equilibrium points. More elaborate
determinist models with lagged variables may lead
to undamped regular oscillations about a single
equilibrium point, but these will not be further
discussed in this enfiy.
In the past, economists were largely concerned
with the study of equilibrium positions towards
which market competition and other social and
economic forces would drive the economic
individuals and conglomerations involved. The
3000
particular concerns of the statisticians and econo-
metricians were more often with the movements of
those equilibrium points and with the dispersion of
the individuals or groups around these points. In the
simple cases where there was just a single
equilibrium they might for example study the
shapes of the frequency distributions of the one or
more coordinates of the point, and suggest and test
various theories to explain how such shapes could
arise, as well as what caused the movement of the
equilibrium point itself. Thus there have been
theories to account for and predict the age-
distributions of the populations of various
territories, the size-distributions of their cities and
the distribution of the shares of votes cast for a
particular party in the various constituencies, and
again the distribution of income, wealth and other
measures of prosperity, both between individuals
and between various groups of persons.
However, if stochastic disturbances can interfere
with the equilibrating forces or even shift the three
equilibrium points, there may be preserved a
considerable spread of distribution around each
stable equilibrium point, and if the stochastic
disturbances are strong, there may still be
movement between the two groups across the
unstable equilibrium point. In the former case we
should expect two separated equilibrium
distributions, whose relative sizes would depend on
the nature of the initial distribution, but in the latter
case a single continuous but probably bimodal
equilibrium distribution whose shape could well be
independent of the initial distribution.
In those cases where a considerable valley
between the two peaks of the long-term equilibrium
distribution is preserved, the stochastic mechanism
is quite different from the simple determinist
explanation for such a bimodal equilibrium
distribution: this determinist explanation is simply
that two quite distinct populations have been
juxtaposed and counted together as one population.
For example, if a wealthy island were to annexe an
impoverished island with roughly the same
population and then compiled wealth-or income-
distribution figures for the two combined, one
might expect a fairly stable bimodal wealth-or
income-distribution. However, with good
Divided Populations and Stochastic Models
transport between the two islands one might expect
eventually that the later generations sprung from the
impoverished island would acquire gradually some
of the cultural and other advantages of the
descendants of the wealthy islanders and, vice
versa, some of the descendants of the wealthy
islanders and, would be impoverished as a result of
the competition of the more gifted immigrants from
the other island. Thus in the long run the stochastic
intermingling of the two races might make the
stochastic model more relevant than the determinist
analysis of the equilibrium distribution to be
expected.
In section “
” we shall explain how to
introduce a stochastic variable into our determinist
model of lines of bequests so as to change it into a
stochastic model of the distribution of bequests in
successive generations and in the following section
will provide some numerical examples to suggest
some questions which such models might be helpful
in answering if they were suitably elaborated. These
questions will be related to situations featuring an
apparent contrast between two overlapping groups,
‘poor’ and ‘rich’, where the ‘persons’ to whom the
distributions refer may be individuals or households
or localities or larger groups such as whole
economies.
Easy Rules for a Stochastic Model of a
Divided Population
In our stochastic model we shall consider the
distribution of bequests in each generation over a
series of value-ranges of equal proportionate extent
g. We shall suppose the top of range 0 to be P, and
we shall number the ranges so that for each integer
i, positive or negative, the top of range i is Pgl. We
assume that initially all bequests are at the centres
of the ranges, and we impose rules to ensure that the
same is true in every ensuing generation. This
simplification makes rather narrow ranges desirable
so as to avoid introducing considerable inaccuracy,
but in illustrative examples we shall have to use
wide intervals with g = 10.2 = 1.585 or 10.1 = 1.26
so as to be able to set out the results in the space
available.
Divided Populations and Stochastic Models
The stochastic model differs from the deter-
ministic one of section “
”, in modifying the value
ofB, ,, there calculated from_Bf. Denote that value
by Bt+ i(i) when B, is at the centre of range i, then in
the stochastic model we multiply it by a stochastic
multiplier m„ which when N 2 scatters the bequests
of B, ](i) to the centres of 4 (or more generally, of N
- 2) consecutive ranges, in proportions that leave
their arithmetic mean equal to
In section “
” the rules for choosing the
ranges and the proportions of bequests moved to
each of them will be set out but nonspecialists may
prefer to skip that section and be content with Figs,
and below, which illustrate typical effects of the
multiplier with (i) g 1.585, N 2 and (ii) g = 1.26,
N= 7.
We may then work out for each range i from 0
upwards, the following bequest Bt+1(i) by the
formulae ( )and( )ofthedeterminist model and apply
the stochastic multiplier to the bequests in lines
from each range i, so as to split them into sets
going, when N = 2 to the 4 (or more generally, N +
2) appropriate consecutive ranges. By using the
information for every range containing at least one
bequest where we round off to the nearest integer,
assuming a total number of one million bequests in
each generation, it is then a matter of arithmetic, to
find the size-distribution of bequests in generation /
+ 1 from that of bequests in ranges with non-
negative i in generation t.
Divided Populations
and Stochastic Models,
Fig. 3 S t o c h a s t i c
disturbance about 400
Kssa N = 2 g =1.58 post mult Value of bequests
3001
We sti 11 have to explain how to handle
bequests in the ranges with ; < 1, namely the ranges
below the level P. It is again assumed that all
bequests in such ranges are subsidized up to the
level P. Indeed, some such egalitarian measure as
this is needed if we are to avoid all bequest lines
eventually becoming permanently zero or in a range
well below P, except possibly for a wealthy set all
considerably above the crisis, level (C — E)/R. So
in the calculations we merely have to lump all the
bequests in ranges with i less than 1 into range 0.
This need not prevent there being bequests before
subsidy in each generation in other ranges below P,
and it is the distribution of ranges before subsidy
that we shall calculate in examples and which are
relevant to the distributions of wealth and dividends
which are all available towards the retiring age.
We shall give a very few numerical examples of
such distributions in section “
” to
which those uninterested in the details of the mles
for the stochastic multiplier are advised to skip.
Those rules will now be outlined in section “
Rules Setting the Probabilities of the N +
2 Values of mi
These mles will be illustrated by the case N = 2;
where normally m, may take 4 values g7' g*+ \g'+2,
g'+i, where j is an integer. There are, however, two
simple special cases where only three consecutive
3002
Divided Populations (Thousands)
and Stochastic Models,
Fig. 4 D i s t u r b a n c e a b o u t
400 (N = 7; g = 1.26)
integers are taken, the fourth value having no
bequests dispersed to it. The rules ensure that these
two cases give probabilities 0.25, 0.5, 0.25, 0 and 0,
0.25, 0.5, 0.25; the three non-zero terms are those
of the binomial expansion (0.5 + 0.5) 2. Similarly,
where N is any positive integer there are two
special cases where N+ 1 ranges only may be
occupied with probabilities given by the A ,r - 1
terms of the binomial expansion (0.5 + 0.5) w
Returning to the special case N = 2, the rales
further provide that the value of j and the pro-
portions of the bequests should be so chosen that
the four proportions are a weighted average with
weights 1 p and p (0 < p < 1) of the two special
cases with three terms each, and that the arithmetic
mean of the bequests should equal the value B,-i(i)
obtained in the determinist model. This entails that
the four proportions should be the following: (1
-p)/4, (2 p)t4, (1 + p)/4 andp/4. The arithmetic
mean of the bequests must then be (l + pg). Pg1 (1 +
g)’/4 so that our rales require
2 with this set-up will depend on how widely the
(1 + p(g - 1)) • Pg>( 1 + g) = 4Bt+1 (0
where the right-hand side is known. This uniquely
determines j and p and they may easily be derived.
In the general case where N is any positive
integer the main modifications are that the binomial
expansions in the special cases are now (0.5 + 0.5) A'
and that in equation ( ) and the preceding line, 4
must be replaced by 2 V.
Divided Populations and Stochastic Models
Model Generating 2-Peaked- Distribution:
Illustrative Cases
In this section we shall illustrate the kinds of two-
peaked distributions that are generated by such
simplified stochastic models and the ways one
might use them, by a few numerical exercises
involving an imaginary set of a million lines of
bequests. We shall mainly use arithmetic and dia-
grams for the exposition.
Let us start with the standard values for the
parameters given in section “
” equation ( ) as
R = 2.5%, T = 25, C = 12, E = 10,
TAX = 2/3 and WEX = 250
and in section “
” as
p = 50, g = 10.2 (8)
The long-run equilibrium distribution obtained
(7)
stochastic multiplier disperses the bequests from
each range in a single generation, and this is set by
the parameter N. Figures , and show that widening
the dispersion in this example through the first four
values, N - 0, 1, 2 and 3 already exhibits a wide
variety of types of solution. Case N 0. Two
separated distributions: in ranges 2
Divided Populations and Stochastic Models 3003

Value of bequests (log-scale)

Divided Populations and Stochastic Models, Fig. 5 E q u i l i b r i u m b e q u e s t d i s t r i b u t i o n s

Divided Populations
and Stochastic Models,
Fig. 6 E q u i l i b r i u m
bequest distributions

Divided Populations
and Stochastic Models,
Fig. 7 E q u i l i b r i u m P a r e t o
Curve (non-cumulative)

Level of bequest (log-scale)

3004
Divided Populations
and Stochastic Models,
Fig. 8 E f f e c t o f h i g h e r
tax-exempt limit
Level of bequest (log-scale)
and — 1; and in ranges 10 and 11. The proportion
of bequests in the two distributions will equal the
initial proportions that were in ranges up to and
including 1 and in ranges 2 and above. Cases N = 2
and N = 3. All bequests are in ranges immediately
below P; 4 of them when N = 2 and 5 when N = 3.
Cases N = 4 and over. All bequests are in a single
distribution extending over many ranges from well
below P up to well above WEXthe tax exemption
limit. These are the most interesting cases. Some
have a pair of peaks separated by a valley, but those
with N taking higher values have only a single peak.
Figure illustrates the case N = 0 where we have
assumed that half the bequests in the initial
generation were in ranges 3, 4, 5 etc. ... and half in
the ranges 0, 1,2 etc. The result, due to the
minimum value of A, is a very divided distribution
little different from the complete division that
would be found in the determinist model: the case
with N L, not shown, is less stark, allowing
a spread over five ranges in the upper peak and over
three in the lower peak. Figure shows the unusual
cases where in the long term there are no bequests
in any range above P. With the particular values we
took for the other parameters this unusual feature
only occurs when N = 2 and N 3. f igure
compares the typical bimodal
form when N = 4 with the typical unimodal form
when N = 18. The logarithmic scale used may given
the deceptive impression that when N 4 the valley
between the peaks is not deep
Divided Populations and Stochastic Models
and therefore easily crossed, but a more careful
inspection will reveal that it is very deep, since the
valley floor indicates a range with roughly 10,000
bequests, whereas even the lower peak indicates
one containing roughly 100,000 bequests.
Figures and are mainly concerned with a
potentially instructive use of stochastic models for
investigating the effects of altering one or more of
the policy parameters. They do this for two
examples of reflationary fiscal policies: raising
WEX the tax-emption level from 250 to 400 (Fig. )
and altering TAX from 3/5 and 2/3 to 50 per cent
(Fig. ). The effects of the higher exemption level are
to deepen and widen the valley and shift the tail to
the right along with the level of the exemption limit,
without affecting its Pareto slope. The effect of the
tax reduction is mainly to lessen the Pareto slope of
the tail. Each measure greatly increases the total
number of bequests above 5000; the tax reduction
by four or fivefold and the higher exemption level
by more than tenfold. This is perhaps the right
moment to reiterate the warning that such examples
are not meant to be more than indications of
elementary methods for testing hunches of what are
the probable logical effects of such changes given
any set of artificial rales being mechanically
obeyed.
All our discussion has been concerned with
equilibrium distributions. However, as in so many
branches of economic theory, knowledge of the
eventual equilibrium corresponding to the
Divided Populations and Stochastic Models
Divided Populations
and Stochastic Models,
Fig. 9 E f f e c t o f l o w e r
tax on bequests
Divided Populations and
Stochastic Models, Fig.
10 S h o r t - t e r m m o v e s
to equilibrium
1E0 1E1
Level of bequest (log-scale)
present state of the economy and current policy
decisions is virtually useless unless one knows how
rapidly that equilibrium will be approached and
particularly what will happen during the reasonably
near future.
This also can be illustrated with our elementary
example. It is to be expected that with low values of
N, approach to equilibrium may be very slow and
that is only too well confirmed by a wide variety of
examples not further reported here. It is more
interesting to give the stochastic disturbances
considerable scattering influence by choosing quite
a high value for N and then following the pace of
approach towards equilibrium from an initial
distribution of bequests chosen so
3005
Eventual
change
Change in
250 years
Change in
125 years
1E2 1E3 1E4
as to differ considerably from that equilibrium
distribution.
But in Fig. we have taken our standard example,
with N = 7, and shown for each range the difference
between the equilibrium number of ranges that
would result from altering TAX from 3/5 to 2/3: we
also show by the two intermediate curves how much
of the approach to the new equilibrium would in
each range be achieved after five and after ten
generations of 25 years each. It will be seen how far
from completed the transition is even after the 250
years.
Space forbids showing further examples of the
short-run effect of altering policy or other param-
eters in such models, by methods similar to those
3006
used in studying long-run effects: naturally such
short-term enquiry can be more important for Dividend Policy
obtaining conclusions remotely relevant to the real
world, yet the theoretical approach to such David J. Denis and John J. McConnell
investigations need differ little from that to the
long-term ones discussed above.
Concluding Observations
In the later sections of this entry we have provided
an illustration of how, without introducing any of
the more detailed causes of divided, i.c. of bimodal,
frequency distributions, one can obtain a skeletal
model of the development of such distributions by
merely including the dispersive elements: (a) an
ability' and willingness on the part of the richer, to
save a higher proportion of then- income than on
the part of the poorer, and (b) stochastic
disturbances; both (a) and (b) tending to increase
inequality; and the egalitarian elements: (c) to
curtail (a) and (b) on the part of the very rich; and
(d) subsidies to set a limit on the poverty of the very
poor. We have hinted how models including at least
these four basic elements, and thereby generating
biomodal distributions, can provide some insight
into possible long- run and short-run effects of
altering policy parameters: in particular we have
argued that considerations of long-run equilibrium
can be very poor guides to short-run effects.
The next step is, naturally, to introduce into the
model the more obvious and significant other
causes acting to modify distributions of wealth,
income, health, nourishment and other measures of
well-being. That step is far too long for inclusion in
an entry of this nature. Moreover, since although
situations of the critical and divisive kinds, on
which such research aims to throw some light
become more frequent every year, statistics of these
phenomena are patchy and unre liable. The
development of stochastic methods to make
allowance for the unreliability and incompleteness
of information is by no means a minor step in such
enquiries. It was this belief that prompted the
submission of this methodological entry.
Dividend Policy
Abstract
Dividends represent the primary means by
which invested capital is returned to common
stockholders. In this article wc summarize the
development of academic thinking on dividend
policy, focusing on three primary perspectives:
(a) the effect of dividend policy on common
stock value and firm performance, (b) the
determinants of dividend policy, and (c) mac-
roeconomic trends in the propensity of firms to
pay dividends.
Keywords
Agency costs; Asymmetric information; Capital
gains; Capital gains taxation; Dividend change;
Dividend policy; Dividend taxation; Dividends;
Manager-shareholder conflict; Modigliani-
Miller theorem; Regular vs special dividends;
Stock repurchases; Stockholders
JEL Classifications
G35
There are two major ways in which a firm can
distribute cash to its common stockholders. The
firm can either declare a cash dividend which it
pays to all its common stockholders or it can
repurchase shares. Stock repurchases may take the
form of registered tender offers, open market
purchases, or negotiated repurchases from a large
shareholder. In a share repurchase, shareholders
may choose not to participate. In contrast, divi-
dends are direct cash payments to shareholders and
are distributed on a pro rata basis to all
shareholders.
Most firms pay cash dividends on a quarterly
basis. The dividend is declared by the firm’s board
of directors on a date known as the ‘announcement
Dividend Policy
date’. The board’s announcement states that a cash
payment will be made to stockholders who are
registered owners on a given ‘record date.’ The
dividend checks are mailed to stockholders on the
‘payment date,’ which is usually about two weeks
after the record date. Stock exchange mles gener-
ally dictate that the stock is bought or sold with the
dividend until the ‘ex-dividend date’, which is a
few business days before the record date. After the
ex-dividend date, the stock is bought and sold
without the dividend.
Dividends may be either labelled or unlabelled.
Most dividends are not given labels by manage-
ment. Unlabelled dividends are commonly referred
to as ‘regular dividends’. When managers label a
dividend, the most common label is ‘extra’.
A Historical Perspective
Prior to 1961, academic treatments of dividends
were primarily descriptive, as, for example, in
Dewing ( ). To the extent that economists con
sidered corporate dividend policy, the commonly
held view was that investors preferred high divi-
dend payouts to low payouts (see, for example,
Graham and Dodd ). The only question was how
much value was attached to dividends relative to
capital gains in valuing a security (Gordon ). This
view was concisely summarized with the saying
that a dividend in the hand is worth two (or some
multiple) of those in the bush. The only question
was: what is the multiple?
In 1961, scientific inquiry into the motives and
consequences of corporate dividend policy shifted
dramatically with the publication of a classic paper
by Miller and Modigliani. Perhaps the most
significant contribution of the Miller and
Modigliani paper was to spell out in careful detail
the assumptions under which their analysis was to
be conducted. The most important of these include
the assumptions that the firm’s investment policy is
fixed and known by investors, that there are no
taxes on dividends or capital gains, that individuals
can costlessly buy and sell securities, that all
investors have the same information, and that
investors have the same information as the
3007
managers of the firm. With this set of assumptions,
Miller and Modigliani demonstrate that a firm’s
stockholders are indifferent among the set of
feasible dividend policies. That is, the value of the
firm is independent of the dividend policy adopted
by management.
Because investment policy is fixed in the Miller-
Modigliani set-up, all feasible dividend policies
involve the distribution of the full present value of
the firm’s free cash flow (that is, cash flow in
excess of that required for investment) and are,
therefore, equally valuable. If internally generated
funds exceed required investment, the excess must
be paid out as a dividend so as to hold investment
constant. If internally generated funds are
insufficient to fund the fixed level of investment,
new shares must be sold. It is also possible for
managers to finance a higher dividend with the sale
of new shares.
The key insight from the Miller-Modigliani
analysis is that investors will be indifferent among
the feasible dividend choices because they can
costlessly create their own dividend stream by
buying and selling shares. If investors demand
higher dividends than the amount paid by the firm,
they can sell shares and consume the proceeds,
leaving themselves in the same position as if the
firm had paid a dividend. Alternatively, if
shareholders prefer to reinvest rather than to con-
sume, they can choose to purchase new shares with
any dividends paid. In this instance, shareholders
would be in the same position that that they would
have been in had no dividends been paid. Thus,
regardless of corporate dividend policy, investors
can costlessly create their own dividend position.
For this reason, stockholders are indifferent to
corporate dividend policy, and, as a consequence,
the value of the firm is independent of its dividend
policy.
After a brief flurry of debate, the Miller-Mo-
digliani irrelevance proposition was essentially
universally accepted as correct under their set of
assumptions. There nevertheless remained an
underlying notion that dividend policy must ‘mat-
ter’ given that managers and security analysts spend
time worrying about it. If so, and if the Miller-
Modigliani proposition is accepted, it
3008
must be due to violation of one or more of the
Miller-Modigliani assumptions in the real world.
Since the early 1960s, the dividend debate has
been lively and interesting. Economists have
analysed theoretically whether the relaxation of the
various Miller-Modigliani assumptions alters their
irrelevance proposition. In addition, economists
have analysed the data from several perspectives.
First, they have undertaken an array of analyses to
determine the effect, if any, of dividend policy on
stock value and firm performance. Second, they
have sought to identify the characteristics associated
with dividend payments (or the lack thereof) by
individual firms. Third, they have attempted to
characterize macroeconomic trends in the level and
propensity of firms to pay dividends, and in the
form of the payout. Our discussion of these issues
focuses primarily (though not exclusively) on
studies of US firms since these are the studies most
accessible to us.
Relaxing the Miller-Modigliani
Assumptions
Taxes
Perhaps the obvious starting point for an inves-
tigation into the effect of relaxing the Miller-
Modigliani assumptions is to introduce taxes. In the
United States, dividend payments by a corporation
do not affect that firm’s taxes. However, at least
historically, dividends have been taxed at a higher
rate than capital gains at the personal level. Thus,
superficially, the US tax code appears to favour a
low dividend payout policy, with payouts occurring
primarily through share repurchases.
Under the assumption that dividends and capital
gains are taxed differentially, Brennan ( ) capital gains affects both ex-dividend day stock
derives a model of stock valuation in which stocks
with high payouts have higher required before-tax
returns than stocks with low payouts. As a coun-
terpoint to this proposition, Miller and Scholes (
) argue that under the US tax code there
exist sufficient loopholes so that investors can
shelter dividend income from taxation, thereby
driving the effective tax rate on dividends to zero.
Early studies of the association between
Dividend Policy
stock returns and dividend yield (for example,
Black and Scholes ; Litzenberger and Ramaswamy ;
Miller and Scholes ) yielded mixed results using
different definitions of dividend yield. Subsequent
studies indicated that the correlation between
dividend yield and stock returns (if any) appeared to
be due to omitted risk factors that were correlated
with dividend yield. For example, Chen et al. ( )
report that
dividend yield and risk measures are cross-
sectionally correlated. Similarly, Fama and French (
) show that, when a three-factor model for
expected returns is used, there is no significant
relation between dividend yields and stock returns.
Other studies have analysed the potential effects
of the differential taxation of dividends and capital
gains by studying the behaviour of stock prices and
trading volume around ex-dividend days. The logic
of these studies is that, in order for investors to be
indifferent between selling a stock just before it
goes ex dividend and just after, stocks should be
priced so that the marginal tax liability would be the
same for each strategy. Thus, if dividends are taxed
more heavily than are capital gains, stock prices
should fall by less than the size of the dividend on
the ex-dividend day. Evidence consistent with a tax
effect in stock price behaviour around ex-dividend
days is provided in Elton and Gruber ( ),
Eades et al. ( ), Green and
Rydqvist ( ), Bell and Jenkinson ( ), and
Elton et al. ( ). In addition, evidence of tax-
motivated trading around ex-dividend days is pro-
vided in Lakonishok and Vermaelen ( ),
Michaely and Vila ( ) and Green and
Rydqvist ( ).
Collectively, the evidence in these studies indi-
cates that the differential taxation of dividends and
returns and trading activity. This conclusion has
been reinforced in studies that examine changes in
tax laws (for example, Poterba and Summers ;
Barclay ; Michaely ). Nonetheless, the fact that
individual investors in high tax brackets receive
large amounts of taxable dividends each year (Allen
and Michaely ) casts doubt on taxes being a first-
order determinant of dividend policy.
Dividend Policy
Agency Costs
A second real-world violation of the Miller-Mo-
digliani assumptions is the existence of agency
costs associated with stock ownership. In particular,
managers of firms maximize their own utility,
which is not necessarily the same as maximizing
the market value of common stock. The costs
associated with this potential conflict of interest
include expenditures for structuring monitoring and
bonding contracts between shareholders and
managers, and residual losses due to imperfectly
constructed contracts (Jensen and Meckling )•
Several authors have argued that dividends may
be important in helping to resolve manager-
shareholder conflicts. If dividend payments reduce
agency costs, firms may pay dividends even if these
payments are taxed disadvantageously.
Easterbrook ( ) and Rozeff ( ) argue
that establishing a policy of paying dividends
enables managers to be evaluated periodically by
the capital market. By paying dividends, managers
are required to tap the capital market more
frequently to obtain funds for investment projects.
Periodic review by the market is one way in which
agency costs are reduced, which in turn raises the
value of the firm. Similarly, Jensen ( ) argues
that establishing a policy of paying dividends
reduces agency problems of overinvestment by
reducing the amount of discretionary cash con-
trolled by managers.
An implication of the agency models is that
dividends will be more valuable in mature firms
with substantial cash flow and poor investment
opportunities. Early tests of this implication
focused on the stock price reaction to dividend
change announcements and produced mixed results.
Lang and Litzenberger ( ) find that
firms with less valuable growth opportunities
exhibit a larger stock price reaction to dividend
increase announcements than firms with more
valuable growth opportunities. Although this
finding is consistent with the agency cost hypoth-
esis, Denis et al. ( ) find that when they con
trol for other factors, particularly the change in
dividend yield, they find no difference in the stock
price reaction to dividend changes between firms
3009
with good growth opportunities and those with poor
growth opportunities. Moreover, they find no
evidence that increases in dividends reduce
corporate investment.
More recent tests of the agency models have
focused on the cross-sectional determinants of
dividend policy. Fama and French ( ) find
that the propensity to pay dividends is positively
related to firm size and profitability, and negatively
related to the value of future growth opportunities.
DeAngelo et al. ( ) find that the
propensity to pay dividends is strongly associated
with the proportion of the firm’s equity that comes
from retained earnings. These findings support the
primary prediction of the agency models that div-
idends are more valuable for mature firms with high
cash flow and poor growth opportunities.
La Porta et al. ( ) and Faccio and Lang
( ) provide further support for the agency
models of dividend policy by analysing interna-
tional evidence. La Porta et al. hypothesize that
agency conflicts will differ across countries because
of differences in the extent of investor protection. In
a sample of 33 different countries, they find that
dividend payments are higher in countries with
better investor protection. This indicates that when
investors are better able to monitor managers, they
are able to force higher dividend payouts. Faccio
and Lang ( ) show
that in western Europe and in Asia dividend pay-
ments are higher when controlling shareholders
have a higher ratio of voting rights to cash flow
rights - that is, those situations in which minority
shareholders are otherwise at greatest risk of
expropriation by the controlling shareholder.
Asymmetric Information
Contrary to the Miller-Modigliani assumption that
investors have the same information as managers, a
large number of studies assume that managers
possess more information about the prospects of the
firm than individuals outside the firm, and that
dividend changes convey this information to
outsiders. This idea was suggested by Miller and
Modigliani and has roots in Lintner’s ( )
classic study on dividend policy. Lintner
interviewed a sample of corporate managers. One of
the primary findings of the interviews is that a
3010
high proportion ofmanagers attempt to maintain a
stable regular dividend. In Lintner’s words, man-
agers demonstrate a ‘reluctance (common to all
companies) to reduce regular rates once established
and a consequent conservatism in raising regular
rates’ ( , p. 84).
If managers change regular dividends only when
the earnings potential of the firm has changed,
changes in regular dividends are likely to provide
some information to the market about the firm’s
prospects. More formal models in which dividends
convey information to outsiders include
Bhattacharya ( ), John and Williams
( ), and Miller and Rock ( ). The common
assumption in these models is that managers have
information not available to outside investors.
Typically, the information has to do with the current
or future earnings of the firm.
Empirical evidence on the information content
of dividends has taken three forms. First, a large set
of studies has analysed whether dividend changes
are associated with abnormal stock returns of the
same sign. Second, studies have analysed whether
dividend changes are associated with subsequent
earnings changes. Third, studies have analysed the
association between dividend changes and changes
in investor expectations regarding future earnings.
Studies have consistently documented that stock
returns around the announcement of a dividend
change are positively correlated with the change in
the dividend (Aharony and Swary ; Asquith and
Mullins ; Brickley ; Healy and Palepu ; Grullon et
al. ; Michaely et al. ; Pettit ). These studies
are robust over time and are robust to controls for
contemporaneous earnings announcements.
Moreover, in general, the studies indicate that the
market reacts more strongly to a dividend decrease
than to a dividend increase.
The findings described above indicate that div-
idend announcements provide information to the
market. Subsequent studies have investigated
whether this information is correlated with current
or future earnings. On this issue, the evidence is
more mixed. In a study of dividend initiations and
omissions, Healy and Palepu ( ) find that the
initiation of dividends follows a period of
Dividend Policy
abnormal earnings growth and that earnings con-
tinue to grow in subsequent years. For omissions,
however, earnings decline in the year of omission,
then rebound in the following years. Using a
comprehensive sample of dividend changes,
Benartzi et al. ( ) find no evidence that divi
dend changes are associated with subsequent
earnings changes of the same sign. Miller’s inter-
pretation of the evidence ( ) is that dividends
appear to be better described as lagging earnings
than as leading earnings.
One difficulty in testing whether dividend
changes ‘signal’ unexpected future earnings is that
it is difficult to identify what level of earnings
would be expected by the market if the dividend
change did not take place. To address this issue,
Ofer and Siegel ( ) study how analysts alter
their estimates of current year earnings when firms
announce dividend changes. They find that analysts
revise their earnings estimates in the direction of
the dividend change and that the size of the
earnings revision is positively associated with the
stock price reaction to the dividend change.
Similarly, Fama and French ( ) report a
positive association
between dividends and firm value after controlling
for past, current and future earnings, as well as
investment and debt. They conclude that dividends
contain information about value that is not
contained in earnings, investment and debt.
The accumulated empirical evidence thus indi-
cates that dividend announcements provide infor-
mation to the market. Whether they convey
information about future earnings is less clear.
Moreover, other findings indicate that information
signalling is unlikely to be a first-order determinant
of dividend policy. For example, as noted earlier,
dividends are paid primarily by larger, more mature
firms with higher cash flow and poorer growth
opportunities. These types of firm would seem to be
least in need of signalling their true value to the
market.
Firm Value and the Form of the Payout
As with increases in regular cash dividends, spe-
cially labelled cash dividends and share repurchases
have been shown to be accompanied
Dividend Policy
by permanent increases in stock prices (Brickley ;
Dann ; Vermaelen ). However, there is little
agreement on the factors that lead managers to
choose one method over another.
Given the Miller-Modigliani assumptions, the
choice of the payout mechanism, like the choice of
dividend policy itself, does not affect the value of
the firm. Therefore, if the form of the payout is to
matter, it must be due to violation of one or more of
the Miller-Modigliani assumptions. To develop a
theory to explain the choice of payout mechanism,
it must be that there are differential costs or benefits
associated with the alternative payout methods.
Furthermore, the relative benefits or costs must be
especially significant because, in general, dividends
have been tax-disadvantaged (at the personal level)
relative to share repurchases.
Economists have explored several possible
explanations as to why a particular form of payout
is chosen, including adverse selection effects
(Barclay and Smith ; Miller and McConnell ), the
impact on equity ownership structure (Stulz ;
Denis ), the signalling power of alternative payout
mechanisms (Ofer and Thakor ; Jagannathan et al. ),
and the impact of executive stock options (Fenn and
Liang ). The evidence indicates that share
repurchases are more likely when recent earnings
increases are temporary, when earnings are riskier,
when firms make heavy use of stock options in
executive compensation contracts and when firms
seek to protect themselves from a hostile takeover.
As regards the choice between regular cash
dividends and specially labelled cash dividends,
reasonable explanations have been relatively scarce.
Brickley ( ) does provide evidence
that specially labelled dividends convey a less
positive message about firm value than do increases
in regular cash dividends. Nonetheless, it is unclear
why this is so. Moreover, there has been little
examination of the choice between special
dividends and share repurchases.
What Managers Say
Lintner’s ( ) classic empirical study began
with a survey of corporate executives. The results
3011
of that survey and the accompanying evidence laid
the foundation for much of the empirical and
theoretical work that has followed over the
succeeding half century. Brav et al. ( ) have
conducted a new and more extensive survey of chief
financial officers (CFOs) regarding their views of
corporate payout policy. Their survey yields further
insights into what managers think about dividend
policy, and complements the existing empirical
evidence.
Brav et al. report that CFOs view dividends as
inflexible in that, once a dividend level has been
established, any dividend cut is likely to have a
significantly adverse impact on the company’s stock
price. Thus, consistent with Lintner’s ( )
original observation, managers tend to be
conservative when adjusting dividends upward in
order to avoid having to cut the dividend at a later
date. Rather than establishing a target payout ratio,
managers set a per share payment that is
downwardly inflexible. According to the survey,
managers do not explicitly view dividends as a
mechanism for signalling information that would
distinguish their companies from competitors, and
they consider tax effects only as an afterthought.
These observations accord with the conclusions
drawn from empirical studies in that both imply that
taxes and signalling are not first-order determinants
of dividend policy.
In contrast to dividends, repurchases are viewed
by managers as a parallel but more flexible way to
distribute cash to shareholders in that they can be
initiated and discontinued as funds are available.
This observation is consistent with the empirical
evidence cited earlier that repurchases tend to be
associated with temporary increases in earnings,
while dividends are associated with earnings
changes that are more permanent. Whether the
modem survey of Brav et al. leads to the volume of
additional empirical work that followed Lintner’s
study remains to be seen.
Summary and Recent Trends
Since the mid-1960s, rigorous consideration has
added considerably to progress in what is known
3012
about dividend policy. We know that firms pay out
to stockholders substantial amounts of cash annu-
ally in the form of regular cash dividends, share
repurchases and specially labelled dividends. We
also know that stock prices increase permanently
when regular dividends are increased, when special
dividends are declared, and when shares are
repurchased, and that stock prices decline when
regular dividends are reduced. While these findings
imply that dividend changes reflect information
available to managers that is not otherwise available
to outside investors, it is still not clear what
information is being conveyed through the dividend
payment. Moreover, although we now know a
considerable amount about the empirical
determinants of the size of payout and the form of
payout, there is little agreement as to whether the
level of cash payout affects the value of the firm or
and whether the choice of the payout method
matters.
We conclude by outlining several recent trends
that pose additional challenges to our understanding
of dividend policy. First, Fama and French ( )
document that the propensity
to pay dividends has declined substantially since the
late-1970s. Second, despite this decline in the
propensity to pay dividends, aggregate dividends
have not declined (DeAngelo et al. ). Rather,
dividends and earnings have become increasingly
concentrated among larger firms. Third, specially
labelled dividends have nearly disappeared
(DeAngelo et al. ). Fourth, share repurchases have
increased substantially so that aggregate payouts
through share repurchases now exceed those
through regular dividends (Grullon and Michaely ).
These trends are difficult to explain given our
current understanding of dividend policy.
Undoubtedly, therefore, economists will continue to
devote substantial effort to understanding the
puzzles of dividend policy.
See Also
► inane 111 w E e' opments)
Dividend Policy
Bibliography
Aharony, J., and I. Swary. 1980. Quarterly dividend and
earnings announcements and stockholders’ returns: An
empirical analysis. Journal of Finance 35: 1 12.
Allen, F., and R. Michaely. 2003. Payout policy. In Handbook
of the economics of finance: volume la, ed. G.
Constantinides, M. Harris, and R. Stulz. Amsterdam:
North-Holland.
Asquith, P., and D. Mullins. 1983. The impact of initiating
dividend payments on shareholders’ wealth. Journal of
Business 56: 77-96.
Barclay, M. 1987. Dividends, taxes, and common stock
prices: The ex-dividend day behavior of common stock
prices before the income tax. Journal of Financial
Economics 14: 31-44.
Barclay, M., andC. Smith. 1988. Corporate payout policy:
Cash dividends versus open-market repurchases. Journal
of Financial Economics 22: 61-82.
Bell, L., and T. Jenkinson. 2002. New evidence of the impact
of dividend taxation and on the identity of the marginal
investor. Journal of Finance 57: 1321-1346.
Benartzi, S., R. Michaely, and R. Thaler. 1997. Do changes in
dividends signal the future or the past? Journal of
Finance 52: 1007-1043.
Bhattacharya, S. 1979. Imperfect information, dividend
policy, and ‘the bird in the hand’ fallacy. Bell Journal of
Economics 10: 259-270.
Bhattacharya, S. 1980. Nondissipative signaling structures
and dividend policy. Quarterly Journal of Economics 95:
1-24.
Black, F., and M. Scholes. 1974. The effects of dividend yield
and dividend policy on common stock prices and returns.
Journal of Financial Economics 1: 1-22.
Brav, A., J. Graham, R. Michaely, and C. Harvey. 2005.
Payout policy in the 21 st century. Journal of Financial
Economics 77: 483-527.
Brennan, M. 1970. Taxes, market valuation and financial
policy. National Tax Journal 23: 417-429.
Brickley, J. 1983. Shareholders wealth, information signaling,
and the specially designated dividend: An empirical
study. Journal of Financial Economics 12: 187-209.
Chen, N., B. Grundy, and R. Stambaugh. 1990. Changing
risk, changing risk premiums, and dividend yield effects.
Journal of Business 63: S51-S70.
Dann, L. 1981. Common stock repurchases: An analysis of
returns to bondholders and stockholders. Journal of
Financial Economics 9: 113-138.
DeAngelo, H., L. DeAngelo, and D. Skinner. 2000. Special
dividends and the evolution of dividend signaling.
Journal of Financial Economics 57: 309-354.
DeAngelo, H., L. DeAngelo, and D. Skinner. 2004. Are
dividends disappearing? Dividend concentration and the
consolidation of earnings. Journal of Financial
Economics 72: 425-456.
DeAngelo, H., L. DeAngelo, and R. Stulz. 2006. Dividend
policy and the eamed/contributed capital mix: A test of
Dividend Policy
the life-cycle theory. Journal of Financial Economics 81:
227-254.
Denis, D. 1990. Defensive changes in corporate payout
policy: Share repurchases and special dividends. Journal
of Finance 45: 1433-1456.
Denis, D., D. Denis, and A. Sarin. 1994. The information
content of dividend changes: Cash flow signaling, over-
investment, and dividend clienteles. Journal of Financial
and Quantitative Analysis 29: 567-587.
Dewing, A. 1953. The Financial policy of corporations. Vol.
2. 5th ed. New York: Ronald Press Co.
Eades, K., P. Hess, and E. Kim. 1984. On interpreting
security returns during the ex-dividend period. Journal of
Financial Economics 13: 3-34.
Easterbrook, F. 1984. Two agency-cost explanations of
dividends. American Economic Review 74: 650-659.
Elton, E., and M. Gruber. 1970. Marginal stockholders’ tax
rates and the clientele effect. Review of Economics and
Statistics 52: 68-74.
Elton, E., M. Gmber, and C. Blake. 2005. Marginal stock-
holder tax effects and ex-dividend day behavior: Evidence
from taxable versus nontaxable closed-end fluids. Review
of Economics and Statistics 87: 579-586.
Faccio, M., and L. Lang. 2002. The ultimate ownership of
western European corporations. Journal of Financial
Economics 65: 365-395.
Fama, E.F., and K. French. 1993. Common risk factors in the
returns on stocks and bonds. Journal of Financial
Economics 33: 3-56.
Fama, E.F., and K. French. 1998. Taxes, financing decisions,
and firm value. Journal of Finance 53: 819-843.
Fama, Eugene, and K. French. 2001. Disappearing dividends:
Changing firm characteristics or lower propensity to pay?
Journal of Financial Economics 60: 3-43.
Fenn, G., andN. Liang. 2001. Corporate payout policy and
managerial stock incentives. Journal of Financial Eco-
nomics 60: 45-72.
Gordon, M. 1959. Dividends, earnings and stock prices.
Review of Economics and Statistics 41: 99-105.
Graham, B., and D. Dodd. 1951. Security analysis: Principles
and technique. New York: McGraw-Hill.
Green, R., and K. Rydqvist. 1999. Ex-day behavior with
dividend preference and limitation to short-term arbitrage:
The case of Swedish lottery bonds. Journal of Financial
Economics 53: 145-187.
Grullon, G., and R. Michaely. 2002. Dividends, share
repurchases and the substitution hypothesis. Journal of
Finance 57: 1649-1684.
Grullon, G., R. Michaely, and B. Swaminathan. 2002. Are
dividend changes a sign of firm maturity? Journal of
Business 75: 387—424.
Healy, P., and K Palcpu. 1988. Earnings information con-
veyed by dividend initiations and omissions. Journal of
Financial Economics 21: 149-176.
Jagannathan, M., C. Stephens, and M. Weisbach. 2000.
Financial flexibility and the choice between dividends and
stock repurchases. Journal of Financial Economics 57:
355-384.
3013
Jensen, M. 1986. Agency costs of free cash flow, corporate
finance, and takeovers. American Economic Review 76:
323-329.
Jensen, M., and W. Meckling. 1976. Theory of the firm:
Managerial behavior, agency costs and ownership
structure. Journal of Financial Economics 3: 305-360.
John, K, and J. Williams. 1985. Dividends, dilution, and
taxes: A signaling equilibrium. Journal of Finance 40:
1053-1070.
La Porta, R., F. Lopez-De Silanes, A. Shleifer, and
R. Vishny. 2000. Agency problems and dividend policy
around the world. Journal of Finance 55: 1-33.
Lakonishok, J., and T. Vermaelen. 1986. Tax induced trading
around ex-dividend dates. Journal of Financial
Economics 16: 287-319.
Lang, L., and R. Litzenberger. 1989. Dividend announce-
ments: Cash flow signaling vs. free cash flow hypothesis.
Journal of Financial Economics 24: 181-192.
Lintner, J. 1956. Distribution of incomes of corporations
among dividends, retained earnings, and taxes. American
Economic Review 46: 97-113.
Litzenberger, R., andK. Ramaswamy. 1979. The effects of
personal taxes and dividends on capital asset prices:
Theory and empirical evidence. Journal of Financial
Economics 1: 163-195.
Michaely, R. 1991. Ex-dividend day stock price behavior:
The case of the 1986 tax reform act. Journal of Finance
46: 845-860.
Michaely, R., and J. Vila. 1995. Investors’ heterogeneity,
prices and volume around the ex-dividend day. Journal
of Financial and Quantitative Analysis 30: 171-198.
Michaely, R., R. Thaler, and K Womack. 1995. Price
reactions to dividend initiations and omissions: Over-
reaction or drift? Journal of Finance 50: 573-608.
Miller, M. 1987. The information content of dividends. In
Macroeconomics: Essays in honor of Franco
Modigliani, ed. J. Bossons, R. Dombush, and
S. Fischer. Cambridge, MA: MIT Press.
Miller, J., and J. McConnell. 1995. Open-market share
repurchase programs and bid-ask spreads on the NYSE:
Implications for corporate payout policy. Journal of
Financial and Quantitative Analysis 30:365-382.
Miller, M., and F. Modigliani. 1961. Dividend policy, growth
and the valuation of shares. Journal of Business 34: 411-
433.
Miller, M., andK. Rock. 1985. Dividend policy under asym-
metric information. Journal of Finance 40: 1031-1051.
Miller, M., and M. Scholes. 1978. Dividends and taxes.
Journal of Financial Economics 6: 333-364.
Miller, M., and M. Scholes. 1982. Dividends and taxes:
Empirical evidence. Journal of Political Economy 90:
1118-1141.
Ofer, A., and D. Siegel. 1987. Corporate financial policy,
information, and market expectations: An empirical
investigation of dividends. Journal of Finance 42: 889-
911.
Ofer, A., and A. Thakor. 1987. A theory of stock price
responses to alternative corporate cash disbursement
3014
methods: Stock repurchases and dividends. Journal of
Finance 42: 365-394.
Pettit, R. 1972. Dividend announcements, security perfor-
mance, and capital market efficiency. Journal of
Finance 27: 993-1007.
Poterba, J., and L. Summers. 1984. New evidence that taxes
affect the valuation of dividends. Journal of Finance 39:
1397-1415.
Rozeff, M. 1982. Growth, beta and agency costs as deter-
minants of dividend payout ratios. Journal of Financial
Research 5: 249-259.
Stulz, R. 1988. Managerial control of voting rights: Financing
policies and the market for corporate contro I. Journal of
Financial Economics 20: 25-54.
Vermaelen, T. 1981. Common stock repurchases and market
signaling: An empirical study. Journal of Financial
Economics 9: 139-183.
Divisia Index
Charles R. Hulten
Abstract
The Divisia index, it its modem application, is a
continuous-time index related to an underlying
economic structure via a potential function.
Under certain conditions, the index can retrieve
important characteristics of the underlying
structure using prices and quantities alone, with-
out Ml knowledge about the structure itself. The
Divisia index is widely used in theoretical
discussions of productivity analysis, and has
important applications elsewhere. In practice, it
is approximated by discrete-time superlative
indexes, like the Tomqvist, or by chain indexes.
Older applications of the Divisia stressed its
discrete-time axiomatic properties.
Keywords
Aggregation; Chain indexes; Continuous-time
indexes; Discrete-time indexes; Divisia index;
Divisia, F.; Duality; Path dependence; Produc-
tion functions; Productivity (measurement prob-
lems); Solow, R.; Tomqvist index
JEL Classifications
C43; C80; E01
Divisia Index
The Divisia index is a continuous-time index
number formula due to Francois Divisia ( -6)
that has been widely used in theoretical
discussions of data aggregation and the measure-
ment of technical change. It is defined with respect
to the time paths of a set of prices [Pft), ... , PJVO)]
and commodities [X; (t),..., Xv (t)] • Total
expenditure on this group of commodities is given
by:
Y( t ) = P , ( t ) X 1 ( t ) + .. . + P N ( t ) X N ( t ).
(1)
With dots over variables indicating derivatives with
respect to time, total differentiation of ( ) yields:
r( t ) ^ P i W ^ P i j t )
Y( t ) hi Y
(0 ^-(0 hi *.•(')'
(2)
The growth rates of the Divisia price and quantity
indexes are the respective weighted averages of the
growth rates of the individual Pt (t) and Xt (t), where
the weights are the components’ shares in total
expenditure. The levels of these indexes are
obtained by line integration over the trajectory
followed by the individual prices and quantities over
the time interval [0, T\. For the quantity index, the
line integral has the following form:
A Pj(t)Xj(t) Xj(t)
I q ( 0, T ) = e x p - ! hi Y,Pj WMUt )
(p ( X) d
X
= exp
(3)
where ip is a vector-valued function whose argu-
ments are Pt (t)/Y (t), prices are assumed to be a
function of the Xt, and F is the curve described by
XL A similar expression characterizes the Divisia
price index (for a more extensive discussion of
Divisia line integrals, see Richter ; Hulten ;
Samuelson and Swamy ).
The value of the index defined by ( ) depends on
the solution of the line integral. This can be
obtained by identifying a ‘potential function’ <I>
Divisia Index
whose partial derivatives are the vector-valued
function <p, that is, = VO. Writing O = log F
function, the value of the index can be shown to
equal F[X(T)]/F[X(0)], implying that the index is
unique only up to a scalar multiple.
In economic terms, the solution to ( ) is asso-
ciated with some underlying economic relationship
among the variables being indexed. Assume, for
example, there is a constant returns to scale
production function F(X) and F, = A.P, (Fi denotes
the partial derivative of F with respect to Xi and A
is a factor of proportionality). Then the function log
F can serve as the requisite potential function for ( ),
and in this particular case, the Divisia index of
inputs can be interpreted as the ratio of output at
time T to output at time zero.
If the form of the potential function is known a
priori, the value of the index could be computed
directly from the function F. However, the rationale
for the Divisia index is that it provides a way of
obtaining the ratio F(X(T))/F(X(0)) by using data
on prices and quantities alone, without direct
knowledge of F. Intuitively, this is possible because,
under sufficiently restrictive assumptions,
information about the slope of the function F (as
estimated by relative prices) over the path followed
by the inputs is sufficient to characterize F up to a
scalar multiple.
When the objective is to form an index of a
subset of inputs - aggregate labour input, for
example - the required potential function is a ‘piece’
of a production fimction. Specifically, if one wants
to form a Divisia index of the first M inputs, the
production fimction needs to be weakly separable
into a fimction of these inputs, that is,
F{G[Xi(t),.. .,XM(t)], XM+i(f),.. .AGvO)}- The
function log G serves as the potential fimction for
the line integration (see also Balk ).
These considerations apply to Divisia price
indexes as well. The relevant potential fimction is
now the factor price frontier vF[P1(t),..
A basic result of duality theory shows that the
partial derivatives of *P are proportional to the
corresponding Xi(fi).
The discussion suggests that the existence of the
Divisia index is closely linked to the conditions for
consistent aggregation. Furthermore, the required
existence of a potential function implies
3015
that aggregation cannot proceed with just any set of
prices or quantities. There must be an a priori
reason for supposing that the variables to be
indexed are theoretically related. This is an impor-
tant characteristic of the Divisia index, one which it
shares with the broader class of economic index
numbers (in contrast to the non-structural axiomatic
approach associated with Irving Fisher ; see also
Balk ). The potential function theorem establishes
the conditions under which the Divisia index is an
‘exact’ index number (to use the terminology of
Diewert ) for some underlying economic structure.
Divisia indexes have the desirable property that
they are invariant when the path of integration lies
entirely in the same level set of the potential
function. That is, ifone input is substituted for
another along a given isoquant, the value of the
index will not change. However, there is no guar-
antee of invariance when the path of integration lies
across several level sets. This reflects the
mathematical property that line integrals are, in
general, path dependent.
Path dependence means that the index ( ) will
generally have a different value for a path Pit) e r i
than path y.(t) e T , even though the beginning and
end points of Fi and T are identical. This can lead to
the following situation: the economy moves along 1
, from X from X' (which is on a different isoquant);
the economy then returns along F to the original
point X; because of path dependence, the vector of
quantities represented by the vector X will have a
different Divisia index value after the trip around
the composite path, and subsequent circuits will
produce still different values. The value of the
Divisia index at any point X is thus arbitrary under
path dependence. The uniqueness of the Divisia
index thus involves path independence.
The condition for path independence is the
existence of a homothetic potential fimction, log F,
such that tp = VlogF, where <p is defined in ( ).
Given the existence of the potential fimction, the
value of ( ) is F(X(T))/F(X(0)), implying path
independence since ( ) depends only on the end
points of the path, ATO) and X( T). Conversely, if
( ) is path independent, there exists a potential
function log F such that Vlog F = <p. In some
3016
applications in productivity analysis, the homo-
theticity condition must be strengthened to linear
homogeneity, but this can be weakened depending
on data availability (Hulten , pp. 11-12).
We note, finally, that the Divisia index is
defined using time as a continuous variable. Data
on prices and quantities typically refer to discrete
points in time, and the indexes constructed from
them must therefore have a discrete-time form. The
continuous-time Divisia index is nevertheless use-
ful, both for informing the structure of these
discrete-time indexes (for example, for the
determining which variables are conceptually
related), and for interpreting the results. The Divisia
framework is also appropriate for the theoretical
analysis of many economic problems, such as the
use of Divisia indexes by Solow ( ) in growth
accounting.
One approach to linking discrete and continuous
index numbers is to approximate the continuous
variables of ( ) with their discrete time counterparts.
Under the Tomqvist (
approach, the growth rates of prices and quantities
are approximated by logarithmic differences, and
the continuous weights by two period arithmetic
averages. The Tomqvist approximation to the
growth rate of the Divisia quantity index can then
be written:
Pi.tX, Pi . t- i Xi . t ~ i
[log X,-, t - logA,, t
A similar approximation applies to the growth rate
of the Divisia index of prices.
While the Tomqvist index may be regarded as
approximate, Diewert ( ) has shown that it is David E. R. Gay
exact when the underlying potential function has the
(continuous ) translog form. This result is very
important in its own right, but can also be regarded
as an important conceptual link between the discrete
and continuous-time families of index numbers,
given the exact properties of the Divisia index in
continuous time.
The continuous Divisia index can also be
approximated by using chain indexing procedures
(the Divisia index is sometimes regarded as a chain
whose links are defined over infinitesimal time
periods). Other numerical approximation techniques
can also be employed.
Divisia, Francois Jean Marie (1889-1964)
See Also
► U.c.exl
Bibliography
Balk, B. 2005. Divisia price and quantity indices: 80 years
after. Statistica Neerlandica 59: 119-158.
Diewert, W. 1976. Exact and superlative index numbers.
Journal of Econometrics 4 (2): 115-145.
Divisia, F. 1925-6. L’indice monetaire et la theorie de la
monnaie. Revue d'Economie Politique 39(4): 842-864;
(5): 980-1008; (6): 1121-1151; 40(1): 49-81. Also sep-
arately: Paris: Societe Anonyme du Recueil Sirey, 1926.
Fisher, I. 1921. The making of index numbers. Boston:
Houghton Mifflin Co.
Hulten, C. 1973. Divisia index numbers. Econometrica 41:
1017-1025.
Hulten, C. 2001. Total factor productivity: A short biography.
In New developments in productivity analysis, Studies in
income and wealth, ed. C. Hulten, E. Dean, and M.
Harper, vol. 63. Chicago: University of Chicago Press for
the NBER.
) Richter, M. 1966. Invariance axioms and economic indexes.
Econometrica 34: 739-755.
Samuelson, P, and S. Swamy. 1974. Invariant economic
index numbers and canonical duality: Survey and syn-
thesis. American Economic Review 64: 566-593.
Solow, R. 1957. Technical change and the aggregate pro-
duction function. Review of Economics and Statistics
39:312-320.
Tomqvist, L. 1936. The bank of Finland’s consumption price
index. Bank of Finland Monthly Bulletin 10: 1-8.
(4)
Divisia, Francois Jean Marie (1889-1964)
Keywords
American Statistical Association; Divisia Index;
Econometric Society; International Econometric
Society.; Mathematical economics; Statistics
and economics
JEL Classifications
B31
Division of Labour
Divisia was bom in Tizi-Ouzou, Algeria. He
received baccalaureate degrees in mathematics and Division of Labour
philosophy at Algiers. After two years in the Ecole
Polytechnique he worked for the government as a Peter Groenewegen
civil engineer (Ponts et Chaussees). His graduate
engineering work at the Ecole Nationale des Ponts
et Chaussees was completed in 1919 after the
interruption of the First World War. After nearly ten Abstract
years as a government engineer he joined the
ministry of national education to continue research
and teaching economics. He became a professor of
applied economics at the Ecole Nationale des Ponts
et Chaussees (1932-50), the Conservatoire National
des Arts et Metiers (1929-59), and the Ecole
Polytechnic (1929-59). He was a founding member
of the Econometric Society and its president in
1935. Subsequently he was also president of the
Paris Statistics Society (1939) and of the
International Econometric Society. He was a Fellow
of the American Statistical Association and of the
American Association for the Advancement of
Science.
His major contributions to economics can be
found centred in several books on economics and
applied statistics. The Divisia Index, a variable-
weight price index, was developed in L’indice
monetaire et la theorie de la monnaie (1926). His
Economique rationnelle (1928) was widely
acclaimed in mathematical economics and was
awarded prizes by the Academy of Sciences and by
the Academy of Moral Sciences and Politics. Using
a microeconomic perspective he cautioned against
uncritical acceptance of macroeconomic research in
Traitement econometrique de la monnaie, l ’interet,
1 ’emploi (1962).
Selected Works
1 9 2 6 . L’indice monetaire et la theorie de la
monnaie. P a r i s .
1 9 2 8 . Economique rationnelle. P a r i s .
1 9 3 1 . L’epargne et la richesse collective.
Paris. 1951-65. Exposes
d’economique. P a r i s .
1 9 6 2 . Traitement econometrique de la monnaie,
Tinteret, Temploi. P a r i s .
3017
Division of labour has been a very important
topic for economic writings from the earliest
times, and was treated in great detail by major
economists, including especially Adam Smith
and Alfred Marshall. This article surveys the
development of ‘division of labour’ from its
beginnings in the writings of Greek philosophers
through the centuries and up to the 21st century.
It therefore also reflects on its offshoots:
international division of labour, sexual division
of labour and its contemporary revival as an
essential adjunct to the theory of economic
growth, labour productivity, inter-firm
cooperation, and its modem limits in coordina-
tion and communication costs.
Keywords
Agricultural productivity; Aristotle; Austrian
capital theory; Babbage, C.; Beccaria, C. B.;
Bagehot, W.; British Association; Capital
accumulation; Capitalism; Carl, E. L.; Carlyle,
T.; Clustering; Communication and coordination
costs; Comte, A.; Concentration; d’Alembert,
C.; Dexterity; Diderot, D.; Division of labour;
Durkheim, E.; Economies of scale; Engels, F.;
Engineering; Equilibrium analysis; Ferguson,
A.; Firm, theory of; Foreign direct investment;
Hegel, G. W. F.; Hodgskin, T.; Housework;
Human capital; Hutcheson, F.; IbnKhaldun;
Increasing returns; Industrial organization;
Industrial psychology; Industrial revolution;
Inequality; Inter-firm cooperation; International
division of labour; Kaldor, N.; Labour
productivity; Mandeville, B.; Manufacturing;
Manufacturing division of labour; Marshall, A.;
Marshall, M. P.; McCulloch, J. R.; New
classical microeconomics; New growth
economics; Nicholson, J. S.; Petty, W.; Plato;
Productivity growth; Putting-out system;
Quesnay, F.; Rae,
3018
J. (1796-1872); Robbins, L. C.; Ruskin, J.;
Segmented labour markets; Senior, N.; Sexual
division of labour; Sidgwick, H.; Social division
of labour; Specialization; Spencer, H.; Taussig,
F. W.; Technical change; Torrens, R.; Tucker,
J.; Turgot, A. R. J.; Ure, A.; Verdoom’s Law;
Women’s work; Xenophon; Young, A. A.
JEL Classifications
B1
Division of labour, or specialization, may be
defined as the division of a process or employment
into parts, each of which is carried out by a separate
person, or any system of production in which tasks
are separated to enable specialization to occur. This
includes the separation of employments and
professions within society at large or social division
of labour as well as the division of labour which
takes place within the walls of a factory building or
within the limits of a of a single industry, the
manufacturing division of labour. Division of
labour as a form of specialization can also be
practiced by small firms which all contribute to the
production of parts (inputs) for the manufacturing
of a complex output, as in the case of aircraft
production or sophisticated electronic equipment.
This form of business organization requires
excellent coordination and communication between
its various parts to ensure continuous supply of the
necessary parts for the manufacturer of the final
output. It is a geographical form of division of
labour, developed from the notion of clustering
related firms in a particular area or industrial district
(for a survey, see Dosi ).
Division of labour and its consequences for
productivity were analysed as early as the time of
the Greek philosophers, including Plato, Aristotle
and Xenophon. Early analysis of the manufacturing
division of labour had to await industrial
developments of the 17th and 18th centuries and
underwent further qualitative change in the 19th,
20th and 21st centuries. Hence manufacturing and
more detailed division of labour should not be seen
as a simple continuum of the social division of
labour. By the end of the
Division of Labour
Middle Ages, social division of labour was exten-
sively practiced; manufacturing division of labour,
generally speaking, came with the Industrial
Revolution. Under modem capitalism, social
division of labour remains largely a market
influenced phenomenon but manufacturing division
of labour is enforced by those who plan and control
the manufacturing process. Furthermore, the one
divides society: the other human activity within the
workshop, or within an industry: labour generally
enhances ‘the individual and the species, [a
manufacturing division] of labour, when carried on
without regard to human capabilities and needs, is a
crime against the person and against humanity’
(Braverman , p. 73). Division of labour was first
practiced within the household, a sexual division of
labour between women’s activities in or near the
house, and those of men further afield. When
applied to local specialization of industries both
nationally and internationally, it has produced a
variety of conceptions of the territorial or
international (global) division of labour
Adam Smith ( ) placed the division of
labour at the forefront of his discussion of economic
growth and progress. Neither in its social nor in its
manufacturing forms did the idea originate with
him. It retained a varying, but often very prominent,
place in 19th-century writings (particularly those of
Senior, Babbage, John Stuart Mill, Marx and
Marshall).‘About 1890, Schmoller, Semmel,
Bucher, Durkheim and Maunier all wrote on
religious and sociological aspects of specialization’
(Salz , p. 284). For much of the 20th century,
division of labour and specialization virtually
disappeared as a major topic from economic texts.
Reasons for this varied. Some economists believed
such discussions were more appropriate to technical
handbooks of production engineering and factory
management. Other writers wished to confine
analysis of its effects to sociological studies
assessing the general impact of division of labour
on society. The return of economic growth as an
important part of the economist’s research
programme from the 1950s onwards, and earlier the
work of Young ( ), brought renewed interest in the
division
of labour in its wake, as did growing
Division of Labour
dissatisfaction with the narrow view confining
economics to studying ‘the disposal of scarce
commodities’(Robbins , p. 38). Global orga
nization of manufacturing made possible by
improvements in transport and communication
implies modem adaptations of the division of labour
which economists cannot ignore. An example is the
formation of industrialized districts, first observed
and analysed by Alfred Marshall ( ), to be
rediscovered and adapted to the
post-Second World War Italian situation by
Becattini (for example , ) and his col
leagues (for a survey, see Goodman and Pamford ).
The various dimensions of division of labour raised
in these introductory paragraphs suggest that a
broad-based treatment ofthe subject is warranted by
featuring highlights within its continuous
development.
The Greeks
Many of the major Greek philosophers discussed
aspects of the division of labour in their writings. In
Book 2 of the Republic, Plato stated the necessity
for a division of labour or specialization in
occupations for social well-being and the adequate
satisfaction of primary wants linking the
phenomenon with exchange, the requirements of ‘a
market, and a currency as a medium of exchange’
(Plato 380 BC, pp. 102-6). Aristotle, though very
conscious of the social need for a division of labour,
did not depart much from Plato’s earlier discussion
(see Bonar , p. 34). More importantly, Xenophon
linked division of labour and specialization to great
cities, because they provided a substantial demand
for individual products while the subdivision of
work raised the skill of individual workers. Extracts
from the work of these Greek pioneers on the
division of labour have been often reprinted (see,
for example, Sun , chs. 2-4). Knowledge of these
Greek texts among Arabian Islamic scholars during
the middle ages enabled them to produce
sophisticated treatments ofthe division of labour.
Examples are the writings of Islamic theologian, al-
Ghazali (1058-1111) and, more importantly, the
writings of fourteenth century Islamic
3019
philosopher and historian, IbnKhaldun, whose
Muqaddima contains a detailed account of the
division of labour (Sun , pp. 7-8, eh. 5).
Subsequent Pre-Smithian Developments
Towards the end of the seventeenth century,
English economic literature rediscovered the con-
cept of the division of labour and began to analyse
the more modem manufacturing forms, linking
them to productivity growth, cost reduction,
increased international competitiveness and asso-
ciating its scope with the more extensive markets
made possible through urbanization. For example,
Petty’s Political Arithmetick written in 1671 com-
pared the benefits of division of labour in textile
production with specialization in ship building:
For as Cloth must be cheaper made, when one Cards,
another Spins, another Weaves, another Draws,
another Presses and Packs; than when all the
Operations above-mentioned, were clumsily
performed by the same hand; so those who command
the Trade of Shipping [need] to build.. .a particular
sort of Vessels for each particular Trade. (Petty , pp.
260-1)
T e n y e a r s l a t e r , i n Another Essay on
Political Arithmetick Concerning the Growth of the
City of London ( , p. 473), Petty
showed that a
major gain from a vast city like London came from
the improvement and growth of manufactures it
encouraged:
For in so vast a City Manufacturers will beget one
another, and each Manufacture will be divided into
as many parts as possible, whereby the Work of each
Artisan will be simple and easy; As for example in
the making of a Watch, if one Man shall make the
Wheels, another the Spring, another shall Engrave the
Dial-plate, and another shall make the Cases, then
the Watch will be better and cheaper, than if the
whole Work be put upon any one man.
In continuing this argument Petty also suggested
that specialization benefits could be achieved from
concentrating certain manufactures on a particular
location, partly because of the savings in transport
and communication costs such concentration
entailed (Petty , pp. 471-2). The anonymous author
of Considerations on the East India Trade ( , pp.
590-2) illustrated
3020
productivity gains from the division of labour by
examples drawn from cloth making, watch making
and shipbuilding. He clearly indicated that sufficient
demand and regular trade were a precondition for
such improvements, which lowered manufacturing
labour costs without the need to lower wages.
During the 18th century, examples of authors aware
of the benefits and preconditions for a division of
labour become more common. Practical writers like
Patrick Lindsay (
Richard Campbell ( ) and Joseph Harris
( ) tended to concentrate on manufacturing
division of labour using examples from linen and
pin production as well as from the familiar watch
making. Those writing from the position of moral or
political philosophy, like Mandeville ( ),
Hutcheson ( ), Ferguson ( ) and
Josiah Tucker ( , ) concentrated more on
aspects of the social division of labour.
Discussion of the division of labour was of
course not confined to English economic literature.
A treatise on wealth published in the 1720s by Ernst
Ludwig Carl discussed the benefits of the division
of labour, applying them also to demonstrate the
gains from free trade through an international
division of labour based on different climates,
resource availability and locational advantages
(cited in Hutchison
pp. 161-2). Among the Physiocrats, Quesnay dealt
briefly with the social aspects of the division of
labour in his article ‘Natural Right’ ( p. 51). Turgot
developed the subject more thoroughly, making it
the starting point of his Reflections, subsequently
associating it with the introduction of money, the
extension of commerce and the accumulation of
capital ( , pp. 44-6,
64, 70). Earlier, Turgot ( , pp. 242-3) had
linked the spread of social division of labour to
inequality, arguing that this particular consequence
of inequality improved living standards for even the
humblest members of society and made possible
cultivation of the arts and sciences. Among the
general principles with which Beccaria ( , pp.
387-8) commenced the argu
ment of his Elementi, the division of labour and its
benefits in terms of increased skills and dexterity
are clearly set out. Finally, it may be noted that the
Encyclopedic of Diderot and d’Alembert in its
Division of Labour
article ‘Art’ discussed the essentials of the
manufacturing division of labour, listing its con-
sequences as improvements in skill, better quality
products, saving of time and of materials, and ‘of
making the time or the labour go further, whether
by the invention of a new machine or the discovery
of a more suitable method’. In its article on pins
(‘Epingle’ ) their manufacture is described as being
generally subdivided into eighteen separate
), operations and thereby a prime example of the
manufacturing division of labour (see Cannan , pp.
94-5).
Adam Smith's Treatment of the Division
of Labour
Adam Smith’s discussion of the division of labour
deserves separate treatment not because of its
‘originality’ or ‘completeness of exposition’
(Cannan , p. 96) but because ‘nobody either before
or after [him], ever thought of putting such a burden
upon division of labour. With A. Smith, it is
practically the only factor in economic progress’
(Schumpeter , p. 187). The first three chapters of
the Wealth of Nations were devoted to its analysis
because it provided one of the two causes
, explaining increases in per capita output by which
Smith defined the wealth of the nation. Although
therefore only one of two causes, the other being
‘the proportion between the number of those who
are employed in useful labour, and that of those
who are not so employed’ (Smith , p. 10), it is the
dominant one. Smith seems to have believed that
scope for substantial increases in the proportion of
the labour force to productive activities was limited.
Using the equation, g=(k. p/w) — 1, developed by
Hicks ( p. 38) to summarize the Smithian growth
progress, if a change in k, the proportion of produc-
tive laboin in the laboin force, is more or less ruled
out, a substantial growth rate (g), given the real
wage (w), depends exclusively on rising pro-
ductivity (p) through extensions of the division of
laboin. Smith’s emphasis on the division of laboin
as a factor in growth via its enormous influence on
productivity makes his treatment of the subject so
novel. Surprisingly, this aspect of his contribution
Division of Labour
was taken up by few 19th-century writers and had
to be largely rediscovered in the work of Young (
) and Kaldor ( ) who reiterated dynamic
aspects of the phenomenon Smith was analysing.
Even though it was the most frequently revised
part of his economics (see Meek and Skinner ),
Smith’s basic account of the division of labour
contains a number of weaknesses. First, Smith
failed to develop aspects of the manufacturing
division of labour with which he ought to have been
familiar. Marglin ( ) points out that
Smith ignored organizational features from a divi-
sion of labour taking place within the one building
of relevance to some well-established industries like
textiles and the manufacture of metal implements.
These organizational features which Smith omitted
were associated with growing labour discipline
problems, wasting time and materials, inherent in
the putting-out system, then the dominant form of
manufacturing organization. In fact it can be
suggested that if this aspect of the division of labour
is more fully taken into account, its important role
in explaining economic growth so much
emphasized by Smith is more easily integrated as a
major factor explaining the industrial revolution
(see Groenewegen ). Marglin ( ) also
questioned the force of ‘the three
different circumstances’ by which Smith ( p. 17)
explained the productivity gains from the division
of labour: increased dexterity, saving of time, and
invention of machinery. Although increased
dexterity is clearly a product of a division of labour
in a manufacturing process, its scope there is rather
limited when compared to that of the continual
practice of surgeons, concert pianists and opera
singers, to give some examples. Time saved in
eliminating time lost in passing from job to job is
trivial and not the ‘very considerable’ benefit Smith
( , pp. 18-19) had
suggested. Savings in materials and time through
transforming a putting-out to a factory system, an
organizational feature of the division of labour
Smith had ignored, was more important, particularly
through eliminating losses from pilfering. Rae (
, pp. 164-5) saw savings in the use of
tools as far more significant than time saved, and
for him (pp. 352-7) this provided the basic reason
for extending the division of labour. Other 19th
3021
century writers, particularly Babbage ( ),
expanded further on this aspect of the matter.
Smith’s association of division of labour with
inventions ( , pp. 19—22) covered both ‘on
the job improvements’ and scientific inventions by
specialists originating from within a more
sophisticated division of professions. It ignores, as
Hegel ( , p. 129) was one of the first to
point out (cf. Stewart -75, vol. 8, pp. 318-19), that
as division of labour makes ‘work more and more
mechanical,... man is able to step aside and install
machines in his place’. This feature of the process
was subsequently noted by Babbage ( ,
pp. 173—4), Ure ( ,
p. 21) and developed by Marx ( ). In short, the
three circumstances Smith saw as explaining the
productivity consequences from the division of
labour derive their basic validity from reasons
different to those Smith advanced. Further, Smith’s
remarks ( , pp. 16-17) on the smaller
benefits from applying the division of labour to
agriculture than to manufacturing can be contrasted
with his quite different and controversial analysis of
the primacy of agricultural investment in terms of
its employment of productive labour. Agriculture’s
more substantial contribution to gross revenue as
Smith ( , Book II,
ch. 5) subsequently argued, was used by him to
define the ‘natural’ course of economic develop-
ment (Book II, ch. 1) and recommended as superior
practice for newly settled regions like the American
colonies. Perelman ( , p. 185)
explained this seeming contradiction in Smith by
suggesting Smith was the ‘first theorist of neo-
imperialism’ because his strategy of development
forces developing regions to specialize in raw
material production whose terms of trade with
manufactures are invariably poor. More likely,
Smith’s views on the productivity of agriculture
relative to manufacturing are posed in terms of
different yardsticks: agricultural activity by the very
nature of its processes is less amenable to division
of labour, even though its ability to employ
productive labour is greater than produced by equal
investments in manufacture and trade. However,
growing mechanization of agriculture, especially in
the 20th century, together with the greater scope for
exporting agricultural
3022
surplus with modem transport, encouraged spe-
cialization in agriculture and very large scale
farming (Salz , p. 283).
A final controversial issue from Smith’s treat-
ment of the division of labour concerns its social
consequences, an argument he placed in the context
of public education. The ‘few simple operations’
which under a division of labour most ordinary
labouring people are asked to perform, renders
them ‘as stupid and ignorant as it is possible for a
human creature to become’ and increased ‘dexterity
at his own particular trade’ is purchased with a
reduction in ‘intellectual, social and martial virtues.,
.unless government take some pains to prevent it’
through providing general education (Smith , pp.
781-5 ). Smith was not alone in presenting this
disadvantage in an extensive division of labour:
similar views were put by Ferguson ( , p. 280) and
Karnes
( ). Ferguson described ‘ignorance as the
mother of industry’ and argued that prosperous
manufactures arise ‘where the mind is least
consulted, and where the workshop may., .be con-
sidered as an engine, the parts of which are men.
‘At the turn of the century, and after, German
philosophers (for example, Schiller ; Hegel and the
young Marx ) developed this into a humanist
critique of industrial society, suggesting like Smith
that these detrimental consequences were
removable by education, especially aesthetic
education. Such sentiments were resurrected in
mid-19th century England by Carlyle ( ) process (pp. 170-1), as well as economy in tool
and Ruskin ( -3, pp. 197-8). For
others, Smith’s remarks were an aberration, ‘as
unfounded [a statement] as can well be imagined’
(McCulloch , p. 350) or even a contradiction with
the division of labour’s ability to inspire inventive
faculties in labourers (West ).
Despite its deficiencies, Smith’s account of the
division of labour proved particularly hardy and
was invariably praised in most general terms by
major textbook writers of the 19th century and
after, though few followed the emphasis Smith gave
it as the key factor explaining growth. Cannan (
, p. 97) ascribed this success to
‘the popularity of its form’. It can also be attributed
to the striking productivity increase inherent in the
pin example (cf. Mill , p. 215) and the
Division of Labour
unambiguous connection Smith drew between
increased division of labour, extending the market
and human proclivities ‘to truck and barter’
(McCulloch , pp. 54-5). The account of the division
of labour is undoubtedly one of Smith’s best
remembered performances in economics.
19th-Century Developments
With the growth of the factory system and more
extensive use of increasingly sophisticated
machinery, the manufacturing form of division of
labour was considerably expanded. Consequently,
some economic writers focused on a number of new
aspects of the phenomenon, linking the division of
labour with developments in the machine tool
industry, large scale production and its advantages,
and hence, on a more theoretical level, with
increasing returns to scale and explicit recognition
of a different pattern of productivity growth in
manufacturing from that in agriculture.
Charles Babbage was in many respects the
pioneer in presenting the division of labour as ‘the
most important principle on which the economy of a
manufacture depends’ ( , p. 169). He
therefore carefully revised the advantages of a
division of labour as first expounded by Adam
Smith. In this discussion, time (and cost) savings
were also related to time saved in learning a skill
and reduced waste of materials during the learning
using (p. 172), while the association between divi-
sion of labour, dexterity and the introduction of new
machines was developed more precisely and
rigorously. More significantly, Babbage pointed to a
hitherto ignored additional advantage of the division
of labour he had derived from observation. This had
earlier been discussed by Gioja ( -17) whose
interesting contribution on this
subject was analysed by Scazzieri ( , ch. 3).
By dividing the work to be executed into different
processes of skill or of force, .. .the master
manufacturer... can purchase exactly that precise
quantity of both which is necessary for each pro-
cess; whereas, if the whole work were executed by
one workman, that person must possess sufficient
skill to perform the most difficult, and sufficient
Division of Labour
strength to execute the most laborious, of the operations
into which the art is divided. (Babbage ,
pp. 175-6; emphasis in original)
This economy of skill, Babbage demonstrated
from a pin example, not only reinforced the cost
advantages traditionally associated with division of
labour, but was also a major cause of establishing
large factories: ‘When the number of processes into
which it is most advantageous to divide it, and the
number of individuals to be employed in it, are
ascertained then all factories which do not employ a
direct multiple of this number, will produce the
article at a greater cost’ (Babbage , p. 213). Detailed
division of labour, Babbage also argued, as in its
manufacturing form, can also be appbed to mental
labour (p. 191). An illustration of its application to
mining highlights these control and information
gathering features, two aspects of the division of
labour to which Babbage paid particular attention.
His analysis of the division of labour is even more
important because the process as he described it is
made interdependent with machine production,
increased factory size, lower costs and prices from
such concentration of industry and hence induces
growth in demand and an extended market (see
Corsi ).
Ure’s ( ) contribution must also be noted. It
likewise linked development of the factory system
to division of labour, summarizing ‘the principle of
the factory system... as substituting mechanical
science for hand skills, and the partition of a process
into its essential constituents’ ( , farther the division of labour may be carried., .as
p. 20). Ure commented on two other consequences
of the division of labour in modem factories:
deskilling of the workforce when workers become
‘mere overlookers of machines’ and the
development of mechanical engineering since the
‘machine factory displayed the division of labour in
manifold gradations’ and facilitated the substitution
of skilled hands by ‘the planning, the key- groove
cutting, and the drilling machines’
(pp. 20-1).
Accounts of the division of labour by econo-
mists of the middle of the century were generally
less innovative than those of Babbage and Ure,
though they did occasionally provide some new
points of departure. Senior ( , pp. 74-5, 77),
3023
after classifying division of labour as one major
advantage from the use of capital, concentrated on
listing its benefits additional to those given by
Smith. Illustrating from the post office, he argued
that the fact that ‘the same exertions which are
necessary to produce a single given result are often
sufficient to produce many hundreds or many
thousands similar results’ was one aspect of the
division of labour omitted by Smith. The
development of retailing as a separate profession
was likewise something Smith had failed to con-
sider adequately. More importantly, for a number of
reasons, but particularly the division of labour,
Senior suggested ‘additional Labour when
employed in Manufactures is MORE, when
employed in Agriculture is LESS efficient in pro-
portion’, linking manufacturing activity implicitly
to increasing returns to scale ( , pp. 81-2).
Mill ( ) treated division of labour as an impor
tant aspect of cooperation, arguing that irrespective
of its well-known productivity advantages, without
this complex cooperation in the modem division of
labour ‘few things would be produced at all’ (Mill ,
p. 118). In discussing the productivity advantages,
Mill cited the modification and additional
advantages provided by Babbage ( ) and Rae (
), adding little to
their discussion. However, in Chapter 9 dealing
with large scale and small scale production, he
highlighted the point, so ‘ably illustrated by Mr
Babbage... [that] the larger the enterprise, the
one of the principal causes of large manufactories’
(Mill , p. 131), thereby bringing the argument
firmly into the corpus of economics. Mill’s account
was largely followed by Fawcett ( )
and in most of its essentials by Nicholson ( ).
Marx’s account ( , chs. 13-15) combines
much of this discussion, endowing it in the process
with sharper analytical insights derived from his
study of both the technical literature and his
appreciation of the significance of the qualitative
changes underlying the evolution of the division of
labour. To Marx is owed the important distinction
between manufacturing and social division of
labour, as well as the precise assessment of the
organizational features of its application to modem
manufacture, derived from his careful study of
3024
Babbage, Ure and many other sources. No wonder
that Nicholson ( , p. 105) described Marx’s
treatment as ‘both learned and exhaustive and...
well worth reading’. More recently, Rosenberg (
) expressed regret that Marx’s close study
of ‘both the history of technology, and its newly
emerging forms’ has had so few imitators among
contemporary economists.
Marshall is another economist from the second
half of the 19th century who fully appreciated the
importance of the division of labour and revealed it
in its more modem forms. In 1879, the Economics
of Industry, written with his wife (Marshall and
Marshall ), devoted Chapter 8 of Book I to the
division of labour, immediately after its Chapter 7
on organization of industry. It distinguished the
opportunity to apply a division of labour as inherent
in the nature of the work, as dependent on direction
and control by an entrepreneur as earlier indicated
by Bagehot, and as applied to firms: ‘If there are
any producers, large and small, all engaged in the
same process, Sub- sidiary Industries will grow up
to meet their special wants.’ These include special
machine tool makers for the industry, improved
transport to enhance communication between
related firms, as well as auxiliary enterprises in
banking and credit provision (Marshall and
Marshall p. 52 ). Localization of industry also
fosters ‘education of skills and taste’ and ‘diffusion
of linked knowledge’, and encourages large firms.
Hence division of labour is closely related to
economies of scale, where size has enabled
specialization to grow more and more. Marshall
also devoted no less than three chapters to division
of labour in his Principles ( , Book IV, chs. 9-11),
not only
covering points traditionally dealt with under this
heading, but often introducing subtle modifications.
For example, Marshall ( , p. 263)
discounted detrimental social consequences from
monotonous work by pointing to the mental stim-
ulus from the ‘social surroundings of the factory’
and the view that factory work was not inconsistent
with ‘considerable intelligence and mental
resources’. Likewise, he extended Babbage’s
principle of ‘economy of skill’ to economy of
machinery and materials ( , p. 265), used it
as a major explanatory factor for the localization
Division of Labour
of specialized industry (p. 271) and made it the
chief advantage of large scale production in his
famous discussion of economies of scale (p. 278).
Later, Marshall applied these aspects of his work to
his detailed study of industry and trade to explain
such things as America’s leadership in standardized
production (seen by Marshall p. 149, as an
‘unprecedented’ application of Babbage’s ‘great
principle of economical production’), the successful
specialization of plant during the First World War,
and new issues concerning the growth of the firm. It
is therefore paradoxical that Marshall’s work in
other respects induced the demise of the division of
labour in theoretical literature. This arose from the
incompatibility of increasing returns to scale with
stable demand and supply equilibrium (Marshall ,
Appendix H). Apart from this, modem equilibrium
analysis found it difficult to come to grips with the
dynamic features of the division of labour process,
and it is presumably at least partly for this reason
that division of labour was dropped as an important
subject from the economic textbooks (see Kaldor ).
However, the locational aspects of the division of
labour were further addressed by Becattini (for
example , ) in his development of the notion of
industrial districts as a concentration of related
firms. Marshall had discovered this aspect of
industrial organization through the factory tours in
the British midlands and Scotland he engaged in
from the late 1860s, on which he first reported in
1879. When division of labour for technical reasons
could not take place within the same building, small
firms spring up specializing in part of the
manufacturing process, thereby generating a
division of labour among firms concentrated in a
particular geographical area (for a survey, see
Goodman and Pamford 1988).
International Division of Labour
Torrens ( ) appears to have been the first
economist to distinguish the territorial division of
labour from the mechanical division, suggesting
that the former is inspired by ‘different soils and
climates [being] adapted to the growth of
Division of Labour
different production’ thereby inducing regional
specialization in those products which best suit ‘the
varieties of their soil’ and climate. Taking
advantage of territorial division of labour through
regional and international trade enhances produc-
tivity and increases the wealth of nations as much as
a manufacturing division of labour. Senior ( , p.
76) also drew attention to this aspect of
the division of labour, attributing its discovery to
Torrens. Marshall ( , pp. 267-77) covered
territorial division of labour under localization of
industry while Taussig ( , pp. 41-7) called it
‘the geographical division of labour’ with gains
arising from ‘the adaptation of different regions to
specific articles’ for climatic and resource endow-
ment reasons as well as from the general increase in
proficiency which all specialization brings. During
the 1970s a new dimension of the international
division of labour was analysed, concentrating on
its direct foreign investment aspects. Its novel
features were a tendency to ‘undermine the
traditional bisection of the world into a few indus-
trialized countries on the one hand, and a great
majority of developing countries integrated into the
world economy solely as raw material producers on
the other, and [secondly, to compel] the increasing
subdivision of manufacturing processes into a
number of partial operations at different industrial
sites throughout the world’ to take advantage of
favourable labour market circumstances, relatively
cheap transport opportunities, tax breaks and other
government inducements for foreign investors
(Frobel et al.
, p. 45). This multinational dimension to
application of the division of labour is a direct
descendent from the concept as understood by
Smith, Babbage, Ure and Marx.
The characteristics of the contemporary global
division of labour have been well captured by
Hobsbawm ( , pp. 65-6):
Thus, while the global division of labour was once
confined to the exchange of products between par-
ticular regions, today it is possible to produce across
the frontiers of states and continents. This is what the
process is founded on. The abolition of trade barriers
and liberalization of markets is, in my opinion, a
secondary phenomenon. This is the real difference
between the global economy before 1914 and today.
Before the Great War, there was pan
3025
global movement of capital, goods and labor. But the
emancipation of manufacturing and occasionally
agricultural products from the territory in which they
were produced was not yet possible. When people
talked about Italian, British and American industry,
they meant not only industries owned by citizens of
these countries, but also something that took place
almost entirely in Italy, Britain, or America, and was
then traded with other countries. This is no longer the
case. How can you say that a Ford is an American
car, given that it is made of Japanese and European
components, as well as parts manufactured in
Detroit?
Sexual Division of Labour
The first explicit reference to a sexual division of
labour in economic literature I could find is
Hodgskin ( , pp. 111—12). He argued that
There is no state of society, probably, in which
division of labour between the sexes does not take
place. It is and must be practiced the instant a family
exists. Among even the most barbarous tribes, war is
the exclusive business of the males; they are in
general the principal hunters and fishers ... the
woman labours in and about the hut... In modem as
well as in ancient times,... we find the men as the
rule taking the out-door work to themselves, leaving
the women most of the domestic occupations The
aptitude of the sexes for different employments, is
only an example of the more general principle, that
every human being ... is better adapted than another
to some particular occupation.
Marx and Engels ( -6, pp. 42-3) ascribed
beginnings of the division of labour ‘originally [to]
nothing but the division of labour in the sexual act’
and only later to that ‘spontaneously’ or ‘naturally’
derived from predisposition, needs, accidents, and
so on. Engels ( , esp. p. 311)
elaborated further on the matter presenting the
sexual division of labour in the family as a barrier
to the ‘emancipation of women’. Such an eman-
cipation, he argued, was ‘possibly only as a result of
modem large- scale industry [which] actually called
for the participation of women in production and
moreover, strives to convert private domestic work
also into a public industry’. Both aspects of the
sexual division of labour to which Engels referred
in the context of women’s emancipation have been
taken up in more recent research. The role of
domestic labour has been
3026 Division of Labour

analysed by contemporary writers (see, for example, also based on other grounds. Robbins ( pp. 32-8)
Himmelweit and Mohun ; Gershuny ) while argued that study of the ‘technical arts of
attention has also been drawn to the shift in the production’ belonged to engineering and not to
provision of services from domestic production to economics or, in the case of ‘motion study’, to
production for the market (laundromats, take-away- industrial psychology even if this meant removal of
food) as a result of the gradual break-down of the traditional topics like division of labour from
traditional sexual division of labour within the economics. Robbins’s approach followed
family (Gouvemeur ). Sexual division of labour Sidgwick’s ( , pp. 104-7) treatment, removing
issues have also been applied in segmented labour all technical aspects from the topic, leaving only
market analysis, thereby enriching this particular what he called the pure economics side. Others
aspect of labour economics. suggested it was better to leave discussion of divi-
Becker ( ) has analysed the sexual divisionsion of labour to sociologists because Durkheim,
of labour in the context of human capital investment and before him Comte and Herbert Spencer, had
and allocating the work load of parties within the absorbed it within this emerging discipline. How-
household. Thus both the allocation of effort within ever, some economists in the 20th century objected
a household, and the advantages of investing in to removal of the division of labour from econom-
specific human capital are designed to enhance the ics. In particular, this would reduce understanding
social division of labour and its benefits without of the dynamics of economic progress.
necessarily diminishing the exploitative aspects of Allyn Young ( ) was one of these econo
such arrangements (Becker p. S41). Social factors mists. He made Adam Smith’s theorem that the
are, however, equally important. Increasing returns division of labour is limited by the extent of the
by itself cannot explain the traditional division of market the central theme of his address to section F
labour within the household; a division of labour of the British Association, arguing this was ‘one of
itself subject to change. The increased contribution the most illuminating and fruitful generalizations
to housework by men during the 1970s is one which can be found in the whole literature of
observed aspect of this social change (Becker , p. economics’ (Young , p. 529). Rather than covering
S56). Furthermore, as Posner ( all aspects of the division of labour, Young
, pp. 54, 129)
has noted in concentrated on two interdependent matters:
particular, women were not fully brought into the ‘growth of indirect and roundabout methods of
work place on a large scale until the two world production and the division of labour [or increased
wars, and this only became a dominant pattern in specialization] among industries’ (Young , p. 529)
employment from the 1950s onwards. Cigno ( ) but the former, as Kaldor ( , pp. 355-6) pointed
discusses many of these issues as part of out, was not to be
his economics of the family. confounded with the Austrian capital theoretic
notion. From this he deduced division of labour as a
cumulative, self-reinforcing process, because every
Decline and Rehabilitation of Division of re-organization of production, sometimes described
Labour in the 20th and 21st Centuries as a new invention, involves fresh application of
scientific progress to industry,
The association between division of labour and
increasing returns, the consequent possibility of alters the conditions of industrial activity and initi-
falling supply and cost curves, created problems for ates responses elsewhere in the industrial structure
which in turn have further unsettling effects ... The
equilibrium analysis already noticed as a factor
apparatus of supply and demand in their relation to
explaining decline in emphasis on the division of prices does not seem to be particularly helpfid for
labour and induced its virtual elimination from the purpose of an inquiry into these broader aspects
much of the theoretical literature. Attempts to of increasing returns. (Young , p. 533)
remove division of labour from economics were However, apart from this damaging conclusion for
competitive price theory, the ‘possibility of
Division of Labour
economic progress’ could not really be grasped by
ignoring these factors of greater specialization,
better combinations of advantages of location, and a
consequent increased number of specialized
producers between basic raw materials and final
producers (Young , pp. 538-40).
Kaldor was a major economist who took up
Young’s challenge in both its critical (Kaldor , ) and
more constructive aspects
(Kaldor , ).The major thrust of Kaldor’s positive
argument proclaimed that faster growth is derived
from faster growth in the manufacturing sector,
partly from the cumulative features linking the
growth of manufacturing to growth of labour
productivity via static and dynamic economies of
scale, or the notion of increasing returns as devel-
oped by Young from the division of labour. This
strong and powerful interaction of productivity
growth and manufacturing growth is also posited in
Verdoom’s Law ( ) but its association with
aspects of the division of labour is what is relevant
here. Faster manufacturing growth draws labour
from other sectors of the economy, inducing faster
productivity growth, but as the scope of transferring
such labour from lower productivity sectors like
agriculture dries up, the growth process slows down
(see Thirlwall ). A key feature of the process, as
Rowthom ( , p. 899, n. 1) noticed
in one his skirmishes with Kaldor on the subject, is
that it is an interdependent, cumulative historical
process where ‘higher productivity means more
exports which means greater industrial output
which via its effects on investment, innovation and
scale of production reacts back on productivity
growth’. The importance of such a process was
given detailed empirical examination in a discus-
sion of the Taiwan machine tools industry in the
1970s as an application of the division of labour,
envisaged as increases in output increasing pro-
ductivity, with ‘technological change, broadly
defined, sandwiched in between’ (Amsden , p. 271).
Writers in the new growth economics, who
emphasized the impact of increasing returns from
specialization on growth performance (Romer )
drew in part for their inspiration on the literature of
the division of labour, in Romer’s case as
represented by Marshall (
and possibly Young ( ).
3027
Research from the 1990s has particularly
stressed the importance of communication and co-
ordination costs of the division of labour. Becker
and Murphy ( ) portray these costs
as setting limits on the division of labour more
important than that exerted by the extent of the
market so heavily emphasized by Adam Smith.
Subsequent, Camacho ( ) has studied this
aspect in more detail, drawing a clear and direct
relationship between increases in the division of
labour and rises in both communication and co-
ordination costs, as an essential extension to the
modem theory of the firm and the market. Pemin (
) has treated inter-firm cooperation
and its benefits from a similar angle, assessing the
benefits for production from such cooperation as an
economy of conventions and inter-firm agreements.
This analysis thereby treats division of labour once
again as part of the organizational theory of the firm
or a production unit in which much emphasis is
placed on the potential tradeoffs between the
economies reaped from specialization and the
transaction costs it generates (Yang and Ng ). In this
way, division of labour has also become an
important part of the foundations for a new classical
micro-economic analytic framework.
Conclusion
Viewed dynamically within the context of economic
growth, as Smith ( ) and others had
intended the division of labour to be contemplated,
it continues to be a powerful tool for understanding
the process of growth and development. On this
ground alone it can therefore not be jettisoned from
economics as unwanted baggage, as Robbins ( )
mistakenly suggested. When
its importance for understanding aspects of the
labour process, the labour market, the theory of
production and the theory of the firm contemplated
at the plant and the industry level are included, this
argument is even stronger. As mentioned in the
previous paragraph, on these grounds division of
labour is making a definite come-back as part of the
) theory of a new classical microeconomics. Last, but
not least, the importance of
3028
the division of labour for economics is underlined Engels, F. 1884. Origin ofthe family, private property and the
by the fact that some ofthe major economic minds state. In MarxEngels selected works, vol. 2. Moscow:
Progress Publishers.
from both past and present have invariably included Fawcett, H. 1863. Manual of political economy. London and
it as an important part of their economic analysis. Cambridge: Macmillan & Co.
Bibliography
Arnsden, A.H. 1985. The division of labour is limited by the
rate of growth of the market; The Taiwan machine tool
industry in the 1970s. Cambridge Journal of Economics
9: 271-284.
Anonymous. 1701. Considerations on the East-India trade. In
A select collection of early English tracts on commerce, ed. J.
R. McCulloch. London, 1856; Cambridge: Cambridge
University Press, 1954. Babbage, C. 1832. On the economy of
machinety and manufactures. London. Fourth enlarged
edition of 1835; New York: Augustus M. Kelley, 1963.
Becattini, G. 1990. The Marshallian industrial district as a
socio-economic notion. In Industrial districts and inter-
firm co-operation in Italy, eds. F. Pyke, G. Becattini, and
W. Sengenberger. Geneva: International Institute of
Labour Studies.
Becattini, G. 2001. The caterpillar and the butterfly. Flor-
ence: Felice de Monier.
Beccaria, C. 1771. Elementi di economiapublica. In Opere,
ed. S. Romagnoli. Florence: Sansoni, 1958. Becker, G.S.
1985. Human capital, effort, and the sexual division of labour.
Journal of Labour Economics 3(1): S33-S58.
Becker, G.S.. and K.M. Murphy. 1992. The division of
labour, coordination costs, and knowledge. Quarterly
Journal of Economics 107: 1137-1160.
Bonar, J. 1893. Philosophy and political economy, 3rd ed.
London: Allen and Unwin, 1967.
Braverman, H. 1974. Labour and monopoly capital: The
degradation of work in the twentieth century. New York:
Monthly Review Press.
Camacho. A. 1996. Division of labour, variability, coordi-
nation and the theory of the firm and markets. Dor-
drecht: Kluwer.
Campbell, R. 1747. The London tradesman. London. Cannan.
E. 1929. Review of economic theory. London: P.S. King &
Son.
Carlyle, T. 1843. Past and present. London: Chapman &
Hall. Another ed. London: G. Routledge & Sons. 1893.
Cigno. A. 1991. Economics of the family. Oxford: Oxford
University Press.
Corsi, M. 1984. Dsistema di fabbrica e la divisione del
lavoro, ilpensiero di Charles Babbage. Quaderni di
storiadell’economiapolitica 3: 111-123.
Dosi. G. 1988. Sources, procedures, and micro-economic
effects of innovation. Journal of Economic Literature 26:
1120-1171.
Division of Labour
Ferguson, A. 1767. An essay on the history of civil society.
Edinburgh.
Frobel, F., J. Heinrichs, and O. Kreye. 1980. The new
international division of labour. Trans. P. Burgess.
Cambridge: Cambridge University Press.
Gershuny, J. 1983. Social innovation and division of labour.
Oxford: Oxford University Press.
Gioja, M. 1815. Nuoveprospettodellescienzeeconomiche.
Milan: G. Pirotta.
Goodman, E., and J. Pamford, eds. 1989. Small firms and
industrial districts in Italy. London: Routledge.
Gouvemeur, J. 1978. (Contemporary capitalism and marx- ist
economics. Trans. R. le Famu. Oxford: Robertson, 1983.
Groenewegen, P. 1977. Adam Smith and the division of
labour: A bi-centenary estimate. Australian Economic
Papers 16: 161- 174.
Harris, J. 1757. An essay upon money and coins: Part I.
London: G. Hawkins.
Hegel, G.W.F. 1821. Philosophy of right. Trans. T.M. Knox.
Oxford: Clarendon Press, 1982.
Hicks, J.R. 1965. Capital and growth. Oxford: Clarendon
Press.
1 limmelweit, S., and S. Mohun. 1977. Domestic labour and
capital. Cambridge Journal of Economics 1: 15-31.
Hobsbawm, E. 2000. The new century. London: Abacus.
Hodgskin. T. 1827. Popular political economy. London New
York: C. Tait A.M. Kelley, 1966.
Hutcheson. F. 1755. A system of moral philosophy. Glasgow.
Hutchison. T.W. 1988. Before Adam Smith: The emergence of
political economy 1662-1776. Oxford: Basil Blackwell.
Kaldor. N. 1966. Causes ofthe slow rate of economic growth
in the United Kingdom. Cambridge: Cambridge
University Press.
Kaldor. N. 1967. Strategic factors in economic development.
Ithaca: State School of Industrial and Labour Relations.
Cornell University.
Kaldor. N. 1972. The irrelevance of equilibrium economics.
Economic Journal 82: 1237 1255.
Kaldor. N. 1975. What is wrong with economic theory?
Quarterly Journal of Economics 89: 347-357.
Kames. H.H. 1774. Sketches of the history of man. Edinburgh
London: W. Creeeh/W. Strahan and
T. Cadell.
Lindsay, P. 1733. The interest of Scotland considered.
Edinburgh: R. Fleming & Co.
McCulloch, J.R. 1825. Principles of political economy.
London: Murray, 1870.
McCulloch, J.R. 1850. Introduction and notes to Adam Smith.
In An inquiry into the nature and causes ofthe wealth of
nations, 4th ed. Edinburgh London: A. & C. Black
Longman.
Division of Labour
Mandeville, B. 1729. The fable of the bees. London: A.
Roberts.
Marglin, S. 1974. What do bosses do? The origins and
functions of hierarchy in capitalist production. Review of
Radical Political Economics 6: 60-112.
Marshall, A. 1890. Principles of economics, 8th ed. London:
Macmillan & Co., 1920.
Marshall, A. 1919. Industry and trade, 4th ed. London:
Macmillan & Co., 1923.
Marshall, A., and M.R Marshall. 1879. The economics of
industry. London: Macmillan and Company.
Marx, K. 1844. Economic and philosophic manuscripts.
Moscow: Foreign Languages Publishing House, 1959.
Marx, K. 1867. Capital, vol. 1. Moscow: Foreign Languages
Publishing House, 1959.
Marx, K., and F. Engels. 1845-6. The German ideology.
Moscow: Progress Publishers, 1964.
Meek, R.L., and A.S. Skinner. 1973. The development of
Adam Smith’s ideas on the division of labour. Economic
Journal 83: 1094-1116.
Mill, J.S. 1821. Elements of political economy. In The
selected economic writings of James Mill, 3rd ed, ed.
D.N. Winch. Edinburgh: Oliver & Boyd for the Scottish
Economic Society, 1966.
Mill, J.S. 1848. Principles of political economy. In The
collected works of John Stuart Mill, vols. 2 and 3, ed.
J.M. Robson. Toronto: Toronto University Press, 1965.
Nicholson, J.S. 1893. Principles of political economy, 2nd ed.
London: A. & C. Black.
Perelman, M. 1984. Classical political economy: primitive
accumulation and the social division of labour. London:
Rowrnan & Allenheld.
Pemin, J.L. 1993. La cooperation entrefirmes:uneapproche
par T economic des conventions. Economic applique 46:
105-126.
Petty, W. 1671. Political arithmetick. In Economic writings of
Sir William Petty, ed. C.H. Hull. New York: A.M. Kelley,
1963.
Petty, W. 1683. Another essay on political arithmetick
concerning the growth of the city of London. In Eco-
nomic writings of Sir William Petty, ed. C.H. Hull. New
York: A.M. Kelley, 1963.
Plato. 380 BC. The republic. Harmondsworth: Penguin
Classics, 1955.
Posner, R.A. 1992. Sex and reason. Cambridge, MA: Harvard
University Press.
Quesnay, F. 1765. Natural right. Extracts. Trans. R.L. Meek.
The economics of physiocracy. London: Allen &Unwin,
1962.
Rae, J. 1834. Statement of some new principles on the subject
of political economy. New York: A.M. Kelley, 1964.
Robbins, L. 1932. An essay on the nature and significance of
economic science. London: Macmillan, 1935.
Rorner, P.M. 1987. New theories of economic growth:
Growth based on increasing returns due to specialization.
American Economic Review 77: 56-62.
3029
Rosenberg, N. 1976. Marx as a student of technology.
Monthly Review 28: 56-77. In Inside the black box:
Technology and economics, ed. N. Rosenberg.
Cambridge: Cambridge University Press, 1982.
Rowthom, R.E. 1975. A reply to LordKaldor’s comment.
Economic Journal 85: 897-901.
Ruskin, J. 1851-1853. The stones of Venice.In The complete
works of John Ruskin, eds. E.T. Cook and A.
Wedderbum. London: George Allen, 1904.
Salz, A. 1934. Specialisation. In International encyclopaedia
of the social sciences, vol. 13, ed. E.R.A. Seligman. New
York: Macmillan Company, 1948.
Scazzieri, R. 1981. Ejficienza, produttivita e livelli di
attitivita. Bologna: II Mulino.
Schiller, F. 1793. On the aesthetic education of man. Trans.
R. Snell. New York: Ungar, 1980.
Schumpeter, J.A. 1954. History of economic analysis.
London: Oxford University Press.
Senior, N. 1836. An outline of the science of political
economy. London: George Allen & Unwin, 1938,1951.
Sidgwick, H. 1883. Principles of political economy, 2nd ed.
London: Macmillan & Co.
Smith, A. 1776. An inquiry into the nature and causes of the
wealth of nations, ed. R.H. Campbell and A.S. Skinner.
Oxford: Clarendon Press, 1976.
Stewart, D. 1858. In Collected works of Sir Dugald Stewart,
ed. Sir William Hamilton. Edinburgh.
Sun, G.Z. 2005. Readings in the economics of the division of
labour. Singapore: World Scientific Publishing
Company.
Taussig, F.W. 1911. Principles of economics, 3rded. New
York: Macmillan, 1936.
Thirlwall, A.P. 1983. A plain man’s guide to Kaldor’s growth
law. Journal of Post Keynesian Economics 5: 345-358.
Torrens, R. 1808. The economists refuted. Sydney: Depart-
ment of Economics, University of Sydney, Reprints of
Economic Classics, 1984.
Tucker, J. 1755. The elements of commerce and the theoiy of
taxes. London.
Tucker, J. 1774. Four tracts on political and economic
subjects, 2nd ed. Gloucester: R. Raikes.
Turgot, A.R.J. 1751. Lettre a Madame de Graffignysur les
letters d’un Peruvienne. In Oeuvres de Turgot et docu-
ments le concernant, vol. 1, ed. G. Schcllc. Paris: F.
Alcan, 1913.
Turgot, A.R.J. 1766. Reflections on the production and distri-
bution of wealth. In The economics ofA.RJ. Turgot, ed.
P.- D. Groenewegen. The Hague: Nijhoff 1977.
Ure, A. 1835. The philosophy of manufactures. London: C.
Knight/Frank Cass, 1967.
Verdoom, P.J. 1949. Fattori che regolano lo sviluppo della
produttivita del lavoro. L’industria 1: 45-53.
West, E.G. 1964. Adam Smith’s two views on the division of
labour. Economica 31: 23-32.
Yang, X.-K., and Y.-K. Ng. 1993. Specialisation and economic
integration: A new classical micro-economic framework.
New York: Elsevier Science.
Young, A. 1928. Increasing returns and economic progress.
Economic Journal 38: 327-342.
3030
Dmitriev, Vladimir Karpovich (1868-1913)
D. M. Nuti
Vladimir Karpovich Dmitriev was the first Russian
mathematical economist. His Economic Essays on
Value, Competition and Utility ( , production prices gained early recognition
; English edition ) are a classic text in
economic literature.
Vladimir Karpovich Dmitriev was bom on 24
November 1868 on the Rai Estate in Smolensk
Gubernia, Smolensk Uezd. On completing his
classical education at the Tula Classical Gymna-
sium he went to Moscow University to study
medicine but subsequently transferred to the Law
Faculty where he began his studies in Political
Economy. After graduation in 1896 he married T.A.
Vatatsi and left to take the post of excise controller
in the small town of Von’kovitsy in Podol’sk
Gubernia. He served there for three years but
contracted lung tuberculosis and had to leave the
service. He was in great need all his life and his
chronic illness eventually aggravated and caused his
death on 30 November 1913.
Dmitriev’s First Essay on The theory of value
ofD. Ricardo was published in 1898, followed in
1902 by a Second Essay on The competition theory
of A. Cournot and a Third Essay on The theory of
marginal utility (published together and with a
Conclusion); the three essays were reprinted and
issued together in 1904. He also published a large
volume on the consumption of alcohol in Russia
(with an introduction by P.V. Strove) in 1911 and
half a dozen articles on the same topics as his
books. He was planning at least three further Essays
on rent, on industrial crises and on monetary
circulation, which apparently were never written or
at any rate published.
Dmitriev’s contributions to economic theory
include: (i) the development of an input-output
method for the determination of the quantity of
labour directly and indirectly embodied in com-
modities; (ii) a theory of production prices based on
dated labour, similar to that of Piero Sraffa; (iii)
Dmitriev, Vladimir Karpovich (1868-1913)
a statement of ‘wage-profit frontier’ derived from
technology and alternative level of real wage; (iv) a
theory of non-productive costs in competition
between firms. While these contributions do not
amount quite to the ‘organic synthesis of the labour
theory of value and the theory of marginal utility’
promised in the title page of the 1902 and 1904
editions of the Essays, they are highly original and
remarkable in their anticipation of subsequent work.
Dmitriev’s propositions on labour values and
(Chuprov 1905; Bortkiewicz , who praised and used
extensively ‘this remarkable work’; Struve 1908,
who hailed Dmitriev as a ‘logically and
mathematically thought-out Ricardo’; and
Shaposhnikov’s memorial lecture a year after his
death, ). Until shortly after the October revolution
Dmitriev was widely mentioned in Russian
economic literature, then he was entirely forgotten
until the Soviet school of mathematical economists
brought him out of his official oblivion circa 1960
(Nemchinov ; Belkin, Grobman, Lunts, in
Aganbegyan and Belkin 1961) attracting the
attention of Western scholars (Nove and Zauberman
; Zauberman )•
In his first Essay Dmitriev considers the ques-
tion ‘how is it possible to calculate the amount of
labour expended for the production of a given
economic good from the very beginning of history,
when man managed without capital, down to the
present time’ (p. 43 of the English edition, to which
all page references are made here). He answers that
there is no need for ‘historical digressions’ ofthis
kind; the quantity of labour AC* which goes
directly and indirectly into the production of
commodity A is expressed by the equation
N a = nA H--------H-----------No + ■ ■ ■ H------Nm (1)
m\ mo mM
where nA is the direct labour input of a unit of
commodity^: 1/mj is the amount of the i th com-
modity used up in the production of commodity A,
where i = 1, 2, ..., M; and N, is the labour directly
and indirectly embodied in the /th commodity (this
is Eq. in the First Essay, p. 44). The coefficient 1
lm\ here is to be interpreted either as
Dmitriev, Vladimir Karpovich (1868-1913)
the intermediate inputs requirement for the pro-
duction of the A commodity, or as the straight-line
amortization of the i th fixed capital good (assuming
uniform productiveness over its lifetime); some of
these coefficients may be equal to zero, as in
Dmitriev’s system of Eq. in the First Essay. For
each of the Mother commodities there is an
equation of the same form, relating labour (directly
and indirectly) embodied to input coefficients and
the labour embodied in the inputs (p. 44). We obtain
a system of (M + 1) equations in (M + 1) unknowns,
which is always adequate for the determination of N,
giving the required sum of the labour expended on
the production of product A. Therefore, without any
digressions into the prehistoric times of the first
inception of technical capital, we can always find the
total sum of the labour directly and indirectly
expended on the production of any product under
present day production conditions, both of this
product itself and of those capital goods involved in
its production (p. 44, emphasis in the text).
This is clearly a full-fledged input-output system,
where N, are the lull coefficients of labour, the N)
are the direct labour inputs, and the l/m are identical
with Leontief’s input-output coefficients. The
analytical apparatus provided by Leontief four
decades later adds two things: (i) a method for the
actual computation of the solution, namely the
inversion of the matrix (I-A'), where I is the identity
matrix and A' is the transpose of the matrix of
technical coefficients; and (ii) the generalization of
notion of full input (i.e. direct and indirect input
requirements) from labour to other production
inputs. In Leontief’s type of notation, if we call the
amount of i th product required per unit of the j th
product, A the [af\ matrix: a„j the direct labour input
of product j, and a the column vector f«OJ]; and /y
the full-input coefficient, i.e. the element of the (I-
A!) 1 matrix, we obtain
n
./'.k X! "v/* +
;=i
where i, k,j = 1 , 2 , . . . , a n d fa is Kronecker’s
delta, i.e. is equal to zero except for / k when it is
equal to unity. If we indicate full labour inputs
3031
(i.e. Dmitriev’s Ns) by /ok Leontief’s approach gives
n
/ok = a l
° f}k
j=
1 or
f0 = (I-A/)-1a (4)
where F0 = [/<)k|. Dmitriev’s formulation of full
labour inputs is
n
/ok = «ok + 53/ojtfjk (3')
1=1
or
f0 = a + A'f0 (4')
which is just another way of rewriting Leontief s
Eq. .
The importance of Dmitriev’s approach for
socialist planning was already understood in the
1920s; A.V. Chayanov (1926) developed
Dmitriev’s scheme into an input-output table for
agriculture. In the 1960s the ability to claim Russian
priority in the d iscovery of input-output equations
in the work of Dmitriev was an important step in
the struggle for the use of mathematical methods in
socialist planning. In 1962 the Central Statistical
Administration produced an 83 x 83 intersectoral
balance of labour outlays in the Soviet economy for
1959-60, using the first ex-post input-output tables
for the Soviet economy, compiled for 1959. This
balance shows, in terms of labour, the inter-
industrial flows, the formation of the final bill of
goods, the formation of national product and cost
incurred in the non-productive sphere (see
Eidel’man ; Zauberman ). This calculation
corresponds exactly to the Dmitriev-Leontief full
(2) labour coefficients.
Dmitriev also had a theory ofprices of produc-
tion which is a reformulation and development of
Ricardian price theory and corresponds to Marxian
production prices. Dmitriev starts from the
3032
refutation of the criticism levied in his time (for
instance by Walras) against the ‘classical’ theory of
price determination based on production costs, ‘that
it defines price from prices, that it defines one
unknown from other unknowns’ (p. 41). This
allegation, Dmitriev argues, can be levied against
Adam Smith, who did not deal with the problem of
the determination of the profit rate, except for a
vague reference to the demand for and supply of
capital, i.e. going outside the sphere of production.
But Ricardo is not subject to this criticism; indeed
‘The most important point in Ricardo’s theory is
undoubtedly his theory of the conditions defining
the “average” profit rate ...’ and ‘Ricardo’s immortal
contribution was his brilliant solution of this
seemingly insoluble problem’ (pp. 50 and 58, First
Essay).
For the study of p r i c e s (or v a l u e s , in his terminology)
Dmitriev uses a framework slightly different from
that employed for the study of l a b o u r v a l u e s (or labour
embodied in commodities). Instead of extending his
p o i n t i n p u t - p o i n t o u t p u t framework, whereby commodities are
produced by means of labour and other commodities
(Eq. ), he uses an Austrian-type model where
commodities are produced by dated labour, i.e. a f l o w
input-point output framework, whereby commodities are
produced by dated labour. For each commodity
Dmitriev formulates a price equation of the type:
XA = nAaXa( 1 + r) + niaXa(\ + r) + L L + n a X f l 1 + r )
m
where X is the price of commodity A , a is the amount
A
of wage good (say, com) consumed by workers, X is
the unit price of the wage good; nA, nx,..., nm are the
labour inputs required respectively t , t \ , ..., /Am time
A A
units before the output of commodity A becomes
available (this is Eq. 25, p. 54). If there are M
commodities in addition to the wage good, we have
(M + 1) equations; there are M relative prices to be
determined, in terms of an arbitrary commodity
whose price is taken as unit of account, plus the
profit rate; the system is complete and can
simultaneously determine relative price and the
profit rate.
Dmitriev, Vladimir Karpovich (1868-1913)
It is to Ricardo’s credit that he was the first to note
that there is one production equation by means of
which we may determine the magnitude of r directly
(i.e. without having recourse for assistance to the
other equations). This equation gives us the produc-
tion conditions of the product a to which in the final
analysis the expenditure on all the products, A, B,C,
..., is reduced (p. 59).
For the wage good, with labour inputs N],
= 1 + + JVt(l + • • • +Aq(l +
Xa [Na( rfl rfl + r)'fl.
(6)
From this (Eq. 44, First Essay) we can obtain
1
I>‘'(1+r)' (7)
which today is familiar as the ‘wage-profit frontier’:
Dmitriev writes it instead in the implicit form
r = F(Na,Nu ... ,Nq; ta, ta l , . . . , taq; a).
(8)
Dmitriev then extends this analysis to the case
where workers consume not a single commodity but
a number of commodities in fixed proportions. The
condition for a positive profit rate to arise is that
‘we can obtain a l a r g e r quantity of the same product
(5) within some finite period of time as a result of the
production process’ (p. 62).
Dmitriev, in sum, considers ‘production of
a
commodities by means of dated labour’, not
‘production of commodities by means of com-
modities’ (at least when discussing the determi-
nation of the profit rate), with wages being
advanced, not ‘posticipated’ as in Sraffa ( ).
Their similarity descends from the common
Ricardian root. Although Dmitriev’s approach is
close to Marx, he goes out of his way to d e n y the
Marxian theory of exploitation and to show,
‘proceeding from Ricardo’s analysis, that the origin
of industrial profit does not stand in any “special”
relationship to the human labour used in production’
(p. 64), In order to do this, Dmitriev investigates the
properties of an imaginary system where work is
performed exclusively by animals and machines.
The conditions
Dmitriev, Vladimir Karpovich (1868-1913)
for a positive profit rate are shown to be quite
general; however, the fact that we do not usually
talk of ‘exploitation’ of animals and machines does
not in any conceivable sense rule out the proposition
of human exploitation when human labour is
actually used in production.
Having formulated and developed Ricardian
propositions on prices of production Dmitriev
proceeds to show that these propositions hold only
under the most restrictive assumptions. Among
these are constant returns to scale, i.e. zero rents,
and perfect competition of a kind that brings prices
down to the (constant) necessary costs of
commodities (including profit at a rate determined
by technology and the real wage). He decidedly
parts company from Ricardo and shows that
whenever at least one of these conditions is not
satisfied prices depend on demand conditions as
well, and not even ‘long-run’ equilibrium prices can
be obtained purely from the knowledge of
technology and the real wage.
Already at the end of the First Essay, Dmitriev
shows that a demand price based exclusively on
production conditions cannot handle the cases of
monopoly prices and of positive rent. But the
greatest blow to the Ricardian theory of price
determination is given in the Second Essay, where
Dmitriev most emphatically argues that demand
conditions contribute to price determination also for
‘goods which are infinitely reproducible by labour
under conditions excluding the possibility of the
occurrence of rent’ (p. 92) even under competitive
conditions. In order to do this, Dmitriev challenges
the proposition that ‘competition lowers prices’ (p.
93) and starting from Cournot’s analysis of
competition he constmcts a theory of unrestricted
but not-so-perfect competition.
Dmitriev argues that the assumption that supply
= production contradicts not only economic reality,
but also the other basic hypothesis of competitive
analysis, ‘that every individual tends to pursue the
greatest advantage’ (p. 118). He relaxes the
assumption to allow for stocks and unused capacity,
representing potential supply. Dmitriev postulates
that for a given volume of production rational
behaviour of producers leads
3033
them to a tacit collusion on price, i.e. joint profit
maximisation as in the monopoly case, but (i) such
collusion is enforceable only because of the
existence of a potential threat in the form of a
potential supply greater than the collusion sales
level, and (ii) competition between producers takes
the form of expanding the level of potential supply,
with sales lagging behind. For a given number n of
producers there is an equilibrium potential supply
such that the price corresponds to what would be
charged by a monopolist. For n tending to infinity,
the cost of the potential supply tends to equal the
revenue from actual sales; profit (over and above the
interest component of production costs) is zero, as in
the customary competitive equilibrium, not because
price is equal to the necessary production cost of the
output sold, but because the additional cost of hold-
ing stocks or installing unused capacity brings the
total cost of potential output up to the level of actual
sales revenue and wipes out profits completely (p.
134).
A further instance of unproductive expenditure
is mentioned by Dmitriev in his Conclusion, namely
‘advertising’ to expand sales of an individual
entrepreneur when the total sales level remains the
same. In a notable passage Dmitriev compares the
role of commodity stocks with the strategy
of‘intensified armament ofthe Powers in peace time’
(pp. 148-9). It follows from this analysis that
unrestricted competition has a cost for the economy,
i.e. a social cost of wasted output, excess
inventories, unused capacity or redundant
advertising. This is only partly compensated by
consumers’ gain from prices lower than monopoly
prices.
A most important implication of Dmitriev’s
analysis is his account of the economic conse-
quences of technical progress (Section 7), which
raises the level of potential supply at which the
temporary profit, obtained by individual producers
breaking their tacit price-collusion, disappears.
‘Therefore an expansion of output following a
reduction of production costs will, in general,
extend not only to an expansion of supply but also to
an increase in excess commodity inventories (p.
171). The building up of excess commodity
inventories following technical progress
3034
gives rise to fluctuations in the levels of output
capacity, capacity utilization, and inventory levels
(pp. 173—8 ). When technical progress takes place,
‘over-production’ periodically occurs, and this ‘ i s i n n o
s e ns e a result of errors of economic
consequence of the inability of production to adapt
to excessively variable demand ... but is a direct
result of the straggle of competing entrepreneurs, e a c h
o f w h o m is m o t i v a t e d i n h is o w n a c t i o n s b y q u i te c o r r e c t e c o n o m i c j u d g e m e n t ’ (p.
117).
The only way of eliminating wasted output,
excess inventories and unused capacity, and the
non-productive costs which these involve, is the
establishment of forward markets ( T e r m i n h a n d e l ):
‘forward contracts make nonproductive “reserve
stocks” unnecessary since they make it possible to
sell goods which have still not been produced but
merely can he produced ...’ (p. 178, footnote 1).
Dmitriev relegates this qualification to a footnote,
but this is really a central point in his argument,
except that - we now know - forward markets would
have to be not only complete but also exclusive (i.e.
no future spot markets could reopen), which is
neither practical nor advisable.
For Dmitriev the short-run equilibrium of an
economic system is determined by the given levels
of supply and the demand functions. He concludes
that if prices of commodities happen to coincide
with their necessary reproduction costs, actual
prices will correspond to the solution of the
Walrasian system. But if the supply level of a
commodity is such that its price exceeds its
necessary reproduction costs, the question of the
distribution of the extra-normal profit lies, for
Dmitriev, ‘outside the sphere of economic
research’, because it is the result of a ‘straggle’ and
is taken as a question of fact by economic theory.
There may be ‘a general sociological solution’ (p.
207); ‘Otherwise we should have to admit that the
question cannot have any general solution at all’ ( i b i d ) .
Ultimately, price theory becomes the theory of the
self-defeating attempts, by economic agents, to gain
from a social struggle which is rational by the
standards of individuals though not of society, and
the theory of the ensuing waste and fluctuations.
Dmitriev, Vladimir Karpovich (1868-1913)
Bibliography
Belkin, V.D. 1961. Natsionalnyi dokhod i mezhotraslevoy
balans | National income and intersectoral balance]. In
Primenenie matematiki i elektronnoy tekhnihi v
judge ment, i.e. it is not a planirovanii [The use of mathematics and electronic
techniques in planning], ed. A.G. Aganbegyan and V.D.
Belkin. Moscow.
Bortkiewicz von I.. 1906. Wertrechnung und Pre- isrechnung
ini Marxschen Sistem, (in three parts). Archiv fur
Sozialwissenschaft und Sozialpolitik 23(1) 1907, 25(1);
1907, 25(2). The second and the third parts are translated
into English, as ‘Value and Price in (lie Marxian System’,
International Economic Papers, 1952, no. 2.
EideTman, M.R. 1962. Pervyi mezhotraslevoi balans zatrat
truda v narodnom khoziaistve SSSR [The first
intersectoral balance of labour expenditures in the
national economy of the USSR]. Vestnik Statistiki, no.
HI, 1962.
Leontief, WAV'. 1941. The structure of the American econ-
omy 1919 1939. New York: Oxford University Press.
Leontief, WAV'., et al. 1953. Studies in the structure of the
American economy: Theoretical and empirical explo-
rations in input-output analysis. White Plains: Interna-
tional Arts and Science Press.
Nemchinov, V.S. 1959. The use of mathematical methods in
economics. In The use of mathematics in economics
(English trans: Nove, A.)., ed. V.S. Nemchinov Moscow.
London. 1964.
Nemchinov. V.S. 1961. A model of an economic region.
Moscow. Trans. Mathematical studies in economics and
statistics in the USSR and Eastern Europe, c ol. 1. 1964.
p. 14.
Nemchinov. V.S. 1963. Basic elements of a model of planned
price formation. Voprosy Ekonomiki, no. 12. Socialist
economics (trans: Nove. A.. Nuti. D.M.).
Harmondsworth: Penguin. 1972.
Nove. A.. and A. Zauberman. 1961. A resurrected Russian
economist of 1900. Soviet Studies 13: 96-101.
Shaposhnikov. N.N. 1914. Pervyi Russkii ekonomist-
matematik Vladimir Karpovich Dmitriev, Doklad v
posvvashchennom pamyati Dmitrieva zasedanii O-va im.
A.I. Chuprova [The first Russian mathematical economist
V.K. Dmitriev, a lecture as a meeting of the A.I. Chuprov
Society, held in memory of Dmitriev], Moscow.
Sraffa, P. 1960. Production of commodities by means of
commodities. Cambridge: Cambridge University Press.
Zauberman, A. 1962. A few remarks on a discovery in Soviet
economics. Bulletin of the Oxford Institute of Economics
and Statistics 24: 437-445.
Zauberman, A. 1963. A note on the Soviet inter-industry
labour input balance. Soviet Studies 15: 53-57.
Selected Works
Dmitriev, Vladimir Karpovich. 1898. Ekonomicheskie
Ocherki, Vyp, I, ‘Teoriya tsennosti D. Ricardo (opyf
tochnago analyzaf [Economic Essays, Issue I, ‘The
Dobb, Maurice Herbert (1900-1976)
theory of value of D. Ricardo, an attempt at a rigorous
analysis’]. Moscow.
Dmitriev, Vladimir Karpovich. 1902. Ekonomicheskie
Ocherki, ‘Chast’ 1-aya (opyt’ organicheskago sinteza
trudovoi teorii tsennosti i teorii predeTnoi poleznosti)’,
Vypuski 2-i i 3-i. Ocherk 2-i: ‘Teoriya konkurrentsii Og.
Kumo (Velikago “zabytago” ekonomista)’. Ocherk 3-i:
‘Teoriya predeTnoi poleznosti’ [Economic Essays, Part I,
Attempt at an organic synthesis of the labour theory of
value and the theory of marginal utility, Issues 2 and 3.
Second Essay: The theory of competition of A. Cournot
(the great ‘forgotten’ economist). Third Essay: The theory
of marginal utility], Moscow.
Dmitriev, Vladimir Karpovich. 1904. Ekonomicheskie
Ocherki (Seriya I-aya: ‘opyt’ organicheskago sinteza
trudovoi toerii tsennosti i teorii predeTnoi poleznosti’)
| Economic Essays, First Series: Attempt at an organic
synthesis of the labour theory of value and the theory of
marginal utility]. Moscow.
Dmitriev, Vladimir Karpovich. 1911. Kriticheskie
izsledovaniya o potreblenii alkogolya v Rossii, s pre- disl.
P.B. Struve, Issledovaniya i raboty po polit. ekonomii i
obshchestv. znaniyam, izd. pod red. P.B. Struve, Vyp. I
[Critical Studies on the consumption of alcohol in Russia,
with an introduction by P.V. Struve: Studies and works in
political economy and social sciences, edited by P.V.
Struve, Issue I]. Moscow. (With an English translation of
the table of contents).
Dmitriev, Vladimir Karpovich. 1974. Economic essays on
value, competition and utility. Edited with an introduction
by D.M. Nuti. Cambridge: Cambridge University Press.
Dobb, Maurice Herbert (1900-1976)
Amartya Sen
Keywords
Dobb, M.; Economic development; Equality;
Exploitation; Feudalism; Labour theory of
value; Langer-Lemer price mechanism; Market
socialism; Marxism; Planning; Revealed
preference theory; Saving and investment;
Socialist pricing theory; Sraffa, P.; Sweezy, P.
M.; Utility theory of value
JEL Classifications
B31
3035
Maurice Dobb was undoubtedly one of the out-
standing political economists of this century. Ele
was a Marxist, and was one of the most creative
contributors to Marxian economics. As Ronald
Meek put it, in his obituary of Dobb for the British
Academy, ‘over a period of fifty years [Dobb]
established and maintained his position as one of the
most eminent Marxist economists in the world’.
Dobb’s Political Economy and Capitalism (1937)
and Studies in the Development of Capitalism
(1946) are his two most outstanding contributions to
Marxian economics. The former is primarily
concerned with economic theory (including such
subjects as value theory, economic crises,
imperialism, socialist economies), and the latter
with economic history (particularly the emergence
of capitalism from feudalism). These two fields -
economic theory and economic history - were
intimately connected in Dobb’s approach to
economics. He also wrote an influential book on
Soviet economic development. This was first
published under the title Russian Economic
Development since the Revolution (1928), and later
in a revised edition as Soviet Economic
Development since 1917 (1948).
Maurice Dobb was bom on 24 July 1900 in
London. His father Walter Herbert Dobb had a
draper’s retail business and his mother Elsie Annie
Moir came from a Scottish merchant’s family. He
was educated at Charterhouse, and then at Pembroke
College, Cambridge, where he studied economics.
This was followed by two postgraduate years at the
London School of Economics, where he did his
Ph.D. on ‘The Entrepreneur’. The thesis formed the
basis of his book Capitalist Enterprise and Social
Progress (1925). Dobb returned to Cambridge at the
end of 1924 on being appointed as a lecturer in
economics. He taught in Cambridge until his
retirement in 1967. He was a Fellow of Trinity
College, and was elected to a University Readership
in 1959. He received honorary degrees from the
Charles University of Prague, the University of
Budapest, and Leicester University, and was elected
a Fellow of the British Academy. After retirement
he and Ids wife, Barbara, stayed on in the
neighbouring village of Fulboum. He died on 17
August 1976.
3036
Dobb was a theorist of great originality and
reach. He was also, throughout his life, deeply
concerned with economic policy and planning. His
foundational critique of ‘market socialism’ as
developed by Oscar Lange and Abba Lemer,
appeared in the Economic Journal of 1933, later
reproduced along with a number of related contri-
butions in his On Economic Theory and Socialism
(1955). His relatively elementary book Wages
(1928) presented not merely a simple introduction
to labour economics, but also an alternative outlook
on these questions, including their policy
implications, leading to interesting disputations with
John Hicks, among others. In later years Dobb was
much concerned with planning for economic
development. In three lectures delivered at the Delhi
School of Economics, later published as Some
Aspects of Economic Development (1951), Dobb
discussed some of the central issues of development
planning for an economy with unemployed or
underutilized labour, and his ideas were more
extensively developed in his later book, An Essay
on Economic Growth and Planning (1960).
Maurice Dobb also published a number of
papers on more traditional fields in economic
theory, including welfare economics, and some of
these papers were collected together in his Welfare
Economics and the Economics of Socialism (1969).
In his Theories of Value and Distribution since
Adam Smith: Ideology and Economic Theory
(1973), he responded inter alia to the new
developments in Cambridge political economy,
including the influential ‘Prelude to a Critique of
Economic Theory’ by Piero Sraffa ( ). Mau
rice Dobb’s association with Piero Sraffa extended
over a long period, both as a colleague at Trinity
College, and also as a collaborator in editing Works
and Correspondence of David Ricardo, published in
11 volumes between 1951 and 1973 (on the latter,
see Pollitt ).
In addition to academic writings, Maurice Dobb
also did a good deal of popular writing, both for
workers’ education and for general public
discussion. He wrote a number of pamphlets,
including The Development of Modern Capitalism
(1922), Money and Prices (1924), An Outline of
European Histoiy (1926), Modern Capitalism
Dobb, Maurice Herbert (1900-1976)
(1927), On Marxism Today (1932), Planning and
Capitalism (1937), Soviet Planning and Labour in
Peace and War (1942), Marx as an Economist: An
Essay (1943), Capitalism Yesterday and Today
(1958), and Economic Growth and Underdeveloped
Countries (1963), and many others. Dobb was a
superb communicator, and the nature of his own
research was much influenced by policy debates and
public discussions. Dobb the economist was not
only close to Dobb the historian, but also in constant
company of Dobb the member of the public. It
would be difficult to find another economist who
could match Dobb in his extraordinary combination
of genuinely ‘high-brow’ theory, on the one hand,
and popular writing on the other. The author of
Political Economy and Capitalism (from the
appearance of which - as Ronald Meek ( )
rightly notes - ‘that future historians
of economic thought will probably date the emer-
gence of Marxist economics as a really serious
economic discipline’: was also spending a good deal
of effort writing pamphlets and material for labour
education, and doing straightforward journalism. It
is not possible to appreciate fully Maurice Dobb’s
contributions to economics without taking note of
his views of the role of economics in public
discussions and debates.
Another interesting issue in understanding
Dobb’s approach to economics concerns his
adherence to the labour theory of value. The labour
theory has been under attack not only from
neoclassical economists, but also from such anti-
neoclassical political economists as Joan Robinson
and, indirectly, even Piero Sraffa. In his last major
work, Theories of Value and Distribution since
Adam Smith (1973), Maurice Dobb speaks much in
support of the relevance of Sraffa’s ( ) major
contribution, which
eschews the use of labour values (on this see
Steedman ), but without abandoning his insistence
on the importance of the labour theory of value. It is
easy to think that there is some inconsistency here,
and it is tempting to trace the origin of this alleged
inconsistency to Dobb’s earlier writings, which
made Abram Bergson remark that ‘in Dobb’s
analysis the labour theory is not so much an analytic
tool as excess baggage’ (Bergson , p. 445).
Dobb, Maurice Herbert (1900-1976)
The key to understanding Dobb’s attitude to the
labour theory of value is to recognize that he did not
see it just as an intermediate product in explaining
relative prices and distributions. He took ‘the
labour-principle’ as ‘making an important
qualitative statement about the nature of the
economic problem’ (Dobb 1937, p. 21). He rejected
seeing the labour theory of value as simply a ‘first
approximation’ containing ‘nothing essential that
cannot be expressed equally well and easily in other
terms’ (Dobb 1973, pp. 148-9). The description of
the production process in terms of labour
involvement has an interest that extends far beyond
the role of the labour value magnitudes in providing
a ‘first approximation’ for relative prices. As Dobb
(1973, pp. 148-9) put it,
there is something in the first approximation that is
lacking in later approximations or cannot be
expressed so easily in those terms (e.g., the first
approximation may be a device for emphasising and
throwing into relief something of greater generality
and less particularity).
Any description of reality involves some selec-
tion of facts to emphasize certain features and to
underplay others, and the labour theory of value
was seen by Dobb as emphasizing the role of those
who are involved in ‘personal participation in the
process ofproductionper se’ in contrast with those reach to his own Marxist theorizing. He could
who do not have such personal involvement.
As such ‘exploitation’ is neither something ‘meta-
physical’ nor simply an ethical judgement (still less communicate with economists belonging to other
‘just a noise’ ) as has sometimes been depicted: it is schools. Dobb’s honesty and lack of dogmatism
a factual description of a socio-economic relation-
ship, as much as is Marc Bloch’s apt characterisation
of Feudalism as a system where feudal Lords ‘lived
on the labour of other men’. (Dobb 1973, p. 145.
The possibility of calculating prices without
going through value magnitudes, and the greater
efficiency of doing that (on this see Steedman ),
does not affect this descriptive relevance of the
labour theory of value in any way. Maurice Dobb
also outlined the relationship of this primarily
descriptive interpretation of labour theory of value
with evaluative questions, for example, assessing by Maurice Dobb’s willingness to engage in spirited
the ‘right of ownership’ (see especially Dobb 1937). but courteous debates with economists of other
3037
The importance for Dobb of descriptive rele-
vance is brought out also by his complex attitude to
the utility theory of value. While he rejected the
view that the utility picture is the best way of seeing
relative values (‘by taking as its foundation a fact of
individual consciousness’), he lamented the
descriptive impoverishment that is brought about by
replacing the subjective utility theory by the
‘revealed preference’ approach.
If all that is postulated is simply that men choose,
without anything being stated even as to how they
choose or what governs their choice, it would seem
impossible for economics to provide us with any
more than a sort of algebra of human choice. (Dobb
1937, p. 171.
Indeed, as early as 1929, a long time before the
‘revealed preference theory’ was formally inau-
gurated by Paul Samuelson, Dobb (1929, p. 32) had
warned:
Actually the whole tendency of modem theory is to
abandon such psychological conceptions: to make
utility and disutility coincident with observed offers
in the market; to abandon a ‘theory of value’ in
pursuit of a ‘theory of price’. But this is to surrender,
not to solve the problem.
Maurice Dobb’s open-minded attitude to non-
Marxian traditions in economics added strength and
combine Marxist reasoning and methodology with
other traditions, and he was eager to be able to
were important for the development of the Marxist
economic tradition in the English- speaking world,
because he occupied a unique position in Marxist
thinking in Britain. As Eric Hobsbawm ( , p.
1) has noted,
for several generations (as these are measured in the
brief lives of students) he was not just the only
Marxist economist in a British university of whom
most people had heard, but virtually the only don
known as a communist to the wider world.
The Marxist economic tradition was well served
schools. Dobb achieved this without compromising
the integrity of his position. The distinctly Marxist
quality of his economic writings was as important
3038
as his willingness to listen and dispassionately ana-
lyse die claims of other schools of thought with
which he engaged in systematic disputation. The
gentleness of Dobb’s style of disputation arose from
strength rather than from weakness.
Dobb’s willingness to appreciate positive ele-
ments in other economic traditions while retaining
the distinctive qualities of his own approach is
brought out very clearly also in his truly far-
reaching critique of the theory of socialist pricing as
presented by Lange, Lemer, Dickinson and others in
the 1930s. Dobb noted the efficiency advantages of
a price mechanism, especially in a static context. He
was, however, one of the first economists to analyse
clearly the conflict between the demands of
efficiency expressed in the equilibrium conditions
of the Langer-Lemer price mechanism (and also of
course in a perfectly competitive market
equilibrium), and the demands that would be
imposed by the requirements of equality, given the
initial conditions. In his paper called ‘Economic
Theory and the Problems of a Socialist Economy’
published in 1933, Maurice Dobb argued thus:
If carpenters are scarcer or more costly to train than
scavengers, the market will place a higher value
upon their services, and carpenters will derive a
higher income and have greater ‘voting power’ as
consumers. On the side of supply the extra ‘costli-
ness’ of carpenters will receive expression, but only
at the expense of giving carpenters a differential
‘pull’ as consumers, and hence vitiating the index of
demand. On the other hand, if carpenters and
scavengers are to be given equal weight as con-
sumers by assuring them equal incomes, then the
extra costliness of carpenters will find no expression
in costs of production. Here is the central dilemma.
Precisely because consumers are also producers, both
costs and needs are precluded from receiving
simultaneous expression in the same system of
market valuations. Precisely to the extent that market
valuations are rendered adequate in one direction
they lose significance in the other. (1933, p. 37)
The fact that given an initial distribution of
resources the demands of efficiency and those of
equity may - and typically will - conflict is, of
course, one of the major issues in the theory of
resource allocation, with implications for market
socialism as well as for competitive markets in a
private ownership economy. As a matter of fact,
Dobb, Maurice Herbert (1900-1976)
Marx had inter alia noted this conflict in his
Critique of Gotha Programme, but in the discussion
centring around Langer-Lemer systems, this deep
conflict had attracted relatively little attention,
except in the arguments presented by Maurice
Dobb. The fact that even a socialist economy has to
cope with inequalities of initial resource distribution
(arising from, among other things, differences in
inherited talents and acquired skills) makes it a
relevant question for a socialist economy as well as
for competitive market economies, and Dobb’s was
one of the first clear analyses of this central
question of resource allocation.
The second respect in which Maurice Dobb
found the literature on market socialism inadequate
concerns allocation over time. In discussing the
achievements and failures of the market mechanism,
Maurice Dobb argued that the planning of
investment decisions
may contribute much more to human welfare than
could the most perfect micro-economic adjustment,
of which the market (if it worked like the textbooks,
at least, and there were no income-inequalities) is
admittedly more fitted in most cases to take care.
(Dobb 1960, p. 76)
In his book An Essay in Economic Growth and
Planning (1960), Dobb provided a major investi-
gation of the basis of planned investment decisions,
covering overall investment rates, sectoral divisions,
choice of techniques, and pricing policies related to
allocation (including that over time).
This contribution of Dobb relates closely to his
analysis of the problems of economic development.
In his earlier book Some Aspects of Economic
Development (1951), Dobb had already presented a
pioneering analysis of the problem of economic
development in a surplus-labour economy, with
shortage of capital and of many skills. While, on tiie
one hand, he anticipated W.A. Lewis’s ( )
more well-known investigation of economic growth
with ‘unlimited supplies of labour’, he also went on
to demonstrate the far-reaching implications of the
over-all savings rates being socially sub-optimal
and inadequate. Briefly, he showed that this requires
not only policies directly aimed at raising the rates
of saving and investment, but it also has
implications for the choice of techniques, sectoral
balances, and price fixation.
Dobb, Maurice Herbert (1900-1976)
In such a brief note, it is not possible to do
justice to the enormous range of Maurice Dobb’s
contributions to economic theory, applied eco-
nomics and economic history. Different authors
influenced by Maurice Dobb have emphasized
different aspects of his many-sided works (see, for
example, Feinstein, 1967, and the Cambridge
Journal of Economics ’ Maurice Dobb Memorial
Issue (1978)). He has also had influence even
outside professional economics, particularly in
history, especially through his analysis of the
development of capitalism.
Dobb argued that the decline of feudalism was
caused primarily by ‘the inefficiency of Feudalism
as a system of production, coupled with the growing
needs of the ruling class for revenue’ (1946, p. 42 ).
This view of feudal decline, with its emphasis on
internal pressures, became the subject of a lively
debate in the early 1950s. An alternative position,
forcefully presented by Paul Sweezy in particular,
emphasized some external developments, especially
the growth of trade, operating through the relations
between the feudal countryside and the towns that
developed on its periphery No matter what view is
taken as to ‘who won’ the debates on the transition
from feudalism to capitalism, Dobb’s creative role
in opening up a central question in economic history
as well as a major issue in Marxist political
economy can scarcely be disputed. Indeed, Studies
in the Development of Capitalism (1946) has been a
prime mover in the emergence of the powerful
Marxian tradition of economic history in the
English-speaking world, which has produced
scholars of the eminence of Christopher Hill,
Rodney Hilton, Eric Hobsbawm, Edward Thompson
and others.
It is worth emphasizing that aside from the
explicit contributions made by Maurice Dobb to
economic history, he also did use a historical
approach to economic analysis in general. Maurice
Dobb’s deep involvement in descriptive richness (as
exemplified by his analysis of ‘the requirements of a
theory of value’), his insistence on not neglecting
the long-mn features of resource allocation
(influencing his work on planning as well as
development), his concern with observed
phenomena in slumps and depressions in examining
theories of ‘crises’, and so on, all relate to the
3039
historian’s perspective. Dobb’s works in the
apparently divergent areas of economic theory,
applied economics and economic history are, in
fact, quite closely related to each other.
Maurice Dobb was not only a major bridge-
builder between Marxist and non-Marxist economic
traditions (aside from pioneering the development
of Marxist economics in Britain and to some extent
in the entire English-speaking world): he also built
many bridges between the different pursuits of
economic theorists, applied economists and
economic historians. Dobb’s political economy
involved the rejection of the narrowly economic as
well as the narrowly doctrinaire. He was a great
economist in the best of the broad tradition of
classical political economy.
Selected Works
1 9 2 5 . Capitalist enterprise and social progress.
London: Routledge.
1 9 2 8 . Russian economic development since the
revolution. L o n d o n : R o u t l e d g e .
1928. Wages. London: Nisbet; Cambridge: Cam-
bridge University Press.
1929. A sceptical view of the theory of wages.
Economic Journal 39: 506-519.
1933. Economic theory and the problems of a
socialist economy. Economic Journal 43: 588-
598.
1 9 3 7 . Political economy and capitalism: Some
essays in economic tradition. L o n d o n :
Routledge.
1946. Studies in the development of capitalism.
London: Routledge.
1 9 4 8 . Soviet economic development since 1917.
London: Routledge.
1950. Reply (to Paul Sweezy’s article on the tran-
sition from feudalism to capitalism). Science
and society 14(2): 157-167.
1950. Some aspects of economic development.
Three lectures. Delhi: Ranjit Publishers, for the
Delhi School of Economics.
1 9 5 5 . On economic theoiy and socialism.
London: Routledge.
1 9 6 0 . An essay on economic growth and plan-
ning. L o n d o n : R o u t l e d g e .
3040
1 9 6 9 . Welfare economics and the economics of
socialism. C a m b r i d g e : C a m b r i d g e
University Press.
1 9 7 3 . Theories of value and distribution since
Adam Smith: Ideology and economic theory.
Cambridge: Cambridge
University Press.
Bibliography
Bergson, A. 1949. Socialist economics. In A survey of
contemporary economics, ed. H.S. Ellis. Philadelphia:
Blakiston.
Cambridge Journal of Economics. 1978. Maurice Dobb
memorial issue, vol. 2(2), June.
Hobsbawm, E.J. 1967. Maurice Dobb. In Socialism, cap-
italism and economic growth: Essays presented to
Maurice Dobb, ed. C. Feinstein. Cambridge: Cambridge
University Press.
Lewis, W.A. 1954. Economic development with unlimited
supplies of labour. Manchester School 20 (2): 139-191.
Meek, R. 1978. Obituary of Maurice Herbert Dobb. Pro-
ceedings of the British Academy 1977 (53): 333-344.
Pollitt, B.H. 1990. Clearing the path for ‘Production of
Commodities by Means of Commodities’: Notes on the
collaboration of Maurice Dobb in Piero SrafFa’s edition
of ‘The Works and Correspondence of David Ricardo’. In
Essays on Piero Sraffa: Critical perspectives on the
revival of classical theory, ed. K. Bharadwaj and B.
Schefold. London: Unwin Hyman.
Sraffa, P. 1960. Production of commodities by means of
commodities: Prelude to a critique of economic theoty.
Cambridge: Cambridge University Press.
Sraffa, P. with the collaboration of M.H. Dobb. 1951-73.
Works and correspondence of David Ricardo, 11 vols.
Cambridge: Cambridge University Press.
Steedman, I. 1977. Marx after Sraffa. London: New Left
Books.
Dollarization
Roberto Chang
Abstract
This article focuses on dollarization, a situation
in which a foreign currency (often the US dollar)
replaces a country’s currency in performing one
or more of the basic functions of money. The
distinction between official dollarization
Dollarization
and endogenous dollarization is discussed, as
are the concepts of currency substitution and
liability dollarization. Implications for monetary
and exchange rate policy are emphasized.
Keywords
Aggregate demand; Capital asset pricing model;
Currency substitution; Dollarization; Euro;
Financial dollarization; Inflation; Lender of last
resort; Liability dollarization; Monetary policy;
Net worth effects; Portfolio balance; Search
theory; Seigniorage; Stabilization; Transaction
costs
JEL Classifications
F3
Dollarization is a situation in which a foreign
currency (often the US dollar) replaces a country’s
currency in performing one or more of the basic
functions of money.
Thus in Ortiz ( ) the term ‘dollarization’
refers to the widespread usage of US dollars for
transaction purposes in Mexico. More recently, Ize
and Levy-Yeyati ( ) use ‘financial dollar
ization’ for episodes in which domestic financial
contracts are denominated in dollars or another
foreign currency.
In some countries, dollarization has been the
outcome of official government policy. Examples
include Ecuador in 2000 and El Salvador in 2001,
where the domestic currency was retired from cir-
culation and the US dollar became the official
currency. An immediate implication of such ‘offi-
cial dollarization’ is that domestic prices of tradable
goods are tied to world prices, so domestic inflation
is closely related to US inflation. Hence official
dollarization has been advocated for countries suf-
fering from chronic, high, and volatile inflation.
On the other side of the ledger, official dollar-
ization implies the surrender of independent mon-
etary policy, leaving only fiscal policy available as
a stabilization tool. In addition, the domestic gov-
ernment gives up seigniorage, or the revenue from
money creation, which accrues to the US Federal
Reserve. While both effects are widely regarded as
costly for the domestic economy, their welfare
Dollarization
implications depend on details about the
policymaking process and, in particular, on whether
the monetary authorities can credibly commit to
implement optimal policy (see Chang and Velasco ,
for a discussion).
Finally, official dollarization implies that the
domestic central bank is no longer available as a
lender of last resort, which may be conducive to
financial fragility and crises. Calvo ( ) argues,
however, that last resort lending can be provided by
alternative arrangements.
Impetus for official dollarization as a policy
alternative was greatest at the turn of the millen-
nium, as emerging economies had to cope with a
sequence of financial and exchange rate crises while
several European countries were abandoning their
national currencies in favour of the newly created
euro. Support for official dollarization appears to
have subsided since, however.
More frequently, dollarization has emerged as a
spontaneous response of domestic agents to
inflation. The special case in which such a process
has resulted in the dollar becoming a widespread
medium of exchange is known as ‘currency sub-
stitution’. Currency substitution has been the subject
of a large literature, much of it focused on the
determinants of the relative demand for domestic
vis-a-vis foreign currencies and on implications for
monetary management. Early research followed
Girton and Roper ( ) in postulating
ad hoc aggregate demand functions for domestic
and foreign currency, in the portfolio balance tra-
dition. Somewhat later, Calvo ( ) derived sim
ilar demand functions from an optimizing model in
which domestic and foreign currencies entered the
representative household’s utility function. Those
approaches emphasized the possibility that
increasing substitutability between the domestic and
the foreign currencies would lead to monetary and
exchange rate instability. However, they did not
identify the basic determinants of substitutability,
which was buried in the specification of the
postulated demand function for foreign currency or
the properties of the representative agent’s utility
function. Hence the early studies were of little use
in understanding how to cure the ills associated with
dollarization, and, in particular, they failed to trace
the consequences of
3041
common policies designed to deal directly with
currency substitution, such as outright prohibitions
on the holdings of foreign currency.
Subsequent studies have attempted to address
these shortcomings by modelling more explicitly
the fundamental frictions underlying currency
substitution. Thus Guidotti and Rodriguez ( )
developed a cash-in-advance model of
currency substitution on the assumption that using
foreign currency entailed fixed transaction costs,
while Chang ( ) studied the implica
tions of a similar assumption in an overlapping
generations setting. These models still left
unexplained where the assumed transaction costs
were coming from. Therefore, recent work on this
area models currency substitution entirely from first
principles, in the search theoretic tradition (see, for
instance, Craig and Waller ).
Another focus of recent literature has been the
increased use of the dollar as the currency of
denomination of the debts of domestic residents in
emerging economies, a problem that Calvo ( )
terms ‘liability dollarization’. A substantial
degree of liability dollarization places an economy i
u a vulnerable situation, since presumably many of
the agents with dollar debts have assets
denominated in domestic currency. Such a currency
mismatch situation means that a depreciation of the
domestic currency reduces the net worth of
domestic agents. Ifi in turn, aggregate demand
depends on net worth (as would be the case in the
presence of financial imperfections), a currency
depreciation may lead to a reduction in income and
employment. In other words, liability dollarization
may render depreciations contractionary, not
expansionary as assumed by conventional analysis
(Aghion et al. ; Cespedes et al. ). The combination
of liability dollarization and net worth effects has
been blamed for the severity of the income and
output contractions in recent emerging markets
crises.
At this point, no consensus exists as to the
causes of liability and financial dollarization,
although research on this question is rather active.
Ize and Levy-Yeyati ( ), in particular, have
examined the choice of currency denomination of
assets and liabilities from a capital asset pricing
model (CAPM) perspective, while Jeanne ( )
 3042 Domar, Evsey David (1914-1997)

models liability dollarization as the private sector ed. B. Eichengreen and R. Hausmann. Chicago: Uni-
response to the lack of credibility in monetary versity of Chicago Press.
Ortiz, G. 1983. Currency substitution in Mexico: The dol-
policy. Finally, several studies estimate how mea- larization problem. Journal of Money, Credit, and
sures of financial dollarization depend empirically Banking 15: 174-185.
on other characteristics of an economy. For exam-
ple, Arteta ( ) has found that the dollarization
of bank deposits is empirically more frequent in
countries with a higher degree of exchange rate Domar, Evsey David (1914-1997)
flexibility.
E. Cary Brown

See Also

Keywords
► Domar, E. D.; Economic growth; Harrod-Domar
growth model; Portfolio theory; Proportional
Bibliography income tax; Slavery; Technical change

Aghion, P., P. Bachetta, and A. Banajee. 2001. Currency

crises and monetary policy in an economy with credit
constraints. European Economic Review 45: 1121-1150.
Arteta, C. 2005. Exchange rate regimes and financial dol-
larization: Does flexibility reduce currency mismatches in
bank intermediation? Topics in Macroeconomics 5(1):
1226-1246.
Calvo, G.A. 1985. Currency substitution and the real
exchange rate: The utility maximization approach. Jour-
nal of International Money and Finance 4: 175-188.
Calvo, G.A. 2005. Capital markets and the exchange rate with
special reference to the dollarization debate in Latin
America. In Emerging capital markets in turmoil, ed. G.
Calvo. Cambridge, MA: MIT Press.
Cespedes, L., R. Chang, and A. Velasco. 2004. Balance sheets
and exchange rate policy. American Economic Review
94: 1183-1193.
Chang, R. 1994. Endogenous currency substitution, infla-
tionary finance, and welfare. Journal of Money, Credit,
and Banking 26: 903-916.
Chang, R., and A. Velasco. 2002. Dollarization: Analytical
issues. In Dollarization, ed. E. Levy-Yeyati and F.
Sturzenegger. Cambridge, MA: MIT Press.
Craig, B., and C. Waller. 2004. Dollarization and currency
exchange. Journal of Monetary Economics 51: 671-689.
Girton, L., andD. Roper. 1981. Theory and implications of
currency substitution. Journal of Money, Credit, and
Banking 13: 12-30.
Guidotti, P.E., and C.A. Rodriguez. 1992. Dollarization in
Latin America: Gresham’s law in reverse? IMF Staff
Papers 39: 518-544.
Ize, A., and E. Levy-Yeyati. 2003. Financial dollarization.
Journal of International Economics 59: 323-347.
Jeanne, O. 2005. Why do emerging economies borrow in
foreign currency? In Other people's money,
JEL Classifications
B31
Domar (Domashevitsky) was bom in 1914 in Lodz,
Russia (now Poland), spent most of his early life in
Harbin, Manchuria, and moved permanently to the
United States in 1936. His undergraduate degree in
economics (1939) was from the University of
California (Los Angeles); his graduate work was at
the Universities of Michigan (MA, Mathematical
Statistics) and Harvard (Ph.D., 1947), where he
studied with Alvin Hansen, the leading American
Keynesian and most important single intellectual
influence on Domar. Domar is best known for his
leadership role, along with Roy Harrod, in the
initiation of modem growth theory.
His first position was with the research staff of
the Board of Governors of the Federal Reserve
System, where he worked on fiscal problems from
1943 to 1946. His subsequent academic career took
him briefly to the Carnegie Institute of Technology,
the Cowles Foundation and the University of
Chicago, the Johns Hopkins University in 1948 for
ten years, and the Massachusetts Institute of
Technology in 1958, from which he retired in 1984.
An avid traveller, he held more than a dozen
visiting professorships in universities at home and
abroad.
Domar, Evsey David (1914-1997)
While the claim to the earliest statement of the
famous Harrod-Domar growth model was clearly
Harrod’s ( ), Domar arrived independently at a
structurally similar model but from a different point
of view (1946, 1947). By incorporating into static
Keynesian analysis the capacity changes associated
with investment, he found that steady-state capacity
growth required investment to grow at a rate equal
to the savings rate multiplied by the capital-output
ratio. From this simple beginning, growth theory
took off to become a major focus, one might almost
say obsession, of the profession in the 1950s and
1960s. Domar also made important contributions to
some of its conceptual and measurement problems,
such as the proper treatment of depreciation (1953)
and the measurement of technological change
(1961), and he coined the term ‘residual’ for the
fraction of expanding output unexplained by the
contribution of factors of production.
In fiscal theory, his early investigation, with
Richard Musgrave (1944a), of the effect of a
proportional income tax, with and without loss
offsets, on portfolio choice was very similar in style
and approach to portfolio theory of a decade later.
Given individual preferences, the portfolio decision
was modelled as a choice between alternative
portfolios weighing their expected net returns
against their risks (expected losses). The
unconventional conclusion was reached that, given
risk aversion, the imposition of a proportional
income tax with symmetrical treatment of gains and
losses would induce individuals to adjust their
portfolios towards riskier assets. The reminder that
expected risks and yields are both reduced by an
income tax was an important correction to a
simplistic focus on yields alone.
As an applied theorist, Domar had the knack of
getting important results with simple theory. At a
time when deficit finance was harshly criticized for
increasing the debt burden and tax rate, Domar
showed (1944b) that in a growing economy even
continuous deficit finance resulted in only limited
debt-income ratios and tax rates. Second, he made a
fertile historical hypothesis (1970) - that the
economic basis for the introduction of serfdom (or
slavery) was a low land-to-labour cost. Third, he
ingeniously modified the administrative rales that
guided the behaviour of collective farms
3043
(1966) or that determined the compensation of
socialist managers (1974) to induce them towards
more efficient price-output decisions.
Domar’s work was informed by a rare combi-
nation of historical, empirical and theoretical
breadth. His profound scholarship, in several lan-
guages, periods, and areas, often resurrected
important findings of earlier writers previously
overlooked.
Selected Works
1944a. (With R.A. Musgrave.) Proportional income
taxation and risk-taking. Quarterly Journal of
Economics 58: 388-422.
1944b. The burden of the debt and the national
income. American Economic Review 34: 798-
827. Reprinted in Domar (1957).
1946. Capital expansion, rate of growth, and
employment. Econometrica 14: 137—
47.
Reprinted in Domar (1957).
1947. Expansion and employment. American
Economic Review 37: 34-55. Reprinted in
Domar (1957).
1948. The problem of capital accumulation.
American Economic Review 38: 777-94.
Reprinted in Domar (1957).
1952. Economic growth: An econometric
approach. American Economic Review, Papers
and Proceedings 42: 479-95. Reprinted in
Domar (1957).
1953. Depreciation, replacement and growth.
Economic Journal 63: 1-32. Reprinted in Domar
(1957).
1 9 5 7 . Essays in the theory of economic growth.
New York: Oxford University Press.
1961. On the measurement of technological change.
Economic Journal 71: 709-29.
1966. The Soviet collective farm as a producer
cooperative. American Economic Review 56:
734-57.
1970. The causes of slavery or serfdom: A
hypothesis. Journal of Economic History 30: 18-
32.
1974. On the optimal compensation of a socialist
manager. Quarterly Journal of Economics 88: 1-
18.
3044
Bibliography
Harrod, R.F. 1939. An essay in dynamic theory. Economic
Journal 49: 14 33, Errata, 377.
Harrod, R.F. 1948. Towards a dynamic economics. Some
recent developments of economic theory and their
applications to policy. London: Macmillan.
Domesday Book
F. W. Maitland
Domesday Book is the name which, at least since
the 12th century, has been borne by the record of the
great survey of England made by order of William
the Conqueror. Apparently the decree for the survey
was issued at a moot held at Gloucester at the
midwinter of 1085-86, and the work was completed
in the course of the following year. Royal
commissioners (legati) were sent into each shire
with a list of interrogatories, to which they were to
obtain sworn answers from local juries. Their
procedure seems to have been this - they held a
great shire moot, at which every hundred or
wapentake of the shire was represented by a jury,
while every vill was represented by a deputation of
villagers. From each hundred-jury they obtained a
verdict about all the land in the hundred, the
villagers being at hand to correct or supplement
verdicts, while ‘the whole shire’ was also present,
and from time to time appeal could be made to its
testimony. The statement thus supplied was
reduced into writing and duly transmitted to the king.
It was afterwards methodised and abstracted, and
fairly transcribed in the great volume of Domesday
and deposited in the royal treasury at Winchester,
amongst the other muniments of the realm. It still
exists, fresh and perfect as when the scribe put pen to
parchment, the oldest cadastre, or survey of a
kingdom, now existing in the world (Palgrave,
History of Normandy and England, vol. iii, p. 575).
Oiu best information about the form of the
original verdicts is contained, not in Domesday
Book itself, but in a document known as the
Domesday Book
'Inquisitio Comitatus Cantabrigiensis'. This seems
to be a copy made in the 12th century of the
verdicts delivered by the juries which represented
some of the hundreds of Cambridgeshire. The
verdicts having been obtained, they were sent to the
king’s treasury, and a digest was made of them by
the royal officers. This digest is Domesday Book. If
we may draw a general inference from
Cambridgeshire, the materials supplied by the
commissioners were subjected to a process of
rearrangement. A scheme that was wholly geo-
graphical gave way to one which was partly geo-
graphical, partly proprietary. Domesday book deals
with each shire separately, but within the shire it
collects, under the name of each 'tenant in chief, all
the estates that he holds, no matter in what hundred
they may be. For example, the Cambridgeshire
verdicts showed that Count Alan had lands in many
hundreds. In the original verdicts the entries relating
to his estates were therefore scattered about; in
Domesday Book they are all collected together.
Domesday Book consists of two volumes,
sometimes called ‘Great Domesday' and 'Little
Domesday'. The latter deals with Essex, Norfolk,
and Suffolk; the former with so much of the rest of
England as was surveyed. A document in the
keeping of the cathedral chapter of Exeter, and
known as ‘the Exon Domesday’, contains an
account of a large part of the south-western shires,
which is veiy closely connected with that given by
what, for distinction’s sake, is sometimes called ‘the
Exchequer Domesday’. Seemingly this Exon
Domesday is independent of the Exchequer Record,
and goes back by a different route to the original
verdicts. The same may perhaps be said of the
‘Inquisitio Eliensis’, an account of the estates held
by the church of Ely. This Ely inquest must not be
eon- fused with the Cambridgeshire inquest.
Domesday Book was printed and published in
1783 in two folio volumes. A third volume
containing indexes was published in 1811, and this
was followed in 1816 by a fourth volume containing
the Exon Domesday, the Ely Inquest, and some
other matters. Of late years useful facsimiles have
been published by the Ordnance Survey Office of
various parts of the great Exchequer Record, and
can be obtained at moderate
Domesday Book
prices. The important Cambridgeshire Inquisition
was first published by N.E. Hamilton in 1876.
A large literature has gradually been collecting
round Domesday Book. Among the older books
Robert Kelham’s Domesday Book Illustrated ( ) and
the essays of Phibp Carteret Webb
deserve to be mentioned. Sir Henry Ellis, in his
General Introduction to Domesday Book ( ),
supplied valuable indexes, and summed up the older
learning. In the fifth volume of E.A. Freeman’s
Norman Conquest good use has been made of all
that bears on political history, on the history of great
men, great churches, great events. James F.
Morgan’s England under the Norman Occupation (
) is a good introduction
to the study of Domesday, and the like may be said
of W. de Gray Birch’s Domesday Book ( ).
A new epoch in the scientific exploration of the
record is marked by the various works of R.W.
Eyton dealing with Dorset, Somerset, Lincoln, and
Stafford, especially by the key to Domesday Book.
Two volumes of essays by various writers, called
Domesday Studies ( -91),
contain two valuable papers by J.H. Round, besides
other matters. In some county histories Domesday
has been well used, but here it is possible to name
only the books of general importance. F. Seebohm’s
English Village Community has done much to
awaken a new and an economic interest in our
oldest statistics.
Much remains to be done. The student who
approaches Domesday from the economic side will
at once see that he has before him a vast mass of
detailed statistics which ought to tell him much
about agriculture, prices, rents, and the like. At the
same time he will feel that he is debarred from
making use of these precious materials by the
difficulty of discovering the meaning of the crabbed
formulas which are repeated on page after page. The
difficulty is a very real one. Domesday Book stands
alone. It is so far removed in time from the
documents which most nearly resemble it, the
extents of manors which are found in monastic
cartularies, that we have to explain it out of itself or
not at all, for we shall look in vain for help
elsewhere. Then again the terms that it employs as
technical terms are, we may say, derived from two
different languages which have
3045
only of late come into contact with each other.
About half of them have been introduced by the
Norman conquerors, while the other half are words
which were in use in England under Edward the
Confessor. Hence many puzzles; for example, what
word did English juries say when French clerks
wrote down villanusl Then again, the more our
record is studied, the more plainly do we see that
one main purpose governs both its form and its
matter. King William is not collecting miscella-
neous information in the spirit of a scientific
inquirer. He is in quest of geld. Domesday Book is a
geld book, a tax book. Geldability, actual or
potential, this is its main theme. If then we are to
understand its statistics, the first thing necessary is a
theory of geld, of the manner in which the great tax
has been and is assessed and collected. Towards the
construction of such a theory not a little has been
done by modem writers, especially by Eyton and
Round, but until the work has been completed,
speculations about rents and values seem doomed to
failure. Everywhere, for example, the question
meets us whether we are reading of real areal units
of land or of units which are the results of a mde
system of taxation, and a great deal of labour must
yet be spent on the book before this question will
have been adequately answered.
References
Brich, W. 1887. Domesday book. New York/London: E. &
J.B. Young & Co/Society for Promoting Christian
Knowledge.
Dove, P.E. (ed.) 1888-91. Domesday studies. London/New
York: Longmans, Green & Co.
Ellis, H. 1833. General introduction to Domesday book.
London.
Eyton, R.W. 1878. A key to Domesday. London.
Freeman, E.A. 1869-79. The history of the Norman Conquest,
vol. 5. Oxford.
Kelham, R. 1788. Domesday hook illustrated. London.
Maitland, F.W. 1897. Domesday book and beyond.
Cambridge: Cambridge University Press.
Morgan, J.F. 1858. England under the Norman occupation.
London.
Palgrave, F. 1851-64. The history of Normandy and of
England, vols. 1 and 2. London: J.W. Parker; vols. 3 and
4, London: Macmillan.
Seebohm, F. 1883. The English village community. London:
Longmans & Co.
Webb, PC. 1756. A short account of some particulars
concerning Domes-day. London.
3046
Domestic Labour
S. Himmelweit
The term domestic labour entered economic
vocabulary in the early 1970s as a result of feminist
interest in criticizing and expanding economic
categories to incorporate women’s activities. Both
mainstream and critical traditions in economics
tried to grapple with the problem of how to account
for the difference between men’s and women’s
position on the labour market. One approach was to
relate women’s lesser training and skills in paid
employment to competing demands made on a
(married) woman’s time by domestic commitments,
with a tacit, though unexplained, acceptance that for
women paid employment has to fit into time left
over after the allocation of that needed for domestic
labour, while for men it is the other way round. It is
only by the addition of such an assumption that the
analysis of domestic labour can be said to have had
anything to say about women.
Neoclassical economists have seen domestic
labour as one of three competing claims on people’s
time, the others being paid work and leisure. A
household maximizes ‘its’ utility, which is a
function of the consumption goods bought with
income received from paid work by members of the
household, the direct consumption of the products
of time spent in domestic labour and a variety of
ways of spending remaining leisure time (Becker ).
Women have a comparative advantage to men in
domestic labour over paid work and so one or other
partner should specialize; either a woman should
not take paid employment or her husband should do
no housework. Even if there is no intrinsic
difference between men and women initially,
specialized human capital can be acquired in each
type of labour, so it makes sense for a division of
labour to take place and for at most only one
member of a household to work both in the home
and outside. This is taken to explain both why the
majority of domestic labour is performed by women
and also why
Domestic Labour
women work shorter hours in paid employment than
men, accumulate less market-oriented training and
skills, and have broken employment histories.
Two criticisms can be mounted of this approach.
The first is that the comparative advantage itself
needs explanation. At an individual level it can be
accounted for by the lower relative earnings of
women. But the outcome of individual household
choices cannot, without circularity, then in turn be
used to explain women’s inferior earnings by the
lesser time spent in the labour market acquiring
appropriate human capital. At best such an approach
can account for the division between houseworkers
and paid workers, and if combined with an
assumption that sex is used as a screening
mechanism by employers, a form of rational
statistical discrimination, why one sex as a whole
will be more likely to constitute the houseworkers
and the other the employees. But to explain why sex
presents itself as a variable by which to screen, and
why it is the female sex that constitutes the
homeworkers, recourse must be made to biological
differences in aptitude, an acceptable fail-back to
some, but not to those who wish to show the power
of the neoclassical economic approach to explain
everything, nor to the feminist movement whose
claims that a woman’s place was socially rather
than naturally in the home had led to the initial
interest in the question.
The second criticism poses more fundamental
problems for this type of analysis. The concept of a
household’s ‘utility function’ is a very shaky one.
Individualism, upon which neoclassical economics
is based, takes individuals as the only actors and
decision makers, rejecting thereby, for example, the
marxist notion of class interests and forces. The
idea of a household utility function cannot therefore
be entertained unless either all members of the
household have identical preferences concerning the
allocation of resources and leisure time among
themselves or some rule for aggregating diverse
preferences is adopted. Quite apart from the
difficulty of devising such a rule which satisfies
fairly minimal criteria to ensure that household
preferences represent some meaningful aggregate of
those of its members
Domestic Labour
(Samuelson ; Arrow ), there is little evidence that
households, rather than individuals, make decisions
at all. Indeed, feminists would argue that such an
approach obscures one of the key questions it was
supposed to illuminate, differential power and thus
an unequal division of labour within the household
(Pahl ).
This problem can be overcome by assuming one
member of the household is sufficiently powerful,
well-endowed and altruistic that all other members
of the household are or aspire to be ‘his’
beneficiaries (Becker ). Then the interest of all
family members are serviced by the maximization
of family income and the whole family can behave
as one single decision-making unit. The
assumptions of this model, which might seem
appropriate only to an idealized picture of a
Victorian patriarchal family, are necessary in order
to avoid oligopolistic decision making, and even
then care has to be taken to ensure that the
paterfamilias is not driven into a comer solution,
whereupon the unity of the family breaks down.
Marxist approaches criticize neoclassical ana-
lyses for failing to take account of the different
social relations involved in wage and housework.
The categories ofmarxist analysis are particularly
appropriate to the analysis of unequal power rela-
tions, making marxism seem to some feminists
more likely to offer a useful approach. The marx- ist
notion of exploitation is based upon the char-
acterization of specific forms of surplus extraction.
The attempt to analyse domestic labour in these
terms would therefore illuminate power relations
within the household, without falling into the trap of
conflating housework with paid labour, by
recognizing its relations of production, not just its
product, to be specific.
Accounts which characterized domestic labour
as a separate mode of production came both from
writers claiming to be orthodox marxists, extending
rather than revising Marx’s work, and from others
who saw themselves more as using parts of Marx’s
mode of analysis to criticize and reformulate
orthodox Marxism. Of the latter group Christine
Delphy, for example, argued that there is a
transhistorical family mode of production in which
wives’ labour power is exploited by their husbands
which has coexisted with and
3047
outlasted the modes of production Marx described
(Delphy ). Harrison, on the other hand, sees the
domestic mode of production as a specific
subordinate counterpart to a capitalist mode of
production unchanged from that of traditional
marxist analysis.
Other accounts rejected the characterization of
domestic labour as a separate mode of production
on the grounds that a mode of production must be
capable of independent self-perpetuation, since the
term was used for the characterization of whole
societies. The notion of a social formation
encompassing two or more modes of production
articulated with each other, while appropriate to the
analysis of transition between modes, was not
appropriate to the continued mutually dependent
symbiotic relationship which exists between
housework and wage work for capital. The alter-
native was to extend the notion of the capitalist
mode of production to include housework (Gardiner
et al ). That extension was needed because the
transformation of the wage into reproduced labour
power is a process requiring labour and taking place
under specific relations of production, and not the
unproblematic natural process that Marx took it to
be (O’Brien ).
The effect within marxist theory of character-
izing housework as a separate mode of production is
to make housewives a class, exploited through
performing surplus labour above the amount needed
to reproduce their own labour power. This surplus
was appropriated, according to different versions,
either by their husband directly or transferred
through lowering the value of his labour power to
the capitalist who employed him. But if housework
was seen as part of the capitalist mode of
production, a housewife’s class position, like that of
anyone else, would be determined by her access to
the means of production, and for most women that
would put them in the working class along with
their husbands.
Another area of dispute was whether domestic
labour should be seen as value and/or surplus value
producing. Some argued that it did produce value,
because it produced the commodity labour power
(Dalla Costa ). In so far as the housewife worked
longer hours than that needed to reproduce her own
labour she also produced
3048
surplus value. Against this it could be argued that
the housewife by producing use-values needed to
reproduce labour power did not thereby make
labour power her product, any more than the baker,
butcher or obstetrician did (Seccombe ). Labour
power is an attribute of a living human being and is
not, pace Marx, a commodity like any other in that
it is not directly produced by labour at all. In that
case the labour that a housewife expands is use-
value but not value creating, and therefore a fortiori
not surplus value creating.
The dispute as to whether domestic labour
counted as productive labour turned upon the same
issue, since within the capitalist mode of production
labour is productive, according to the Marxist
definition, if and only if it produces surplus value.
Those who argued that domestic labour produced
surplus value could therefore also claim that it was
productive labour. But against this could be put
Marx’s own demonstration that productive labour
must, to produce surplus value, take place between
two exchanges: in the first labour power is bought
for a wage, in the second the product is sold.
Domestic labour requires neither exchange and
therefore is technically outside the classification
into productive and unproductive labour, which
applies only to wage labour (Fee ).
The ‘domestic labour debate’ as it became
known failed to answer the question to which it was
addressed: what is the material basis of women’s
oppression? To do so, it would have had to do more
than classify domestic labour using the existing
categories of Marxist analysis. By using those
developed for the study of wage labour for capital it
fell into a similar trap to the neoclassical approach.
The neoclassical approach failed to recognize
that the different social relations under which
domestic labour went on rendered the use of the
theory of utility maximization developed to model
market decision-making inappropriate. The
assumptions needed in order that the division of
labour within the home could be set up as a soluble
decision-making problem had to turn the gender-
divided household into a homogeneous single
decision-making unit. Divisions within
Domestic Labour
the household disappeared and its individual
members became indistinguishable by anything that
could be remotely related to gender except by
recourse to some form of biological reduction- ism.
Circularity is a common problem with utility
analysis and in this case the only way to avoid it
was by appeal to supposed biological differences,
the very suppositions which feminists had rejected
as insufficient to explain the social construction of
gender-divided work patterns. Marxism can escape
the charge of circularity because its method is a
historical one. Circularity thus becomes recast as the
reproduction through time of the conditions which
give rise to a gender- divided society. But ultimately
marxism fell into the same trap. Although it did
recognize that domestic labour and wage labour go
on under different relations of production, it failed
to give those different relations any constmctive
effect, seeing domestic labour as simply labour that
did not have all the attributes of waged labour for
capital. To have got further it would have been
necessary to relate the analysis of domestic labour
to the sex of those who performed it and to its
fundamental characteristic of being labour involved
in reproduction rather than just another form of
production (Himmelweit ).
See Also
► r.ily
► . ib vork
► . e.bour Supply oi
Bibliography
Arrow, K.J. 1951. Social choice and individual values,
Cowles Commission Monograph No. 12. New York:
John Wiley & Sons.
Becker, G. 1965. A theory of the allocation of time. Eco-
nomic Journal 75: 493-517.
Becker, G. 1974. A theory of social interactions. Journal of
Political Economy 82(6): 1063-1093.
Dalla Costa, M. 1973. Women and the subversion of the
community. In The power ofwomen and the subversion of the
community, 2nd ed. Bristol: Falling Wall Press. Delphy, C.
1970. The main enemy. Partisans (Paris), Nos. 54-55.
Donisthorpe, Wordsworth (1847-1914)
Fee, T. 1976. Domestic labour: an analysis of housework and
its relation to the production process. Review of Radical
Political Economy 8(1): 1-8.
Gardiner, J, Himmelweit, S., and M. Macintosh. 1975.
Women’s domestic labour. Bulletin of the Conference of
Socialist Economists 4(2). Reprinted in On the political
economy of women, CSE Pamphlet No. 2. London: Stage
One, 1976.
Harrison, J. 1973. The political economy of housework.
Bulletin of the Conference of Socialist Economists 3(1):
35-52.
Himmelweit, S. 1984. The real dualism of sex and class.
Review of Radical Political Economics 16(1): 167-183. '
O’Brien, M. 1981. The politics of reproduction. London:
Routledge & Regan Paul.
Pahl, J. 1980. Patterns of money management within mar-
riage. Journal of Social Policy 9(3): 313-335.
Samuelson, PA. 1956. Social indifference curves. Quarterly
Journal of Economics 70(1): 1-22.
Seccombe, W. 1975. Domestic labour - reply to critics. New
Left Review 94: 85-96.
Donisthorpe, Wordsworth (1847-1914)
Peter Newman
Wordsworth Donisthorpe was bom on 24 March
1847 in Harrogate, graduated from Trinity College,
Cambridge in 1869 and was called to the Bar at the
Inner Temple in 1879. Thereafter he lived and
practised law in London. What is apparently the last
of his many books and pamphlets was published in
1913, the year before he died, his habitual
enthusiasm as yet undimmed.
He is of interest to economists because of his
first book, Principles of Plutology (reviewed in the
Saturday Review, 9 September 1876, pp. 331-2). In
it his vigorous and eccentric style, reminiscent of
that of Fleeming Jenkin’s last two papers in
economics (Colvin and Ewing Vol. II, pp. 122-54),
is already there in full spate. While showing traces
of Caimes and Jev- ons, the book is for the most
part subjectively original and objectively mediocre.
Chapter IX on ‘The Laws of Value’ is an interesting
exception.
3049
His clear understanding of the importance of the
Law of One Price (‘[This] proposition is the fun-
damental one’, p. 133) is refreshing and unusual for
its time, as is his Wicksteedian insistence on the
reservation price of the seller, so that ‘sellers and
buyers are not two classes, but one class’ (p. 132).
In the same chapter he discusses substitutes and
complements (‘co-elements’). Although attempting
no rigorous definitions he does lay down various
Taws’ concerning them, i.e. propositions of
comparative statics. Thus, the Third Law reads:
‘Other things equal, a rise in the value of a co-
element is followed by a fall in the values of its co-
elements, and a fall by a rise, but not necessarily at
the same rate’ (p. 153), while the Fourth Law is:
‘Other things equal, a rise in the value of any
commodity is followed by a rise in the value of its
substitutes, and a fall by a fall, but not necessarily at
the same rate’ (ibid.). He points out that, depending
on the circumstances, two commodities may stand
in both relations to each other, e.g. wool and cotton
as inputs to cloth are often co-elements while as
individual consumer goods they are usually
substitutes.
Perhaps disappointed at the book’s reception (I
know of no economist’s reference to it) Donisthorpe
soon left economics for political philosophy and
became a leading pamphleteer for anarchic
Individualism, a libertarian movement that perhaps
owed as much to fear of Henry George as to
admiration for Herbert Spencer. It took the
complacent view that the state should interfere with
individual activities only when deemed necessary to
protect the rights of private property, a view which
aroused such derisive epithets as ‘tomtits of
Anarchy’ from its opponents (see Westminster
Gazette, 3 and 11 August 1894).
However, a streak of Yorkshire shrewdness and
wit kept Donisthorpe from becoming quite as
doctrinaire and saintly an Individualist as, say,
Auberon Herbert. For example, the last chapter of
his Law in a Free State (1895 ) contains an uproar-
ious but penetrating account of the problems posed
to the Individualist polity, by what we would now
call externalities of various kinds, economic,
political, social and moral.
3050 Dorfman, Joseph (1904-1991)

Selected Works later. Dorfman was a student of Clarence Ayres at

1876. Principles ofplutology. London: Williams &
Norgate.
1880. The claims of labour; or, serfdom, wagedom
and freedom. London: Tinsley.
1889. Individualism: A system of politics. London:
Macmillan.
1891. The limits of liberty. In A plea for liberty, ed.
T. Mackay. New York: D. Appleton. 1893. L o v e a n d
law: An essay on marriage. London: W. Reeves.
1895a. L a w i n a f r e e s ta t e . London: Macmillan.
1865b. A system of measures of length, area, bulk, weight, value, force &c.
London: Spottiswoode. 1898. Down the stream of civilization.
London:
G. Newnes.
1913. Uropa. A new philosophically-constructed language based on Latin roots. Guildford:
W. Stent & Sons.
References
Colvin, S., and J.A. Ewing, oils. 1887. Papers, literary,
scientific &c., by the late Fleeming Jenkin, F.R.S., LL. D.
Willi a memoir by Robert Louis Stevenson. 2 vols.
London: Longmans, Green.
Dorfman, Joseph (1904-1991)
Henry W. Spiegel
Keywords
Dorfman, History of economic thought; Veblen,
T. Selected Works
Reed, and of Wesley C. Mitchell and John Maurice
Clark at Columbia. Mitchell in turn had been a
student of Thorstein Veblen. These four economists,
all with institutional leanings, stand out among the
formative influences that affected Dorfman’s early
career. He made Veblen the subject of his doctoral
dissertation, which was published under the title
Thorstein Veblen and His America in 1934. This was
at the time the only book-length appraisal of a
modern economist that gave close attention not only
to the subject's writings but also to biographical
detail, the contemporary climate of opinion, and the
general social and cultural setting of the work.
This type of holistic approach is characteristic
also of Dorfman's monumental The Economic Mind
in American Civilization, a five-volume work that he
published from 1946 to 1959. It is dedicated 'To the
pioneering spirit of Thorstein Veblen and the first-
born of his intellecUial heirs, Wesley C. Mitchell’.
The work is a detailed history of American
economic thought from colonial times to 1933, the
first of its kind and not likely to be replaced for
many years. It is based on extensive research and in
many instances provides the first comprehensive
account of a writer’s life and work. Dorfman sees a
break of emphasis in the history of American
economic thought at the time of the Civil War: it
was commerce before, and industry later. He notes
with respect the achievements of the past, and is a
critical but tactful chronicler of past foibles. He was
a pioneer in exploring not only the printed page but
also archival material made up of 'papers’, ‘letters’,
and similarly elusive sources of information, the
first writer to do so on a large and systematic scale
in the history of economic thought.

1934. Thorstein Veblen and his America. New York:

JEL Classifications Viking Press.
B31 1935. (With R.G. Tugwell.) William beach Law-
rence: Apostle of Ricardo. New York, reprinted
Historian of American economic thought, Dorfman
from Columbia University Quarterly, September,
was born in Russia in 1904 and educated at Reed
1935.
College and at Columbia University, where he
earned a Ph.D. degree in 1935 and taught from
1931 until his retirement 40 years
Dornbusch, Rudiger (1942-2002)
1940. The economic philosophy of Thomas Jeffer-
son. New York: Academy of Political Science.
1 9 4 6 — 5 9 . The economic mind in American
Civilization. 5 v o l s . N e w Y o r k :
Viking Press.
1954. Introduction to Adams, H.C., Relation of the
state to industrial action, and economics and
jurisprudence. Two Essays, ed. J. Dorfman. New
York: Columbia University Press.
1960. (With R.G. Tugwell.) Early American policy:
Six Columbian contributors. New York:
Columbia University Press.
1967-9. Introduction to W.C. Mitchell, Types of
economic theory: From mercantilism to
institutionalism, ed. J. Dorfman. New York:
A.M. Kelley.
Dornbusch, Rudiger (1942-2002)
Kenneth Rogoff
Abstract
Rudiger Dornbusch was one of the leading
researchers in international macroeconomics in
the late 20th century. He introduced the
influential concept of exchange rate ‘over-
shooting’ to explain the excessive volatility of
exchange rates after the break-up of the Bretton
Woods system of fixed exchange rates in the
early 1970s. Along with Stanley Fischer and
Paul Samuelson, he revived the Ricardian theory
of international trade whereby trade was driven
by differences in technology; their simple
tractable framework became similarly influential
in the study of international trade.
Keywords
Comparative advantage; Dornbusch, R.;
Exchange rate dynamics; Exchange rates
(floating vs. fixed); Fischer, S.; Hecksher- Ohlin
framework; New open economy macro-
economics; Non-traded goods; Overshooting;
Ricardian trade theory; Samuelson, P.; Sticky
prices; Value at risk (VAR)
3051
JEL Classifications
B31
Rudiger Dornbusch was bom in Germany on 8 June
1942. He received his Licence es Sciences
Politiques from the University of Geneva in 1966,
and his Ph.D. in Economics from the University of
Chicago in 1971. He was an assistant professor at
the Department of Economics at the University of
Rochester from 1972 to 1974, an associate professor
at the Graduate School of Business at Chicago
University from 1974 to 1975, and a member of the
MIT Department of Economics from 1975 to 1978.
He became a Professor of Economics at MIT in
1978. From 1984 until his death from cancer on 25
July 2002, he was Ford International Professor of
Economics at MIT.
Dornbusch was, by any measure, one of the
giants of late 20th century international macroeco-
nomics. His celebrated Journal of Political Econ-
omy paper ‘Expectations and exchange rate
dynamics’ (1976), which introduced the concept of
exchange rate ‘overshooting’, became the
workhorse of international macroeconomics over
the ensuing two decades. His American Economic
Review paper (with Stanley Fischer and Paul Sam-
uelson) ‘Comparative advantage, trade and pay-
ments in a Ricardian model with a continuum of
goods’ (1977) introduced a simple tractable
framework that became similarly influential in the
study of international trade.
This entry begins by reviewing Dombusch’s two
most important scientific contributions, and goes on
to give a brief sketch of his broader influence on the
profession through students (he served as an advisor
on over 125 doctoral dissertations), through his
leading intermediate textbook Macroeconomics
(written with Stanley Fischer), and through his role
as an important voice in the public policy debate.
Exchange Rate Overshooting
Dombusch’s overshooting model of exchange rates
(1976) captured the imagination of policymakers
and academics alike during the
3052
early years of floating exchange rates. The model
attracted enormous attention because, after the
break-up of the Bretton Woods system of fixed
exchange rates in the early 1970s, exchange rates
seemed far too volatile relative to the underlying
fundamentals. Although subsequent empirical work
has undermined the model’s original bold claim to
explain floating exchange rates (see Meese and
Rogoff ), the model is still viewed as relevant,
especially during episodes of major shifts in
monetary policy. In fact, an informal survey
conducted by Alan Deardorff of eight top
economics departments found that, as late as 1990,
Dornbusch’s overshooting model was the only
paper taught in every one of their graduate
international finance courses.
The idea of overshooting is so simple and
elegant that the small-country version can be illus-
trated with just a couple of equations (the analysis
here draws on Rogoff ). The assumption of
‘uncovered interest parity’ relates the home nominal
interest rate to the exogenous foreign nominal
interest rate and the expected rate of depreciation of
the exchange rate:
i t = i * + E ( e , + i - e ,) (1)
where i, is the level home nominal interest rate and
et is the logarithm of the exchange rate (the home
currency price of foreign currency), so that E(et +] —
et) is the expected rate of change in the exchange
rate. The second key relationship is a money
demand equation that relates the real balances to the
nominal interest rate.
m, -p, = -Xi t + rj y, (2)
where y denotes the log of output, m is the nominal
money supply andp is the price level. Higher
interest rates lower the demand for real balances,
and an increase in output raises it. Dornbusch posed
the question of what would happen if there were a
one-time permanent increase in the money supply,
m. If prices were fully flexible, it would be possible
to maintain equilibrium in the above two equations
by having prices and exchange rates all rise
Dornbusch, Rudiger (1942-2002)
permanently in proportion to the increase in the
money supply. In this case, money would be neutral
and have no real effects.
In reality, however, while asset markets
(including the exchange rate) adjust very quickly,
goods markets adjust more slowly partly due to
temporary price rigidities. Therefore, in this set-up
money is neutral only in the long run (in which the
price level rises proportionately to the money sup-
ply). But with goods markets clearing only slowly,
what is the impact of a money shock on exchange
rates and interest rates? Assume that output, y, is
also fixed. If domestic prices are constant, then a
rise in the money supply implies a rise in real
balances, m - p. But this means that the home
nominal interest rate i must fall, so there is a
corresponding rise in the demand for real balances.
Then, however, the uncovered interest parity equa-
tion (Eq. (1) above) implies that e, must fall, or
depreciate, relative to expectations of et+\.
That is, after any initial movement of the
exchange rate in response to an unexpected shock,
the currency must subsequently be expected to
appreciate. But recall that in the long run, even with
sticky prices, money is still neutral, so the exchange
rate has to depreciate by the same amount as the
rise in the domestic price level, thus producing no
real effect.
How is all this possible? The answer,
Dornbusch deduced, is that the initial money shock
must cause the exchange rate to depreciate by more
in the short ran than it does in the long run. It
‘overshoots’. Therefore, Dombusch’s model offered
a highly plausible explanation of why exchange
rates seem to be so volatile relative to fundamentals.
At one level of abstraction, of course,
‘overshooting’ is an application of Paul
Samuelson’s ‘Le Chatelier’s principle’ theorem:
when prices in some markets are inflexible in the
short run, prices in others may overreact in the short
run. But Dombusch’s model did much more than
innovatively contrast the fast adjustment of asset
markets with the slow adjustment of goods markets
(an insight that any realistic short-run dynamic
macroeconomic model should take into account). It
offered a concrete and coherent analysis of an
extremely important
Dornbusch, Rudiger (1942-2002)
practical phenomenon. Over the decades since
Dombusch’s article appeared, the term ‘over-
shooting’ has become deeply woven into the
popular economic lexicon.
Modem research has advanced considerably
beyond the overshooting model, of course, and the
Mundell-Fleming-Dombusch model has largely
been supplanted by ‘new open economy
macroeconomics’ (see Obstfeld and Rogoff ). And
the notion of looking at money shocks via a money
demand equation has increasingly been supplanted
by frameworks which view the overnight interest
rate as the key instrument of monetary policy.
Nevertheless, these newer frameworks typically
include sticky prices - perhaps the most
fundamental, and controversial, element of
Dombusch’s model - and hence can all replicate a
similar phenomenon to ‘overshooting.’
Although Dombusch’s overshooting paper was
his best-known work, with over 900 citations in
refereed journals, he published numerous other very
well-known articles, including his 1973 American
Economic Review paper that was among the first to
incorporate non-traded goods in a monetary model
(see also his elegant 1974a contribution to the
collection edited by Robert Aliber), his 1983
Journal of Political Economy paper that illustrated
how changes in the real interest rate could affect
exchange rates and current accounts, and his 1987
American Economic Review paper that
demonstrated a link between market structure and
the adjustment of relative prices to exchange rate
movements. Without doubt, however, his other
extremely influential paper was not in international
finance but in trade.
Ricardian Model of Trade
Dombusch’s 1977 American Economic Review
paper with MIT colleagues.
Stanley ischer and Paul Samuelson almost
single-handedly revived the analysis of Ricardian
trade; a ‘Ricardian’ model of trade is one with only
one factor of production (usually taken to be
labour). Trade is driven by differences in technol-
ogy. The Ricardian model is contrasted with the
3053
Hecksher-Ohlin framework, where countries have
identical technologies but different relative endow-
ments of the factors of production (labour and cap-
ital, in the simplest canonical case). Prior to
Dombusch-Fischer-Samuelson (DFS), the Ricardian
approach had been dormant for years, having been
largely supplanted by the Hecksher-Ohlin
framework. The Ricardian model had lost out not so
much because of poor empirical results but because
it had come be viewed as intractable for all but
illustrative purposes. By introducing a continuum of
goods (rather than a discrete number), DFS were
able to analyse elegantly a broad range of
comparative static questions that had previously
seemed unapproachable. DFS showed, for example,
how to mobilize the combination of comparative
advantage and trade costs to cndogenize the
dividing line between ‘traded’ and ‘non-traded’
goods, and how to analyse the classic ‘transfer’
problem where one country owes debt to another.
Although at first only a trickle of papers followed
DFS, the power oftheir continuum specification has
led to a recent explosion of related research. DFS
have become the stalling point for a number of
applied papers (see, for example, Copeland and
Taylor ). In addition, DFS form the basis for a broad
range of empirical papers (see, for example, Eaton
and Kortum ; Kehoe and Ruhl ; Kraay and Ventura ;
Kei-Mu Yi ; Ghironi and Melitz ; see also Feenstra
and Hanson ). As the empirical work following DFS
deepens, it is fair to say that trade economists have
increasing faith in the fundamental underpinnings of
the model.
Broader Contributions
Aside from his path-breaking research, Dornbusch
made important contributions to economics in a
number of other dimensions. His intermediate
undergraduate textbook with Stanley Fischer, Mac-
roeconomics, written in the mid-1970s, became a
worldwide best-seller. The book was really the first
to integrate modem supply-side economics into the
standard demand-driven framework of the day. As
such, students were able to gain a far deeper
3054
understanding of problems such as the effects of oil
price shocks.
Dornbusch was enormously influential as a
graduate teacher at MIT. At his regular early-
morning international economics ‘breakfasts’,
Dornbusch would dissect recent models and serve
up provocative questions in a fast-paced freewheel-
ing style; many students remember these unique
seminars as their most influential experiences as
Ph.D. students. Dornbusch served as thesis advisor
to scores of economists (as noted earlier, more than
125 in all), including Jeffrey Frankel, Paul
Krugman, Maurice Obstfeld and Kenneth Rogoff.
His dynamic, Socratic lecturing style also attracted
students from outside MIT to his advanced graduate
classes, including the likes of Jeffrey Sachs and
Lawrence Summers. Many Dornbusch students
went on to become finance ministers and heads of
central banks throughout the world.
Through clear and incisive policy analysis
embodied in editorials, speeches, and private meet-
ings, Dornbusch exercised an enormous influence
on global macroeconomic policy. He was a frequent
guest of leading government officials throughout
the world, who greatly valued and respected his
advice. Arguably, no other recent economist has
had so great an impact on the global macroeco-
nomic policy debate, especially in emerging mar-
kets such as Brazil, Korea and Mexico, but also in
more advanced countries such as Italy and Ger-
many. Notably, in his later writing he succeeded in
drawing ever more concrete insights from contem-
porary academic research, displaying a magnificent
ability to translate complex theoretical models into
ideas of immediate practical relevance. For exam-
ple, his 1994 Brookings paper (with Alejandro
Werner) argued that Mexico’s pegged exchange rate
had become overvalued to an extent that was
unsustainable. Dornbusch’s comments on markets
prior to the currency collapse at the end of 1994
were highly influential. He also advanced a number
of innovative ideas for dealing with international
debt problems. His policy analysis was notable in
that he managed to adopt strong views while con-
tinuing to be perceived as an independent and
objective thinker. Over the last ten years of his life,
Dornbusch became especially well-known for his
monthly ‘Economic Perspectives’ newsletter,
Dornbusch, Rudiger (1942-2002)
which covered with panache a broad range of top-
ical global economic problems. One innovative
idea, first developed in the newsletter and then
formally published in his ‘Primer on Emerging
Market Crises’ (2002) was to apply ‘value at risk’
analysis to the balance sheet of a country. In his
primer, he wrote:
.. .the right answer to crisis avoidance is controlling
risk. The appropriate conceptual framework is value
at risk (VAR) - a model-driven estimate of the
maximum risk for a particular balance sheet situation
over a specified horizon. There are surely genuine
issues with the specifics of VAR surrounding
modelling as has been widely discussed with respect
to bank risk models used for meeting BIS
requirements. But just as surely there is no issue
whatsoever in recognizing that this general approach
is the right one. If authorities everywhere enforced a
culture of risk-oriented evaluation of balance sheets,
extreme situations such as those of Asia in 1997
would disappear or, at the least, become a rare
species. (2002, pp. 743-54)
In this short space it has not been possible to do
frill justice to the range and breadth of Dombusch’s
contributions. But I hope the reader has gained some
perspective on why he will have a lasting influence.
See Also
► Comparati ; vantage
► ixchang /olatility
► ixtremal Q ;utiles and Value-at-I
► i co-Rican momics
Selected Works
1973. Devaluation, money and nontraded goods.
American Economic Review 63, 871-880. 1974a.
Real and monetary aspects of the effects of
exchange rate regime changes. In National
Monetary Policies and the International Financial
System, ed. R. Aliber. Chicago: University of
Chicago Press.
1974b. Tariffs and nontraded goods. Journal of
International Economics 4, 177-185.
1976. Expectations and exchange rate dynamics.
Journal of Political Economy 84, 1161-1176.
Double-Entry Bookkeeping 3055

1977. (With S. Fischer and P. Samuelson.) Com-

parative advantage, trade and payments in a Double-Entry Bookkeeping
Ricardian model with a continuum of goods.
American Economic Review 67, 823—839.
1980. Open economy macroeconomics. New York:
Basic Books.
1983. Real interest rates, home goods and optimal
external borrowing. Journal of Political Econ-
omy 91, 141—153.
1987. Exchange rates and prices. American Eco-
nomic Review 77, 93—106.
1990. (With S. Fischer.) Macroeconomics, 5th
edition. New York: McGraw-Hill.
1994. (With A. Wemer.) Mexico: Stabilization,
reform and no growth. Brookings Papers on
Economic Activity, 1994: 1, 253-315.
2002. A primer on emerging market crises. In
Preventing currency crises in emerging markets,
ed. S. Edwards and J.A. Frankel. Chicago: The
University of Chicago Press.
Bibliography
Basil S. Yamey
Abstract
Double-entry bookkeeping is a system for
arranging and organizing accounting informa-
tion. It requires that each transaction (or other
change) recorded in the accounting system must
be recorded twice, and for the same money
amount, once in debit form and once in credit
form. Because it is concerned with the
organization of information rather than with the
scope and detail of that information, the system
of double-entry bookkeeping is highly
adaptable. It neither generates nor requires any
particular set of valuation rales or profit con-
cepts, and it is compatible with different treat-
ments for changes in the value of money.

Copeland, B., and M. Scott Taylor. 1994. North-South trade Keywords

and the environment. Quarterly Journal of Economics
109: 755-787.
Eaton, J., and S. Kortunr 2002. Technology, geography, and
trade. Econometrica 70: 1741-1779.
Feenstra, R., and G. Hanson. 1996. Globalization,
outsourcing, and wage inequality. American Economic
Review 86: 240-245.
Ghironi, F., and M. Melitz. 2005. International trade and
macroeconomic dynamics with heterogeneous firms.
Mimeo: Harvard University.
Kehoe, T.J., and K.J. Ruhl. 2002. How important is the new
goods margin in international trade? Mimeo: University
of Minnesota.
Kei-Mu, Yi. 2003. Can vertical specialization explain the
growth of world trade? Journal of Political Economy
111:52-102.
Kraay, A., and J. Ventura. 2002. Trade integration and risk
sharing. European Economic Review 46: 1023-1048. ""
Meese, R., and K. Rogoff. 1983. Empirical exchange rate
models of the seventies: Do they fit out of sample?
Journal of International Economics 14: 3-24.
Obstfeld, M., and K. Rogoff. 1996. Foundations of
international macroeconomics. Cambridge, MA: MIT
Press.
Rogoff, K. 2002. Dombusch’s overshooting model after 25
years: IMF Mundell-Fleming Lecture. International
Monetary Fund Staff Papers 49 (Special Issue), 1 35
(including remarks by Rudiger Dombusch).
Accounting; Assets and liabilities; Balance-
sheet equation; Double-entry bookkeeping;
National accounting; Pacioli, L.; Sombart, W.;
Spengler, O.; Transaction analysis; van Gezel,
W.
JEL Classifications
M4
Firms of all kinds need, in different degrees, to
maintain records of their transactions with other
firms and persons, of the debts they owe or are
owed, and of their assets. The records they keep for
this purpose constitute their accounting records.
Traditionally they have consisted of account-books
of various kinds, but they can take the form also of
magnetic tapes and so on. If the records are kept on
a systematic basis, one can speak of an accounting
system. From the accounting records one can
prepare a variety of accounting statements in which
the detailed accounting information is rearranged,
regrouped and presented in summary form. The
balance
3056
sheet and the profit-and-loss (or income) account or
statement are important examples of such
accounting statements.
Double-entry bookkeeping is a system or
method for the arrangement and classification of
accounting information. It developed in Italy, pos-
sibly in the second half of the 13th century. A
description of the system was first published in
Venice in 1494 as one part of a famous compen-
dium of mathematical and commercial information:
Luca Pacioli’s Suninia de Arithmetica Geometric/
Proportioni et. Proportionality. Knowledge of the
double-entry system spread gradually from Italy to
the rest of I-’ll rope by way of commercial contacts,
schools and published treatises. It is not possible to
establish how widely the system was used by
merchants and others, say, in the 18th century. But
by the late 19th century it had become the standard
system for accounting records. Today it is used by
virtually all corporate enterprises and many other
firms as well as non-profit-making organizations in
the West and also elsewhere. It has also proved
suitable to serve as a useful scaffolding for the
construction of the national income and related
accounts for countries or regions.
Double-entry bookkeeping is no more than a
system for arranging and organizing accounting
information. It does not itself define the scope and
detail of that information. Thus, for example, the
double-entry system does not require that all
transactions with third parties should be recorded,
although it is the convention now to record all of
them. What is more important, it does not prescribe
which occurrences or changes that do not involve
external transactions should be recorded in the
accounts. Thus it does not prescribe whether
changes in the value of the film's assets should be
recorded, how they should be determined, or how
they should be recorded. Double entry neither
generates nor requires any particular set of valua-
tion rules or profit concepts. Different valuation
bases or conventions, and different treatments for
changes in the value of money, are all compatible
with the use of the double-entry system. The system
itself is highly adaptable, since it is concerned with
arrangement and organization rather than with
scope and content. Its adaptability has made
Double-Entry Bookkeeping
it possible for it to serve as the basis for arranging
the records needed by the relatively small-scale
merchants in the early modem period of economic
expansion as well as for those of the largest cor-
porate enterprises operating today. But this does not
mean that asset values were recorded and profits
calculated in the same way by 17th-century
merchants as they are by today’s corporate enter-
prises. In fact, 17th-century merchants used several
alternative bases for recording changes in asset
values. And some of these would not be used by
companies today.
Moreover, although all the companies within the
same jurisdiction are subject to the same laws and
the same institutional constraints (for example,
those imposed by the stock-market authorities and
those reflecting professional accounting standards),
there is still scope for considerable variation in the
determination and statement of accounting profits
and asset values. However, because of
developments in legislation and in the other
constraining forces operating on corporate
enterprises, it is no longer the case that a company
chairman in the United Kingdom would be able to
say (as Arthur Chamberlain, chairman of Tube
Investments said in 1935) that he ‘would almost
undertake to draw up two balance-sheets for the
same company, both coming within an auditor's
statutory certificate, in which practically the only
recognizable items would be the name and the
capital authorised and issued’.
Double entry requires that each transaction (or
other event) recorded in the accounting system must
be recorded twice, and for the same money amount,
once in debit form and once in credit form. In
double cntiy, as Pacioli expressed it, ‘all the entries
placed in the ledger must be double, that is if you
make a creditor (cntiy) you must make a debtor
(cntiy)’. The debit and credit entries arc made in the
ledger, on the basis of the information entered in
preliminary records. The ledger, which may for
convenience be subdivided into a series of
specialized ledgers, consists of a number of ledger
accounts, pertaining, for example, to particular
debtors or creditors, particular assets or particular
categories of expenditure. It is the convention that
the debit entry is made on the
Double-Entry Bookkeeping
left-hand (debit) side of the appropriate ledger
account, and the corresponding off-setting credit
entry on the right-hand (credit) side of the other
appropriate ledger account.
The duality of entries for each transaction (or
other recorded event) ties together the ledger
accounts into an interlocking system of recorded
information. Moreover, as each transaction gives
rise to two equal but opposite entries, the system of
accounts (if properly kept) is always in balance or
equilibrium. The total of debit entries must be equal
to the total of credit entries. Similarly, the total of
the balances on all ledger accounts that have debit
balances must be equal to the total of the balances
on all the remaining ledger accounts that have credit
balances. (If debit balances are taken as positive
amounts and credit balances as negative amounts,
the algebraic sum of the balances on all ledger
accounts is zero.) The equality of debits and credits
is the basis for the trial balance. This is a list of the
balances on all open (that is, unbalanced) accounts
in the ledger, distinguishing between debit and
credit balances. If the trial balance does not balance,
there is some error in the ledger. Postlethwayt in his
Dictionary (1751) wrote of the ‘agreeable
satisfaction’ of getting a trial balance to balance,
and said that the trial balance will ‘shew you that
this [double entry], of all methods, is the most
excellent’. The fact that a trial balance does not
balance is proof that the ledger does contain some
error. The converse is, of course, not correct.
Roger North, son of the prominent Turkey
merchant Sir Dudley North, wrote in 1714 as
follows: ‘The making true Drs. (debtors) and Crs.
(creditors) is the greatest Difficulty of Accompting,
and perpetually exerciseth the Judgment; being an
Act of the Mind, intent upon the Nature and Truth
of Things. ’ Writers of instructional books on
bookkeeping and accounts through the centimes
have devised various lists, rales or approaches to
help the accountant decide which debit and credit
entries he should make for the various categories of
transaction.
An early rale, widely used, was as follows
(taken from a verse, ‘Rules to be Observed’, in a
book of 1553 by James Peele):
3057
To make the thinges Received, or the receiver,
Debter to the thinges delivered, or to the deliverer.
This rule is obviously readily applicable to
many categories of transaction. If cash is received
from a debtor, debit the cash account; and credit the
debtor’s account. If office furniture is bought on
credit, debit the furniture account; and credit the
supplier’s account. If the owner withdraws cash
from the business, debit the capital (that is,
owner’s) account; and credit the cash account. But
it is evidently a straining of the language to say,
when an amount is written off the book value of,
say, a ship, in order to reflect diminution of value
due to wear and tear, that the profit-and-loss
account, which is to be debited, ‘receives’ some-
thing that has been ‘delivered’ to it by the ship
account. Teachers and textbook writers not sur-
prisingly looked for a rale that is robust enough to
cover comfortably all transactions and events to be
recorded, and to indicate unambiguously in each
case where the debit and where the credit are to be
placed.
The most common rale or approach adopted
today in transaction analysis in the double-entry
system derives from the so-called balance-sheet
equation. The earliest formulation of this approach
can be traced to the work of a Dutchman, Willem
van Gezel, published in 1681.
The basic balance-sheet equation is:
Owner sEquity(orthefirm snetworth) = Assets -
Liabilities Net Assets ; or Owner s Equity +
Liabilities Assets.
The ledger contains accounts for the various
assets and liabilities; and there are accounts in it for
the capital contributed or withdrawn by the
owner(s) and for any increases (decreases) in ‘net
worth’ resulting from the activities of the firm. In
the double-entry system, increases in assets are
indicated by debits to an asset account - the extent
to which assets are subdivided into separate ledger
accounts is for each firm to decide. Conversely,
decreases in assets are recorded as credits to asset
accounts. The total of a firm’s assets is represented
by the total of claims on those assets; namely, its
liabilities (that is, its debts to third parties) and its
owner’s equity. The total of these claims must be a
credit amount that equals the debit amount
3058
representing the assets. An increase (decrease) in a
claim is therefore represented by a credit (debit) in
a liability account or an equity' account. (Again, the
extent to which claims are subdivided into various
ledger accounts is a matter for each film to decide.
As regards the equity element, it is common for a
ledger to contain separate accounts for each major
category of business expenditure and income, a
trading account, perhaps subdivided by type of
activity', for showing the gross profit, and a profit-
and-loss account to bring together the results from
all the subordinate ledger accounts.)
Transaction analysis follows readily. The pay-
ment of salaries reduces the asset 'cash' and reduces
the owner’s equity', since the payment, taken by
itself, represents a loss to the film: hence, debit the
salaries (eventually, profit-and- loss) account; and
credit the cash account. The depreciation of an asset
likewise reduces an asset and reduces the equity':
debit the depreciation account (eventually profit-
and-loss) account; and credit the ship account.
As has already been emphasized, the double-
entry system does not itself dictate whether or in
what circumstances increases or decreases in assets
are to be recognized in the accounts. Neither does
the system dictate the basis on which, or the
circumstances in which, assets are to be revalued in
the accounts. Decisions of these kinds are
accounting decisions; and whenever such decisions
are taken, the double-entry system of recording will
accommodate them in accordance with its own
logical structure. It follows from this that, although
the value of the owner’s equity in the ledger will
always be equal to the value of the firm’s net assets
(that is, assets minus liabilities to those outside the
firm) as stated in the accounts, those two values
depend on the bases on which the values of assets
are stated in the accounts.
Subject to this crucial qualification, it follows
from the equilibrium feature of the double-entry
system that the change (increase or decrease) in the
value of the net assets of a firm over a period will
be reflected as entries in the various ledger accounts
that represent the owner’s equity. Those entries in
the various equity accounts that relate to
Double-Entry Bookkeeping
the film's operations, when they are brought
together in the profit-and-loss account, yield a
balance that is equal to the change in the value of
the net assets over the period. It is the profit (loss)
for the period. This profit is equal to the change in
the value of the net assets over the period
(allowance being made for any contributions or
withdrawals of assets by the owner). It may be
noted that the same profit figure would be
established if one took the difference between the
totals of two inventories of the film's net assets
taken, respectively, at the beginning and at the end
of the period, provided that the same valuations
were used and the same allowance made for the
owner’s contributions and withdrawals. The method
of profit calculation by means of successive
inventories of assets and liabilities was widely used
in the past. The surviving 16th- century records of
the large-scale commercial, financial and mining
enterprise of the Fugger family of Augsburg
provide examples of this procedure.
The equality' - Profit (Loss) = Change in Net
Assets - evidently holds only if all the changes
recorded in asset and liability accounts (other than
the owner’s contributions or withdrawals) are also
recorded in equity accounts that, in Uim, are closed
into the profit-and-loss account. In contemporary
corporate financial accounting it is permissible to
allow the counter-entries representing certain
changes in asset values, depending upon the
circumstances, to bypass the profit-and-loss account
(for example, by recording these changes as debits
or credits to one or other reserve account). This
practice breaks the nexus between changes in net
asset values and profits. It does, however, allow
more ‘realistic’ values to be used in asset accounts
where, otherwise, their use might produce
‘distortions’ in the profit figures that could mislead
users such as investors and investment advisers.
Both ‘realistic’ and ‘distortions’ are words that give
rise to much debate in accounting circles. The
double-entry recording system can accommodate
the practice of bypassing the profit-and-loss account
as comfortably as it can the alternative. The system
itself imposes no discipline or constraint upon
accountant or management - except the constraint
that
Douglas, Clifford Hugh (1879-1952) 3059

for each transaction or change recorded in the firm’s system for recording and classifying accounting
accounting system, equal but offsetting debit and data that has been used increasingly over a period of
credit entries have to be made in accounts in the six centuries must indeed have substantial practical
ledger. merit. Double entry is a useful and versatile method
The German economic historian, Werner for organizing accounting data, its value increasing
Sombart, claimed that ‘capitalism without double- with the volume and complexity of the data to be
entry bookkeeping is simply inconceivable’, and organized. In turn, the efficient organization of data
that double-entry was one of the most significant helps management at various levels in many ways,
inventions or creations of the human spirit. In more notably in large organizations. But its
similar vein, Oswald Spengler asserted that the contribution to efficiency does not proceed along
creator of double-entry bookkeeping could take his the lines emphasized by Sombart.
place worthily beside his contemporaries Columbus
and Copernicus. These scholars evidently attributed See Also
to the double-entry system a role that goes well
beyond what one might think appropriate to ascribe ► accounting and Economics
to a system of organizing and arranging accounting ► vssets and Liabilities
data. In a nutshell, Sombart argued that, historically, ► ombart, Werner (1863
the double-entry system opened up possibilities and
provided stimuli that enabled capitalism to develop Bibliography
fully. It clarified the acquisitive ends of commerce
and provided the rational basis on which this Yamey, B.S. 1964. Accounting and the rise of capitalism.
acquisition could be carried on. It provided the basis Journal of Accounting Research 2: 117-136. for a
discussion of Sombart’s views on double-entry book-
for the continued rational pursuit of profits, and
keeping and capitalism.
virtually compelled its users to pursue the
acquisition of wealth. It also enabled the firm or
enterprise to be separated from its owners, thus
facilitating the development of corporate
enterprises. Douglas, Clifford Hugh (1879-1952)
These views are in their details either untenable
or grossly exaggerated. To note only a few points: David Clark
the profits of an enterprise and its capital employed
can be calculated without doubleentry bookkeeping;
joint-stock companies, such as the Dutch East India
Company, have existed and flourished without
Major Douglas, the founder of the Social Credit
double-entry bookkeeping; 16th- and 17th-century
movement, was bom in Stockport, Cheshire, in
merchants, like the Fugger, who did not use the
1879. After a period at Pembroke College, Cam-
system do not seem to have been any less
bridge, he trained as an engineer and then served
acquisitive, rational and successful than those who
with the Royal Flying Corps. He died at Dundee,
did use the system; and the adoption of the double-
Scotland, in 1952.
entry system could not have changed, or even have
Major Douglas is best known for his A + B
reinforced, the temperament, commercial acumen,
theorem, which his followers used to impress
motivation or goals of those who adopted it for
laypersons and exasperate academic economists. It
organizing their accounting records.
was based on the claim that all productive
To reject grandiose claims made for doubleentry
organizations make two kinds of payments: Group
bookkeeping is not to deny the more workaday
A payments, made up of wages, salaries and
usefulness of the system. A method or
dividends; and Group B payments, made up of all
other payments to banks and suppliers of materials.
In his own words (Douglas 1924):
3060
Since all payments go into prices, the rate of flow of
prices cannot be less than A plus B. Since A will not
purchase A plus B, a proportion of the product at
least equivalent to B must be distributed by a fomr of
purchasing power which is not comprised in the
description grouped under A.
By first reducing prices below cost to the indi-
vidual consumer and then making up this difference
between pri ce and cost by a Treasury issue to the Douglas, Paul Howard (1892-1976)
producer, Douglas argued that such an issue of
‘Social Credit’ would enable underconsumption to Colin G. Clark
be eliminated without inflation. The antisocialist
Douglas appeared oblivious to the fact that his
scheme would have required an army of inspectors
to fix and supervise the huge number of individual
price reductions involved.
Social Credit ideas had the largest following in
the Dominion economies of Canada, Australia and
New Zealand. The province of Alberta had a Social
Credit government between 1935 and 1971 and
British Columbia one from 1952 to 1972. In other
countries, his followers ranged from the ‘Red Dean’
(the Very Reverend Hewlett Johnson, Dean of
Canterbury) through to the neo-fascist author Ezra
Pound.
See Also
► A-..;--.':., C- AO:
► iA; v::./ T: v-::
Selected Works
1920. E c o n o m i c d e m o c r a c y ’ . London: Stanley Nott.
1921. C r e d i t p o w e r a n d d e m o c r a c y . London: Cecil Palmer.
1924. S o c i a l c r e d i t . London:
Spottiswoode.
1931. W a r n i n g d e m o c r a c y ’ . London: Stanley Nott.
References
Dobb, M.H. 1936. Social credit discredited: Being an to make an estimate of the movement of real wages.
examination in terms of political economy of the much
advertized nostrums of Major Douglas, which are subject
to a devastating analysis and found wanting as a solution
to the troubles of our time. London, no publisher given.
Douglas, Paul Howard (1892-1976)
Mairet, P. (ed.). 1934. The Douglas manual. London: Stanley
Nott.
McConnell, W.K. 1932. The Douglas credit scheme: A simple
explanation and criticism. Sydney: Angus & Robertson.
Keywords
Agricultural economics; Climacteric (of 1896-
1914); Cobb, C. W.; Cobb-Douglas functions;
Douglas, P. H.; Phelps Brown, H.; Real wage
growth
J EL Classifications
B31
Bom in 1892 in Salem, Massachusetts, Paul
Douglas attended Bowdoin College in Maine (BA,
1913) and Columbia University (Ph.D., 1921). After
holding a number of teaching posts between 1916
and 1920, he joined the faculty of the University of
Chicago where he remained (apart from service in
the Second World War) until 1948, when he
became a United States Senator from Illinois. After
his retirement from the Senate in 1966, he taught at
the New School for Social Research for two years
(1967-9).
Paul Douglas first became well known for his
Eyre & massive theoretical and factual studies (for
example, 1930) of all the available information on
wages in the United States from 1890. This work
required laborious following up of old, obscure
records, and repairing gaps in the available
knowledge, such as domestic service wages.
Douglas also collected information on prices so as
In Britain there was almost complete cessation
of the growth of real wages between 1896 and 1914.
Understandably, it was a period of growing social
tension. Sir Henry Phelps Brown called it
Douglas, Paul Howard (1892-1976)
the ‘climacteric’. We still do not really understand
its cause; there was some sociological evidence
about the deterioration of the quality of business-
men. D.H. Robertson found at least a partial
explanation in economic causes, namely, that, of the
two leading British export industries, cotton was
produced under constant returns and coal under
diminishing returns.
This problem remains of primary interest to
economic historians, and naturally they enquire
whether there is any evidence of a similar ‘climac-
teric’ in other countries. In Germany there was a
slowing down of the rate of rise in real wages, but
not very marked. Douglas’s American data likewise
do not show such a ‘climacteric’. Recent research,
however, has thrown some doubt not on Douglas’s
wage data, but on his price data; and perhaps there
was some slowing down of the rate of growth of
real wages.
Douglas became famous to the whole economic
world through the ‘Cobb-Douglas function’ (for
example, 1934). Working in conjunction with
Charles W. Cobb, a mathematician from Amherst
College, and using Massachusetts State annual
factory returns, Douglas in 1928 established the
following relation: Let product be P, labour input L,
capital input C, and k a constant. Then P kLaCb.
(The same formula, with land in place of capital,
had already been used by Wicksell - for example,
1900 - but he gave it neither theoretical nor
empirical development.)
We may, if we wish, constrain a and b to add up
to 1; but we get much the same results
unconstrained. If a and b add up to more than 1 this
is an indication of economies of scale (increasing
returns) - a uniform increase in the quantities of
inputs giving a more than proportionate increase in
product.
Annual data, which many economists have been
using, give results mainly dependent on fluctuations
in the short-period business cycle - which is not
what we want at all. It is only when we have data
for such a long period as to make it possible to
average out the business cycle that we can draw
conclusions about productivity. This has been done
by Solow in the United States, Aukrust in Norway,
and Niitamo in Finland. In each case it was found,
in the long ran, that the
3061
product was rising much more rapidly than expected
from inputs and their exponents. This difference is
generally held to be due to technical advance,
though some look for economies of scale. Some
difficult but promising work by Denison further
analyses the labour input by numerous categories,
male and female, adult and juvenile, and various
levels of education. These methods reduce the
unknown factor - but it does not disappear.
Differentiating the Cobb-Douglas formula to
obtain marginal productivities, then aggregate
earnings of the factors should be proportional to a:b
- assuming that each factor is remunerated
according to its marginal productivity. When he first
made this calculation (so he told me), Douglas fully
expected the aggregate income of labour to be
below that indicated by its marginal productivity.
He was surprised, however, to find that it was
almost exactly what was to be expected - about 75
per cent of the product.
The Cobb-Douglas formula has had abundant
application in agricultural economics, especially for
cross-section studies, where each farm may be
considered an independent piece of evidence. Land
is introduced as a factor, and also data for other
inputs - fertilizers, insecticides, and so on - even (in
one study in Sweden) the age of the farmer - a
negative factor.
Douglas was very much a political economist.
Organized labour in the United States did not
attempt to form a political party of its own as in
Britain, but instead played the two existing parties
off against each other in demanding concessions.
But in the 1920s this was not frilly agreed. The
other element in the population with a grievance
against the current state of affairs was the farmers,
and an attempt was made to form a Farmer-Labour
political party. Douglas took an active part in these
negotiations, and was national treasurer of the
organization. But with the Roosevelt reforms of the
1930s the prospects of a Farmer-Labour party died
away.
Chicago had acquired a worldwide reputation
for corruption and crime; and the ruling Democratic
Party considered that its ‘image’ would be improved
by an upright professor of economics on the city
council. Douglas assured me that some
3062 Du Pont de Nemours, Pierre Samuel (1739-1817)

improvement had taken place, though less than was 1939b. (With H.G. Lewis.) Some problems in the
hoped for. Later, the despotic Mayor Daley measurement of income elasticities.
achieved a real reduction in crime. But once I asked Econometrica 7: 208—220.
Douglas whether, if I wished to set up a milk 1939c. (With M. Bronfenbrenner.) Cross-section
distribution business in Chicago, he could guarantee studies in the Cobb-Douglas function. Journal
my safety. He replied that, ‘regrettably’, he could of Political Economy 47: 761—785.
not. 1947. (With E.H. Schoenberg.) Studies in the
Douglas was a Quaker, and in the First World supply curve of labour; the relation in 1929
War applied for exemption from military service on between average earnings in American cities
religious grounds. But in the Second World War he and the proportions seeking employment.
felt very differently. In spite of his age, he obtained Journal of Political Economy 45: 45-79.
a commission in the marines through President 1948. Are there laws of production? American
Roosevelt’s personal intervention, and took part in Economic Review 38: 1-41.
the bloody landing on Iwojima. He sustained an 1 9 5 2 . Ethics in government. The Godkin
injury to his hand which was with him for the rest lectures at Harvard University, 1951.
of his life. Cambridge, MA: Harvard
From city councillor he advanced to become University Press.
Senator for Illinois. On the very day that he arrived 1 9 7 2 . In the fullness of time: The memoirs of
in Washington he found a vanload of furniture Paul
which had been offered to him as a gift He sent it H. Douglas. New York: Harcourt Brace
back. This episode prompted him to write a little Jovanovich.
book, Ethics in Government (1952). He saw no 1976. The Cobb-Douglas production function once
harm in the small presents customarily exchanged again: Its history, its testing and some new
among businessmen and politicians - calendars, empirical values. Journal of Political Economy
cigars, and so on - but instructed his staff to return 84: 903-915.
any present valued at over four dollars.
Bibliography
See Also
Wicksell, K. 1900. Marginal productivity as the basis of
distribution in economics. Ekonomisk Tidskrift. English
► Cobb-Douglas Functions trans. in K. Wicksell: Selected papers in economic theory,
ed. E. Lindahl. London: Allen &
Unwin, 1958.
Selected Works
1928. (With C.W. Cobb.) A theory of production.
American Economic Review 18(Suppl): 139-
165. 1930. Real wages in the United States, 1890- Du Pont de Nemours, Pierre Samuel
1926. Boston/New York: Houghton Mifflin (1739-1817)
Company.
1934. The theory of wages. New York: Macmillan. Peter Groenewegen
1 9 3 6 . Social security in the United States: An
analysis and appraisal of the federal social
security act. N e w York/London:
Whittlesey House/McGraw-
Keywords
Hill.
Advisers; Assignats; Du Pont de Nemours, P.
1939a. The effect of wage increases upon employ-
S.; Excise taxes; Mathematical economics;
ment. American Economic Review 29(Suppl):
Mercier de la Riviere, P.-P; Mirabeau, V. R.,
138-157.
Marquis de; Physiocracy; Quesnay, F.; Silver;
Turgot, A. R. J
Du Pont de Nemours, Pierre Samuel (1739-1817) 3063

of diagrams in economic argument and, most

JEL Classifications
importantly, as the editor of Quesnay and Turgot,
B31
whose works he helped to preserve. An assessment
of his work as economist needs to take all facets of
Economic writer and editor. Bom in Paris, he
his career into account, as the one full-length
trained for various occupations including medicine
attempt at this (McLain ) has in fact done.
and watch making. A pamphlet on taxation (1763)
Virtually all Du Pont’s economic work is char-
brought him in contact with Mirabeau and Quesnay,
acterized by dogmatic adherence to the Physiocracy
under whose guidance he wrote a work on the grain
developed by Quesnay and codified by Mercier de
trade (1764). He also befriended Turgot, with whom
la Riviere. Turgot criticized this ‘servitude to the
he diligently corresponded until Turgot’s death.
ideas of the master’ as totally inappropriate in mat-
From 1766 to late 1768 he edited the Journal de
ters of science (Schelle -23, vol. 2, p. 677). Despite
{’Agriculture in the Physiocratic cause, then the
such criticism Du Pont allowed his dogmatism to
Ephemerides until 1772. During this period he also
colour excursions into the history of economics (Du
published Quesnay’s economics under the title
Pont 1769) and, more importantly, his preparation
Physiocratie (Du Pont 1767) and summarized
of Tuigot’s works for the press (see Groenewegen ),
Mercier ( ), adding material on
particularly his editions of the Reflections (Turgot ).
the history of the new science (Du Pont 1768).
Two examples of his more novel contributions to
From the early 1770s he developed a career as
economics can be given. One is his use of diagrams
economic adviser through correspondence with the
in explaining economic policy, which Theocharis (
King of Sweden and the Margrave of Baden; the
, p. 60)
correspondence with the latter was subsequently
described as the first use of a diagram by a profes-
published (Knies ). In 1774 he was appointed tutor
sional economist for ‘illustrating an economic argu-
to the Polish royal family. On becoming controleur-
ment set out in essentially dynamic time’, thereby
general, Turgot required his friend’s assistance and
making Du Pont (1774) ‘the earliest French contri-
Du Pont was back in Paris by early 1775. Financial
bution of importance in mathematical economics'.
compensation for loss of his royal tutorship enabled
The problem analysed is the price effects of an
him to purchase landed property near Nemours.
excise reduction, the benefits of which are argued to
Turgot’s dismissal from office in 1776 did not end
accrue ultimately to the landowning class. The
Du Pont’s career in giving official economic advice;
excise reduction’s initial income effect on manufac-
a highlight of which is his influence on the 1786
turers and merchants allows them either to reduce
Anglo-French Commercial Treaty. Du Pont was
their own prices or to pay higher prices for raw
politically active in the French Revolution, serving
materials. By assuming this increased competition
from 1789 as Deputy for Nemours in the National
for raw materials to raise their price in each period
Assembly and becoming its President during 1790;
by three-fourths of the increase in the preceding
in 1794 to 1797 he was imprisoned for short
period, Du Pont shows how a new equilibrium price
periods. He migrated to the United States in 1799
will be reached which transfers the benefits from
but returned to Paris in 1802. From 1803 to 1810 he
excise reduction to the rural sector. His proof relies
served in the Paris Chamber of Commerce, and in
on the properties of diminishing geometrical
addition edited Turgot’s works (Du Pont 1808-11).
progressions which also formed the basis for much
In 1815 he returned to the United States and settled
of the analysis of the Tableau economique. Du
in Delaware, the town where his son Irenee had
Pont’s analysis of the inflationary consequences
started the gunpowder factory from which the Du
from issuing assignats is a second example.
Pont chemical conglomerate developed, and where
Although much of this is similar to Turgot’s ( )
he died in 1817. Du Pont is now mainly remem-
analysis, some of it is of interest in
bered as a major propagator of Physiocracy, an
explaining Smith’s version of the specie mechanism
early historian of economics, a pioneer in the use
to which Du Pont (1790, p. 28) explicitly refers.
Issuing paper money by assignats makes silver
3064
superfluous as a circulating medium; this drives die
metal out of die country because its only other use
is to be sold abroad (Du Pont 1790, p. 42), a specie
mechanism like Smith’s ( , pp. 293—4) that is
independent of relative price movements. Both
examples of his more original economics relate to
matters of economic policy and add force to the
claim by McLain ( , p. 255) that Du Pont
represents ‘the first important case of a professional
economist turned government policy-maker, a tra-
dition in which he would be followed by [many]
others
Selected Works
1763. Reflexions sur Vecrit intitule: Richesses de
Vetat. P a r i s .
1764. De Vexportation el de Importation des
grains. S o i s s o n s a n d P a r i s .
1767. Physiocratie, on Constitution Naturelle du
Couvernement le plus avantageux au genre
humain. L e y d e n a n d P a r i s .
1768. De l’origine et des progres d’une science
nouvelle. In Physiocrates, ed. E. Daire. Paris,
1846.
1769. Notice abregee des differents ecrits
modemes qui ont concours en France a former
la science de l’economie politique. In Oeuvres
Oeconomiques et Philosophiques de Frangois
Quesnay, ed. A. Oncken. Frankfuit/Paris, 1888.
1774. On economic curves, ed. H.W. Spiegel. Bal-
timore: Johns Hopkins. Reprints of Economic
Tracts, 1955.
1790. The dangers of inflation. Trans. E.E. Lincoln.
Boston: Kress Library Publications, 1950.
1808-11. Oeuvres de Turgot. 9 vols. Paris.
Bibliography
Groenewegen, P.D. 1977. The economics of A.R.J. Turgot.
The Hague: Nijhoff.
Rnies, K. 1892. Carl Friedrichs von Baden Brieflicher
Verkehr mit Mirabeau und Du Pont. Heidelberg: Carl
Winter.
McLain, J.J. 1977. The economic writings of Du Pont de
Nemours. Newark/London: University of Delaware Press.
Mercier de la Riviere, P.P. 1767. L’Ordre Naturel et
Essentiel des Societies politiques. Paris.
Dual Economies
Schelle, G. 1913-23. Oeuvres de Turgot et documents le
concernant. 5 vols. Paris: F Alcan.
Smith, A. 1776. In An inquiry into the nature and causes of
the wealth of nations, ed. R.H. Campbell and A.S.
Skinner. Oxford: Clarendon. 1976.
Theocharis, R.D. 1961. Early developments in mathematical
economics. London: Macmillan.
Turgot, A.R.J. 1749. Letter on paper money. In Groenewegen
(1977).
Turgot, A.R.J. 1766. Reflections on the production and
distribution of wealth. In Groenewegen (1977).
Dual Economies
David Vines and Andrew Zeitlin
Abstract
Dual economies have asymmetric sectors, the
interaction between which influences the path of
development. These are typically a rural,
traditional, or agricultural sector on one hand,
and an urban, modem, or industrial sector on the
other. The relevant asymmetries are not merely
technological but also include institutional,
behavioural, and informational aspects. Modem
treatments have grown out of the work of W.
Arthur Lewis, whose model was based on the
existence of surplus labour in agriculture.
Subsequent authors have considered the
implications of alternative assumptions for the
development of a dual economy.
Keywords
Agriculture and economic development;
Asymmetric information; Capital accumulation;
Credit markets in developing countries;
Development economics; Disguised unem-
ployment; Dual economies; Engel’s law; Harris-
Todaro hypothesis; Hecksher-Ohlin trade theory;
Immigration; Kuznets curve; Labour surplus
economies; Lewis, W. A.; Malthus’s theory of
population; Neoclassical growth theory;
Outsourcing; Pure surplus labour; Rural-urban
migration; Scissors problem; Subsistence; Terms
of trade; Urban unemployment; Wage
differentials
Dual Economies 3065

order to facilitate the development of urban

JEL Classifications
manufacturing. These policies led to famine and to
03
the deaths of approximately 30 million people. Thus
both theorists and policymakers have long
Dual economies have asymmetric sectors, the
recognized that, in an economy with two very
interaction between which influences the path of
different sectors, growth prospects hinge on how
development. W. Arthur Lewis introduced this idea
these sectors interact.
in his paper, ‘Economic Development with Unlim-
In his Nobel Prize autobiography, Lewis ( )
ited Supplies of Labour’ (Lewis ), which earned him
writes that his interest was in the ‘fundamental
the Nobel Prize for Economics in 1979. That paper
forces determining the rate of economic growth’.
contains two theoretical models, both designed to
But he was not satisfied with the neoclassical model
explain the intrinsic problems of underdevelopment
of growth that was emeiging at the time (Solow ;
When the prize was awarded, Ronald Findlay wrote
Swan ), out of the work of Roy F. Harrod ( ) and
that ‘a large part of... development economics ... can
Evsey D. Domar
be seen as an extended commentary on the meaning
(1945). That neoclassical framework aimed to pro-
and ramifications [of this article]’ (Findlay , p. 64).
vide a general theory of growth. But to Lewis it
Here we focus primarily on the first of Lewis’s two
seemed inadequate because it did not deal with
models of dualism - that of a single underdeveloped
interactions between the industrial and the agricul-
economy. We describe that model, trace the
tural sectors: in Lewis’s words, this model
evolution of the ideas which grew from it, and
contained no discussion ‘of what determines the
discuss the continuing importance of these ideas in
relative price of steel and coffee [namely, of indus-
the study of economic development.
trial goods and agricultural goods]. The approach
Long before Lewis wrote his article, there had
through marginal utility made no sense to me. And
been much thinking about ‘dual’ economies, con-
the Heckscher-Ohlin framework could not be used,
ceived of as economies with both an industrial
since that assumes that trading partners have the
sector and an agricultural sector. Adam Smith and
same production functions, whereas coffee cannot
David Ricardo both focused on the interaction
be grown in most of the steel-producing countries.’
between these sectors during the Industrial Revo-
Furthermore, the neoclassical theory seemed
lution; for Ricardo the outlook for industrial growth
inadequate to him for historical reasons:
was ‘dismal’ because of diminishing returns in
‘[apparently, during the first fifty years of the
agriculture (see Hicks ; Pasinetti ). In the early 20th
industrial revolution, real wages in Britain remained
century, there was an extended discussion in the
more or less constant while profits and savings
Soviet Union of the ‘scissors problem’, concerning
soared. This could [also] not be squared with the
the determination of the terms of trade between
neoclassical framework, in which a rise in
these two sectors. Evgeny Preobrazhensky ( )
investment should raise wages and depress the rate
argued that a
of return on capital’ (Lewis ).
decrease in the relative price of agricultural goods
Then, Lewis continues:
could be used to stimulate industrial investment; One day in August, 1952, walking down the road in
others replied that sufficient agricultural goods Bangkok, it came to me suddenly that both problems
would not be available at lower relative prices and have the same solution. Throw away the neoclassical
that these goods would need to be seized by force, assumption that the quantity of labour is fixed. An
‘unlimited supply of labour’ will keep wages down,
something which the collectivization of agriculture producing cheap coffee in the first case and high
made possible (see Sah and Stiglitz ). And during profits in the second case.
the Great Leap Forward in China in the 1950s, The result is a dual (national or world) economy,
Chairman Mao attempted to confiscate an where one part is a reservoir of cheap labour for the
other. The unlimited supply of labour derives
increasing quantity of primary goods from the ultimately from population pressure, so it is a phase
Chinese countryside in in the demographic cycle. (Lewis p. 397)
3066
This key insight launched Lewis on the journey
that led to his famous article. Spelling out the
implications of his insight led him to use the term
‘dualism’ to describe economies in which there are
differences between industrial and agricultural
sectors that cannot be adequately explained by
differences in production technologies or in factor
endowments, in the manner normally used by
economists.
The Lewis Model
Lewis identified three such differences between
industry and agriculture, which we term
‘asymmetries’ in this article (following Kanbiu and
McIntosh ).
First, there are technological differences
between the sectors. Labour is used in each sector.
In agriculture it is combined with land in produc-
tion, whereas industrial goods are produced by
combining labour with reproducible capital.
Moreover, industrial goods can be consumed or
invested, whereas agricultural goods can only be
consumed.
Second, there are organizational differences
between the sectors. The large, rural agricultural
sector functions on traditional lines and is primarily
based on subsistence; industrial production happens
in a modem, market-oriented sector, located in
towns and cities. There is ‘an unlimited supply of
labour, available at [a] subsistence wage’ (Lewis ,
p. 139) to both sectors. Lewis interprets the word
‘subsistence’ broadly. The level of the wage is
determined in some way by conventions in the
underdeveloped agricultural sector. Lewis is non-
committal as to whether wages in this sector are set
according to actual subsistence needs, or living
standards, or workers’ average product. The central
idea is that workers are paid above their marginal
product. Labour can be transferred from agricultural
sector to the industrial sector by the migration of
workers to towns and cities. The overall stock of
labour in the economy is normally fixed in supply
(though Lewis, like Ricardo, did sometimes allow
for Malthusian features). Workers in the cities are
paid not
Dual Economies
much more than the subsistence wage, although
there may be a gap, as discussed below.
Third, and finally, there are differences in the
behaviour of the actors in the two sectors.
Capitalists in the industrial sector save all their
profits, because they are ambitious. Workers save
nothing, in cither sector, because they are poor
(Lewis describes them as not belonging to the ‘the
saving class’ - ,
p. 157). And landlords in agriculture are assumed to
consume all their income, which comes to them to
the extent that agricultural workers receive a wage
below their average product.
The general story is this: the profits in the
modem, capitalist, sector create a growing supply of
savings. This finances the formation of an
increasing stock of capital, which is used to employ
more and more labour in the urban workforce.
We can explain the story in detail, using a
simplified version of the model. To do this we make
four sets of extreme assumptions, (a) There is
‘pure’ surplus labour, by which we mean that the
marginal product of workers withdrawn from
agriculture is zero. Wages initially consist only of
agricultural goods, the level of wages per worker is
exogenous, and workers are indifferent between
working in industry and in agriculture at the same
wage, (b) When one individual worker leaves
agriculture and no longer needs to be rewarded
there, then all the increase in the agricultural surplus
(that is, all the increase in the total of food produced
minus the total of wages paid to agricultural
workers) accrues to landlords and is spent by them
on consumption of industrial goods, (c) Industrial
capitalists employ labour up to the point at which
the marginal physical product of labour is equal to
the cost of the wage, measured in industrial goods.
(d) All industrial profits are saved and then invested
in industrial production.
Given these assumptions, there are two steps to
the argument. First, given assumptions (a), (c) and
(d), the rate of growth depends negatively on the
relative price of agricultural goods in terms of
industrial goods. This is because an increase in food
prices raises the cost of the wage per worker in
terms of steel, causing less labour-intensive
methods of production to be adopted, that is,
Dual Economies 3067

causing production to become more capital inten- increasing inequality is a frequent, if not necessary,
sive. As a result of this, any given amount of correlate to this rising savings share, at least in the
savings, and the accumulation of capital that it early stages of development (see, for example, Fei
causes, will ‘go less far’ in employing labour in et al. ). This story thus also provides an explanation
industry, and, as a result, industrial output will grow of the ‘Kuznets curve’.
less rapidly. Second, assumptions (a) and (b)
determine the relative price of agricultural goods, in
the following way: the accumulation of capital in Generalizations
industry increases the demand for industrial
Lewis does sometimes enlist the extreme simpli-
workers, who must be transferred from agriculture.
fications made above. They correspond most
The relative price of agricultural goods will need to
closely to those made by Gustav Ranis and John C.
be high enough to induce the workers’ landlords to
H. Fei ( ), who used them to explain, more
offer up those agricultural goods that they would
formally than Lewis did, what they call the ‘first
have paid to the transferred workers but now
phase’ of economic development - a phase in which
receive as surplus, so as to receive industrial goods
there is ‘pure’ surplus labour. But Lewis also hints
for consumption in exchange. Such trade enables
at many ways in which these assumptions could be
workers to be paid in industry, where they now
relaxed. Ranis and Fei, along with Dale W.
work. As Lewis ( , p. 188)
Jorgenson (and many others), went on to consider
says, ‘the capitalists need the peasants’ food, and ...
the implications of dualism when there are sectoral
the demand for food is inelastic’.
asymmetries different from those outlined above. In
Clearly, the relative prices of industrial and
what follows, we consider a number of these
agricultural goods, and the growth rate of the
extensions.
economy, are jointly determined in this process - as
The first, and most fundamental, generalization
Lewis’s intuition had suggested to him. And it will
of Lewis’s model was made by Ranis and Fei ( ),
clearly be true that the relative price of agricultural
who demonstrated that the dualistic
goods will need to be less high - and so the rate of
framework continued to give insight into the pro-
growth will be higher - the lower is the price of
cess of economic growth even when the condition
agricultural goods required for landlords to release
of pure surplus labour does not hold. They initiated
their surplus in exchange for consumable industrial
a large body work on this question by examining
goods.
the microeconomic foundations of surplus labour
Note that the share of income that accrues to
and exploring what occurs when these conditions
industrial capitalists will increase during the growth
come to an end. This occurs when a sufficient
process, as the capitalist sector grows in size. This
number of workers have been removed from
suggested to Lewis ( , p. 155 ) that a
agriculture for the marginal productivity of the
growth process of this kind might help to solve
remaining agricultural workers to become positive.
what he called the ‘central problem’ of develop-
As a result, agricultural output declines as further
ment: the need to raise the savings rate enough to
workers leave. (This may happen even if there is
enable rapid growth to take place. In this model it is
technological progress in agriculture, providing that
necessary to transfer labour into industry, in order
this progress is not sufficient to fully compensate
to increase the overall savings rate of the economy.
for lost labour.) Consequently, the marginal
This is due to the behavioural assumption that
agricultural surplus per worker, which accrues to
agricultural income is not saved; we revisit this
landlords as each worker leaves - and which is
assumption below. Interestingly - from today’s
traded by landlords for industrial goods - begins to
point of view - Lewis thought that a savings rate of
decline, even if the wage per worker (measured in
ten to twelve per cent might be sufficient to achieve
terms of agricultural goods) is exogenous. This
the ‘rapid capital accumulation’ that he believed
means that the cost of labour to industry, measured
integral to the process of development (Lewis , p.
in terms of industrial goods, will
155). Note also that
3068
begin to rise above the level described in the sketch
above - thereby constraining the rate of growth.
This is the ‘first turning point’ identified by Ranis
and Fei. It corresponds to the onset of Ricardo’s
‘dismal’ diminishing returns. Ranis and Fei label
what happens beyond this point as the ‘second
phase’ of economic development. In that phase the
economy is characterized by ‘disguised
unemployment’, since labour in agriculture is still
paid more than its marginal product.
Lewis himself was accused of not allowing for
this possibility, even though he had written that the
existence of zero marginal product is ‘not... of
fundamental importance to our analysis’ (Lewis , p.
142). This accusation led to what Lewis later called
an ‘irrelevant and intemperate controversy’ about
the existence, or not, of ‘pure’ surplus labour
(Lewis , p. 77). Ranis ( , p. 8)
agrees with Lewis’s self-defence: in a retrospective
assessment, he describes the postulation of a ‘pure’
labour surplus as a red herring. Amartya Sen ( )
helpfully clarifies the debate about this
issue.
Growth becomes more difficult in this second
stage of development. Recall that Lewis argues that
the real wages per worker, and the level of welfare
per worker, do not fall as growth proceeds. But
growth is driven by the transfer of labour from
agriculture to industry, which, in this second phase,
causes agricultural output to fall. As a consequence
of this the relative price of agricultural goods rises,
and real wages can remain constant only if workers
are able to substitute towards industrial goods in
such a way as to avoid any damage to their welfare.
The Agricultural Sector as a Constraint
on Growth
To highlight the essential role of such substitution,
Mukesh Eswaran and Ashok Kotwal ( )
assume an extreme version of Engel’s law. Con-
sumers are assumed to spend all their income on
food until they reach a particular threshold level of
consumption, when they become sated with food.
Beyond this point all further increases in con-
sumption are devoted to industrial goods. At the
Dual Economies
same time they assume that labour always has a
positive marginal product in agriculture. Under
these assumptions, if workers remain so poor that
they are not sated with food, then the transfer of
labour across sectors - and therefore accumulation
of industrial capital - becomes impossible. The
inability of the poor to ‘eat shirts’ - an extreme
version of what Ranis ( ) describes
as the ‘product’ dimension of dualism - becomes a
constraint on whether savings can lead to devel-
opment. (And this constraint will bind quite inde-
pendently of how high the marginal physical
product of labour is in industry.) Any attempt to
increase savings rates, in the manner desired by
Lewis, so as to draw labour out of agriculture,
would fail in these circumstances. The withdrawal
of labour would lead to a reduction in the supply of
food per worker - the only thing that matters for
workers’ real wages - and so to a shortage of food.
That shortage would turn the terms of trade against
industry, depressing industrial profits and savings
until the downward pressure on the supply of food
had been removed, or until growth has ceased. As a
result, all the gains from any increase in industrial
production would accme, in the form of lower
prices, to those who consume industrial goods,
rather than enabling growth, as in the Lewis model.
It is thus clear that an important influence on
whether development can proceed under dualism is
the ability to shift workers’ demands away from
agricultural goods.
Of course, in a small, open, economy, the rel-
ative prices of tradables will be tied down, and the
economy can respond to any developing shortage of
food simply by exporting manufactures and
importing food. That was Ricardo’s insight, over
100 years earlier, about the gains to Britain from the
abolition of the Com Laws; Lewis’s model of
dualism in the world economy also incorporates
such trade. But Lewis ( , p. 94) cautions that
there may be limits to this if export prices are not
really exogenous, and if, instead, the county needs
to cheapen its exports to pay for the imports of food
- and other goods - that it will need as it grows.
Perhaps partly because of this, Lewis ( , p.
176) argues that a country which
exhausts its surplus labour supply might instead
export its savings, investing in industrial
Dual Economies
development in countries where the surplus labour
condition continues to hold, and so enabling the
output of manufactured goods to grow without
driving down the rate of profit. In addition, the
country might import labour from these countries.
In this way Lewis’s early contributions anticipated,
and fed into, debates about the roles of outsourcing
and immigration in contemporary globalization.
Jorgenson ( ) further develops the study of
the dynamics of a dualistic economy in this second
phase of development - when there is a positive
marginal product of labour in agriculture and
disguised unemployment. He incorporates a
Malthusian perspective, by supposing that popu-
lation growth is increasing in the amount of food
consumed per capita, up to a biological ceiling that
corresponds to the food-consumption threshold of
Eswaran and Kotwal. This has the consequence that
too rapid a rate of growth of population can cause a
Malthusian trap by preventing the emergence of any
significant agricultural surplus. Growth of
manufacturing activity, such as that analysed by
Lewis, can then be sustained only if technological
progress in agriculture enables food production to
outstrip population growth. (Capital accumulation
in agriculture could have a similar effect in a model
more general than that used by Jorgenson.) Only
then can an agricultural surplus emerge, and grow,
and so only then can labour progressively move
away from agriculture. If this does not happen, then
any increases in profits, savings and capital
accumulation in industry become self-defeating,
since they turn the terms of trade against industry
and so bring down profits and savings, and bring
growth to an end, in the way described two
paragraphs above.
As stressed by Avinash Dixit ( , p. 346),
such a model focuses on ‘the constraint on growth
imposed by the rate of release of labour from
agriculture’, whereas in Lewis’s model the focus
had been on the ability of capital accumulation in
industry to soak up the surplus labour force in
agriculture. Nevertheless, as Dixit notes, growth
paths in the two models will produce similar out-
comes. In particular, in both models one would
observe an endogenous rise in the savings rate as
3069
development proceeds. And in both models, it may
be the case that any attempt to foster growth in
industry, by a ‘big push’ to save more, is self-
defeating. (This can be true in Jorgenson’s model,
and as we saw above, it can also be true beyond the
‘first stage’ of growth in the Lewis model, if it is
not possible to induce workers to substitute away
from agricultural goods.) This is why Jorgenson
thought of increases in savings rates as an outcome
of development, not as a policy tool which can be
used to promote development (Jorgenson , p. 328).
It is worth contrasting this view of potential
‘development traps’ with that which had been put
forward in the 1940s by Paul Rosenstein-Rodan (
), who built on his experience of eastern
Europe. Rosenstein-Rodan’s viewpoint also came
from thinking about the interaction between
agriculture and industry; like Lewis, he argued that
development could only come to an agricultural
economy through a process of industrialization.
This, he argued, is because only industrial
capitalists could afford to pay for the large fixed
costs that are necessary to enable them to produce
goods in a modern way, with low marginal costs.
But if most people five in an impoverished agri-
cultural sector then this would constrain their
incomes, and so would limit their demand for
modem industrial goods. That might make it
unprofitable to make the required investment, and
so might thwart the process of development. Here,
just as for Lewis, a shortage of savings can be the
problem of development. But by contrast with
Lewis, a big push might fix it, since, roughly
speaking, if all capitalists invested at once and paid
their workers higher wages, then the demand for
industrial goods would grow, making the
investment worthwhile. This insight gave birth to
the other great analytical engine of development
economics, subsequently formalized by Kevin M.
Murphy, Andrei Shleifer and Robert W. Vishny (
) and Kuninori Matsuyama
( ), and well explained by Paul Krugman
( ). Since the pecuniary externalities that
allow an economy to escape from a development
trap are accessible only in the ‘modem’ sector,
asymmetries between sectors are also central to this
view.
3070
Further Aspects of Labour Transfer
The Lewis model was also generalized to explain
the gap between the wage paid in the rural sector
and that paid in the urban sector and to explore the
consequences of such a gap. Lewis himself ( p. 150)
acknowledged the existence of a wage gap, and
suggested that it may result from the psychological
costs of lifestyle changes, from the need to reward
skills accumulated in the urban sector, or from the
ability of workers in cities to bargain for higher
wages. (This is particularly relevant when we
recognize that the urban sector includes government
employment and some services.) Subsequent
authors took up this question, arguing, for example,
that wage premia may arise because they lead to
greater productivity through effects on health or
employee motivation (for example, Dasgupta and
Ray ; Shapiro and Stiglitz ).
The consequences of such a gap, for the process
of labour transfer from agriculture to industry, were
set out in the celebrated work of John Harris and
Michael R Todaro ( ). It may be
that a wage floor in the urban formal sector pre-
vents the market from clearing there. If a wage
floor operates, then workers who choose to leave
the rural sector face the prospect of receiving an
urban wage which is above that of the rural sector,
if they get employed, but also face some probability
of becoming unemployed. In the simplest version of
this model, equilibrium occurs when labour
migration equalizes expected income across sectors
- an outcome in which the rural wage equals a
weighted average of the incomes received by
employed and unemployed urban workers,
weighted according to the probability of
unemployment in the urban sector. Even without
this extreme outcome, there are important policy
implications in such a model. The more elastic is
labour supply to the urban sector with respect to
expected income there, the greater the amount of
urban unemployment that will be induced by any
policies that increase urban wages. This
incorporation of urban unemployment into the
model also enables one to begin to discuss the
growth of a third sector: the production of services
in cities (see Fields ). Roughly
Dual Economies
speaking, we can say that services get produced by
(some of) those who migrate to cites, but do not get
a job in manufacturing.
It is clear that the expansion of the industrial
sector will ultimately take the economy beyond the
second phase of economic development, in which
there is disguised unemployment. This is because
withdrawal of labour from agriculture will
eventually reach the point at which the marginal
product of the remaining labour rises to equality
with the subsistence wage. Ranis and Fei call this a
‘second turning point’. At this point the marginal
worker, offered a subsistence wage, can now
instead offer his or her labour to a higher bidder.
From then on the wage (measured in agricultural
goods) will begin to rise in both sectors as growth
continues. We can say that the ‘dualistic’ structure
of the economy then comes to an end, in that the
rural economy becomes ‘commercialized’.
(Something similar, too, will happen in any services
sector.) That leads one back to a labour-scarce
economy, the analysis of which is better suited to
neoclassical theory. A two-sector neoclassical
growth model - something like the model of
Hirofumi Uzawa ( , ) - may be
a better way to think about growth in these
circumstances.
One key strand of the story of dualism that we
have been telling is the assumption that capitalists
save, but workers (and landlords) do not. Lewis’s
explanation of this asymmetry is largely
behavioural. But such differences in savings rates
between the traditional and the modem sectors
might also be explained institutionally, by means of
credit-market imperfections. If a technological
asymmetry precludes investment in rural areas, and
if limited financial development means that rural
residents lack access to investment opportunities in
manufacturing, then the agricultural surplus will not
be used directly to finance investment. Moreover,
typical characterizations of credit-market
imperfections highlight the moral hazard problems
that persist in rural areas because the poor there are
unable to provide the kind of collateral required for
formal- sector loans. (Small rural landholdings are
of limited use as collateral.) Such lack of collateral
stands as a barrier to borrowing, even though
Dual Economies
loans might be used to facilitate growth by pro-
moting education, or capital accumulation, or
technical progress in agriculture. (See Ray , for a
summary of these arguments.) By contrast, Abhijit
Baneijee and Andrew Newman ( ) provide an
alternative perspective,
emphasizing a sectoral asymmetry in the infor-
mational dimensions of credit-market imperfec-
tions, and showing how this can affect the
willingness of individual workers to migrate in a
dualistic economy. They present a model in which
there is access to credit for consumption in rural
areas. Given that workers have limited collateral
wherever they live, a crucial determinant of their
access to credit is the amount of information that
lenders have about prospective borrowers. In
contrast with the relative anonymity of urban life,
small communities of the rural sector may provide
superior information about borrowers, and thus
foster lending. Baneijee and Newman show that
dualism, characterized in terms of this differential
severity of information asymmetries, might lead to a
suboptimal allocation of labour across sectors. By
financing consumption in the rural sector, rural
credit might actually provide an incentive for labour
to remain there; this incentive could offset the
relatively high wages of the modem sector and
could thereby impede the development process.
Their paper suggests - at the least - that the lens of
asymmetric information can shed useful light on the
development of such economies.
Defining Characteristics of Economic
Dualism
We conclude by noting that we have described a
number of reasons for differences between the
industrial and agricultural sectors of a developing
economy. Just as in Lewis’s original article, all
these differences go beyond mere asymmetries in
production technologies or factor endowments
between the sectors. This is why, following Ravi
Kanbur and James McIntosh ( ), we would
not normally describe the two-sector growth models
of Uzawa ( , ) as models of dual
ism, even though in those models the two sectors
3071
have different factor intensities. Nor would we say
that that the two-sector Hecksher-Ohlin model of
international trade is a model of a dualistic economy
- even when its two sectors have different factor
intensities, and even when the two sectors are
labelled ‘agriculture’ and ‘industry’. Furthermore,
although the specificity of factors to sectors appears
central to Lewis’s set-up (with land specific to
agriculture and capital specific to industry), this
feature does not seem to be sufficient to merit the
label of ‘dualism’. Thus, for example, we would not
regard the short-run version of the Heckscher-Ohlin
trade model presented by J. Peter Neary ( ),
with factors
specific in each of the two sectors, as portraying a
dualistic economy.
Instead, we would argue that the defining char-
acteristic of modem theories of economic dualism
lies - just as it did in Lewis’s article - in a focus on
sectoral asymmetries that are not simply techno-
logical. For Lewis, and for Ranis and Fei, there were
organizational differences between sectors - in that
wages were assumed to be determined by
institutional factors in the agricultural sector - and
behavioural differences between sectors - in that
those in the rural sector were assumed to be
unwilling to save, while capitalists were assumed to
save everything. A focus on these features might
imply that ‘pull’ factors drive labour transfer, and
hence economic growth, in a dualistic economy. But
since Lewis, economists studying economic
development have explored alternative asymmetries
between sectors and have reached different
conclusions. The model of Eswaran and Kotwal, in
which the defining asymmetries are product
asymmetries - an assumption that all income is spent
on agricultural goods until some threshold -
highlights the need for labour productivity increases
in agriculture to avoid stagnation of real wages. This
is a need that persists even in the presence of rising
productivity in industry. Jorgenson, who coupled
such a view with a demonstration that Malthusian
pressures can prevent income from ever rising
above this threshold, showed clearly that growth can
be constrained unless the ‘push’ factor of growth in
agricultural technology is strong enough. Baneijee
and Newman, by contrast, have emphasized that
3072
informational asymmetries between traditional and
modem sectors can constrain the growth process.
We thus believe that, in the study of any
particular economy, it is important to understand
which asymmetries impose binding constraints on
growth. Different constraints imply the need for
different policies. But identifying the relevant
asymmetries is even more important if we wish to
remove these underlying constraints themselves.
Joseph Stiglitz has proposed that we do just this,
advocating what he calls ‘growth strategies based on
duality’s elimination’ (Stiglitz , p. 56). Much
empirical work is necessary if we are to understand
what such strategies might require.
See Also
► Labour Surplus Economies
► l/i vis, W. Art!
Bibliography
Baneijee, A.V., and A.F. Newman. 1998. Information, the
dual economy, and development. Review of Economic
Studies 65: 631-635.
Dasgupta, P., and D. Ray. 1986. Inequality as a determinant of
malnutrition and unemployment: theory. Economic
Journal 96: 1011-1034.
Dixit, A. 1973. Models of dual economies. In Models of
economic growth: Proceedings of a conference held by the
international economic association at Jerusalem, ed. J.A.
Mirrlees and N.H. Stem. London: Macmillan. Domar, E.D.
1946. Capital expansion, rate of growth, and employment.
Econometrica 14: 137-147.
Eckaus, R.S. 1955. The factor proportions problem in
underdeveloped areas. American Economic Review 45:
539-565.
Eswaran, M., and A. Kotwal. 1993. A theory of real wage
growth in LDCs. Journal of Development Economics 42:
243-269.
Fei, J.C.H., G. Ranis, and S.W.Y. Kuo. 1979. Growth with
equity: The Taiwan case. Oxford: Oxford University
Press.
Fields, G.S. 1975. Rural-urban migration, urban unem-
ployment and underemployment, and job-search activity
in LDCs. Journal of Development Economics 2: 165-187.
Findlay, R. 1980. On W. Arthur Lewis’ contributions to
economics. Scandinavian Journal of Economics 82: 62-
79.
Dual Economies
Harris, J.R., andM.P. Todaro. 1970. Migration, unemploy-
ment and development: A two-sector analysis. American
Economic Review 60: 126-142.
Harrod, R.F. 1939. An essay in dynamic theory. Economic
Journal 49: 13-33.
Hicks, J. 1965. Capital and growth. Oxford: Oxford Uni-
versity Press.
Higgins, B. 1956. The ‘dualistic theory’ of underdeveloped
areas. Economic Development and Cultural Change 4:
99-115.
Jorgenson, D.W. 1961. The development of a dual economy.
Economic Journal 11: 309-334.
Kanbur, R., and J. McIntosh. 1987. Dual economies. In The
new Palgrave: A dictionary of economics, ed. J. Eatwell,
M. Milgate, and P.K. Newman, vol. 1. London:
Macmillan.
Krugman, P. 1993. Toward a counter-counterrevolution in
development theory. In Proceedings of the World Bank
Annual conference on development economics, 1992:
Supplement to the World Bank economic review and the
World Bank research observer, ed. L.H. Summers and S.
Shah. Washington, DC: World Bank.
Lewis, W.A. 1953. Report on industrialisation and the Gold
Coast. Accra: Government Printing Department.
Lewis, W.A. 1954. Economic development with unlimited
supplies of labour. Manchester School 28: 139-191.
Lewis, W.A. 1955. The theory of economic growth. London:
Allen & Unwin.
Lewis, W.A. 1972. Reflections on unlimited labour. In
International economics and development: Essays in
honour of Raul Prebisch, ed. L.E. di Marco. London:
Academic.
Lewis, W.A. 1978. Growth and fluctuations, 1870-1913.
London: Allen & Unwin.
Lewis, W.A. 1979. Autobiography. In Lindbeck (1992). Also
online. Available at
prizes/economics/laureatesS 79/ e v s-E utobio .
Accessed 9 Jan 2007.
Lindbeck, A., ed. 1992. Nobel lectures: Economic sciences
1969-1980. Singapore: World Scientific Publishing.
Matsuyama, K. 1991. Increasing returns, industrialization,
and the indeterminacy of equilibrium. Quarterly Journal
of Economics 106: 617-650.
Murphy, K.M., A. Shleifer, and R.W. Vishny. 1989. Indus-
trialization and the big push. Journal of Political Econ-
omy 97: 1003-1026.
Neary, J.P 1978. Short-run capacity specificity and the pure
theory of international trade. Economic Journal 88:488-
510.
Pasinetti, L. 1974. Growth and income distribution - Essays
in economic theory. Cambridge: Cambridge University
Press.
Preobrazhensky, E. 1924. The new economics. Trans. B.
Pearce. Oxford: Clarendon Press, 1965.
Ranis, G. 2003. Is dualism worth revisiting? Discussion
Paper No. 870. Economic Growth Center, Yale
University.
Ranis, G., and J.C.H. Fei. 1961. A theory of economic
development. American Economic Review 51: 533-565.
Dual Track Liberalization
Ray, D. 1998. Development economics. Princeton: Princeton
University Press.
Rosenstein-Rodan, P. 1943. Problems of industrialisation of
eastern and south-eastern Europe. Economic Journal 53:
202-211.
Sah, R.K., and J.E. Stiglitz. 1984. The economics of price
scissors. American Economic Review 74: 125-138.
Sen, A. 1966. Peasants and dualism with or without surplus
labour. Journal of Political Economy 74: 425-450.
Shapiro, C., and J.E. Stiglitz. 1984. Equilibrium unem-
ployment as a worker discipline device. American Eco-
nomic Review 14: 433—444.
Solow, R.M. 1956. A contribution to the theory of economic
growth. Quarterly Journal of Economics 70: 65-94.
Stiglitz, J.E. 1999. Duality and development: some reflections
on economic policy. In Development, duality, and the
international economic regime: Essays in honor of
Gustav Ranis, ed. G. Saxonhouse and T.N. Srinivasan.
Ann Arbor: University of Michigan Press.
Swan, T.W. 1956. Economic growth and capital accumu-
lation. Economic Record 32: 334-361.
Uzawa, H. 1961. On a two-sector model of economic growth.
Review of Economic Studies 29: 40-47.
Uzawa, H. 1963. On a two-sector model of economic growth
II. Review of Economic Studies 30: 105-118.
Dual Track Liberalization
Yingyi Qian
Abstract
Dual track liberalization is a reform strategy in
which a market track is introduced while the
plan track is maintained at the same time. Dual
track liberation is Pareto improving in the sense
that it makes some people better off without
making anybody worse off. Because prices are
liberalized at the margin, dual track
liberalization can also achieve efficiency. China
used the dual track reform strategy in
liberalizing many markets such as the markets of
agricultural goods, industrial goods, consumer
goods, foreign exchange, and labour, as well as
in creating special economic zones.
3073
Keywords
Allocative efficiency; China, economics in;
Compensatory transfers; Corruption; Dual track
liberalization; Foreign exchange control; Market
liberalization; Pareto efficiency; Planning; Price
control; Price liberalization; Rationing; Rent
seeking; Special economic zones (China)
JEL Classification
P3
Dual track liberalization is a reform strategy of
market liberalization in which a market track is
introduced while the plan track is maintained at the
same time. Under the plan track, economic agents
are assigned rights to and obligations for a fixed
quantity of goods and services at fixed planned
prices as specified in the pre-existing plan. Under
the market track, economic agents can participate in
the market at free market prices, provided that they
fulfil their obligations under the pre-existing plan.
The essential feature of the dual track strategy to
market liberalization is that prices are liberalized at
the margin while inframarginal plan prices and
quotas are maintained for some time before being
phased out. Although the dual track reform strategy
is widely adopted in China during its transition from
plan to market, it is also used in other countries. For
example, when introducing new legislation, a
‘grandfathering’ clause is often adopted to protect
existing interests, which is a form of the dual track
approach to reform.
Analysis of dual track liberalization follows two
lines of approach. The first focuses on its Pareto-
improvement property, that is, dual track liberation
makes nobody worse off while it makes somebody
better off- and therefore it has a political advantage
in implementing reforms. Most efficiency-
improving market liberalization reforms potentially
create winners and losers, despite the fact that, in
theory, efficiency gains should be large enough to
allow the potential losers to be compensated. For
example, the single track approach to liberalization
(that is, where all
3074
the prices are freed at once) in general cannot
guarantee an outcome without losers. Dual track
liberalization means that planned quantity continues
to be delivered at plan price but any additional
quantity can be sold freely in the market. With the
dual track, the surpluses of the rationed users and
the planned suppliers remain exactly the same. The
purpose of maintaining the plan track is to provide
implicit transfers to compensate potential losers
from market liberalization by protecting status quo
rents under the pre-existing plan. On the one hand,
the introduction of the market track provides the
opportunity for economic agents who participate in
it to be better off. At the same time, the new users
and suppliers outside the plan are also better off.
Therefore, the intuitive appeal of dual track
liberalization for reformers lies precisely in the fact
that it represents a mechanism of the
implementation of a reform without creating losers
(Lau et al. ).
The second approach focuses on the efficiency
property of dual track liberalization. Pareto-
improvement property implies that it always
improves efficiency. This is independent of other
assumptions, for example, as to whether the market
is competitive or not. In contrast, the single track
approach to liberalization may improve efficiency
under perfect competition, but may not improve
efficiency if the market is monopolistic (Li ). The
more subtle and deeper point is that the dual track
approach to liberalization may achieve allocative
efficiency, despite the fact that it appears
inefficient, by maintaining the inefficient planned
track. The fundamental reason is that the
compensatory transfers, which are implicitly
embodied in the planned track, are inframarginal,
and thus the distortion can be avoided.
To see this we look at the special case where the
pre-reform status quo features efficient rationing
and efficient planned supply in the sense that the
planned output is allocated to users with the highest
willingness to pay and the planned supply is
delivered by suppliers with the lowest marginal
costs. Nevertheless, the price of the good is fixed at
an artificially low level and the production quota is
fixed below market equilibrium (Fig. ). When the
market track is introduced into this
Dual Track Liberalization
Dual Track Liberalization, Fig. 1 The case of efficient
supply and efficient rationing
setting, it is clear that the market equilibrium
quantity and price would be identical to the case
without the planned price and quota to start with.
Therefore, dual track liberalization achieves effi-
ciency. Notice that efficiency is achieved without
making anyone worse off. Indeed, the rents enjoyed
by the buyers under rationing (area A in Fig. ) are
preserved under dual track liberalization, but would
be lost under single track liberalization.
In a more general case of inefficient rationing
and/or inefficient planned supply, efficiency can stil
l be ach ieved provided market liberalization is full,
in the sense that market resales of plan- allocated
goods and market purchases by planned suppliers
for fulfilling planned delivery quotas are permitted
after the fulfilment of the obligations of planned
suppliers and rationed users under the plan. This
removes any inefficiency associated with the
original planned prices and quotas and makes
imputed rents under planning inframarginal. This
type of transaction takes many forms in practice, for
example, subcontracting by inefficient planned
suppliers to more efficient non-planned suppliers,
and labour reallocation when workers in inefficient
enterprises keep the housing while taking a new job
in more efficient firms. In both examples, after
fulfilling the obligations under the plan (planned
delivery of supply and welfare support through
housing subsidies), the market track functions to
undo the inefficiency of the plan track.
Dual Track Liberalization
This above partial equilibrium analysis can be
generalized to a general equilibrium mode (Lau et
al. ). Efficiency requires full market liberalization
under which market resales, subcontracting, and
market purchases for redelivery are all allowed.
Indeed, the distinction between limited and full
market liberalization is a major difference between
Lau et al. ( ) and Byrd
( ), and others who have studied the dual
track approach.
If such resales and purchases are not allowed or
cannot be achieved, then dual track liberalization is
limited and efficiency in general cannot be
achieved, although it can be improved. Of course,
in the special case discussed above with efficient
supply and efficient rationing, dual track with lim-
ited market liberalization is the same as dual track
with full market liberalization. In general, dual
track limited market liberalization need not be the
same as dual track full market liberalization.
Sometimes dual track liberalization of the
market takes the following sequential form: in a
first stage, limited market liberalization is
implemented, and then in a second stage full market
liberalization is implemented. In the first stage,
going from a centrally planned economy to limited
market liberalization, Pareto improvement is clearly
attained, but efficiency cannot be guaranteed.
Specifically, limited market liberalization generally
leads to inefficient overproduction due to market
entry. In the second stage, when full liberalization
is introduced, efficiency is attained but Pareto
improvement may not be. This is because the
second-stage full market liberalization implies
efficiency, and thus there must be a production
contraction and some people have to reduce
production and are made worse off. Therefore, the
sequential dual track liberalization may result in
some opposition to further reforms after the first
and before the second stage, while the dual track
full market liberalization that is implemented in one
stroke will not. Nevertheless, it is also clear that,
even under the sequential dual track liberalization,
there are no losers at the end of the second stage
compared with the status quo before the reform.
The dual track approach to market liberalization
is an example of reform making the best use
3075
of existing information and institutions. First, it
utilizes efficiently the existing information
embedded in the original plan (that is, existing rents
distribution) and its implementation does not
require additional information. Second, it also
enforces the plan through the existing plan
institutions and does not need additional institu-
tions. Enforcement of the plan track is crucial for
preserving pre-existing rents. However, contrary to
common understanding of the relationship between
state power and reform, state enforcement power is
needed here not to implement an unpopular reform,
but to carry out one that creates only winners,
without losers.
Economists sometimes find dual track liberal-
ization puzzling and counter-intuitive, for several
reasons. First, economists are used to the law of one
price: in a competitive setting, multiple prices entail
inefficiency. However, in dual price liberalization,
the planned price comes together with planned
quantity, when they are fixed, they do not entail
inefficiency, at least not additional inefficiency.
Second, dual track resembles price control, which is
associated with inefficiency and rent seeking. But
dual track is not price control; on the contrary, it is a
move towards price liberalization. An important
difference between the plan track under dual track
and price control is that the plan track embodies
both fixed prices and fixed quantities; it is a
package of price and quantity control, not just price
control. Under pure price control, the government
fixes only prices, but not quantities. Third, to
reformers, dual track seems a partial reform and not
a complete reform. This is true under dual track
with limited market liberalization, but not true with
full market liberalization. Although dual track with
limited market liberalization does not achieve
efficiency, it improves efficiency and makes
nobody worse off.
Dual track liberalization requires enforcement of
the rights and obligations under the plan track. In
fact, enforcement of the plan track alone would
prevent any decline in aggregate output. Can the
plan track be enforced? With a collapsing govern-
ment, it cannot. But enforcing the pre-existing plan
is informationally much less demanding for the
government than drawing up a new plan. Under
central planning, the information
3076
requirement for drawing up a plan is huge.
Enforcing a pre-existing plan is different. In fact,
the dual track approach uses minimal additional
information as compared with other possible com-
pensation schemes that may be used with other
approaches to reform. Compliance with the plan by
economic agents depends on their expectations of
the credibility of state enforcement. If state
enforcement is not credible, then the economic
agents will have no incentive to fulfil their plan
obligations. If people think that they are not going
to receive the plan- mandated deliveries at plan
prices, they will not make the plan-mandated sales
at the fixed plan prices. In that case, dual track
liberalization degenerates to single track
liberalization.
Lack of enforcement of the plan track may
result in supply diversion as analysed by Murphy et
al. ( ). These authors studied a partial
reform model with the following two crucial
assumptions: (i) suppliers are free to sell to all
users, and (ii) buyers who are not covered by the
plan can freely purchase inputs at any price, but
buyers who are covered by the plan are not allowed
to purchase inputs above the plan price. This partial
reform model differs from the dual track
liberalization model in an important respect: there is
no plan delivery quota enforced on the suppliers.
In their model, partial reform may lead to inef-
ficient supply diversion to such an extent that the
outcome can be worse than that without reform.
Therefore, the partial reform is not only not Pareto
improving, but also total welfare reducing. Con-
sider the case where the initial condition is also
characterized by efficient rationing and efficient
supply as shown in Fig. , where the planned price
Pp is below the market clearing level Pm. Then,
after the partial liberalization as defined above,
suppliers can sell the good freely to the highest
bidders. While the firms under the plan are forced
to buy the good at price P p, the firms outside the
plan are free to buy the good at any price. Then
they will bid the good for price P p + e where e is a
positive but small number. Because the firms under
the plan are constrained to pay P p, an amount will
be diverted from them to those not covered by the
plan. Because the willingness to
Dual Track Liberalization
pay from those not covered by the plan is lower
than those covered by the plan (by the assumption
of efficient rationing), this kind of partial reform
induces a net efficiency loss. While the sector not
covered by the plan gains, the sector covered by the
plan loses, and the total welfare effect is
unambiguously negative. Although the assumption
of efficient rationing and efficient supply under
central planning is too strong, the result of
inefficient supply diversion under partial reform
remains valid with weaker conditions about initial
rationing and supply.
So which model is more relevant? It depends on
the quota enforcement capability of the government.
A good enforcement capability makes the dual track
liberalization model of Lau, Qian and Roland more
relevant, while a poor enforcement would make the
partial reform model of Murphy, Shleifer, and
Vishny more relevant. The dual track liberalization
model is motivated mainly by the practice in China,
where enforcement has been reasonably good, while
the partial reform model is mainly motivated by the
experiences of the last years of the Soviet Union,
when the state enforcement power diminished
quickly.
In China’s context, lack of quota enforcement
sometimes takes the following form. The govern-
ment may be unable to freeze the plan by creating
new quotas with (below market equilibrium)
planned price and giving windfall rents to some
people who are politically connected. This may lead
to corruption: firms find it easier to make profits by
lobbying the government for allocating more input
goods delivery at low planned prices, without the
corresponding obligations to deliver low price
outputs as under central planning. They then sell the
goods at the market price to receive the windfall
gains. This type of cormption is often attributed to
the dual track approach to liberalization. Indeed,
without the coexistence of the planned prices and
market prices, the above form of cormption is not
possible. By eliminating the two prices, such form
of corruption would disappear. However, the
essence of the problem is the failure in the
enforcement of the original planned track. If the
planned track is strictly enforced, no new quotas
should be created. (On the other hand, full market
liberalization
Dual Track Liberalization
allows for market arbitrage, which may increase the
welfare of those who were allocated with goods at
below-market prices. This is essential for achieving
efficiency. The difference is that the potential rents
are inherited from the previous regime in this case,
not from a new creation.)
Dual Track Liberalization in Practice
Studies of dual track liberalization focus mostly on
China, although other cases, such as that of
Mauritius, are also mentioned. The origin of the
dual track can be traced to the 1950s when China
had two prices for grain, the official price and
negotiated price. However, dual track approach to
market liberalization as a reform strategy was used
only after 1979, first in the agricultural goods
markets, and then in other markets (Byrd ;
Naughton ; Lau et al. ).
Agriculture Goods The agricultural reform in China
started with a dual track approach to market
liberalization. Under that reform, the commune (and
later the household) was assigned the responsibility
to sell a fixed quantity of output to the state
procurement agency as previously mandated under
the plan at predetermined plan prices and to pay a
fixed tax (often in kind) to the government. It also
had the right (and obligation) to receive a fixed
quantity of inputs, principally chemical fertilizers,
from state-owned suppliers at predetermined plan
prices. Subject to fulfilling these conditions, the
commune was free to produce and sell whatever it
considered profitable, and retain any profit.
Moreover, the commune could purchase from the
market grain (or other) output for resale to the state
in fulfilment of its responsibility. There was thus a
full market liberalization.
Between 1978 and 1988 state procurement of
domestically produced grain remained essentially
fixed, with 47.8 million tons in 1978 and 50.5
million tons in 1988. During that same period, total
grain output increased by almost one-third. But the
dual track approach to liberalization applied to Consumer Goods Prior to the economic reform of
agricultural products other than grain: between 1978 1979, most essential consumer goods and
and 1990, the share of transactions
3077
at plan prices in all agricultural goods fell from 94
to 31%, when the agricultural output in China
doubled. There was a huge supply response to the
introduction of the market track.
Industrial Goods The most noticeable and often
cited application of the dual track approach to
liberalization is to industrial goods (Byrd ;
McMillan and Naughton ). The Chinese gov
ernment issued a document in May 1984 stipulating
that there would be two forms of production in
state-owned enterprises: planned and
non-planned. Correspondingly, there were two
types of material supplies for enterprises, namely,
state allocation and free purchase. Prices of goods
in the former were fixed by the state and prices of
goods above quota quantity could be sold in the
market at price within a range up to 20% higher or
lower than of the planned price. In February 1985,
the 20% price cap was removed and the dual track
for industrial goods was formally in place (Wu and
Zhao ). As a result, the share of transactions at plan
prices, in terms of output value, fell from 100%
before the reform to 45% in 1990.
Coal and steel are the two important industrial
commodities most tightly controlled under central
planning, and both coal and steel markets were
liberalized through the dual track approach. For
coal, China’s principal energy source, the planned
delivery led to some slight increases in absolute
terms during the 1980s, but the market track
increased dramatically from 293 million tons to 628
million tons over the same period - the supply came
mainly from small rural mines run by Township-
Village Enterprises. As a result, the share of the
plan allocation declined from 53% in 1981 to 42%
in 1990. For steel, the plan track was quite stable in
absolute terms during the 1980s, but the share of
plan allocation fell from 52% in 1981 to 30% in
1990. In the cases of both coal and steel, because
the plan track was essentially frozen, the economy
was able to ‘grow out of the plan’ on the basis of
the expansion of the market track (Naughton ).
3078 Dual Track Liberalization

services for urban residents, such as grain, cooking non-state sector (the liberalized sector) pays market
oil, meat, electricity, housing, and the monthly wages and decides on hiring and firing. Between
transport pass, were rationed with coupons at values 1978 and 1994, employment in the non-state sector
lower than corresponding free market prices. With increased by 318.8%, while employment in the state
dual track liberalization, urban residents continued sector (including civil servants in government
to have the right to purchase grain, meat, electricity agencies and non-profit organizations) increased by
and housing at the same pre-reform prices and only 50.5%. Second, even within the state sector
within the limits of the pre-reform rationed there are also two tracks. Beginning in 1980, while
quantities, but, at the same time, they were also free pre-existing employees maintained their permanent
to buy consumer goods from the free market at employment status, most new hires in the state
generally higher prices. The proportion of sector were made under the more flexible contract
transactions at plan prices declined from 97% in system and often at lower effective wage rates.
1978 to only 30% in 1990. Employment in the plan track was virtually
stationary - it declined from 87.14 million in 1983,
Foreign Exchange Under central planning, foreign on the eve of the introduction of economic reform
exchange transactions were strictly controlled by in industry, to 83.61 million in 1994.
the government at the official exchange rate.
Exporters were required to surrender to the state all Special Economic Zones Dual track liberalization
foreign exchange they earned at the official can also have a geographical dimension: special
exchange rate, and importers were allocated with economic zones are such examples. Although
planned quotas of foreign exchange, also at the similar zones for processing exports can be seen in
official exchange rate. Foreign visitors to China other Asian economies, special economic zones had
were required to use 'foreign exchange certificates’, a more profound effect in China because the whole
which were available at the official exchange rate. country was still under central planning when they
Starting from May 1988, China allowed trading of were created. Therefore, the purpose of special
foreign exchange at Foreign Exchange Adjustment economic zones was more than for exporting; it was
Centres (more commonly referred to as ‘swap a strategy for market reform.
centres’) at the rate determined by market supply In 1980, China established four ‘special eco-
and demand, called ‘swap rate’. This was the nomic zones’, Shenzhen, Zhuhai and Shantou in
beginning of the dual track in the foreign exchange Guangdong province and Xiamen in Fujian prov-
market. The swap rate was, not surprisingly, ince. Most transactions relating to activities inside
significantly higher than the official rate. The the zones were on the market track, including prices
supply of foreign exchange in the swap markets was of input and output goods and wages of labour - at a
provided by exporters through the foreign exchange time when the rest of the economy was still
they were allowed to retain from net increases in operating under central planning. The special
their export earnings in relation to the base period. economic zones were insulated from the rest of the
By the end of 1993, transactions at official economy to minimize the impact on and interaction
exchange rates accounted only for about 20% of the with the rest of the economic system. Initially, firms
total; the rest were at the market rate. inside the special economic zones had to import all
their inputs and export all their outputs - thus
Labour As in many other centrally planned econ- creating no disruption to the domestic aggregate
omics, the labour market in China was also supply and demand. The principal purpose of this
distorted: most labour was allocated to approach was to minimize the impact of new
unproductive, state-owned enterprises and few to economic activities on the old-style domestic state-
the non-state sector. Dual track liberalization in the owned enterprises. Thus, once again, there were two
labour market takes two forms. In the first, the tracks and the reform was Pareto improving.
Duality
In order for the special economic zones to work,
merely creating them was not enough. One of the
crucial conditions was the insulation of the non-
liberalized sector from the liberalized sector so that
the latter’s existing rents could be maintained while
the other sector was liberalized. Therefore, creation
of special economic zones is a type of limited
market liberalization. It is Pareto improving and
efficiency enhancing, but cannot be fully efficient.
Phasing Out the Plan Track
With rapid growth, the plan track will become a
matter of little consequence to most potential losers,
which in turn reduces the cost required for
compensating them. In China, the plan track in
product markets was largely phased out during the
1990s. By 1996, the plan track was reduced to
16.6% in agricultural goods, 14.7% in industrial
producer goods, and only 7.2% in total retail sales
of consumer goods. However, this phasing-out of
the plan track was generally accompanied by Duality
compensation. For example, urban food coupons
(grain, meat, oil, and so on) were removed in the Lawrence E. Blume
early 1990s with lump-sum compensation. But the
cost of compensation was much smaller in relative
terms as compared to the potential cost
ofcompensationinthe early 1980s. The dual track
exchange rate ended on 1 January 1994, when the
two exchange rates - the official rate and the swap
rate - were merged into a single, market rate. In this opposed to the usual analysis through first- order
last step of foreign exchange reform, those
organizations that used to receive cheap foreign
exchange were provided with annual lump-sum
subsidies for a period of three years, which was
sufficient for them to purchase the pre-reform
allocation of foreign exchange. Because at that time
the share of centrally allocated foreign exchange
had already fallen to less than 20% of the total, the
cost of compensation was not too large.
See Also
► Chii
3079
Bibliography
Byrd, W.A. 1991. The market mechanism and economic-
reforms in China. Armonk: M.E. Sharpe.
Lau, L.J., Y. Qian, andG. Roland. 1997. Pareto-improving
economic reforms through dual track liberalization.
Economics Letters 55: 285-292.
Lau, L.J., Y. Qian, and G. Roland. 2000. Reform without
losers: An interpretation of China’s dual track approach
to transition. Journal of Political Economy 108: 120-143. D
Li, W. 1999. A tale of two reforms. R.4ND Journal of
Economics 30: 120-136.
McMillan, L, and B. Naughton. 1992. How to reform a
planned economy: Lessons from China. Oxford Review of
Economic Policy 8: 130-143.
Murphy, K.M., A. Shleifer, and R.W. Vishny. 1992. The
transition to a market economy: Pitfalls of partial reform.
Quarterly Journal of Economics 107: 887-906.
Naughton, B. 1995. Growing out of the plan. Cambridge:
Cambridge University Press.
Wu, J., and R. Zhao. 1987. The dual pricing system in
China’s industry. Journal of Comparative Economics
11:309-318.
Abstract
This article surveys duality in producer theory,
consumer theory and welfare economics. As
conditions for optimization, the various dualities
are derived here from convex duality theory,
using Fenchel transforms and subdifferentials.
Keywords
Antonelli, G.B.; Bergson-Samuelson social
welfare function; Convex programming; Con-
vexity; Cost functions; Cyclic monotonicity;
Duality; Envelope th; Fenchel transform; Firm,
theory of the; Hicksian-compensated demand;
Hotelling, H.; Hotelling’s lemma; Hyperplanes;
Indirect utility function; Lagrange multipliers;
Marginal revolution; Monotonicity; Profit
functions; Quasi-
3080 Duality

equilibrium; Saddlepoints; Separation th; description of the set, and by listing the closed half-
Shephard, R.W.; Shephard’s lemma spaces that contain it. A closed (upper) halfspace in
Rn is a set of the form hpa = {x : p ■ x > a}, where p
is another n-dimensional vector, a is a number andp
JEL Classifications ■ x is the inner product. The vectorp is the normal
DO v e c t o r to the half-spaces h . Geometrically speaking,
pa

this is the set of points lying on or above the line p •

x a. The famous separation theorem for convex sets
implies that every closed convex set is the
Introduction intersection of the half-spaces containing it.
Suppose that C is a closed convex set, and that p
The word ‘duality’ is often used to invoke a contrast is a vector in Rn. How do we find all the numbers a
between two related concepts, as when the informal, such that C C hpa? If there is an.r e C such that p ■ x
peasant, or agricultural sector of an economy is < a, then a is too big. So the natural candidate is w
labelled as dual to the formal, or profit-maximizing, in/ e c P'-x- If a > there will be an x e C such that p ■
sector. In microeconomic analysis, however, x < a on the other hand, if a<w, then p-x > a for
‘duality’ refers to connections between quantities all .r C C. So the halfspaces hpa for a < w are the
and prices which arise as a consequence of the closed half-spaces containing C.
hypotheses of optimization and convexity. This construction can be applied to functions. A
Connected to this duality are the relationship concave function on Rn is an [—oo, oo) valued
between utility and expenditure functions (and function / such that the hypograph of / the set hypo /
profit and production functions), primal and dual = {(.r, a) C Rn l : a < fix)}, is convex. If hypo/is
linear programs, shadow prices, and a variety of closed,/is said to be upper semi- continuous (use).
other economic concepts. In most textbooks, the The domain dom/of concave /is the set of vectors in
duality between, say, utility and expenditure Rn for which/is finite-valued. Concave (and convex)
functions arises from a sleight of hand with the functions are very well- behaved on the relative
first-order conditions for optimization. These dual interiors of their effective domains. The relative
relationships, however, are not naturally a product interior ri C of a convex set C is the interior relative
of the calculus; they are rooted in convex analysis to the smallest affine set containing C (see convex
and, in particular, in different ways of describing a programming), and onri dom//(concave or convex)
convex set. This article will lay out some basic is continuous.
duality theory from the point of view of convex Suppose that/is use. The minimal level a such
analysis, as a remedy for the microeconomic theory that /t(p 1)a the hyperplane in Rn+1 with normal vector
textbooks the reader may have suffered. (p, —1), contains hypo/is f*(p) = infYp • x — fix).
Why the normal vector (p, —1)? Because the graph
of the affine function x i—> f*(p) + px is a tangent
line to f the graph of/lies everywhere beneath it, and
no other line with the same slope and a smaller
Mathematical Background
intercept has this property. The function f*(p) is the
Duality in microeconomics is properly understood (concave) Fenchel transform or conjugate off and is
as a consequence of convexity assumptions, such as traditionally denoted f*. The construction of the
laws of diminishing marginal returns. In preceding paragraph can be done just this way: the
microeconomic models, many sets of interest are concave indicatorfunction of a convex set C is the
closed convex sets. The mathematics here is sur- function 5 fix) which is 0 on C and oo otherwise,
veyed in convex programming. The urtext for this and 5*c(p) = in/ s c p ■ x. For any function/ not
material is Rockafellar ( ). necessarily use or
Closed convex sets can be described in two ways:
by listing their elements, the ‘primal’
Duality
concave, the Fenchel transform /* is use and
concave. If / is in fact both use and concave, then /
** / This fact is known as the conjugate duality
theorem. Convex functions with range ( - oo, oo] are
treated identically. The function /is convex if and
only if / is concave, but the definitions are handled
slightly differently in order to preserve the intuition
just described. The set epi /= {x, a) : a > /(x)}, and
the convex Fenchel transform is defined differently:
f*(p)
= supxp ■ x f(x). The convex indicator function of a
convex set C is the function Sc(x) which is 0 on C
and +otherwise; its (convex) conjugate is dc (p) =
supx/? • x. These facts are discussed in convex
programming.
If concave functions have tangent lines, then
they must have something like gradients. A vector p
is a sub gradient off at x if/(x) + p ■ iy — x) < fly).
If/has a unique sub gradient at x, then/is differ-
entiable at x and p = V/(x), and conversely. But the
sub gradient need not be unique: the set df(x) of sub
gradients at x is the sub differential off at x. The
domain off dom/is the set of x such that fix) > - DC.
The sub differential is non-empty
for all x in its relative interior. It follows from the
definition of concavity (and is proved in convex
programming that the subdifferential correspon-
dence is monotonic: if p C_ dflx) and q € dfly),
thenfp — q) ■ (x y) < 0. If/is convex, then the
inequality is reversed, and (p q) ■ (x y) > 0. Finally,
suppose /is use and concave. Then so is its conjugate
/*, and their sub differentials have an inverse
relationship: p C_ dflx) if and only if x e df(p).
Cost, Profit and Production
In the theory of the firm, profit functions and cost
functions are alternative ways of describing the
firms’ technology choices. A technology is
described by a set of vectors F in RN. Each vector Z
C F is an input-output vector. We adopt the
convention that negative coefficients correspond to
input quantities and positive quantities correspond to
theory. It says that the net supply function of good i
outputs. Suppose that the first L goods are inputs
is the derivative of the profit function with respect
and the last M = N—L are outputs, so that F C Rh
to the price of good i. The usual proof is via the
/R™. It is convenient to assume free disposal, so
envelope theorem: the marginal change in
3081
that if (x, y) 6 F, and both x'<x and/ < y (more input
and less output), then (x', y') <E F. Two important
dual representations of the technology are the cost
and profit functions. The profit function is n(p, vv)
= sup<xy) c /. p ■ y + vv ■ x for p e and vv c_ Rl,
which is the conjugate of the convex indicator
function of F. The cost function too can be obtained
through conjugacy. The set F(y) = {x : (x, y) C_ F)
is the set of all input bundles that produce y. Then
C(y, vv) = supx e /,(y)w ■ x, that is, C(y, •) =
SF(y>*.
Immediately the properties of the Fenchel
transform imply that n (p,w) is convex in its
arguments and C(y,w) is concave in w, the profit
function is lsc and the cost function is use. (This
implies that both functions are continuous on the
relative interior of their effective domains.) Cost
and profit functions are also linear homogeneous.
Doubling all prices doubles both costs and reve-
nues. Cost is also monotonic. If wj < w/ for every
input /, then C(y, vv') < C(y, w) and if w\ < w/ for
all /, then C(y, vv) < C(y, vv).
The point of duality is that, if the technology is
closed and convex, then cost profit functions each
characterize the technology F. The conjugate
duality theorem (see convex programming) implies
that n*(x, y) = SF * *(x, y) = SF(x, y), the convex
indicator function of F:
sup p ■ x + w • y — 7i(p, vv)
(P,«-)eRN
0 if(x,y) GF,
Too otherwise.
If F is closed and convex, then each F(y) is convex.
If F is closed then F(y) will also be closed. Then
<5F<-V)is concave and use, so
sup vv • x T C(y, vv) = sup vv • xSF<y>*
weR( w Ry
(w) = ^(x).
Hotelling’s lemma is a famous result of duality
3082
profits from a change in price p is the quantity of
good i times the change in the price plus the price of
all goods times the changes in their respective
quantities. But the quantity changes are second-
order because the quantities solve the profit max-
imization first-order conditions, that price times the
marginal change in quantities in technologically
feasible directions is 0. Every advanced
microeconomics text proves this. A result like this is
true whenever the technology is convex, even if the
technology is not smooth.
The convex version of Hotelling’s lemma is a
consequence of the inversion property of sub dif-
ferentials for concave and convex / that/? c df(x) if
and only if x e df*(p). See convex programming for
a brief discussion.
Hotelling’s lemma (x,y) E 0 n(p,w) if and only if
ix, y) is profit-maximizing at prices (p, w).
Hotelling’s lemma is quickly argued. If (x, y) E
dn(p, w) = ddF*(p, W), then (p, w)
G d8F * *(x, y) = ddF(x, y). Then 8F(x, y) + (P, w) '
((x, /) - (x, y)) < 8F(x, y ) for all (x, y). This implies
that x E F and furthermore that ip, w) • ((*, y) - (x,
y)) < o for all (x, y) E Fin other words, that (x, y) is
profit-maximizing at prices (p,w). Conversely,
suppose that (x,y) is profit maximizing at 8 prices
(/AM ). Then (p,w) satisfies the sub gradient
inequality of 8F at (x,y), and so ip, w) E d&F.
Consequently,
(x,y) E dd F *(p, w ) = dn(p, vr).
The textbook treatment of duality observes that,
if net supply is the first derivative ofthe profit
function, then the own-price derivative of net supply
must be the second own-partial derivative of profit
with respect to price, and convexity ofthe profit
function implies that this partial derivative should
be positive, so net supply is increasing in price. The
same fact follows in the convex framework from the
monotonicity properties of the sub gradients.
Suppose that (u; p) and (w,p) are two price vectors,
and suppose that (x, y) and (x, y) are two profit-
maximizing production plans corresponding to the
two price vectors. Then (w — w', p — p’)(x — x\ y
— y ) > 0. If the two price vectors are identical for
all prices but, say, Pk ¥=Pk’ then (Pk -p'k){yk-y'k) >
o, and net
Duality
supply is non-decreasing in price. As with net
supplies, some comparative statics of conditional
factor demand with respect to input price changes
follows from the monotonicity property of
subgradients.
Another implication of profit function convexity
and (twice continuous) differentiability is symmetry
of the derivatives of net supply:
dyk (Pn d2n dy1.
dp i dpkdp1 dptdpk dpk
The convex analysis version of this is that for
any finite sequences of goods i,...,/,
Pi ■ (y, - yi) +Pj ■ (yk - y ; ) + •••+/>/• (y, - ,v/) < o.
This requirement, which has a corresponding
expression in terms of differences in prices, is
called cyci/c monoion/c/iy. All subdifferential
correspondences are cyclicly monotone. The con-
nection with symmetry is not obvious, but it helps
to know that Rockafellar ( ) leaves as an
exercise (and so do we) that cyclic monotonicity is
a property of a linear transformation corresponding
to an n x n matrix M if and only if M is symmetric
and positive semi-definite. Monotonicity is cyclic
monotonicity for sequences of length 2.
The other famous result in duality theory for
production is Shephard’s lemma, which does for
cost functions what Hotelling’s lemma does for
profit functions: conditional input demands are the
derivatives of the cost functions. This is dem-
onstrated in the same way, since the cost function
and the indicator function for the set of inputs from
which y is produceable are both convex and have
closed hypographs.
Utility and Expenditure Functions
A quasi-concave utility function U defined on the
commodity space R“ has upper contours sets, the
sets Ru of consumptions bundles which have utility
at least u, which are convex. If u is use, these sets
are closed as well.
Duality 3083

The expenditure function gives for each utility Furthermore, if x is demanded at prices p and utility
level u and price vector p the minimum cost of u. and y is demanded at prices q and the same utility
realizing utility u at prices p : e [p, u) = inf (p ■ x : u, then (p — q) • (x — y) < 0.
u(x) > u). If the infimum is actually realized at a The downward-sloping property just restates the
consumption bundle x, then .r is the Hicksian or monotonicity property of the subdifferential
compensated real income demand. correspondence. For the special case of changes in a
In terms of convex analysis, e(p, u) is the single price, the statement is that demand is non-
conjugate of the concave indicator function pu(x) of increasing in its own price.
the set R(u) = {x : U(x) > u j, that is, e{p,u) p*uip) ■ D
Thus e(p, u) will be use and concave inp for each it.
The expenditure fimetion is also linearly Equilibrium and Optimality
homogeneous in prices. If prices double, then the
least cost of achieving u will double as well. The equivalence between Pareto optima and com-
petitive equilibria can also be viewed as an
The duality of utility and expenditure functions
expression of duality. When preferences have
is that each can be derived from the other; they are
concave utility representations, quasi-equilibrium
alternative characterizations of preference. Since the
concave indicator function cp„(x) is closed andemerges from Lagrangean duality. Quasi-
equilibrium entails feasibility, profit maximization,
convex, e(-, uf = ipu{.x). For fixed u, the Fenchel
and expenditure minimization rather than utility
transform of the expenditure fimetion is the concave
maximization. That is, each trader’s consumption
indicator fimetion of R(u); infpp ■ x — e(p. u) is 0 if
allocation is expenditure minimizing for the level of
U(x) > u and —oo otherwise. If .r C R(u), then the
utility it achieves. The now traditional route of
cost of x at any price p can be no less than the
Arrow (
minimum cost necessary to achieve utility u. The ) and Debreu ( ) to the
Second Welfare Theorem first demonstrates that a
gap between the cost of x and the cost of utility
Pareto-optimal allocation can be regarded as a
level u is made by taking ever smaller prices, and so
its minimum is 0. Suppose that x is not in R(u). The
quasi-equilibrium for an appropriate set of prices.
separation theorem for convex sets says there is a
Under some additional conditions, the quasi-
price p such that p ■ x < inly s R(u) p . y; there is a
equilibrium is in fact a competitive equilibrium,
price at which x is cheaper than the cost of u. Now,
wherein utility maximization on an appropriate
by taking ever larger multiples of p, the magnitude
budget set replaces expenditure minimization. Our
of the gap can be made arbitrarily large, and so the
concern here is with the first step on this path.
value of the conjugate is - oo. Thus the conjugate is
Suppose that each of / individuals has prefer-
the concave indicator function of R(u). ences represented by a concave utility function on
Among the most useful consequence of the Ry , and that production is represented, as in section
duality between utility and expenditure functions is
“ ”, by a closed
the relationship between derivatives of the and convex set F of feasible production plans.
expenditure function and the Hicksian, or com- Suppose that 0 C_ F (it is possible to produce
pensated, demand. Hicksian demand. The com- nothing) and that the aggregate endowment e is
pensated demand at prices p and utility u are those
strictly positive. Assume, too, that there is free
consumption bundles in R(ii) which minimize disposal in production. Every Pareto optimum is the
expenditure at prices p. This result is just maximum of a Bergson-Samuelson social welfare
Shephard’s lemma for expenditure functions: function of the form flf-,u, defined on the set of all
consumption allocations. An allocation is a vector
Hicks Compensated Demand Consumption bundle
(x, y) where x e R+1 is a consumption allocation, a
x is a Hicks compensated consumption bundle at
consumption bundle for each individual, and v is a
prices p if and only if x C dpe(p, u).
production plan. The allocation is feasible if y G F
and y + e — ffiXt > 0. A Lagrangean for this convex
program is
3084 Duality

L(x,y,p) condition, and so a saddlepoint (.r*, y*,p*) exists;

Y. i , l ‘ ( x ‘ ) +P- ( y + e -Y .i x ‘) if .veR^.yef andpeRL,
+co that is, sup v; yL(x, y, p*) < L(x*, y*, p*) < L (x*, y\
if .vGR™,yeF and p^R^,
—oo otherwise,
p) for all V c RNI, y € F and p C R1. Then (x*, y*) is
where p is the vector of Lagrange multipliers for the Pareto optimal and p* solves the dual problem
L goods constraints. minA,supxy, L(x, y, p). The interpretation of (x*, y\
The possibility of 0 production and the strict p* ) as a quasi-equihbrium comes from examining
positivity of the aggregate endowment guarantee the dual problem. The dual problem can be
that the set of feasible solutions satisfies Slater’s rewritten as

inf
/> a1

(1)

In the dual problem, the Lagrange multipliers Historical Notes

can be thought of as goods prices. The Second
Welfare Theorem interprets the optimal allocation
as an equilibrium allocation using the Lagrange
multipliers as equilibrium prices. To see this, look
at the second line of ( ). At prices p, a production
plan is chosen fromy to maximize profitsp ■ y, so
the value of this term is n(p). Each consumer is
asked to solve
max A/W/fxo) — p ■ x = —min p ■ x
i The results on consumer and producer theory
= kiU* — m i n p ■ X — d ; ( n ; ( j q ) —
u*)
where u* = Uj(x*). The term being minimized is the
Lagrangean for the problem of expenditure
minimization, and so x- is the Hicksian demand for
consumer i at prices p and utility level u* u^x)).
finally, the optimal allocation is feasible, and so
(x*, y*,p*) is a quasi-equilibrium.
Given the observation about expenditure min-
imization, the saddle value of the Lagrangean is See Also
Duality ideas appeared very early in the marginal
revolution. Antonelli, for instance, introduced the
indirect utility function in 1886. The modem lit-
erature begins with Hotelling ( ), who pro
vided us with Hotelling’s lemma and cyclic
monotonicity. Shephard ( ) was the first mod
em treatment of duality, making use of notions such
as the support function and the separating
hyperplane theorem.
are surveyed more extensively in Diewert ( ),
who also provides a guide to the early
literature. In its focus on Fenchel duality, this
review has not even touched on the duality between
direct and indirect aggregators, such as utility and
indirect utility, and topics that would naturally
accompany this subject such as Roy’s identity.
Again, this is admirably surveyed in Diewert ( ).

- e
*(p*’ ► m
? ) + n(p*) Convex Programming
i ► Convexity
► duality
The planner chooses prices to minimize net sur- ► Lagrange Multipliers
plus, which is the sum of profits from production
and the excess of total Bergson-Samuelson welfare
less the cost of the consumption allocation.
Duhring, Eugen Karl (1833-1921)
► ■ areto cr ./
► >uas r:j
Bibliography
Arrow, K.J. 1952. An extension of the basic theorems of
classical welfare economics. In Proceedings of the sec-
ond Berkeley symposium on mathematical statistics and
probability, ed. J. Neyman. Berkeley: University of
California Press.
Debreu, G. 1951. The coefficient of resource utilization.
Econometrica 19: 273-292.
Diewert, W.E. 1981. The measurement of deadweight loss
revisited. Econometrica 49: 1225-1244.
Hotelling, H. 1932. Edgeworth’s taxation paradox and the
nature of demand and supply. Journal of Political
Economy 40: 577-616.
Rockafellar, R.T. 1970. Convex analysis. Princeton:
Princeton University Press.
Rockafellar, R.T. 1974. Conjugate duality and optimization.
Philadelphia: SIAM.
Shephard, R.W. 1953. Cost and production functions.
Princeton: Princeton University Press.
Duhring, Eugen Karl (1833-1921)
Tom Bottomore
Keywords
Bernstein, E.; Duhring, E. K.; Engels, F.; Marx,
K. H.; Positivism; Private property; Schumpeter,
F. A
JEL Classifications
B31
Duhring was bom on 12 January 1833 in Berlin and
died on 21 September 1921 at Nowawes bei
Potsdam. The son of a Prussian state official,
Duhring studied law, philosophy and economics at
the University of Berlin and practised law until
blindness obliged him to abandon this career. He
then became a Privatdozent at the University of
Berlin, where he taught philosophy and economics
from 1863 to 1877, and began to write
voluminously on a wide range
3085
of subjects, from the natural sciences to philosophy,
social theory and socialism, his aim being to
construct a system of social reform based upon
positive science. His system was expounded in a
series of books on capital and labour (1865), the
principles of political economy (1866), a critical
history of philosophy (1869), a critical history of
political economy and socialism (1871), and courses
in political economy and philosophy (1873, 1875).
Duhring was an adherent of positivism, concerned
in his philosophical works to expound a ‘strictly sci-
entific world outlook’, in opposition particularly to
the Hegelian dialectic. His economic writings
emphasize the role of political factors in the
development of capitalism, and he argued that
social injustice is not caused primarily by the
economic system, but by social and political
circumstances, the remedy being to control the
misuse of private property and capital (not abolish
them) through workers’ organizations and state
intervention.
Schumpeter ( , pp. 509-10), praised
Duhring’s history of mechanics (1873), which was
awarded an academic prize, suggested that he would
retain a prominent place in the history of anti-
metaphysical and positivist currents of thought, and
noted that he made an important criticism of
Marxist theory in his argument that political causes
had played a major part in constituting the property
relations of capitalist society. In other respects,
however, Schumpeter considered that Duhring had
made no significant contribution to economic
theory.
Engels, in his well-known book (originally
published as a series of articles), Herr Eugen
Duhring s Revolution in Science [Anti-Diihring\ (
-8), which has done more than anything
else to keep Duhring’s name alive, took a much
more critical view, deriding his work as a prime
example of the ‘higher nonsense’ which infected
German academic life. His philosophical views
were dismissed by Engels as ‘vulgar materialism’
and compared unfavourably with the ‘revolutionary
side’ of Hegel’s dialectics; and in the chapter of
Anti-Diihring devoted to the history of political
economy (largely written by Marx, but not
published in full until the
3086 Dummy Variables

third edition of the book in 1894 ), Diihring was

castigated for his superficiality and theoretical Dummy Variables
misconceptions. It was, however, the concern with
Duhring’s programme of social reform, and its Pietro Balestra
possible baleful effect on the developing labour
movement (Eduard Bernstein, for example, was
initially impressed by Duhring’s Cursiis of 1873,
though soon repelled by his anti-Semitism) that Abstract

originally provoked Engels’s articles, and was The dummy-variable method is a useful device
countered in the final section of the book for introducing, into a regression analysis,
(frequently reprinted later as a separate text under information contained in qualitative or categor-
the title Socialism, Utopian and Scientific) by an ical variables, that is, in variables that are not
exposition of Marxist socialism which became conventionally measured on a numerical scale,
enormously influential. such as race, sex, marital status, occupation, or
It seems doubtful that Diihring occupies more level of education. It is a means for considering
than a minor place in the history of economic and a specific scheme of parameter variation, in
social thought, except for this encounter with Marx which the variability of the coefficients is linked
and Engels, though Schumpeter ( p. 509) called him to the causal effect of some precisely identified
a ‘significant thinker’ and the entry in the qualitative variable. But when the qualitative
Encyclopedia of the Social Sciences (1931, vol. 5, effects are generic, as in the cross- section time-
p. 273) described his writings as ‘among the series model, an interpretation in terms of
important intellectual achievements of the random effects may seem more appealing.
nineteenth century’.

Keywords
Selected Works Covariance model; Cross-section time-series
model; Dummy variables; Engel curve; Error
1 8 7 1 . Kritische Geschichte der Nationaldkomie
component model; Qualitative variables; Ran-
und des Sozialismus. Berlin: T.
dom coefficient model
Grieben.
1873. Cursus der National- und
Sozialokonomie einschliesslich der Hauptpunkte JEL Classifications
der Finanzpolitik. Berlin: T. Cl
Grieben.
1 8 7 5 . Cursus der Philosophic als streng In economics, as well as in other disciplines, qual-
wissenschaftlicher Weltanschauung und itative factors often play an important role. For
Lebensgestaltung. Leipzig: E. instance, the achievement of a student in school
Koschny. may be determined, among other factors, by his
father’s profession, which is a qualitative variable
Bibliography having as many attributes (characteristics) as there
are professions. In medicine, to take another
Albrecht, G. 1927. Eugen Diihring: ein Beitrag zur example, the response of a patient to a drug may be
Geschichte der Sozialwissenschaften. Jena: G. Fischer. influenced by the patient’s sex and the patient’s
Engels, F. 1877-8. Anti-Diihring. Herr Eugen Diihring s smoking habits, which may be represented by two
revolution in science. Moscow: Progress Publishers,
1947. qualitative variables, each one having two attri-
Schumpeter, J.A. 1954. A history of economic analysis. butes. The dummy-variable method is a simple
London: Allen & Unwin.
Dummy Variables 3087

and useful device for introducing, into a regression Inserting these two dummy variables in the Engel
analysis, information contained in qualitative or curve, we obtain the following expanded regression:
categorical variables; that is, in variables that are
not conventionally measured on a numerical scale.
Such qualitative variables may include race, sex, Specification I
marital status, occupation, level of education,
region, seasonal effects, and so on. In some appli-
cations, the dummy- variable procedure may also be y , = a[dy + a2d2i + bxi + n ,
fruitfully applied to a quantitative variable such as which may be estimated by ordinary least- squares.
age, the influence of which is frequently U-shaped. Alternatively, noting that du + d2i = 1 for all i, we
A system of dummy variables defined by age can write:
classes conforms to any curvature and consequently
may lead to more significant results.
The working of the dummy-variable method is
Specification II
best illustrated by an example. Suppose we wish to
fit an Engel curve for travel expenditure, based on a
sample of n individuals. For each individual i, we y , = a2 + {a\ - a2)dy + bxi + u,
have quantitative information on his travel
which, again, may be estimated by ordinary least-
expenditures (y,) and on his disposable income (.r ; ),
squares.
both variables being expressed in logarithms. A
It is easy to see how the procedure can be
natural specification of the Engel curve is:
extended to take care of a finer classification of
yt = a + bxi + Ui education levels. Suppose, for instance, that we
actually have 5 education levels (s attributes). All
where a and b are unknown regression parameters we require is that the attributes be exhaustive and
and Uj is a non-observable random term. Under the mutually exclusive. We then have the two following
usual classical assumptions (which we shall adopt equivalent specifications:
throughout this presentation), ordinary least-
squares produce the best estimates for a and b.
Suppose now that we have additional informa-
tion concerning the education level of each indi- Specification I
vidual in the sample (presence or absence of college
education). If we believe that the education level yt = a\du + a2d2i + . . . + asdsi + by + u , -
affects the travel habits of individuals, we should
explicitly account for such an effect in the Specification II
regression equation. Here, the education level is a
qualitative variable with two attributes: college
= a s + (fli — a i ) d y + • • • + (as_i — a s )d s _ + bxi
education; no college education. To each attribute,
+ Ui.
we can associate a dummy variable which takes the
following form:
Obviously, the two specifications produce the same
results but give rise to different inputs. Specification
J 1 if college education I includes all the s dummy variables but no constant
'' | 0 if no college education term. In this case, the coefficient of dn gives the
specific effect of attribute j. Specification II
^ f 1 if college education ( 0 if no includes .v - 1 dummy variables and an overall
college education constant term. The constant term represents the
3088
specific effect of the omitted attribute, and the
coefficients of the different djr represent the contrast
(difference) of the effect of the /th attribute with
respect to the effect of the omitted attribute. (Note
that it is not possible to include all dummy variables
plus an overall constant term, because of perfect
collinearity.)
It is important to stress that by the introduction
of additive dummy variables, it is implicitly
assumed that the qualitative variable affects only
the intercept but not the slope of the regression
equation. In our example, the elasticity parameter,
b, is the same for all individuals; only the intercepts
differ from individual to individual depending on
their education level. If we are interested in
individual variation in slope, we can apply the same
technique, as long as at least one explanatory
variable has a constant coefficient over all
individuals. Take the initial case of only two
attributes. If the elasticity parameter varies
according to the level of education, we have the
following specification:
yt = a\du + a2d2i + b\d\iXi + b2d2lx, + ut.
Simple algebra shows that ordinary least-squares
estimation of this model amounts to performing two
separate regressions, one for each class of
individuals. If, however, the model contained an
additional explanatory variable, say z„ with con-
stant coefficient c, by simply adding the term cz, to
the above equation, we would simultaneously allow
for variation in the intercept and variation in the
slope (for x).
The dummy variable model also provides a
conceptual framework for testing the significance of
the qualitative variable in an easy way. Suppose we
wish to test the hypothesis of no influence of the
level of education on travel expenditures. The
hypothesis is true if the s coefficients a, are all
equal; that is, if the s 1 differences a2 - as , / 1 titative explanatory variables. The interested reader
... , s' - 1, are all zero. The test therefore boils down
to a simple test of significance of the v-1
coefficients of the dummy variables in Specification
II. If S 2, the z-tcst applied to the single coefficient
of dli is appropriate. If s > 2, we may conveniently
compute the following quantity:
Dummy Variables
(SSC - S S ) / ( s - 1)
SS/(n-s-l)
which is distributed as an F-variable with s 1 and n-
s- 1 degrees of freedom. In the above expression,
SS is the sum of squared residuals for the model
with the dummy variables (either Specification I or
II), and SSC is the sum of squared residuals for the
model with no dummy variables but with an overall
constant term.
In some economic applications the main
parameter of interest is the slope parameter, the
coefficients of the dummy variables being nuisance
parameters. When, as in the present context, only
one qualitative variable (with s attributes) appears
in the regression equation, an easy computational
device is available which eliminates the problem of
estimating the coefficients of the dummy variables.
To this end, it suffices to estimate, by ordinary
least-squares, the simple regression equation:
Specification III
y*=bx*+u*.
where the quantitative variables (both explained and
explanatory) for each individual are expressed as
deviations from the mean over all individuals
possessing the same attribute. For the dichotomous
case presented in the beginning, for an individual
with college education, we subtract the mean over
all individuals with college education and likewise
for an individual with no college education. Note,
however, that the true number of degrees of free-
dom is not n— 1 but n— 1 —s. The same procedure
also applies when the model contains other quan-
may consult Balestra ( ) for the conditions
under which this simple transformation is valid in
the context of generalized regression.
The case of multiple qualitative variables (of the
explanatory type) can be handled in a similar
fashion. However, some precaution
 Urban Rural
Dummy Variables
College education a a
l+ 2+ 3a
a\ + a2
No college education fll + «3 «i
must be taken to avoid perfect collinearity of the
dummy variables. The easiest and most informative
way to do this is to include, in the regression
equation, an overall constant term and to add for
each qualitative variable as many dummy variables
as there are attributes minus one. Take the case of
our Engel curve and suppose that, in addition to the
education level (only two levels for simplicity), the
place ofresidence also plays a role. Let us
distinguish two types of place of residence: urban
and rural. Again, we associate to these two
attributes two dummy variables, say e\i and e2z-. A
correct specification of the model which allows for
both qualitative effects is:
yz = a\ + d2<i\i + + bx \ +
Given the individual’s characteristics, the measure
of the qualitative effects is straightforward, as
shown in the following table:
The specification given above for the multiple
qualitative variable model corresponds to Specifi-
cation II of the single qualitative variable model.
Unfortunately, when there are two or more quali-
tative variables there is no easy transformation
analogous to the one incorporated in Specification
HI, except under certain extraordinary circum-
stances (Balestra ).
One such circumstance arises in connection with
cross-section time-series models. Suppose that we
have n individuals observed over t periods of time.
If we believe in the presence of both an individual
effect and a time effect, we may add to our model
two sets of dummy variables, one corresponding to
the individual effects and the other corresponding to
the time effects. This is the so-called covariance
model. The number of parameters to be estimated is
possibly quite large when n or t or both are big. To
avoid this, we may estimate a transformed model
(with no dummies and no constant term) in which
each quantitative variable (both explained and
explanatory) for individual i and time period j is
3089
transformed by subtracting from it both the mean of
the z'th individual and the mean of the /th time
period and by adding to it the overall mean. Note
that, by this transformation, we lose n+t-l degrees of
freedom.
To conclude, the purpose of the preceding
expository presentation has been to show that the
dummy-variable method is a powerful and, at the
same time, simple tool for the introduction of
qualitative effects in regression analysis. It has
found and will undoubtedly find numerous
applications in empirical economic research.
Broadly speaking, it may be viewed as a means
for considering a specific scheme of parameter
variation, in which the variability of the coefficients
is linked to the causal effect of some precisely
identified qualitative variable. But it is not, by any
means, the only scheme available. For instance,
when the qualitative effects are generic, as in the
cross-section time-series model, one may question
the validity of representing such effects by fixed
parameters. An interpretation in terms of random
effects may seem more appealing. This type of
consideration has led to the development of other
schemes of parameter variation such as the error
component model and the random coefficient model.
A final remark is in order. In the present dis-
cussion, qualitative variables of the explanatory type
only have been considered. When the qualitative
variable is the explained (or dependent) variable, the
problem of these limited dependent variables is far
more complex, both conceptually and
computationally.
Bibliography
Balestra, P. 1982. Dummy variables in regression analysis. In
Advances in economic theory, ed. Mauro Baranzini.
Oxford: Blackwell.
Goldberger, A.S. 1960. Econometric theory, 218-227. New
York: Wiley.
Maddala, G.S. 1977. Econometrics, chap. 9. New York:
McGraw-Hill.
Suits, D.B. 1957. Use of dummy variables in regression
equations. Journal of the American Statistical Associ-
ation 52: 548-551.
3090
Dumping
Wilfred J. Ethier
The term ‘dumping’ has been used for centuries in a
general way to refer to export sales at a price low
enough to cause significant harm to some interests
in the importing country. Beginning early in this
century, many countries instituted anti-dumping
laws, and this has required a more precise definition
of the term. The most common definition, both in
the law and among professional economists, is
export sales at a price below that at which similar
goods are sold in the domestic market of the
exporting country, taking into account differences
in quality, attendant services and the like. However,
an alternative definition, export sales at a price
below the cost of production, is also incorporated
into many of the laws, and this alternative has in
recent years become of increasing practical
importance.
Anti-dumping laws typically define the practice,
prohibit it, provide for a penalty in cases where it
nonetheless occurs, and establish an administrative
procedure for determining in specific cases whether
it has occurred and what penalty to impose. The
penalty is usually in the form of an import levy
related to the ‘dumping margin’, or difference
between the export price and the source-country
domestic price (or cost of production).
Such anti-dumping duties, though inherently at
odds with most-favoured-nation treatment, are
internationally accepted. The General Agreement on
Tariffs and Trade (GATT) does not outlaw
dumping, but it does countenance anti-dumping
laws. The Tokyo Round of trade negotiations
produced a code of conduct for antidumping
legislation.
Numerous instances of alleged dumping have
characterized recent tariff debates within and among
the industrial countries. Recent changes in the
administration of US anti-dumping and
countervailing-duty statutes will likely further
increase their use. More generally, the marked
postwar reduction in tariffs on manufactured goods
within the GATT framework on international
Dumping
convergence of cost structures in the industrial
economies, prompt the conjecture that antidumping
and anti-subsidy statutes will be a principal battle-
ground for the ‘new protectionism’ concerning trade
in manufactures among the developed economies. If
so, the theory of dumping must become a major part
of the positive theory of protection relevant to such
trade.
The early literature generally defined dumping
as price discrimination between national markets.
This was the definition adhered to by Viner ( ) in
his
classic treatment and followed by most major
authors (see Yntema ; Robinson ; Haberler ). Indeed,
much of the early theory of price discrimination was
developed in this context. Two problems arise when
the phenomenon is viewed in this light. The first is
why the firm is able to discriminate. This requires
the firm to have some control over price; that is,
imperfect competition is central. It also requires the
firm to be able to segment markets on a national
basis: tariffs or other trade barriers can serve this
purpose. The second problem is why the export
price should be lower than the domestic price rather
than vice versa One possible response is that such
reverse dumping is indeed as common as dumping
but is simply not a policy issue. For example, the
sale of luxury German automobiles in the US at
prices much higher than those in Germany brings
forth not a whimper of an official threat from
Washington, while the sales of low-priced European
automobiles in the US was the occasion for a
celebrated action some years ago. An alternative
response is to hunt for circumstances that allow
dumping to be more than a mere accident. One
possibility is that the trade pattern is unidirectional
and given by other considerations. Exporting firms
thus compete only among themselves at home, but
also with foreign firms in the export market. Thus,
even if the market elasticity is the same in both
countries, the elasticity facing each firm will be
higher in the importing country, because more firms
compete there. Thus, other things equal, exporters
will charge lower prices abroad than at home (see
Eichengreen and van der Ven ).
A second possibility involves transport costs
between markets. Other things equal, such costs
result in a firm having a smaller share, in equilib-
rium, in its export market than in its domestic
Dumping
market This again creates a presumption that the
elasticity facing the firm is higher for foreign sales
than for domestic sales. Furthermore, this reasoning
does not require a unidirectional trade pattern; it is
quite consistent with reciprocal dumping, or cross-
hauling, with each country dumping in the other
(see Brander and Krugman ).
Most of the formal theory of dumping essen-
tially consists only of the theoiy of monopolistic
price discrimination between two markets. But by
contrast the ‘sales at a price below cost of produc-
tion’ criterion has gradually become relatively more
important in recent years, both in practice and in
revisions of anti-dumping laws. Economists have
long recognized that the two criteria are not
inconsistent. A price-discriminating firm might well
price its exports below average cost in a slump as
long as export revenues at least cover the variable
cost of producing those exports {sporadic or
cyclical dumping). Or the firm might permanently
sell its exports below average cost if those exports
allow it to realize sufficient economies of scale.
Predatory dumping, to drive rivals from the market,
has long received much public attention, but
economists have typically minimized its importance
(see Viner ). How anti-dumping laws might in fact
be applied in all these situations has, not
surprisingly, concerned economists for years. More
recently, attention has encompassed cases where
export price does not cover even marginal cost. This
might well occur ex post if the exporting firm must
commit itself before demand conditions in the
export market are fully known. Or the firm might do
so deliberately if, instead of maximizing profit, it
wishes to maximize sales subject to a profit
constraint. More interesting, and closer to recent
work in industrial organization, is the possibility
that export sales, even at a low price, might make it
easier for the firm to maintain excess capacity for
the purpose of deterring entry by potential rivals
(see Davies and McGuinness ).
Even though economists no longer confine
themselves to price discrimination, dumping has
been treated (aside from sporadic instances) usually
either as profit (or sales) maximization by a
discriminating monopolist or as an oligopolistic
tactic to eliminate competition, to deter entry, or to
3091
enforce a cartel. Industries with dumping (or
allegations of dumping) are most often char-
acterized by large fixed costs, factor-market rigid-
ities, susceptibility to demand fluctuations and
downward price rigidity. Though by no means
inconsistent with oligopolistic rivalry in segmented
markets, these characteristics involve much more.
Thus our theory largely excludes those
considerations fundamental to most contemporary
problems: imperfectly adjusting factor markets in
the presence of changing conditions of product
demand. The earlier literature did consider the
related problem of dumping to stabilize production
over the business cycle. Viner ( p. 28), although
conjecturing that ‘it is probable that this is the most
prevalent form of dumping’, basically treated it as
only a distinct motive. The interdependence
between such dumping and factor-market
equilibrium within the relevant industries of trading
countries is critical. The ability to dump aboard
during periods of slack demand allows a firm to
offer its workers greater job security, and thereby
allows that firm to pay lower wages over time than
it would have to do if it did not offer that security.
Thus the possibility of dumping influences the
normal trading equilibrium. Furthermore, that
equilibrium clearly must be sensitive to employment
practices in different countries and to the relations
between the business cycles of the various countries
(see Ethier ).
In addition to the deficient treatment of the
fundamental issues involving factor markets, we
have no theory of anti-dumping laws. The basic
theory of tariffs applies of course to anti-dumping
duties put in place and left there. But an anti-
dumping law is a threat to impose (with consider-
ably less than certainty) a duty in response to certain
behaviour on the part of exporters and so will
influence that behaviour even if not actually
imposed. International trade theory has not yet
addressed this issue. This is just one aspect, though
an especially important one, of our prominent lack
of a contemporary theory of protection. These two
omissions (factor markets and antidumping laws)
are serious, but the first is being addressed, and we
have the technical equipment to deal adequately
with both of them, so one would expect the
deficiencies to be mended soon.
3092
See Also
Bibliography
Brander, J., P. Krugman, and 3/4. 1983. A ‘reciprocal
dumping’ model of international trade. Journal of
International Economics 15(3/4): 313-321.
Davies, S.W., and A.J. McGuinness. 1982. Dumping at less
than marginal cost. Journal of International Economics
12(1/2): 169-182.
Eichengreen, B., and H. van der Ven. 1984. US antidumping
policies: The case of steel. In The structure and evolution
of recent US trade policy, ed. R.E. Baldwin and A.O.
Kreuger. Chicago: University of Chicago Press, for the
National Bureau of Economic Research.
Ethier, W.J. 1982. Dumping. Journal of Political Economy
90(3): 487-506.
Haberler, G. 1937. The theory of international trade with its
applications to commercial policy. New York: Macmillan.
Robinson, J. 1933. The economics of imperfect competition.
London: Macmillan.
Viner, J. 1923. Dumping: A problem in international trade.
Chicago: University of Chicago Press.
Viner, J. 1931. Dumping. In Encyclopedia of the social
sciences, ed. E.R.A. Seligman and A. Johnson. New
York: Macmillan.
Yntema, T.O. 1928. The influence of dumping on monopoly
price. Journal of Political Economy 36: 686-698.
Dunbar, Charles Franklin
(1830-1900)
A. W. Coats
Dunbar’s career illustrates the narrow gap between Richard B. Freeman
practical and academic economics in his lifetime,
for he demonstrated that scholarly instincts,
common sense and knowledge of current affairs
could overcome deficiencies in formal academic Keywords
training. Exactly 20 years after graduating from Business cycles; Dispute resolution; Dunlop, J.;
Harvard in 1851 he returned as Professor of Industrial relations; Labour’s share of income;
Political Economy, having previously worked in a Marginal productivity theory; Mediation; Wage
mercantile business, qualified and practised as contours; Wage determination
Dunbar, Charles Franklin (1830-1900)
a lawyer, and written articles on political questions
for the Boston Daily Advertiser, of which he was
sole proprietor and editor from 1865 to 1869. After
President Eliot’s invitation to Harvard, Dunbar
spent two years travelling and studying in Europe,
and subsequently served as Head of the Department
of Political Economy for nearly 30 years, Dean of
the College (1876-82) and the Faculty of Arts and
Sciences (1890-95), and as editor from 1886 to
1896 of The Quarterly Journal of Economics, the
first English-language scholarly periodical in the
subject. His election as second President of the
American Economic Association in 1892, following
Francis A. Walker, testifies to his standing in the
emerging economics profession. While he
published comparatively little, his works on
currency, finance and banking were widely
respected, and his essays on the history, condition
and methods of economics were wise and balanced
at a time of intense controversy.
Selected Works
Dunbar, Charles Franklin. 1891a. Chapters on the
theory and histoiy of banking. New York: G.P.
Putnam’s Sons. 2nd enlarged, edn, ed. O.M.W.
Sprague, 1901; 3rd enlarged edn, 1917. Dunbar,
Charles Franklin. 1891b. Laws of the United States
relating to currency, finance, and banking from
1789 to 1891. Boston: Ginn & Co. Revised, edn,
1897.
Sprague, O.M.W. (ed.). 1904. Economic essays.
New York: Macmillan Co.
Dunlop, John Thomas (1914-2003)
Dunlop, John Thomas (1914-2003) 3093

problems, which should be analysed as such rather

JEL Classifications
than as a bourse. His mode of analysis was that of a
B31
naturalist, who looks at the world with his own eyes
and experience, with direct knowledge of the
John Dunlop was an extraordinary labour econo-
institutions and practitioners, without trying to force
mist, Professor and Dean of the Faculty at Harvard
observation into a narrow conceptual framework.
University, Secretary of Labor of the United States,
Dunlop’s career spanned a wide variety of
and mentor to students and practitioners in the
activities. Earning his AB (1935) and Ph.D. (1939)
world of labour. He was extraordinary because he
from Berkeley, he rose to become professor of
was more than an economist and because he was
economics at Harvard and Dean of the University
driven by a moral vision of what economists and
(1970-1973), when he helped stabilize the
academics should do to make the world better.
university during a period of student disorders, and
Labour economists and policymakers paid close
Lamont University Professor (1970-2003). He
attention to Dunlop’s thoughts because he
worked for the National War Labor Board (1943-
combined academic research with unparalleled
1954); served as member or chair on various
practical experience in solving problems and
national panels with responsibility for resolving
building institutions. His academic writings, which
labour disputes; led labour-management committees
include several classic articles as well as major
in areas ranging from missile sites to apparel, the
books, reflect Dunlop’s participation in events and
public sector, and health; served as Director of the
direct observations of social behaviour.
Cost of Living Council (1973-1974), and as
Dunlop first attracted academic attention with
Secretary of Labor of the United States (1975-
his 1938 Economic Journal article on the move-
1976). From 1993 to 1994 he chaired the
ment of real and money wages over the business
Commission on the Future of Worker-Management
cycle, which forced Keynes to admit that the
Relations, popularly known as the Dunlop
General Theory was wrong on this issue: real
Commission, which was given the charge ‘to
wages fall in recessions not in booms, contrary to
recommend ways to improve labor-management
simple marginal productivity analysis. Quite an
cooperation and productivity’. The politics and
achievement for a 24-year-old economist. Dunlop
economics of the time were not right, however, for
followed this with Wage Determination Under
bringing management and labour to a consensus on
Trade Unions (1944), in which he modelled unions
modernizing labour relations, so that much of the
as optimizing organizations; with analyses of the
Commission’s recommendations went unheeded.
cyclic variation of labour’s share, with the concept
Dunlop approached his work -advising presi-
of ‘wage contours’ that captured the notion that
dents and cabinet officials and telling academics
product markets influenced wages, and with
about the real world and practitioners about aca-
numerous analysis of wage determination, labour
demic theory - with one goal: to help solve prob-
relations, mediation and dispute resolution.
lems. The moral principle that guided him - that
Dunlop’s book Industrial Relations Systems (1958)
academics should use their knowledge and skill to
sought to develop a broader perspective on how
help solve problems faced by real people, by
labour relations fit into economics.
workers and firms, and governments - represents
In the 1980s, concerned that labour economists
social science at its best.
were limited in their conceptual vision by narrow
optimizing models and in their empirical analysis
by extant government data-sets, Dunlop carped at
them for failing to see what he could see in the
labour market. Dunlop saw the labour market as
pre-eminently a social institution to resolve labour Selected Works

1938. The movement of real and money wage rates.

Economic Journal 48: 413—434.
3094 Dunoyer, Barthelemy Charles Pierre Joseph (1 786-1862)

1944. Wage determination under trade unions. rapports avec la liberte. In 1832 he was elected to
New York: Macmillan. the French Institute, and in 1845 he became pres-
1948a. Productivity and the wage structure. In ident of the Societe d’Economie Politique. He spent
Income, employment and public policy: Essays in two decades in public life, entering the government
honor of Alvin H. Hansen. New York: W.W. Norton. in 1830 under the bourgeois monarchy of Louis-
1948b. The development of labor organizations: Philippe, and withdrawing after the coup d’etat of
A theoretical framework. In Insights into labor 1851. His articles appeared frequently in the
issues, ed. R. Lester, and J. Shister. New York: Journal d’Economie Politique and in other French
Macmillan. journals.
1057. The task of contemporary theory. In New Dunoyer added nothing new to economic theory
concepts in wage determination, ed. G. Taylor, and F. but he was part of a group of French radicals who
Pierson. New York: McGraw-Hill. helped create a powerful means of social analysis
1058. Industrial relation systems. New York: by fusing liberal historicopolitical thought with the
Henry Holt & Co. economic orthodoxy of J.B. Say. Inspired by Turgot
1084. Dispute resolution. Dover: Auburn House and Condorcet, Dunoyer and his cohorts advanced
Publishing Co. an evolutionary theory of history that identified
progress with the gradual disintegration of authority
and its replacement by the quiescent, voluntary
Bibliography relationships of the marketplace. These writers
anticipated the flowering of industrielisme, of a kind
Segal, M. 1986. Post-institutionalism in labor economics: the
forties and fifties revisited. Industrial and Labor apart from Saint-Simon’s insofar as it envisioned
Relations Review 39: 388-403. government as a mere subsidiary institution,
charged mainly with the functions of preserving
order and ministering to the needs of production.
Having thus nested their basic anarchism in an
evolutionary concept of social development, the
group fit surprisingly well into the republican and
constimtional framework of the July Monarchy.
Dunoyer, Barthelemy Charles Pierre
The product of Dunoyer's mature thought was
Joseph (1786-1862)
his three-volume work, De la liberte du travail.
R. F. Hebert Despite its brilliance and good sense, it is more a
history of civilization than a sustained economic
treatise. Dunoyer anticipated Herbert Spencer by
developing the idea that society is an organic
composition of institutions and individuals with
French economist and publicist, born at Carennac
specific functions. Although he regarded govern-
(southwest France) on 20 May 1786, died at Paris
ment's role as minimal, the presence of government
on 4 December 1862. Dunoyer studied law in Paris,
professionals finds justification in his conception of
where he befriended Charles Comte, who shared his
'immaterial wealth’ (i.e. services). Although he
liberalism and joined him in founding and editing
followed classical economics in most things,
Le Censeur. a journal of institutional and legal
Dunoyer rejected Say’s Law, holding that a general
reform. The journal was discontinued in 1820 due
glut could arise due to the ignorance or error of
to increasingly repressive press laws. Subsequently
entrepreneurs, or to the unequal distribution of
Comte went to Switzerland, whereas Dunoyer
wealth. Unlike Sismondi, however, whose ideas had
stayed in Paris and devoted himself exclusively to
a certain allure, he spurned government palliatives,
economics. He became professor of political
trusting the growth of industry to gradually reduce
economy at the Athenee, later publishing his
entrepreneurial error and to smooth
lecrttres under the title L Industrie et la morale
considerees dans lews
Duopoly
the distribution of income. Dunoyer also denied the
classical theory of rent, because he admitted only
one factor of production, labour. On population
matters he was an unregenerate Malthusian, which
tempered his basic faith in progress with a hint of
pessimism.
Selected Works
1825. L'Industrie et la morale consideree dans
leurs rapports avec la liberte. Paris: A. Sautelet.
Revised, enlarged and reprinted as Nouveau
traite d’economie sociale, 2 vols. Paris: A.
Sautelet, 1830.
1 8 4 0 . Esprit et methodes compares de l
’Angleterre et de la France dans les entreprises
de travaux publics et en particulier des chemins
de fer. P a r i s : C a r i l i a n - G o e u r y
et V. Dalmont.
1845. De la liberte du travail, 3 vols. Paris:
Guillaumin.
References
Allix, E. 1911. La deformation de l’economie politique
liberate apres J.B. Say: Charles Dunoyer. Revue cl 'Histoire
Economique et Sociale 4: 115-147. Villey-Desmeserts, E.L.
1899. L'oeuvreeconomique de Charles Dunoyer. Paris: L.
Larose.
Weinburg, M. 1978. The social analysis of three early 19th-
century French Liberals: Say. Comte and Dunoyer.
Journal of Libertarian Studies 2: 45-63.
Duopoly
James W. Friedman
A duopoly is a market in which two firms sell a
product to a large number of consumers. Each
consumer is too small to affect the market price for
the product: that is, on the buyers’ side, the market
is competitive. Therefore, in its essence duopoly is
a two player variable sum game. Each of the two
duopolists is a rational decision-maker
3095
whose actions will affect both himself and his rival.
Although the interests of the duopolists are
intertwined, they are not wholly coincident nor
wholly in conflict. In contrast to the agents in
competitive markets, the duopolists must each
concern themselves with what the other duopolist is
likely to do.
The situation facing the duopolists is non-
cooperative in the sense that they are barred from
making binding agreements with one another. The
relevance of this depends crucially on whether the
model is a static market (i.e. a one-time-only, or
one-shot market) or a market consisting of many
time periods.
The first study of duopoly is the great contri-
bution of Cournot ( ) in which the decision
problem of the firms is posed for a homogeneous
products market in a static setting. The equilibrium
concept proposed by Cournot, variously called the
Cournot equilibrium or the Cournot-Nash
equilibrium, has become a cornerstone of non-
cooperative game theory. To sketch his model let x
and y be the output levels of firms A and B, let /(x +
y) be the inverse demand function for the market,
let C(x) and F (y) be the two firms’ total cost
functions, and let their respective profit functions be
nA = xflx + y) C(x) and
T? = yAx + y) - r (y).
Cournot proposed as an equilibrium a pair of
output levels (x, y) such that neither firm could have
obtained higher profit by having chosen some other
output. Thus it4 is maximized with respect to x (with
y given), while, simultaneously, it8 is maximized
with respect to y (with x given). If xc > 0 and yc > 0
the Cournot equilibrium is a solution to the
simultaneous equations
OTF
-fa = f{x + y) + xf'{x + y) - C'{x) = 0
^=f{x + y)+yf{x + y)-r'{x) = o
The Cournot equilibrium defines consistency
conditions. If firm A contemplates (xc, yc) as an
outcome, and believes firm B is contemplating the
same output pair, then firm A will see (a) that it
cannot do better than to choose xc (given the
3096
expectation that firm B will choose / ) and (b) that,
should firm B go through the same thought process,
it will reach a parallel conclusion.
To translate this model into the language of
game theory, player A chooses a strategy .r from the
set of all allowed output levels, say all .r > 0. This
set, [0, oo], is called the strategy space or strategy
set of the player. Similarly for player B. The
players’ payoff functions are their respective profit
functions. Thus the payoff function of a player gives
his payoff as a function of the strategies of all
players in the game. At a non-cooperative
equilibrium (see Nash ; Owen , or Friedman ) no
player could obtain a higher payoff through the use
of a different strategy, given the strategies of the
other players. Note, finally, that the actual
behaviour of one duopolist cannot affect the actual
behaviour of the other in this static setting, because
they choose their output levels simultaneously.
They do take one another into account in making
decisions by analysing the game using both payoff
functions.
Cournot's contribution went largely unnoticed
for nearly half a century, after which it was
scathingly reviewed by B ertrand ( ). B ertrand
berates Cournot on two grounds. First he says that
the firms will collude to achieve monopoly-like
profits. This possibility is acknowledged by Cournot
who made a conscious choice to explore behaviour
in the absence of collusion. Bertrand’s point was
echoed later by Chamberlin ( ),
although neither of them showed how the duopolists
could be expected to maintain a collusive agreement
nor did they solve the problem of the distribution of
profits between the firms. These issues are
addressed below in connection with recent
developments.
Bertrand’s second criticism is that price, not
output, should be the firm’s decision variable. Then,
using Cournot’s mineral spring example in which
C(x) = T (y) = 0, he sketches the Bertrand
equilibrium, arguing that consumers will buy from
the firm charging the lower price, and showing that
the only prices that can be in equilibrium are zero
for both firms. Bertrand’s equilibrium concept is
precisely that of Cournot, transferred to the price
choosing variant of Cournot’s model. Bertrand’s
analysis was taken
Duopoly
up, elaborated and extended by Edgeworth ( ). He
supposed that the firms have produc
tion capacity limits, each of which is less than the
market demand at zero price. Consequently no pair
of prices is an equilibrium.
While the logic of Bertrand and Edgeworth is
correct, the economic relevance is dubious. Real
world firms choose both prices and output levels;
however, the discontinuity of one firm’s sales with
respect to another firm’s decision variable is not an
obvious feature of economic life. Consequently, the
Cournot formulation seems preferable. A way to
reconcile price choosing firms with an absence of
demand discontinuities is via differentiated products
models.
Edgeworth and many of his contemporaries
thought there was no worthwhile content in
Cournot’s duopoly theory. Edgeworth ( ,
p. Ill), writing forty years after Bertrand, said ‘Now
the demolition of Cournot’s theory is generally
accepted. Professor Amoroso is singular in his
fidelity to Cournot’. Amoroso’s good judgement
was shared by Wicksell ( ). It is now generally
accepted that Cournot was the first to perceive
clearly and enunciate the game theoretic concept of
non-cooperative equilibrium, which received a
general statement from Nash ( ) and is the
cornerstone of one of the main parts of game theory.
The next influential innovation is due to Bowley
( ) who invented the conjectural var
iation (which later received this name from Frisch ).
He wrote the two firms’ first order conditions for
equilibrium as dnA/dx + (dnA/dy) (dy/d x) = 0 for
firm A and dt^/dy + (dnB/dx) (dx/d v) = 0 for firm B.
The dy/dx in firm A’s condition indicates the way
that A thinks B’s output choice will vary according
to the way that A varies his own output choice. A
parallel meaning attaches to dx/dy in B’s first order
condition. The presence of these conjectural
variation terms is indefensible in a static model, but
it shows the underlying concern that writers had
with dynamic models, while, at the same time,
limiting their formal analysis to static models.
Given that the model is static with the two firms
simultaneously selecting outputs, and doing so only
once, there can be no conjectural variation. B’s
output choice will depend on what B expects A to
do, but that expectation will
Duopoly
not vary as A changes his mind about what output to
select. B's expectation depends on B's thought
processes and the information B has about the
structure of the model, and does not depend on As
actual thought processes.
Dynamic elements of reaction of one firm to the
choice of another go back to Cournot who
performed a ‘stability’ analysis. He solved dnA/d .r =
0 to obtain x = v (y) and dnB/dy = 0 to obtain y = U
(A'). He looked for conditions under which, starting
from an arbitrary ,r°, the sequence (x,„ y"+1) for n =
0, 2, 4, ... would converge to (,r c, / ), the Cournot
equilibrium. Bertrand and Edgeworth also wrote of
actions and reactions, and Bowley introduces a new
reactive element with his conjectiual variation
terms. Later Stackelberg ( ) posed the leader-
follower
duopoly in which one firm, say A, chooses x and,
after that choice is communicated to B, y is chosen.
Swill always choosey according toy = w(x) and this
is known to A who maximizes tv4 = xf [.r + u (x)] —
C(.r) with respect to x. Note that a conjectiual
variation term for A makes a legitimate appearance
because B's decision is, in fact, a function of A’s
choice. Wicksell ( ) and
Bowley ( ) anticipate Stackelberg’s
leader-follower equilibrium in their discussions of
bilateral monopoly. All of these treatments strongly
suggest an explicitly multiperiod formulation under
which each firm maximizes a discounted profit
stream and behaves according to a reaction function
under which a firm’s output choice in time t is
selected as a function of the other firm’s output
choice in time t— 1. The last twenty years have seen
such analysis, beginning with Friedman ( ).
The next major step in duopoly was the recog-
nition that, in many industries, the firms sell very
similar, non-identical, products. In such a market, it
is equally easy to represent the firms as price
choosers or as quantity choosers. In either case,
equilibrium can readily involve the firms selling at
different prices. The pioneers here are Hotelling (
) and Chamberlin ( ). To sketch a dif
ferentiated products duopoly, let the firm’s prices be
p and r, and let their demand functions be x = (f>(p,
r) and y = ij/(p, r), respectively. The two firm’s are
assumed to produce gross (but
3097
imperfect) substitutes, so frip, r) > 0 and r) > 0, but
the own-price derivatives ((f)p and ij/r) are negative
and both firms’ total revenues are bounded. Profit
functions are tA4 = pf (p, r) - C[(f> (p, /')] and n8 =
rf (p, r) — T [if (p, /')]. A non-cooperative
(Coumot-Nash) equilibrium occurs at a price pair
(pc, rc) for which tA4 is maximized with respect to p
(given r rc) and
is maximized with respect to r (given/) p ).
Many writers have maintained that the Cournot
equilibrium should not be expected to occur in
practice because it does not lie on the firms’ profit
possibility frontier. In addition to Bertrand and
Chamberlin there is a famous passage in Smith (
) maintaining that people in the same line of
business will attempt to collude whenever they get
together. In response to such observations several
points can be made. Smith’s passage is a comment
in passing that is not made within an analytical
framework, so it cannot be closely judged. Bertrand
and Chamberlin are discussing specific models
within which their remarks do not hold up well,
because the consistency condition embodied in the
Cournot equilibrium is quite compelling and would
be violated by collusive behaviour. Any agreement
between the two firms-in a static setting where
binding agreements cannot be made- will break
down because at least one firm will note that, given
the agreed decision for the rival, it can do better by
violating its agreement. But both firms can perceive
the incentives of either one of them, thus the only
acceptable agreement in such circumstances is for a
price pair (or output pair, if the firms are output
choosers) such that, given the price of its rival,
neither firm can gain by deviating from its
agreement. Such a self-enforcing agreement is
merely a non-cooperative equilibrium. We are back
at Cournot.
However, this is far from the last word on
collusion in the absence of legally binding agree-
ments. Bertrand, Chamberlin, and others who have
made, or agreed with, their assertion probably are
motivated by a belief that voluntary collusion
sometimes occurs in actual markets. They may be
correct in their empirical observation; however, it
remains true that voluntary collusion is not
convincing in the traditional one-shot
3098
models. Therefore, the clear suggestion is that one-
shot models are simply inadequate for analysing
voluntary collusion. Suppose, then, that the model
is changed to have an infinite horizon with each
firm having a discount parameter of a Then, letting
t denote time, player ,4 seeks to maximize
oo oo
Y*'7^ = Ya!\Pi4>ip,A't) — C[(j}(p,,r,)]
t=0 / =0
and the objective function of player B is
OO OO
Y = Y rt) - r[\l/(p„ r,)]
t=0 / =0
Strategy becomes much more complex than in
the static model, because there will be an infinite
succession of price choices by each firm and, prior
to making a price choice any time after t = 0, the
firm will know what past prices have been selected
by its rival. For each t, a firm can choose its price
according to a function (i.e. a rale) that depends on
all past price choices of both of them. The rale for
one period can be different from the rale for
another. A strategy for a firm is a collection of such
rales, one for each period t.
In this model it may be possible to find a non-
cooperative equilibrium that yields an outcome on
the profit possibility frontier. Such an equilibrium is
based on three critical prices for each firm. First
there is (//, f ), the Cournot price pair. Second there
is (p*, r*), chosen so that profits at (p*, r*) are on
the profit possibility frontier and are higher for each
firm than at (pc, f). Third define // as the price for A
that maximizes n p4> (p, r*) — C[4> (p, r*)], and
A
define iJ in a parallel way for B. Now consider the
following strategy for firm A : p = p * , p p for / > 0 if
0 t
(Pk, rk) = (p\ r ) fork = 0,..., t- 1, andpt = p°
otherwise. Imagine a parallel strategy fori?. These
strategies amount to a firm saying ‘I will begin by
cooperating and will continue to cooperate as long
as we both have cooperated in the past. If ever a
lapse from cooperation occurs, I will revert to static
Cournot behaviour’.
Duopoly
Whether this pair of strategies is a non-
cooperative equilibrium depends on the sizes of a
and the profits at (/?*, /-*), (//, /-*), (/?*, r ), and (pc,
rc). A’s choice boils down to comparing (i)
receiving the profit associated with (p*. r*) in all
periods or (ii) obtaining the larger profit associated
with (//, r ) for just one period and the reduced
profit associated with (//, r ) in all subsequent
periods. If a is near enough to one, both firms will
prefer alternative (i). Thus both firms can be better
off following the ‘cooperative’ strategy. Note, how-
ever, that this cooperative outcome is the result of
following non-cooperative equilibrium strategies.
The strategy pair is chosen so that no single firm
can increase its payoff by altering its strategy, given
die strategy of the other firm. The strategies are
designed so that deviating from cooperative behav-
iour is followed by punishment, and the punishment
is carefully crafted so that it will be in the interests
of all players to carry it out when the strategies call
for it. This latter property, that the threats of pun-
ishment are credible because they are incentive
compatible, is called subgame perfection. On the
concept of subgame perfect non-cooperative equi-
libria, see Selten ( ) or Friedman ( ).
The work of Hotelling and Chamberlin raises an
important issue that has received some recent
attention: firms not only choose prices (or output
levels), they decide on the design of their products.
In deciding on how to design a product, the firm
needs to know how design is related to cost of
production and how it is related to consumers’
tastes. The latter has been modelled by Lancaster (
) in terms of inherent charactistics. The
underlying notion is that consumers value certain
attributes of goods that are analogous to the nutri-
ents in foods. A particular product (e.g. a chair of a
given design) is a specific bundle of characteristics.
The product of a rival seller is a somewhat different
bundle of characteristics. A difficulty with this
approach in the most general form that Lancaster
discusses is that it is difficult to define these
characteristics. Less abstract versions are used in
oligopoly models where, following Hotelling,
physical location, is used as the only characteristic
chosen by firms. Any single measurable attribute,
such as sweetness of a bottled dr ink, can also be
used.
Dupuit, Arsene-Jules-Emile Juvenal (1804-1866)
Other topics treated in the duopoly theory lit- von Stackelberg, H. 1934. Marktform und Gleichgewicht.
erature include capital stock decisions, advertising, Vienna: Julius Springer.
Wicksell, K. 1925. Mathematical economics. In K. Wicksell,
and entry. They can be found in Friedman ( ), Selected papers on economic theoiy, ed. Erik Lindahl.
along with a fuller account of the topics Cambridge, MA: Harvard University Press, 1958.
sketched above.
See Also
► Jou
► j. sb Equilibrium
► Oligopoly
► a regie Behavior
Bibliography
Bertrand, J. 1883. Review of Cournot 1838. Journal des
Savants 499-508.
Bowley, A. 1924. The mathematical groundwork of eco-
nomics. New York: Kelley, 1965.
Bowley, A. 1928. Bilateral monopoly. Economic Journal
38:651-659.
Chamberlin, E. 1933. The theory of monopolistic competition,
7th ed. Cambridge: Harvard, 1956.
Cournot, A. 1838. Recherches sur les principes
mathematiques de la theorie des richesses. Trans. N.T.
Bacon. New York: Macmillan, 1927.
Edgeworth, F. 1897. The pure theory of monopoly. Papers
Relating to Political Economy I: 111-142.
Edgeworth, F. 1925. Papers relating to political economy.
New York: Burt Franklin, 1970.
Friedman, J. 1968. Reaction functions and the theory of
duopoly. Review of Economic Studies 35: 257- 272.
Friedman, J. 1983. Oligopoly theory. Cambridge: Cambridge
University Press.
Friedman, J. 1986. Game theory with applications to eco-
nomics. New York: Oxford University Press.
Frisch, R. 1933. Monopole - polypole - la notion de force
dans Teconomie. Festschrift til Harald Westergaard.
Supplement to Nationalekonomisk Tidsskrift.
Hotelling, H. 1929. Stability in competition. Economic
Journal 39: 41-57.
Lancaster, K. 1979. Variety, equity, and efficiency. New
York: Columbia University Press.
Nash, J. 1951. Noncooperative games. Annals of Mathematics
45: 286-295.
Owen, G. 1968. Game theory, 2nd ed. New York: Academic,
1982.
Selten, R. 1975. Reexamination of the perfectness concept for
equilibrium points in extensive games. International
Journal of Games Theory 4: 25-55.
Smith, A. 1776. In An inquiry into the nature and causes of
the wealth of nations, ed. R.H. Campbell, A.S. Skinner,
and W.M. Todd. Oxford: Clarendon Press, 1976.
3099
Dupuit, Arsene-Jules-Emile Juvenal (1804-
1866)
Robert B. Ekelund Jr.
Keywords
Antitrust enforcement; Consumer surplus; Cost-
benefit analysis; Deadweight loss; Diminishing
marginal utility; Dupuit, A.-J.- E. J.; Edgeworth,
F.Y.; Hotelling, H.; Law of demand; Marginal
cost pricing; Menger, C.; Monopoly profit
maximization; Price discrimination; Public
works
JEL Classifications
B31
French engineer and economic theorist, bom at
Fossano, Piedmont, Italy on 18 May 1804, when
this region was part of the French empire; died 5
September 1866 in Paris. After his parents returned
to Paris in 1814, Dupuit continued his education in
the secondary schools at Versailles, at Louis-le-
Grand and at Saint-Louis, where he finished
brilliantly by winning a physics prize in a large
group of competitors. Accepted to the Ecole des
Ponts et Chaussees in 1824, Dupuit soon
distinguished himself as an engineer and, in 1827,
was put in charge of an engineering district in the
department of Sarthe, where he concentrated on
roadway and navigation work. Dupuit’s numerous
and trenchant engineering studies on such topics as
friction and highway deterioration, floods and
hydraulics, and municipal water systems made him
one of the most creative civil engineers of his day.
Decorated for such contributions by the Legion of
Honour in 1843, Dupuit ultimately became director-
chief engineer in
3100
Paris in 1850 and Inspector-General of the Corps of
Civil Engineers in 1855.
No less profound were Dupuit’s contributions to
general economic analysis and to the economic
evaluation of public works (cost-benefit analysis).
In fact, Dupuit was the most illustrious contributor
in the long French tradition of study, teaching and
writing on economic topics at the Ecole des Ponts et
Chaussees, whose professors and students included
Isnard, Henri Navier, Charles Minard, Emile
Cheysson and Charles Ellet.
Led by a desire to evaluate the economic or net
benefits of public provision, Dupuit directed his
considerable analytical gifts to the utility founda-
tion of demand and to its relevance to the welfare
benefits of public works. In three substantial papers
appearing in the Annales des Ponts et Chaussees
(1844, 1849) and the Journal des economistes
(1853), Dupuit became the first non-adventitious
expositor of the theory of marginal utility, of (a
variant of) marginal cost pricing, of simple and
discriminating monopoly theory, and of pricing
principles of the firm where location is a factor in
expressing demand.
The font of Dupuit’s contribution is the con-
struction of a marginal utility curve and the iden-
tification of it with the demand curve or courbe de
consommation (see Fig. ).
Dupuit, Arsene-Jules-Emile Juvenal (1804-1866), Fig. 1
Dupuit, Arsene-Jules-Emile Juvenal (1804-1866)
Arguing in the manner of Carl Menger, who
later elaborated on the point, Dupuit showed that
the maiginal utility that an individual obtained from
a homogeneous stock of goods is determined by the
use to which the last units of the stock are put In
doing so, he clearly pointed out that the marginal
utility of a stock or some particular good diminishes
with increases in quantity and that each consumer
attaches a different marginal utility to the same
good according to the quantity consumed. The
importance of Dupuit’s invention rests in the fact
that the psychological concept of diminishing
marginal utility, and its ramifications, were carried
over to the law of demand. With some, but not all,
of the reservations and qualifications of Alfred
Marshall, Dupuit identified the marginal utility
curve with the demand curve, adding up the utility
curves of individuals to obtain the market demand
curve. Dupuit (1844, p. 106) described his con-
struction (see Fig. ), which applied to all goods,
public and private, as follows:
If... along a line Op the lengths Op, Op', Op" ...
represent various prices for an article, and that... pn,
p’n’, p"n" ... represent the number of articles
consumed corresponding to these prices, then it is
possible to construct a curve Nn'n"P which we shall
call the curve of consumption. ON represents the
quantity consumed when the price is zero, and OP
the price at which consumption falls to zero.
Dupuit, Arsene-Jules-Emile Juvenal (1804-1866)
The identification of marginal utility and
demand, of course, sets up the demand curve as a
welfare tool and Dupuit made specific calculations.
A measiue of the welfare produced by the good
(utilite absolve) at quantity Or is the definite
integral of the demand curve between O and r.
Given that Op is the (average) cost of producing
quantity Or, consumers earn a surplus (utilite rel-
ative) equal to absolute utility (OrnP) less costs of
production (Ornp). (Relative utility (pnP) is none
other than Marshall’s consumers’ surplus without
all the reservations that Marshall attached to the
concept.) Importantly, Dupuit identified area rNn as
lost utility (utilite perdue). Under competitive
conditions this loss was inevitable due to the
opportunity cost of resources. Under a monopoly
structure, for example, if, in Fig. , Op were a
monopoly price with zero production costs
assumed, utilite perdue would be a loss to society -
the ‘deadweight’ loss associated with excise taxes,
tariffs or monopoly. Further, Dupuit advanced the
theorem that the loss in utility was proportional to
the square of the tax of price above marginal cost.
This theorem, with attendant analysis, formed the
base for large areas of neoclassical welfare
economics, including the taxation studies of F.Y.
Edgeworth and the marginal cost pricing argument
of Harold Hotelling.
From this theoretical base, Dupuit investigated
an impressive number of pricing systems and
market models (1849). While Dupuit was an ardent
and stubborn defender of laissez faire in most
markets (1861), he was equally concerned that
public works, provided or regulated by government
as a last resort, should produce the maximum
amount of utility possible. Thus tools such as
marginal cost pricing find their theoretical
foundations in the writings of Dupuit. Although
Dupuit did not provide an explicit formulation of
the principle, one of his bridge pricing examples
and other statements strongly suggest the possi-
bilities of such a technique to maximize welfare, but
as a long-run proposition.
Dupuit analysed, independently of Cournot, who
was apparently unknown to him, the profit-
maximizing behaviour of the simple monopolist. He
saw monopoly at the apex of a range of problems
regarding the production of total welfare,
3101
being unconcerned about the ‘distribution’ of wel-
fare between producers and consumers. His point
was that the amount of ‘absolute utility’ (or what
could be called net benefit) was lessened by
monopoly profit maximization. This led him to
defend the private practice of price discrimination
and to produce an economic theory of discrimina-
tion. Price discrimination could exist, in Dupuit’s
view, with differences in ‘buyer estimates’, with the
ability to segment markets either naturally or
artificially, and with some degree of monopoly
power. The motive was profit maximization, and
although Dupuit discussed the effects of discrim-
ination on price and revenue, he was primarily
interested in the fact, as was Joan Robinson later,
that discrimination could affect the size of the
welfare benefit. This view was expanded to include
the impact of price discrimination of welfare when
buyers were spatially distributed (1849, 1854).
In the matter of policy, Dupuit recommended
that tools be carefully fit to specific problems. If
industries were to be collectivized or regulated by
government, Dupuit proposed the maximization of
net benefit under the constraint of covering total
costs of production. The recovery of total cost
might be achieved through regulated or constrained
price discrimination or through a cost-based single
price technique. However, Dupuit can hardly be
credited with espousing an enlarged role for
government or government intervention. A firm
adherent of Smith’s dictums concerning minimal
government, Dupuit believed that free and open
competition, along with vigorous antitrust or
anticartel enforcement, would ensure optimal
provisions in most cases, including transportation.
Indeed, in the process of analysing the welfare
principles of public works pricing, Dupuit
discovered (in an uncommonly complete manner)
some of the critical welfare- maximizing properties
of a generalized competitive system.
See Also
► Consumer plus
► Public Utility Pricing and Finance
3102
Selected Works
1844. On the measurement of the utility of public
works. Trans. R.H. Barback from the Annales
des Fonts et Chaussees, in International Eco-
nomic Papers, No. 2, London: Macmillan, 1952.
1849. On tolls and transport charges. Trans. E.
Henderson from the Annales des Pouts et
Chaussees, in International Economic Papers,
No. 11, London: Macmillan, 1962.
1853. On utility and its measure - on public utility.
Journal des economistes 36, 1-27.
1854. Peages. In Dictionnaire de VEconomie
Politique, vol. 2. Paris: Guillaumin.
1861. La Liberte Commerciale. Paris: Guillaumin.
1934. De l ’Utilite et sa Mesure: ecrits choisis et
republics, ed. M. de Bernardi.
Turin: La Riforma Sociale.
Bibliography
Ekelttnd, R.B. Jr. 1968. Jules Dupuit and the early theory of
marginal cost pricing. Journal of Political Economy 76:
462-471.
Ekelund, R.B. Jr. 1970. Price discrimination and product
differentiation in economic theory: An early analysis. Durable goods are goods whose useful lifetime
Quarterly Journal of Economics 84: 268-278.
Ekelund, R.B. Jr., and Yeung-Nan Shieh. 1986. Dupuit,
spatial economics, and optimal resource allocation: A
French tradition. Economica 53: 483-496.
Durable Goods Markets
and Aftermarkets
Michael Waldman
Abstract
There is an extensive literature on durable-
goods markets that starts with the work of
Akerlof, Coase, and Swan in the early 1970s. In
this entry I survey the literature by starting with
the three theoretical building blocks of time
inconsistency, adverse selection, and sub-
stitutability between new and used units. I then
Durable Goods Markets and Aftermarkets
focus on oiu understanding of three important
real-world issues. These are whether firms
choose optimal durability levels, whether firms
have incentives to eliminate secondhand
markets, and reasons for leasing. The article also
provides an extensive discussion of aftermarket
monopolization.
Keywords
Adverse selection; Aftermarket monopolization;
Aftermarkets; Akerlof, G.; Asymmetric
information; Bundling; Coase, R.; Commitment;
Complementary goods; Deadweight loss;
Durable goods markets; Durable-goods
monopoly problem; Hold-up theories; Infor-
mation costs; Leasing; Market power; Optimal
durability; Price discrimination; Product-line
pricing problem; Second-hand markets; Time
inconsistency; Tying
JEL Classifications
L13
spans multiple periods.
This article surveys the extensive literature on
durable-goods markets and aftermarkets. I begin
with the main theoretical ideas, then turn to specific
real-world issues such as durability choice and
leasing, discuss aftermarket monopolization and
then end with a brief conclusion. (A more in-depth
survey appears in Waldman .)
Three Theoretical Building Blocks
Much of our understanding of durable-goods mar-
kets derives from three theoretical contributions.
The first is Coase’s ( ) insight concerning time
inconsistency. To see the basic logic, consider
Bulow’s ( ) formalization: a durable-goods
monopolist sells its output in each of two periods
and cannot commit in the first period to second-
period actions. Bulow shows that, because in the
second period the firm does not internalize how its
actions affect the value of used units, its output is
higher than under commitment. First-period
Durable Goods Markets and Aftermarkets
purchasers anticipate this, pay less for new units
and thus lower overall monopoly profitability.
Coase’s insight has spawned a large literature.
One branch of this literature focuses on the Coase
conjecture, that is, the idea that in an infinite- period
setting time inconsistency causes price to drop
immediately to marginal cost. A second branch
identifies tactics such as leasing that firms can
employ to reduce or possibly avoid time
inconsistency. Finally, a third branch applies time
inconsistency to other issues, including new-
product introductions and repurchase prices.
The second major theoretical contribution is
Akerlof’s ( ) adverse-selection argument.
This paper helped start the asymmetric- information
revolution, but was not initially thought of as an
important contribution to durable-goods theory.
However, the paper’s main example concerns
second-hand markets. In Akerlof’s model buyers
have higher valuations than sellers, so efficiency
requires that all units be traded. Further, each seller
is privately informed of his own unit’s quality. The
result is a single price that reflects average quality,
and sellers with high-quality units keep them
because prices do not reflect actual quality, that is,
trade is below the efficient level. (In Akerlof’s
analysis there is no trade, but this result is not
robust.)
A small empirical literature looks for evidence
of adverse selection in durable- goods markets.
Most of these papers find some support. For
example, Bond ( ) considers the used pickup
truck market and finds support for adverse selection
for older trucks, while Genesove ( ) finds
some supporting evidence in used-car dealer auc-
tions. More recently, Gilligan ( ) finds
supporting evidence in business aircraft.
In terms of durable-goods theory, Akerlof’s
contribution was ignored for almost 30 years.
Starting with Hendel and Lizzeri ( ), how
ever, a number of papers have extended Akerlof’s
analysis. There are three basic findings. First,
Akerlof’s main results continue to hold when new
units are incorporated into the analysis. Second,
because adverse selection in the used-unit market
reduces the willingness to pay of new-unit buyers,
firms will market new units in a manner that
reduces adverse selection. Third, as
3103
discussed in detail later, new-unit leasing can be
important for reducing adverse selection.
The third major theoretical contribution is that
there is a close analogy between the product-line
pricing problem and the durable-goods monopoly
problem. This analogy is described in Waldman (
). Consider Mussa and Rosen ( ), which
analyses the product-line pricing problem of a non-
durable-goods monopolist. The monopolist sells
units of varying qualities to consumers who have
heterogeneous valuations on quality. Because the
substitutability between units links the various
prices, the monopolist lowers below efficient levels
the quality level sold to all but the highest-valuation
group.
Now consider a durable-goods monopolist who
controls the quality of a unit at every age. Further,
assume heterogeneity in consumers’ valuations for
quality and a frictionless second-hand market. Then,
if the firm can commit, quality choices are as above.
That is, new-unit quality is efficient. But, because
of the linkages between the various prices, all used-
unit qualities are below efficient levels. As
discussed later, a number of recent papers use this
result to analyse various real-world issues
concerning durable goods.
Three Real-World Issues
Optimal Durability Choice
A much debated issue is whether a durable-goods
monopolist chooses socially optimal durability.
Swan ( , ) considers models that satisfy
the once standard assumption that a unit is a bundle
of ‘service units’, so some number of used units is a
perfect substitute for a new unit. Swan’s steady-
state analysis shows durability choice to be socially
optimal because the firm produces the steady-state
flow of service units at minimum cost. (Swan’s
analysis corrected the conclusions of earlier papers
that had concluded that in such settings the
monopolist would choose inefficiently low
durability levels. )
A large literature investigates the robustness of
Swan’s conclusions. There are two major findings.
The first employs time inconsistency. Bulow ( )
moves away from Swan’s
3104
assumption of steady-state behaviour by consid-
ering a model similar to his earlier one, but now
allows endogenous durability choice. He shows that
time inconsistency provides a rationale for a
durable-goods monopolist to choose less than the
socially-optimal durability level. The logic is that
durability is what leads to time inconsistency, so
reducing durability below the efficient level reduces
time inconsistency and thus increases profitability.
The second major finding appears in Waldman (
) and Hendel and Lizzeri ( ), which
drop the service units assumption and instead
assume that new and used units vary in quality and
that durability choice controls the speed of quality
deterioration. The earlier discussion immediately
translates into an incentive for the firm to choose
less than the socially optimal durability level. That
is, in this setting the incentive for the monopolist to
sell output whose used-unit quality is below the
efficient level translates into durability below the
efficient level. (In Hendel and Lizzeri’s analysis
durability choice can be above, below, or equal to
the first-best level, but it is always below the
second-best level defined by actual outputs.)
Eliminating Second-Hand Markets
Do durable-goods producers with market power
have incentives to eliminate secondhand markets?
For example, do textbook publishers introduce new
editions in order to kill off the market for used
books? Until recently, the standard argument,
found, for example, in Swan ( ), was
that, since the new-unit price reflects prices the
product will sell for on the second-hand market in
subsequent periods, the producer has no such
incentive.
Two recent arguments show that this result is, in
fact, quite limited. The first, which builds on the
discussion above, appears in Waldman (
) and Hendel and Lizzeri ( ). The idea
is that, because substitutability between new and
used units means the price of a used unit on the
second-hand market limits the amount the firm can
charge for new units, the firm sometimes eliminates
the second-hand market or similarly reduces used-
unit availability in order to raise
Durable Goods Markets and Aftermarkets
the new-unit price. In particular, this is more likely
when consumers of used units have low valuations
for the firm’s product. This is both because little
revenue is lost by not serving such consumers and
because serving them means a low used-unit price
and thus a lower new-unit price. (A number of
earlier papers find similar results starting with
demand functions rather than utility maximization.)
The second argument, found initially in
Waldman ( ), employs time inconsistency.
As discussed, the early literature on time incon-
sistency focused on output choice. My 1993 paper
shows time inconsistency also applies to actions
such as new-product introductions that make used
units unavailable because they become obsolete.
The difference between this argument and the one
above concerns commitment. Above it is assumed
the firm can commit, so the firm eliminates the
second-hand market only when it is profitable to do
so. In contrast, here commitment is not assumed, so
the firm may eliminate the secondhand market even
though this lowers overall profitability.
A related empirical analysis appears in Iizuka (
), which shows that the market share of used
textbooks is an important determinant of whether or
not a publisher introduces a new edition. This is
consistent with new editions being used at least
partly to eliminate second-hand markets, although
Iizuka does not distinguish between the two pos-
sibilities described above for why a firm might want
to do this. In future research, it might be possible to
identify which argument is at work by focusing on
how the decision to introduce new editions affects
overall profitability.
Reasons for Leasing
A number of reasons have been identified for why
durable-goods producers frequently lease. (A reason
I do not discuss is that there are sometimes tax
advantages associated with leasing.) One reason,
initially discussed in Coase ( )
and Bulow ( ), is that time inconsistency
lowers profitability when a firm sells output
because it chooses actions in later periods that
inefficiently lower the value of used units. When
the firm leases, however, it retains ownership of
Durable Goods Markets and Aftermarkets
those units so the incentive to take inefficient
actions disappears.
A second reason is also related to a previous
discussion. As discussed in Waldman ( ) and
Hendel and Lizzeri ( ), when used-unit
prices serve as important constraints on the new-
unit price, leasing can be used to eliminate
secondhand markets or at least reduce used-unit
availability. The logic is that leasing allows a firm
to eliminate the second-hand market by allowing the
firm to retire returned used units. My 1997 paper
shows this formally and argues that it is consistent
with classic cases concerning the use of a lease-only
policy such as United Shoe in the shoe machinery
market, IBM in the computer market, and Xerox in
the copier market. (One might argue that leasing is
not needed because a firm can sell and then use high
repurchase prices to purchase and retire used units.
My 1997 paper shows this strategy is inferior to
leasing because of time inconsistency. )
Finally, leasing is a response to adverse selec-
tion. This argument appears in Hendel and Lizzeri (
) and Johnson and Waldman ( ). These
papers show that, whether the new-unit market is
monopolistic or competitive, in a world of asym-
metric information leasing in the new-unit market
can arise because it means used units are returned to
the sellers), which, in turn, avoids or at least reduces
adverse selection in the used-unit market. The two
papers develop different variants of the argument
and show it is consistent with various empirical
findings concerning the automobile market.
Aftermarket Monopolization
Aftermarket monopolization is behaviour that stops
alternative producers from selling aftermarket
products to the firm’s customers. The focus on this
subject started after the US Supreme Court’s 1992
decision in the case Eastman Kodak Company v.
Image Technical Services. Aftermarkets are
common with durable goods, where aftermarkets
refer to markets for complementary products such
as maintenance and upgrades. I consider three
possibilities: (a) hold-up rationales; (b)
3105
price discrimination and efficiency rationales; and
(c) other strategic rationales.
Hold-Up
There are two distinct hold-up arguments, each of
which focuses on aftermarket monopolization by
competitive producers. In both, the firm prohibits
other firms from selling the aftermarket product -
for example, maintenance - and then exploits the
locked-in positions of its customers in pricing the
product. The result is a standard deadweight loss
due to the high aftermarket price, although no
transfer between the consumers and the firm since
competition in the primary market means firms earn
zero profits overall. In the ‘costly-information’
version, consumers ignore the aftermarket price
when purchasing the primary product. In the ‘lack-
of- commitment’ version, developed in Borenstein
et al. ( ),
consumers conectly anticipate the aftermarket price
but, because firms cannot commit, time
inconsistency causes firms to monopolize the
aftermarket and inefficiently raise the aftermarket
price after consumers are locked in. (A third holdup
theory is the ‘surprise’ theory. In this argument
consumers are surprised by the aftermarket
monopolization. Some discussions of this theory
describe a transfer between the consumers and the
firm, but it is unclear why competition does not
result in zero profits, in which case the surprise and
costly-information theories are equivalent.)
Price Discrimination and Efficiency Rationales
For various reasons, such as that many buyers in the
relevant industries are sophisticated firms for which
the costly-information argument is implausible,
attention has shifted towards other arguments many
of which have either neutral or positive social-
welfare implications.
One such argument is the price discrimination
argument that appears in Chen and Ross ( )
and Klein ( ). Suppose the primary-good pro
ducer has market power. Then the firm may
monopolize the aftermarket in order to raise the
aftermarket price and in this way price discriminate
by charging a high aggregate price to the high-
volume/high-valuation consumers. From a social-
welfare standpoint, this argument has
3106
neutral implications since an improved ability to
price discriminate can either raise or lower social
welfare. (Klein argues that this argument applies
even when firms are competitive, although not
perfectly competitive.)
A plausible efficiency rationale follows from
Schmalensee’s ( ) argument that, given a
durable-goods monopolist (which means new units
priced above marginal cost) and a competitive
maintenance market (which means maintenance is
priced at cost), consumers will sometimes
inefficiently maintain rather than replace used units.
Tirole ( ) shows this can
lead to aftermarket monopolization in a durable-
goods monopoly setting because having a monopoly
in both markets allows the firm to avoid the
inefficiency and thus increases its profits.
More recently, Morita and Waldman ( ) and
Carlton and Waldman ( ) show that the argu
ment extends to aftermarkets other than mainte-
nance, and to competitive durable-goods markets
given switching costs. In the latter case the ineffi-
cient substitution problem arises even with compe-
tition because switching costs create market power
at the time of the maintenance/replacement deci-
sion. Interestingly, because competitive sellers earn
zero profits in equilibrium, when aftermarket
monopolization eliminates the distortion, both
social welfare and consumer welfare increase.
Strategic Rationales
There is an extensive literature on strategic ratio-
nales for the tying of complementary products.
Since the tying of primary and aftermarket products
is one potential way to achieve aftermarket
monopolization, much of this literature is relevant
to aftermarket monopolization.
Whinston ( ) shows that, if the primary
good is not essential, tying may force the exit of an
alternative producer of the complementary good and
in this way increase the firm’s profits by
monopolizing the segment of the complementary-
good market for which the primary good is not
required.
In contrast, in Carlton and Waldman ( ) See Also
tying is sometimes used to preserve a monopoly in
the primary-good market. They consider two-period
settings in which a single potential
Durable Goods Markets and Aftermarkets
entrant can enter the complementary market in
either period but the primary market only in the
second. In the presence of fixed costs of entry or
network externalities, the primary-good monopolist
sometimes ties in order to preserve its primary-
good monopoly in the second period. For example,
with entry costs tying stops the alternative producer
from entering the complementary market in the first
period. In turn, because of a possible inability to
cover entry costs, the outcome can be no entry in
either market in either period.
A third argument appears in Carlton and
Waldman ( ). Whinston shows that in
one-period settings there is never an incentive to tie
if the monopolist’s primary product is essential.
Carlton and I show that in durable-goods settings,
given the presence of complementary- good
upgrades and switching costs, tying can be optimal
even when the primary product is essential. The
basic logic is that some profits are realized in later
periods in the sale or lease of the upgraded
complementary good, and the only way the
monopolist can ensure it captures those profits is by
tying and becoming the sole producer of the
complementary good.
Conclusion
Starting in the early 1970s with the work of Aker-
lof, Coase and Swan, significant progress has been
made in our understanding of durable-goods mar-
kets. In this entry I have surveyed this literature as
well as the literature on the related issue of after-
markets. Although I have referred throughout to
various empirical papers, durable-goods markets is
a topic for which theory is far ahead of empirical
investigation. In the future I expect to see work that
extends the theory in various important ways, but
also empirical work that tests the validity of the
various theoretical approaches that have been
explored since the early 1970s.
Durand, David (Bom 1912) 3107

► Bundling and Tying Mussa, M., and S. Rosen. 1978. Monopoly and product
► Coase, Ronald Harry (Bom 1910) quality. Journal of Economic Theoty 18: 301-317.
Schmalensee, R. 1974. Market structure, durability, and
► : : > , I Lifpr/z maintenance effort. Review of Economic Studies 41: 277-
287.
Swan, R 1970. Durability of consumption goods. American
Bibliography Economic Review 60: 884-894.
Swan, P. 1971. The durability of goods and regulation of
monopoly. Bell Journal of Economics 2: 347-357.
Akerlof, G. 1970. The market for ‘lemons’: Quality uncer-
Swan, P. 1980. Alcoa: The influence of recycling on
tainty and the market mechanism. Quarterly Journal of
monopoly power. Journal of Political Economy 88: 76-
Economics 84: 488-500.
99.
Bond, E. 1982. A direct test of the ‘lemons’ model: The
Tirole, J. 1988. The theory of industrial organization.
market for used pick-up trucks. American Economic
Cambridge: MIT Press.
Review 72: 836-840.
Waldman, M. 1993. A new perspective on planned obso-
Borenstein, S., J. Mackie-Mason, and J. Netz. 1995. Antitrust
lescence. Quarterly Journal of Economics 108: 273-283.
policy in aftermarkets. Antitrust Law Journal 63: 455-
Waldman, M. 1996. Durable goods pricing when quality
482.
matters. Journal of Business 69: 489-510.
Bulow, J. 1982. Durable goods monopolists. Journal of
Waldman, M. 1997. Eliminating the market for secondhand
Political Economy 90: 314-332.
goods. Journal of Law and Economics 40: 61-92.
Bulow, J. 1986. An economic theory of planned obsolescence.
Waldman, M. 2003. Durable goods theory for real world
Quarterly Journal of Economics 101: 729-749.
markets. Journal of Economic Perspectives 17(1): 131-
Carlton, D., and M. Waldman. 2002. The strategic use of
154.
tying to preserve and create market power in evolving
Whinston, M. 1990. Tying, foreclosure, and exclusion.
industries. RAND Journal of Economics 33: 194-220.
American Economic Review 80: 837-859.
Carlton, D., and M. Waldman. 2005. Tying, upgrades, and
switching costs in durable goods markets. Mimeo:
Cornell University.
Carlton, D., and M. Waldman. 2006. Competition, monopoly,
and aftermarkets. Mimeo: Cornell University.
Chen, Z., and T. Ross. 1993. Refusals to deal, price dis-
crimination and independent service organizations. Durand, David (Born 1912)
Journal of Economics and Management Strategy 2: 593-
614. J. Fred Weston
Coase, R. 1972. Durability and monopoly. Journal of Law
and Economics 15: 143-149.
Genesove, D. 1993. Adverse selection in the wholesale used
car market. Journal of Political Economy 101: 644-665.
Gilligan, T. 2004. Lemons and leases in the used business
aircraft market. Journal of Political Economy 112: 1157-
Durand was bom in Ithaca, New York. He received
1180. his PhD at Columbia University in 1941. He was a
Hendel, I., and A. Lizzeri. 1999a. Adverse selection in member of the Research Staff of the National
durable goods markets. American Economic Review 89: Bureau of Economic Research from 1946 to 1955
1097-1115.
when he became a Professor at the Massachusetts
Hendel, I., and A. Lizzeri. 1999b. Interfering with secondary
markets. RAND Journal of Economics 30: 1-21. Institute of Technology, where he remained
Hendel, I., and A. Lizzeri. 2002. The role of leasing under throughout his career.
adverse selection. Journal of Political Economy 110: 113- The early contributions of David Durand were
143.
in statistical methodology. His election as a Fellow
Iizuka, T. 2004. An empirical analysis of planned obsoles-
cence. Mimeo: Vanderbilt University. of the American Statistical Association was based
Johnson, J., and M. Waldman. 2003. Leasing, lemons, and on his work in developing and applying statistical
buy-backs. R.4ND Journal of Economics 34: 247-265. analysis in the field of finance, including the
Klein, B. 1993. Market power in antitrust: Economic analysis
construction of historical series on the term
after Kodak. Supreme Court Economic Review 3: 43-92.
Morita, H., and M. Waldman. 2005. Competition, monopoly structure of interest rates.
maintenance, and consumer switching costs. Mimeo: In a National Bureau Conference publication
Cornell University. (1952 ), Durand authored a chapter which laid the
foundation for later developments on cost of capital
theory and measurement. Durand followed the
3108
premise that security appraisal is the key to mea-
suring the cost of capital. He developed two alter-
native methods of appraisal, the Net Income (NI)
method and the Net Operating Income (NOI)
method. In the NI method, the cost of debt interest
is deducted from net operating income and net
income is capitalized at a constant rate; the value of
the firm increases with higher debt leverage until
both the cost of debt and of equity rise substantially.
Under the NOI method a constant capitalization rate
is applied to the net operating income so that the
total value of all bonds and stocks is invariant so the
degree of leverage employed. Durand leaned toward
the NOI method, but recognized circumstances
under which each model had applicability'. Subse-
quent literature predominantly favoured the NOI
approach until the 1980s, when Durand's views
have essentially prevailed.
Durand also made contributions to the theory of
capital budgeting. He emphasized an eclectic
approach including the internal rate of return
method and the net present value procedure. He also
emphasized considering some measure of time to
indicate how fast an investment project will
liquidate itself. He argued that just as portfolio
managers select bonds partly on the basis of term to
maturity, financial managers responsible for capital
budgeting should select investment projects partly
on the basis of their weighted discounted payout
period or duration.
Selected Works
1941. Risk elements in consumer installment
financing. New York: National Bureau of Eco-
nomic Research.
1942. Basic yields of corporate bonds: 1900-1942.
Technical Paper No. 3, New York: National
Bureau of Economic Research.
1948. An appraisal of the errors involved in esti-
mating the size distribution of a given aggregate
income. Review of Economics and Statistics 30:
63-8.
1952. Costs of debt and equity' funds for business:
Trends and problems of measurement. In Con-
ference on research in business finance.
Durbin, Evan Frank Mottram (1906-1948)
215-47. New York: National Bureau of Eco-
nomic Research.
1954. Joint confidence regions for multiple
regression coefficients. Journal of the American
Statistical Association 49: 130-46.
1957. Growth stocks and the Petersburg paradox.
Journal of Finance 1 2 : 3 4 8 - 6 3 .
1971. Stable chaos: An introduction to statistical
control. Morristown: D.H. Mark Publication of
General Learning Press.
1974. Payout period, time spread and duration: Aids
to judgment in capital budgeting. Journal of
Bank Research 5: 20-34.
1981. Comprehensiveness in capital budgeting.
Financial Management 1 0 ( 5 ) : 7 - 1 3 .
Durbin, Evan Frank Mottram (1906-1948)
Elizabeth Durbin
Bom in Devon, Durbin was a Scholar of New
College, Oxford, won the Senior and Junior Webb
Medley Scholarships, first class Honours in Politics,
Philosophy and Economics and the Ricardo
Fellowship to University College, London. Hired as
an economics lecturer at the London School of
Economics in 1930, he was later promoted to senior
lecturer. During the war he was a personal assistant
to Clement Attlee, the Deputy Prime Minister, and
in 1945 he was elected Labour Member of
Parliament for Edmonton. He served as
Parliamentary Private Secretary to Hugh Dalton at
the Treasury, and was appointed junior Minister of
Works in March 1947. He was drowned in
Cornwall in 1948. Durbin is best remembered for
his book The Politics of Democratic Socialism
(1940), an influential statement of the revisionist
case in Britain, of which his close friend and
professional colleague, Hugh Gaitskell, later
commented: ‘it marked the transition from the
pioneering stage to that of responsibility' and power.
’
Durbin-Watson Statistic 3109

As a professional economist, Durbin published See Also

two books on macroeconomic theory and policy,
and a number of articles on economic planning. In
the intellectual turmoil of the early 1930s, he was ► :: v:.:..: v : '7
searching for a theory to explain the trade cycle,
because he believed that its control was essential to Selected Works
die socialist alternative to capitalism. He was
1933a. Purchasing power and trade
strongly influenced by Hayek’s cyclical theory of
depression.
sectoral imbalance, although he argued that the
London: Chapman & Hall.
crisis was precipitated by ‘an excessive supply of
1935a. The problem of credit policy. London:
money’ in die consumers’ sector, not capital scar-
Chapman & Hall.
city in die producers’ sector (Hayek’s view). In
1935b. The social significance of the theory of
further work, he introduced the role of die money
value. Economic Journal 45: 700-710.
market, an important advance on Hayek’s model,
1 9 4 0 . The politics of democratic socialism.
which foreshadowed Keynes’s use of uncertainty in
London: Routledge & Kegan Paul.
The General Theory. Later scholars have also rec-
1949. Problems of economic planning. London:
ognized his contributions to identifying die crucial
Routledge & Kegan Paul.
growth problem of maintaining sufficient savings
without causing sectoral imbalance. Together with
Hugh Gaitskell, J.E. Meade and Douglas Jay, References
Diubin has also been credited with adapting the
Durbin, E. 1985. New Jerusalems: The labour party and the
Keynesian revolution into practical policies for the economics of democratic socialism. London: Routledge
British Laboiu Party. However, he always remained & Kegan Paul.
sceptical about some aspects of The General The-
ory; he did not believe it provided a solution of die
cyclical problem, and he was concerned about the
inflationary potential of continued expansion.
Beginning in 1931 through the Fabian Society,
Durbin-Watson Statistic
Durbin and Gaitskell also organized systematic
research into the theory and practice of socialist James G. MacKinnon
planning and the appropriate criteria for assessing
efficiency in a socialist economy. Thus Durbin was
in the forefront of the planning controversies of the
Thirties, contributing articles to the development of
Keywords
the ‘competitive’ solution for market socialism and
Diubin-Watson statistic; Linear regression
to the marginal cost-pricing debate. He was one of
models; Monte Carlo test; Ordinary least
the first to argue that a mixed economy was
squares (OLS) estimator; Serial correlation;
fundamental to the notion of democratic socialism;
Testing; DW statistic
the market provided individuals freedom to choose
jobs and goods and incentives to innovate, and the
government provided the programme and policies to
JEL Classifications
sustain growth, to allocate resources in the public
Cl
interest and to ensiue social justice.
Diubin’s main achievements were to present a The well-known Durbin-Watson, or DW, statistic,
practical forerunner of the postwar mainstream case which was proposed by Diubin and Watson ( ,
for government intervention and to lay the ), is used for testing the null hypoth
intellectual foundations for the continuing debates esis that the error terms of a linear regression model
about the nature of the socialist vision in Britain. are serially independent.
1 -1 0 0 .. 0 0 0
-1 2 -1 0 .. 0 0 0
0 -1 2 -1 . .. 0 0 0

0 0 0 0 .. -1 2 —
0 0 0 0 .. 0 -1 1

Durbin and Watson ( ) actually considered a

number of statistics that can be written in the form
of ( ) for different choices of the matrix A and chose
to focus on d for reasons of computational and
theoretical convenience. Because both the
numerator and the denominator are proportional to
a2, d is invariant to a.
The exact distribution of d depends on X and the
distribution of the ut. When the error terms are i.i.d.
normal, Durbin and Watson ( )
tabulated bounds on the critical values for tests
based on d against the one-sided alternative that p >
0. These bounds, denoted dL and da, depend on the
sample size and the number of regressors. We can
reject the null hypothesis when d < dL, cannot reject
it when d < d,,, and can draw no firm conclusion
when dL< d < dv. To test against the alternative that
p < 0, we would replace d by 4 - d and use the same
procedure.
The original Durbin-Watson tables have been
extended by various authors, notably Savin and
White ( ). However, since d(J — dL can be
quite large, tests based on the bounds often have
indeterminate outcomes. It is much better to per-
form exact tests conditional on X, and this is easy to
do with modem computing technology. There are
two approaches.
The first approach is to calculate an exact P
value for d using one of several methods for
calculating the distribution of a ratio of quadratic
forms in normal random variables. The method of
Imhof ( ) is probably the best
known of these, but the more recent method of
Ansley et al. ( ) is faster. If a suitable com
puter program is readily available, this approach is
the best one.
An alternative approach is to perform a Monte
Carlo test. As can be seen from ( ), the statistic d
depends only on the vector u and the matrix X,

where A is the n x n matrix

Durbin-Watson Statistic
since u = Mxu, where MA- = I — X(XX)~lX.
Because of its invariance to <r, d does not depend
on any unknown parameters. This implies that a
Monte Carlo test will be exact.
To perform a Monte Carlo test at level a, we
first choose B such that a(B + 1) is an integer (999
is often a reasonable choice) and generate B vectors
u*, each of which is multivariate standard normal.
Each of the uj is regressed on X to calculate a vector
of residuals Mxu*, which is then used to compute a
simulated test statistic d* according to ( ). We can
then calculate simulated P values for a one-tailed
test against either p > 0 or p < 0 or for a two-tailed
test. For example, the simulated P value for a one-
tailed test against p > 0 is
p
*w=-jl2 I ( d j < d )’
j=i
where /(•) is the indicator function that is equal to 1 ►
when its argument is true and equal to 0 otherwise.
We reject the null hypothesis whenever P (d) < a. Bibliography
For more on the calculation of P values for
bootstrap and Monte Carlo tests, see Davidson and Ansley, C., R. Kohn, and T. Shively. 1992. Computing p-
MacKinnon ( ).
Limitations of the DW Statistic
The Durbin-Watson statistic is valid only when all
the regressors can be treated as fixed. It is not valid,
even asymptotically, when Xt includes a lagged
dependent variable or any variable that depends on
lagged values of yt. Because p is biased towards 0
when Xt includes a lagged dependent variable, d is
biased towards 2 in this case. Thus, a test based on
the DW statistic will tend to under-reject when the
null hypothesis is false.
Numerous procedures have been proposed for
testing for serial correlation in models that include
lagged dependent variables. The simplest is to reran
regression ( ), with the addition of the lagged
residuals from that regression. The test statistic is
then the t statistic on the lagged
3111
residuals. This procedure, which is due to Durbin (
) and Godfrey ( ), does not yield an
exact test and should be bootstrapped when the
sample size is small.
Of course, since the finite-sample distribution of
the DW statistic depends on the distribution of the
ut, we cannot expect to obtain an exact test even
when the Xt are exogenous if the normality
assumption is not a good one. In principle, we could
bootstrap d by using re-sampled residuals instead of
multivariate standard normal vectors for the uj. This
would probably work very well in most cases, but it
would not actually yield an exact test.
See Also
values for the generalized Durbin-Watson statistic and
other invariant test statistics. Journal of Econometrics 54:
277-300.
Davidson, R., and J. MacKinnon. 2006. Bootstrap methods in
econometrics. In Palgrave handbooks of econometrics:
volume 1: Econometric theory,
ed. T. Mills and K. Patterson. Basingstoke: Palgrave
Macmillan.
Durbin, J. 1970. Testing for serial correlation in least- squares
regression when some of the regressors are lagged
dependent variables. Econometrica 38: 410-421.
Durbin, J., and G. Watson. 1950. Testing for serial correlation
in least squares regression I. Biometrika 37: 409^128.
Durbin, J., and G. Watson. 1951. Testing for serial correlation
in least squares regression II. Biometrika 38: 159-177.
Godfrey, L. 1978. Testing against general autoregressive and
moving average error models when the regressors include
lagged dependent variables. Econometrica 46: 1293-130L
Irnhof, J. 1961. Computing the distribution of quadratic forms
in normal variables. Biometrika 48: 419-426.
Savin, N., and K. White. 1977. The Durbin-Watson test for
serial correlation with extreme sample sizes or many
regressors. Econometrica 45: 1989-1996.
3112
Durkheim, Emile (1858-1917)
Peter Bearman
Born in Epinal near Strasbourg, Diukheim attended
the Ecole Normale Superieure in Paris, taking his
agregation in 1882. His first important academic
appointment was as Professor of Sociology and
Education at Bordeaux in 1887. The Bordeaux
appointment marked the first sociology
professorship in France. In 1902 Diukheim was
appointed as Professor of Sociology (and Educa-
tion ), at the Sorbonne where he remained until his
death in 1017. OfDurkheim’s foiu major works, The
Division of Labour (1893), The Rules of
Sociological Method (1895), Suicide'. A Study in
Sociology (1897) and The Elementary Forms of
Religious Life (1912 ), the first three were written
wh i I e he was at Bordeaux. At the Sorbonne, Durk-
heim devoted considerable effort towards the
establishment of sociology as a professional dis-
cipline. He founded a journal, L'annee socio-
logique, and was active in supervision of younger
scholars, most notably Granet, Mauss, and
Halbwachs.
The theoretical agenda ofDurkheim’s major
works centred around understanding the collective
bases for social order in modern societies
characterized by increased individuation and
autonomy. For Durkheim, a stable social order was
possible only if the members of a group shared a
common set of beliefs (conscience collective)
governing individual behaviour. With the
progression of the division of labour in society -
which by definition leads to greater individuation
and specialization of persons and roles - the values
that increasingly heterogeneous individuals hold are
seen to become more abstract, and are thus less able
to shape and constrain individual social action. The
social order, in this context, may become weak.
Durkheim recognized that countervailing the weak-
ness of the collective conscience in modem society
was increased functional interdependence of
persons. Yet, in The Division of
Durkheim, Emile (1858-1917)
Labour, Durkheim attacked the notion
(attributed to Spencer as a representative utilitarian)
that such interdependence (the need for individuals
to exchange the products of their labour) was by
itself robust enough to guarantee social stability.
Rather, he asserted that exchange is possible only
because of the existence of shared sentiments which
govern the determination of ‘individual interest’ and
behaviour. Contracts presume prior sentiments
constraining self- interested social action.
Diukheim’s fundamental methodological con-
tribution to sociology is the recognition that macro-
level outcomes cannot be accounted for from the
analysis (or empirical observation) of micro-level
(individual) action. Rather, he argued that social
scientists must recognize that society is a ‘reality sui
generis’, a ‘thing’ greater than, and not reducible to,
its constituent parts. In this framework, sociology is
the positive science of social facts, phenomena
whose own structure can be used as an indicator of
the social solidarity of a group which one cannot
directly apprehend from observation alone.
Durkheim is considered a founder of modem
sociology and anthropology. On both substantive
and methodological grounds, his work can be
considered a sustained attack on economic theory
which typically elides the problem of social order
and assumes that aggregate social outcomes are the
products of individual social action and individual
self-interest.
Selected Works
1893. In The division of labor in society. Trans, and
ed. G. Simpson. New York: Free Press, 1D47.
1895. In The rides of sociological method (trans:
Solovay, S.A. and Mueller, J.H.), ed. G. Gatlin.
New York: Free Press, 1958.
1897. In Suicide: A study in sociology? (trans:
Spaulding, J.A. and Simpson, G.), ed. G.
Simpson. New York: Free Press, 1951.
1912. In The elementary forms of the religious life
(trans: Swain, J.W.). New York: Free
Press, 1954.
Dutch Disease
References
Lukes, S. 1973. Emile Durkheim: His life and work: A
historical and critical study. Harmondsworth: Penguin.
Parsons, T. 1968. Emile Durkheim. \n International ency-
clopedia of the social sciences, ed. D.L. Shils. New York:
Macmillan.
Dutch Disease
Thorvaldur Gylfason
Abstract
This article outlines the ‘Dutch disease’, the fear
of de-industrialization first seen in the
Netherlands in the wake of the appreciation of
the Dutch guilder following the discovery of
natural gas deposits in the North Sea around
1960. It considers its symptoms, and asks
whether it is indeed a ‘disease’ with negative
economic implications. It also briefly reviews Symptoms
some cases of Dutch disease, and the case of
Norway, which appears to have successfully
avoided it.
Keywords
De-industrialization; Dutch disease; Oil
JEL Classifications
F43
Dutch disease, in the original sense of the term,
refers to the fears of de-industrialization that
gripped the Netherlands in the wake of the appre-
ciation of the Dutch guilder following the discovery
of natural gas deposits in the North Sea around
1960. The appreciation of the guilder following the
gas export boom hurt the profitability of
manufacturing and service exports. Total exports
from the Netherlands decreased markedly relative
to Gross Domestic Product (GDP) during the 1960s.
The growth of petroleum exports in the 1960s hurt
other exports disproportionately.
3113
Many feared dire consequences for Dutch
manufacturing. The problem proved short-lived,
however. From the late 1960s onward, exports of
goods and services have increased from less than 40
per cent of GDP to more than 70 per cent, a high
export ratio by world standards. The expected de-
industrialization did not materialize, but the name
stuck. It can be said that, being neither Dutch nor a
disease, the Dutch disease is a double misnomer.
But when a disease bears the name of the first
patient diagnosed with it, it would seem a bit harsh
to require the patient to remain sick for the name to
stick.
Is it a disease? Some view it as matter of one
sector benefiting at the expense of others, without
seeing any macroeconomic or social damage done.
Others view the Dutch disease as an ailment,
pointing to the potentially harmful consequences of
the resulting reallocation of resources - from high-
tech, high-skill intensive service industries to low-
tech, low-skill intensive primary production, for
example - for economic growth and diversification.
An overvalued currency was the first symptom
associated with the Dutch disease, but later several
other symptoms came to light. Figure illustrates
how an oil export boom lifts the equilibrium real
exchange rate at which total exports of goods,
services, and capital match total imports. In the
figure, non-oil exports decline from A to C and
hence by less than oil exports increase, so that total
exports rise from A to B. For total exports to
decline the import schedule would have to shift to
the left (for instance through capital inflow) by an
amount that exceeds the increase in oil exports,
measured by the distance between B and C.
Natural resource discoveries and dependence
tend to go hand in hand with booms and busts: the
prices and supplies of raw materials and related
commodities fluctuate a great deal in world
markets. Fish stocks, for example, are notoriously
volatile. Oil wells are drilled, and then go dry, and
mines are depleted. The
3114
Dutch Disease,
Fig. 1 How an oil export
boom crowds out nonoil
exports
resulting fluctuations in export earnings trigger
exchange rate volatility, perhaps no less so under
fixed exchange rates than under floating rates.
Unstable currencies create uncertainty', which tends
to hurt exports and imports as well as foreign
investment. Further, the Dutch disease can strike
even in countries that do not have a national
currency of their own (as, for instance, in
Greenland, which uses the Danish krone and
depends on fish). In this case, the natural- resource-
based industry is able to pay higher wages and also
higher interest rates than other industries, thus
making it difficult for the latter to stay competitive.
This problem can become particularly acute in
countries with centralized wage bargaining (or with
oligopolistic banking systems, for that matter)
where the natural-resourceintensive industries set
the tone in nation-wide wage negotiations and
dictate wage settlements which other industries can
ill afford. In one or all of these ways, the Dutch
disease tends to reduce the level of total exports or
skew the composition of exports away from
manufacturing and service exports which could be
particularly conducive to economic growth over
time. Exports of capital, including inward foreign
direct investment, may also suffer.
The Netherlands recovered quickly from the
Dutch disease, and has seen a persistent upward
trend in its total exports relative to GDP since the
mid-1960s. On the other hand, in Norway, the
world’s third largest oil exporter after Saudi
Dutch Disease
Arabia and Russia, total exports have risen slowly
relative to GDP, to a level well below that of the
Netherlands (45 per cent in Norway in 2005 com-
pared with 71 per cent in the Netherlands), even if
the Dutch economy is almost three times as large as
that of Norway. Also, the share of manufactured
exports in merchandise exports was 68 per cent in
the Netherlands in 2005 compared with 17 per cent
in Norway. Exports and manufacturing are good for
growth. Openness to trade invigorates imports of
goods and services, capital, technology, ideas and
know-how. The Dutch disease matters mainly
because of its potentially harmful consequences for
economic growth.
Channels
Experience seems to suggest six main channels of
transmission from heavy naUiral resource depen-
dence to sluggish economic growth. At the top of
the list is the Dutch disease. In second place, huge
natural resource rents, especially in conjunction
with ill-defined property rights, imperfect or miss-
ing markets, and lax legal structures, may lead to
rent-seeking behaviour that diverts resources away
from more socially fruitful economic activity. The
struggle for resource rents may lead to a
concentration of economic and political power in
the hands of elites which, once in power, use the
rent to placate their political supporters and thus
Dutch Disease
secure their hold on power, with stunted or weak-
ened democracy and slow growth as a result.
Extensive rent seeking - in other words, seeking to
make money from market distortions - can breed
corruption, thus reducing both economic efficiency
and social equality.
Third, natural resource abundance may imbue
people with a false sense of security and lead
governments to lose sight of the need for growth-
friendly economic management, including free
trade, foreign investment, bureaucratic efficiency
and good institutions, including democracy.
Incentives to create wealth through good policies
and institutions may wane because of the relatively
effortless ability to extract wealth from the soil or
the sea. Fourth, abundant natural resources may
likewise weaken incentives to accumulate human
capital, even if the rent stream from the resources
may enable nations to give a high priority to
education. Fifth, natural resource abundance may
blunt private and public incentives to save and
invest in real capital no less than in human capital,
and thereby weaken financial institutions and
reduce economic growth. Sixth, natural resource
wealth is a fixed factor of production that hampers
economic growth potential by causing a growing
labour force and a growing stock of capital to run
into diminishing returns.
In sum, an abundance of natural capital, if not
well managed, may erode or reduce the quality of
human, physical, social, financial and foreign cap-
ital, and thus stand in the way of rapid economic
growth. Manna from heaven can be a mixed bless-
ing. Consider the attitudes of individuals to then-
own and to other people’s money. A person’s
respect for money tends to vary inversely with his
or her distance from the effort expended to make
the money. For example, loot tends to be invested
with less forethought than honest wages. The same
argument applies to unrequited foreign aid. An
influx of aid tends to increase the real exchange
rate, thereby hurting exports as in Fig. . Import
restrictions exacerbate the appreciation of the
currency, hurting exports further. The figure
suggests that aid needs to be accompanied by trade
liberalization to avoid currency appreciation and its
consequences.
3115
Cases
The list of natural-resource-abundant countries
beset by economic and political difficulties is a long
one. Take Libya. Without its oil export revenues,
Libya (population 6 million) would hardly have had
the means to purchase 700 military aircraft,
submarines and helicopters to pursue the foreign
ambitions of Colonel Gaddafi, in power since 1969.
In Equatorial Guinea, following oil discoveries, the
purchasing power of per capita GDP increased by a
factor of six or seven from 1990 to 2005, while life
expectancy plunged from 46 years to 42. One child
in five dies before reaching its fifth birthday. More
than a half of the population of 500.000 lives on
less than a dollar a day. President Mbasogo has
mled the country with an iron fist since 1979,
usurping the country’s oil wealth for himself and his
family and cronies. The readiness of the rest of the
world to import oil from Equatorial Guinea, and
thus to buy stolen goods, is an integral part of the
problem because a people’s right to its natural
resources is a human right proclaimed in primary
documents of international law and enshrined in
many national constitutions. Article 1 of the
International Covenant on Civil and Political Rights
states that ‘All people may, for their own ends,
freely dispose of their natural wealth and resources.’
Neither Libya nor Equatorial Guinea exports any
manufactures to speak of.
The list of countries afflicted by various symp-
toms of the Dutch disease could be extended to
include Iran, Iraq, Mexico, Nigeria, Russia, Saudi
Arabia, Sudan and Venezuela, among several
others. Some other countries have managed to avoid
such afflictions. A prime example is Norway,
where, before the first drop of oil emerged, the oil
and gas reserves within Norwegian jurisdiction
were defined by law as common property resources,
thereby clearly establishing the legal rights of the
Norwegian people to the resource rents. On this
legal basis, the government has absorbed about 80
per cent of the resource rent over the years, having
learnt the hard way in the 1970s to use a relatively
small portion of the total to meet current fiscal
needs. Most oil revenue is set aside in the state
petroleum fund, recently renamed
3116
the pension fund to reflect its intended use. The
government laid down economic as well as ethical
principles (commandments) to guide the use and
exploitation of the oil and gas for the benefit of
current and future generations of Norwegians. The
main political parties share an understanding that
the national economy needs to be shielded from an
excessive influx of oil money to avoid overheating
and waste. The Central Bank (Norges Bank), which
was granted increased independence from the
government in 2001, manages the fund (currently
around US$400 billion or $85,000 per Norwegian)
on behalf of the Ministry of Finance. This
arrangement maintains a distance between
politicians and the fund. Almost 40 years after
discovering their oil, the Norwegians have a smaller
central government than Denmark, Finland and
Sweden next door.
Norway’s tradition of democracy since long
before the advent of oil has probably helped
immunize the country from the ailments that afflict
most other oil- rich nations. Large-scale rent seek-
ing has been averted in Norway, investment per-
formance has been adequate, and the country’s
education record is excellent. Even so, some (weak)
signs of the Dutch disease can be detected, notably
sluggish exports and foreign direct investment and
the absence of a large, vibrant high-tech
manufacturing sector as in Sweden and Finland.
Norway’s lack of interest in joining the European
Unions can also be viewed in this light.
Then there is Botswana. Having managed its
diamonds quite well and used the rents to support
rapid growth, Botswana has become the richest
country in Africa, measured by the purchasing
power of per capita GDP. Its rapid growth since
1965 has been accompanied by political stability
and a steady advance of democracy. Unlike Sierra
Leone’s alluvial diamonds, which are easy to mine
by shovel and pan, and easy to loot, Botswana’s
kimberlite diamonds lie deep in the ground and can
only be mined with large hydraulic shovels and
other sophisticated equipment. They are therefore
not very lootable. This difference probably helped
Botswana succeed while Sierra Leone failed, and
so, most likely, did South African involvement in
Botswana’s diamond industry.
Dutch Disease and Foreign Aid
See Also
►)
Bibliography
Corden, W.M. 1984. Booming sector and Dutch disease
economics: Survey and consolidation. Oxford Economic
Papers 36: 359-380.
Corden, W.M., and J.P. Neary. 1982. Booming sector and de-
industrialisation in a small open economy. Economic
Journal 92: 825-848.
Ross, M. 2001. Does oil hinder democracy? World Politics
53: 325-361.
Van Wijnbergen, S. 1984. The ‘Dutch disease’: A disease
after all? Economic Journal 94: 41-45.
Wenar, L. 2008. Property rights and the resource curse.
Philosophy and Public Affairs 36: 1-32.
Dutch Disease and Foreign Aid
Christopher Adam
Abstract
Academic and policy debates on aid effec-
tiveness frequently emphasise the vulnerability
of recipients to the Dutch Disease, through
which aid inflows appreciate the real exchange
rate, thereby taxing the tradable export sector
with potentially deleterious effects on growth.
Fear of the Dutch Disease is remarkably
pervasive, even though there is little decisive
evidence that aid-induced Dutch Disease effects
are either large or widespread amongst poor
countries, at least against most plausible
counterfactuals. The lack of strong evidence
reflects a variety of factors, including problems
of measurement, but is primarily due to the fact
that aid flows are often purposive - designed to
address pre-existing distortions in the recipient
economy - and are accompanied by policy
measures specifically designed to mitigate latent
Dutch Disease effects. Although the
conventional macroeconomic transmission
Dutch Disease and Foreign Aid
channels may therefore be weak, the language of
the Dutch Disease continues to be used as a
metaphor for the wide range of political
economy concerns associated with aid surges.
Keywords
Aid absorption; Dutch disease; Export growth;
Foreign aid; Political economy; Real exchange
rate appreciation; Transfer paradox
JEL Classifications
F21; F23; F43
Introduction
An aid-induced ‘Dutch Disease’ occurs when for-
eign aid inflows result in a sharp appreciation of the Disease is also warranted, but in anticipation donors
recipient country’s real exchange rate, undercutting
the international competiveness of its export sector
to such an extent that it degrades an important
driver of economic growth. In principle, the disease
may be so strong as to completely overwhelm the
beneficial effects of the aid transfer. Concerns that may not be particularly strong, the language of the
the Dutch Disease phenomenon, traditionally
associated with natural resource windfalls
(Gylfason ), might also plague foreign aid flows
began to emerge in the 1980s and early 1990s when
donor support to low-income countries was shifting
away from project finance - where the bulk of aid
was used to purchase non-competitive project-
related imports - towards resource transfers that
provided general balance of payments and direct
support to government budgets (van Wijnbergen ;
Younger ). Since the late 1990s, when large-scale
debt relief initiatives to low-income countries were
well under way, the fear of the Dutch Disease
effects of aid has become a deeply entrenched
theme of the macroeconomic policy discourse
between international financial institutions,
dofgylnor agencies and aid-receiving countries (for
example, Heller ).
It would be natural to conclude that such fears are
grounded in robust empirical evidence. But
3117
this is not the case: although the existence of a
Dutch Disease channel is frequently advanced as an
important explanation for the weak measured
growth impacts of foreign aid flows to poor
countries (most notably in Raj an and Sub-
ramanian ), the empirical evidence tends to suggest
that measured aid-induced Dutch Disease effects are
neither widespread amongst poor aid-dependent
economies nor particularly large (Tarp ).
That the Dutch Disease remains so prominent in
policy debates is therefore intriguing. At least three
possible explanations suggest themselves. The first
concerns measurement and argues that while the
effect exists and may be serious in a range of
settings, the fact that aid surges are rare and
relatively modest in scale (certainly compared to
many natural resource booms) means that it is
difficult to identify the effects in the aggregate data.
The second explanation is similar and centres on the
counterfactual: in this case the fear of the Dutch
and recipients put in place mitigation measures that
allow aid to be absorbed without triggering the
disease. Had they not, the damage would have been
large and clearly identifiable in the data. A third
explanation is that while the conventional channel
Dutch Disease has been co-opted to describe a wide
range of political economy and other pathologies
associated with large and rapidly increasing inflows
of external assistance to small economies, the sorts
of problems emphasised by Peter Bauer and other
critics of aid (Dorn ).
The remainder of this entry consists of four
sections. Section “
” lays out in
more detail the basic mechanics of the Dutch
Disease effects of aid and section “ briefly discusses
some of the main themes in the empirical evidence.
Section “
” considers how the particularities of aid flows
modify the standard Dutch Disease story and
Section “ concludes.
3118
Aid Flows and the Dutch Disease: The
Basic Mechanism
The essential nature of the aid-induced Dutch
Disease derives from the observation that while the
growth and diversification of exports play a central
role in the transformation of poor countries, the
mechanics of aid absorption may generate a
negative feedback from aid to export
competitiveness and growth, thereby acting as a
brake on development (Bevan ). The mechanism
works as follows. Foreign aid - which is composed
entirely of traded goods or claims on traded goods -
augments domestic resources, leaving the economy
as a whole better off, at least initially; how much
better off depends on how these resources are
absorbed and with what consequence.
To focus on the basic macroeconomics of the aid
inflow, we abstract from the details of to whom the
aid accrues (whether it is the public or private
sector) and the form of expenditure (consumption or
investment, for example), returning to these
distinctions later. The ex post current account of the
balance of payments must be financed by some
combination of aid inflows and the drawdown of
private and/or official net foreign assets:
04, = A, — KA, — AZ„ (1)
in which At denotes the net aid inflow, KA, the flow
position on the capital account, and AZ t official
reserve accumulation. CAt denotes the current
account deficit before aid and can be expressed as
the excess of domestic expenditure over national
income (inclusive of net factor income), 04, = Et —
Yt = (M, — Xt), where Yt and Et denote Gross
National Income and aggregate expenditure (both
public and private)
' Traded goods, consisting of importables and exportables are
those goods produced and consumed in world markets.
Domestic demand and supply conditions therefore have no
impact on the (world) price of tradables. Non-tradable or
domestic goods, on the other hand are only produced locally;
their price is determined by domestic market conditions. By
definition, donor aid can only consist of traded goods (e.g.
food aid) or a claim over traded goods (i.e. a dollar flow of
aid).
Dutch Disease and Foreign Aid
respectively, Mt. denotes imports and X, exports.
Substituting into ( ) we can re-write the external
balance as
(M, - X,) = A, — KA, — AZ,.
(2)
Equation offers a precise definition of aid
absorption as the extent to which the current
account deficit (net of factor payments) increases in
response to an increase in aid; absorption is thus a
measure of the transfer of resources from (foreign)
donor to (domestic) recipient used to augment
domestic expenditure. Three examples help to
isolate the Dutch Disease. In the first case, the aid
inflow is not absorbed at all but instead is saved in
the form of either public or private foreign asset
accumulation. Thus, using d to denote a change,
d(M, — X,) = 0 and dAt = dKAt + dAZt\ some
fraction of the aid increase is saved on the official
side through net international reserve accumulation
and the remainder on the private side by capital
outflows, either legitimate or not. In this case there
is no pressure on the real exchange rate and no
Dutch Disease, but only because there is no
absorption. This situation is unlikely to occur,
however, at least beyond the very short run. Donors
are generally averse to this sort of outcome: their
motive for providing development assistance is
precisely for it to be absorbed and spent in the
recipient economy, not to be piled up in offshore
bank vaults, even if this might in fact be a prudent
macroeconomic response. The second polar case is
where aid is fully absorbed - so that net imports,
(M, — Xt), rise dollar-for-dollar with the increase in
aid - but where the associated increased expenditure
is entirely in terms of non-competitive final con-
sumption imports such as military hardware,
Ferraris or foreign travel. In this case dM, dAt and
dX, 0: the aid is fully absorbed, but the
2
0n the link between aid and (legal and illegal) capital flight
see Ndikumana and Boyce ( ).
3
If aid flows are temporary a high rate of (official) saving
may be consistent with efficient expenditure smoothing. But
see Buffie et al. ( ) on how ‘use it or lose if
constraints on aid flows affect recipients’ monetary and fiscal
policy choices.
Dutch Disease and Foreign Aid
entire demand impulse leaks offshore through ‘self-
sterilising’ expenditures and hence there is no
transmission of demand pressures to domestic
production.
The more interesting and certainly more com-
mon case is the intermediate one in which the
higher expenditure facilitated by the aid inflow
leads to an increase in the demand for both imports
and domestic goods and services: in this case clA, =
dMt — dXt. Hence the aid is fully absorbed so that
net imports still rise dollar-for- dollar with the aid,
but absorption is represented by some increase in
imports and a fall in exports. This is simply the
‘spending effect’ from the conventional Dutch
Disease theory. Viewed from this perspective, the
absorption of aid inflows becomes a classical small-
country ‘transfer problem’. As a price-taker, the
increased demand for importable goods can be met
via imports at the prevailing world prices. The
increased demand for domestic non-tradable goods,
on the other hand, can only be satisfied if the
domestic supply of non-tradables increases which
requires resources, in particular labour, to be bid
away from the production of tradables (i.e. exports
and import- substitutes). The price of non-tradables
relative to tradable goods must therefore rise in
order to shift demand in favour of tradable goods
and supply in favour of non-tradables. This relative
price movement - i.e. the real exchange rate
appreciation - is what restores both internal balance
(i.e. equilibrates the demand and supply for non-
tradable goods) and external balance (i.e. the
adjustment of net exports, conditional on any
changes in public and private net foreign asset
positions) following the aid inflow. How much the
real exchange rate needs to appreciate and how
large the resource movement will be is case-
specific, determined by consumer preferences and
firms’ production behaviour. The more elastic are
demand and supply in response to movements in the
real exchange rate the easier it is to shift resources
between sectors and the milder the required real
exchange rate appreciation (see Adam and Bevan (
) for a formal
derivation).
The distributional consequences of this adjust-
ment are as follows. Producers of tradable goods
3119
stand to lose as the purchasing power of export
revenues declines relative to the cost of consump-
tion, while profit margins are squeezed by the rising
cost of labour and non-tradable inputs. At the
margin, firms in this sector go out of business.
Producers of non-tradables, on the other hand, gain:
and if the production of non-tradables is labour-
intensive - as is likely to be the case in many low-
income countries - wage earners in aggregate will
gain (as suggested by the Stolper-Samuelson
Theorem).
Up to this point, the real exchange rate appre-
ciation is an efficient macroeconomic response to
the aid inflow, providing the signal for a welfare-
maximising reallocation of resources. For this
adjustment to be harmful to the recipient in
aggregate - in other words for there to be a
‘disease’ - requires something else to be going on.
There are at least two standard ways of motivating
this. The dominant conventional explanation for the
disease is that the tradable sector (typically the
manufacturing sector or commercial agriculture) is
the source of some positive externality - for
example if the export sector is an incubator of
economy-wide productivity growth. If so, the aid-
induced contraction of the sector, relative to the
counterfactual in which aid does not increase, is a
socially inefficient tax on growth. It is commonly
assumed that the relevant externality resides in a
leaming-by-doing mechanism that generates
dynamic economies of scale in the production of
non-traditional exportable goods, whereby
productivity growth is increasing in the cumulative
output (or exports) of the sector (see Beaudry ( ),
derides et al. ( ), and
Martins and Yang ( ), amongst others).
A second perspective emphasises the short-run
volatility of the real exchange rate as opposed to its
long-run level in generating Dutch Disease effects
(for example, Bullr and Harmann ( )
and Eifert and Gelb ( )). Here the driving
force is hysteresis, whereby the short-run temporary
appreciation of the real exchange rate may have
long-run adverse effects on export growth. If credit
market imperfections mean affected firms in the
export sector are unable to borrow against future
expected profits when the real exchange rate
appreciation passes, they will be unable to
3120
sustain the losses caused by cheaper imports and
rising domestic wage costs. At the margin, firms
will exit this sector, and if they face fixed costs of
re-entry (in terms of specific marketing relation-
ships or long-term supply-chain relationships, for
example) or fall behind global technology frontiers,
they will re-engage the export market at a lower
level than before, even when the real exchange rate
returns to its previous level.
For completeness, it is worth briefly mentioning
a third alternative representation of the aid-induced
Dutch Disease phenomenon, one that does not
appeal to conventional growth externalities.
Drawing on the ‘transfer paradox’ literature
originating in the post-First World War debates
between Keynes and Ohlin (Brock ), Yano and
Nugent ( ) argue that for
the small open economy, aid flows can reduce
aggregate welfare if the tradable goods sector is
tariff-distorted and aid transfer takes the form of an
increase in the installed capital stock. They provide
some (weak) evidence in support of this claim, but
as Tokarick ( ) demonstrates, aid
flows will only be welfare-reducing in this frame-
work if the non-tradable goods are strongly com-
plementary to the (tariff-protected) imported good.
When, as is the conventional case, non-traded and
traded goods are substitutes in consumption, aid
cannot reduce welfare.
Two important qualifications are relevant at this
point. First, the conventional Dutch Disease result
does not depend on aid flowing to government but
rather on its impact on the aggregate demand for
non-tradables (and the presence of growth
externalities). Aid typically does accrue to
government in the first instance, but it can be spent
in a variety of ways: directly by government on
current or capital goods (an aid-funded deficit);
transferred to the private sector through income
transfers; used to retire debt; or to finance tax cuts.
As we discuss below, the form of public expendi-
ture will play a crucial role in determining whether
latent Dutch Disease effects are exacerbated or
mitigated.
Second, the foregoing analysis is independent of
whether the country adopts a fixed nominal
exchange rate regime or a float or, indeed, any
hybrid arrangement. What matters, at least over
Dutch Disease and Foreign Aid
the medium term, is how relative prices change and
not how this change is brought about. In a flexible
exchange rate regime, the real exchange rate
appreciation is typically achieved by an
appreciation of the nominal exchange rate; in a
fixed exchange rate regime, adjustment occurs
through rising domestic non-tradable prices relative
to tradable good prices. Over the short run,
however, particularly when domestic prices are
sticky, the dynamics of the real exchange rate do
depend on the nominal exchange rate regime
(Ghosh et al. ; Fielding and Gibson ).
Evidence
Although the empirical question is relatively easy to
define, the evidence on whether aid-induced Dutch
Disease effects exist is mixed and often highly
contested. If the Dutch Disease channel is
important, we expect rising aid inflows to be asso-
ciated with an appreciating real exchange rate in
recipient countries and for the non-tradable sector to
expand at the expense of the exportable sector.
Critically, the contraction of exports should be
associated with lower overall growth. This pattern
should be present in both time-series and the cross-
section or panel evidence.
The identification of this channel is not
straightforward, for a number of reasons. First,
although there are exceptions - the small islands of
the Pacific, some post-conflict countries and a small
number of Sub-Saharan African countries - aid
flows rarely exceed even 10% of GDP, and large
aid surges (the events that allow the econometrician
to statistically identify Dutch Disease effects) are
comparatively uncommon. Second, unlike natural
resource windfalls, aid is rarely exogenous with
respect to the recipient’s economic performance.
Rather, aid is purposive, allocated to countries in
poor economic straits, afflicted by structural and
policy conditions that themselves generate low
growth, over-valued real exchange rates and small
tradable sectors. In other words, aid tends to flow
into countries displaying the symptoms of the Dutch
Disease: controlling for the endogeneity of aid in
these circumstances is a formidable challenge.
Third, policy actions by
Dutch Disease and Foreign Aid 3121

aid recipients, in the short term through fiscal and The empirical evidence suggests that the Dutch
monetary policy and in the longer term by public Disease operates principally through the mis-
investment and other policy reforms, will, if successful, alignment, specifically the overvaluation of the
mitigate the incipient Dutch Disease pressures from aid. short-run real exchange rate, rather than through the
And finally, as with all research that seeks to assess aid appreciation in the equilibrium real exchange rate
effectiveness on the basis of country-level data, the itself (Arellano et al. ; Elbadawi et al. ; Rajan and
profound changes in recent decades in the political Subramanian ; Kang et al. ). This reinforces the
economy and institutional environment shaping aid argument that whether aid flows are likely to trigger
relationship further complicate matters: not only have Dutch Disease effects depends on how recipient
underlying structural conditions changed greatly over governments set the relevant macroeconomic,
the decades, but the geopolitics of aid have funda- public expenditure and structural policy
mentally changed the aid allocation behaviour of donors, instruments, a point that is reinforced by country
particularly since the end of the Cold War. case studies examining specific aid singes. In their
These points notwithstanding, the research lit- study of aid surges in the wake of the low-income
erature does tend to find some evidence supporting country debt-relief initiatives in the early 2000s,
the Dutch Disease channel. In particular, there is Berg et al. (
support for the first two links in the process: aid inflows ) argue that the noticeable absence of Dutch
and aid singes are indeed associated with a tendency Disease effects in the wake of these aid surges often
for the real exchange rate to appreciate, although reflected conscious decisions not to ‘absorb’ the aid
rarely is this effect particularly strong (Werker et al. ; because of an underlying fear of triggering Dutch
Magud and Sosa ; Fielding and Gibson ). Sim- Disease effects, even to the point that the lack of
ilarly, a number of papers identify a link absorption risked undermining the developmental
between aid inflows and the relative size of the objective of the resource transfer.
exportable sector, the most notable contribution
in this field being from Rajan and Subramanian
( ) who Aid and the Dutch Disease: Why the Dog
exploit the within-country cross-industry variation Doesn't (Often) Bark
in growth rates to identify the effect of aid on
manufacturing growth. They find quite sizeable The ambiguity of the empirical evidence hints at
effects: using alternative measures to identify why, in practice, the characterisation of the Dutch
exportable sectors, they suggest that an additional Disease outlined in the simple model described
one percentage point increase in the share of aid to above is rather incomplete. First, the model assumes
GDP results in exportable industries growing a pre-aid equilibrium in which all factors are fully
between 0.5% and 1.4% per annum more slowly employed so that the increased demand for non-
than non-exportable industries. By contrast, Werker tradables necessarily entails a contraction of
et al. ( ), using oil price shocks expe tradable production. In the aid context, however,
rienced by OPEC donors to identify the exogenous recipient countries are typically characterised by
variation in their aid disbursement, find that while unemployed resources, particularly labour. If these
net imports respond strongly to aid inflows - aid is can be brought into use in the production of non-
indeed absorbed - this occurs principally through tradables as demand increases, aid can then be
increased imports, with only a very limited absorbed without driving up real wages and drawing
reduction in exports. resources away from the tradable/exportable sector.
A failure to recognise idle capacity may thus lead to
a systematic ‘over-expectation’ that aid will induce
a Dutch Disease problem (Nkusu ). The key
question, then, is how much effective excess
4
capacity actually exists: measured unemployment
These are, of course, relative growth rates; without con-
trolling for the relative size of sectors it is not possible to
may not be a sufficient statistic if
infer the impact on aggregate output growth.
3122
non-tradable production is intensive in specific fac-
tors such as skilled labour that are not easily
substituted for the abundant unemployed factors.
This highlights the second key feature of aid
inflows, noted above: that aid is intentionally
targeted at countries that are not just poor but often
heavily constrained by distortions and bottlenecks,
so that the economy operates at less than full
capacity given its factor endowments. Moreover,
aid tends to be conditioned on specific policy
actions, directed towards increasing the supply of
key inputs to production, either by the provision of
imports or through technical assistance, for
example, and is bundled with specific conditionality
aimed at removing pohcy distortions or institutional
bottlenecks. Although conditionality is not always
effective, the blending of resources with pohcy
reform and expertise can allow reform-oriented
countries to absorb aid flows and simultaneously
avoid the trade-off implied by the Dutch Disease.
This was certainly the case across a number of
countries in the late 1990s and early 2000s, when
aid-supported macroeconomic stabilisation and
supply-side reforms combined to generate both
rising domestic consumption and rising net exports
as aid helped previously heavily distorted
economies to decompress and grow rapidly: the
cases of Ghana and Uganda in these years stand as
good examples of this process.
Clearly, there are limits to decompression. How
long countries can dodge the bullet of the Dutch
Disease depends both on the depth of reform but
equally on how aid supports the expansion of the
supply side of the economy. In the short ran,
government expenditure patterns that are biased
towards imports can mitigate pressures (albeit at the
cost of potentially distorting otherwise efficient
public spending programmes) while monetary
policy can be geared to matching the path of public
spending to the rate of aid absorption so as to
minimise excess real exchange rate appreciation
(Adam et al. ).
But the obvious mechanism to mitigate or
reverse Dutch Disease effects of aid over the
medium term is to expand aggregate supply through
higher public and/or private investment. One
channel for this is if aid is used to reduce
Dutch Disease and Foreign Aid
domestic taxation and borrowing so that private
investment is crowded-in through higher net of tax
returns and lower domestic interest rates. The
alternative channel is via public investment. Adam
and Bevan ( ), amongst others,
explore the interplay between two dynamic
externalities in the aid-dependent economy. On the
one hand is the demand-side Dutch Disease
channel, in which a short-run exchange rate
appreciation is associated with the contraction of
the exportable sector through a leaming-by- doing
externality, and on the other a growth externality
coming from public infrastructure investment which
delivers increasing returns to private factors of
production. The precise outcome clearly depends on
the relative strength and timing of these offsetting
effects - for example Dutch Disease effects may
dominate if leaming-by-doing effects are
particularly strong, while the returns to public
investment are small and/or take a long time to
realise - but it also depends on the sector-intensity
of returns to public investment. If the productivity-
enhancing effects of public investment are skewed
in favour of production of tradables, the real
exchange rate may still appreciate, as the spending
effect is reinforced by a resource movement into the
now more productive exportable sector, but the
higher productivity of the sector offsets the effects
of the appreciation. Thus the aid inflow is still
associated with an appreciation of the real exchange
rate, but in this instance with an expansion rather
than a contraction of the exportable sector.
Alternatively, if productivity gains are skewed in
favour of the non-tradable sector, expansion of the
exportable sector is driven by the falling relative
price of non-tradables, so that we observe aid
inflows associated, over the medium term, with a
depreciating real exchange rate and an expanding
exportable sector. Adam and Bevan ( ) show that
these effects are
magnified when the non-tradable good is the
principal wage good in the economy. A specific
example of such a process might be public
5
Very similar models appear in Torvik ( ), Agenor
et al. ( ), and Berg et al. ( ).
Dutch Disease and Foreign Aid
investment in the road network, which lowers the than the pure macroeconomics of aid, that keep the
cost of transporting (non-tradable) food to urban Dutch Disease centre stage.
areas, thus helping to drive down the cost of
manufactured goods (Gollin and Rogerson
)• See Also
Conclusions
On the basis of the empirical evidence, the con-
ventionally defined Dutch Disease effects of aid
clearly does exist. It is not merely a ‘theoretical
quirk’, but at the same time even the most robust
evidence cannot point to particularly large effects.
Indeed, well-designed aid programmes, combined
with appropriate macroeconomic and supply-side
policy responses, may be associated with exactly
the opposite effect, with aid supporting increased
investment and export-led growth - what Berg et al.
( ) refer to as
‘Dutch Vigour’.
But this leaves open the question as to why
concerns about the Dutch Disease effects of aid
remain so firmly on the table. The most persuasive
explanation is that the language of the Dutch
Disease - the idea that an unrequited transfer may
be welfare-reducing - continues to be commonly
used as a metaphor for the wide range of political-
economy concerns associated with aid surges.
These may include dysfunctional rent- seeking
behaviour, in which productive resources and
entrepreneurial talent are devoted to the capture and
distribution of rents from aid contracts (see Klein
and Harford ( ), amongst others).
They may also involve skilled personnel within the
public sector being increasingly deployed on the
management of often highly bureaucratic aid
programmes (Brautigam and Knack ). And on the
political side, the ‘Dutch Disease’ label is
frequently used to describe the adverse effects that
visible and vocal aid donors may have on domestic
systems of political accountability - where pol-
iticians and bureaucrats may feel more accountable
to the donors who finance the lion’s share of the
budget than to the domestic electorate - what Peter
Bauer ( ) referred to
as the ‘politicization of daily life’ in aid-receiving
countries. It is predominantly these factors, rather
3123
Acknowledgments I thank Radhika Goyal for excellent
research assistance and the Oppenheimer Fund of the
Department of International Development, University of
Oxford for financial support.
Bibliography
Adam, C., and D. Bevan. 2006. Aid and the supply side:
Public investment, export performance and the Dutch
Disease in low-income countries. World Bank Economic
Review 20(2): 261-290.
Adam, C, E. Buffie, S. O’Connell, and C. Pattillo. 2009.
Monetary policy rules for managing aid surges in Africa.
Review of Development Economics 13(3): 464-490.
Agenor, P, N. Bayraktar, and K. El Aynaoui. 2008. Roads out
of poverty? Assessing the links between aid, public
investment, growth and poverty reduction. Journal of
Development Economics 86(2): 277-295.
Arellano, C., A. Bulir, T. Lane, and L. Lipschitz. 2009. The
dynamic implications of foreign aid and its variability.
Journal of Development Economics 88(1): 87-102.
Bauer, P.T. 1972. Dissent on development. Boston: Harvard
University Press.
Beaudry, P. 2008. Growth and leaming-by-doing. The new
palgrave dictionary of economics. 2nd edn. Basingstoke:
Palgrave.
Berg, A., Aiyar, S., Hussain, M., Roache, S., Mirzoev, T. and
Mahone, A. 2007. The macroeconomics of scaling up aid:
Lessons from recent experience. IMF Occasional Paper
No. 253.
Berg, A., Gottschalk, J., Portillo, R. and Zanna, L.-F. 2010.
The macroeconomics of medium-term aid scaling up
scenarios. IMF Working Paper 10/160.
Bevan, D. 2006. An analytical overview of aid absorption. In
The macroeconomic management of foreign aid:
Opportunities and pitfalls, ed. P. Isard, L. Lipschitz, A.
Mourmouras, and B. Yontcheva. Washington, DC:
International Monetary Fund.
Brautigam, D., and S. Knack. 2004. Foreign aid, institutions,
and governance in Sub-Saharan Africa. Economic
Development and Cultural Change 52(2): 255-286.
Brock, P. 2008. Transfer problem. The new palgrave dic-
tionary of economics. 2nd edn. Basingstoke: Palgrave.
Buffie, E., S. O’Connell, and C. Adam. 2010. Fiscal inertia,
donor credibility and the monetary management of aid
3124
surges. Journal of Development Economics 93(2): 287-
298.
Bullr, A., and J. I lamann. 2008. Volatility of development
aid: From the frying pan into the fire. World Development
36(10): 2048 -2066.
Clerides, S., S. Lach, and J. Tybout. 1998. Is learning by
exporting important? Micro-dynamic evidence from
Colombia, Mexico and Morocco. Quarterly Journal of
Economics 113(3): 903-947.
Dom, J. A. 2008. Peter Bauer. The newpalgrave dictionary of
economics, 2nd edn. Basingstoke: Palgrave.
Eifert, B. and Gelb, A. 2005. Improving the dynamics of aid:
Toward more predictable budget support. World Bank
Policy Research Working Paper 3732.
Elbadawi, I.,L. Kaltani, andR. Soto. 2012. Aid, real exchange
rate misalignment and economic performance in Sub
Saharan Africa. World Development 40(4): 681-700.
Fielding, D., andF. Gibson. 2013. Aid and Dutch disease in
Sub-Saharan Africa. Journal of African Economies 22(1):
1-21.
Ghosh, A., A.-M. Guide, and H. Wolf. 2003. Exchange rate
regimes: Choices and consequences. Cambridge. MA:
MIT Press.
Gollin, D. and Rogerson, R. 2010. Agriculture, roads and
economic development in Uganda. NBER Working Paper
15863.
Gylfason, T. 2008. Dutch disease. Die new palgrave dic-
tionary of economics, 2nd edn. Basingstoke: Palgrave.
Heller, P. S. 2005. Pity the finance minister: Issues in
managing a substantial scaling up of aid flows. IMF
Working Paper 05/180.
Kang, J., A. Prati, and A. Rebucci. 2012. Aid, exports and
growth: A time series perspective on the Dutch Disease
hypothesis. Review of Economics and Institutions 3(2).
Klein, M., and T. Harford. 2005. The market for aid.
Washington, DC: World Bank Publications.
Magud, N. and Sosa, S. 2010. When and why worry about
real exchange rate appreciation: The missing link between
Dutch Disease and growth. IMF Working Paper 10/271.
Martins, P., and Y. Yang. 2009. The impact of exporting on
firm productivity: A metaanalysis of the leaming-by-
doing hypothesis. Review of World Economics 145(3):
431-445.
Ndikumana, L., and J.K. Boyce. 2003. Public debts and pri-
vate assets: Explaining capital flight from Sub-Saharan
African countries. World Development 31(1): 107-130.
Nkusu, M. 2004. Aid and the Dutch Disease in low income
countries: Informed diagnosis for prudent prognosis. IMF
Working Paper 04/49.
Rajan, R., and S. Subramanian. 2011. Aid, Dutch Disease and
manufacturing growth. Journal of Development
Economies 94(1): 106-118.
Tarp, F. 2008. Foreign aid. Die new palgrave dictionary of
economies, 2nd edn. Basingstoke: Palgrave.
Tokarick, S. 2008. Welfare-worsening aid flows to small
countries: The role of non-traded goods. Review of
Development Economics 12(4): 818 827.
Torvik, R. 2001. Leaming-by-doing and the Dutch Disease.
European Economic Review 5: 285-306.
Dynamic Models with Non-clearing Markets
Van Wijnbergen, S. 1984. The ‘Dutch Disease’: A disease
after all? Economic Journal 84(373): 41-55.
Werker, E., F. Ahmed, and C. Cohen. 2009. How is foreign
aid spent? Evidence from a natural experiment American
Economic Journal: Macroeconomics 1(2): 225-244.
Yano, M., and J. Nugent. 1999. Aid, nontraded goods and the
transfer paradox in small countries. American Economic
Review 89(3): 431—149.
Younger. S. 1992. Aid and the Dutch Disease: Macroeco-
nomic management when everybody loves you. World
Development 20( 11): 1587-1597.
Dynamic Models with Non-clearing
Markets
Jean-Pascal Benassy
Abstract
This article studies a new class of models which
synthesize the two traditions of general
equilibrium with non-clearing markets and
imperfect competition on the one hand, and
dynamic stochastic general equilibrium (DSGE)
models on the other hand. This line of models
has become a central paradigm of modern
macroeconomics for at least three reasons: (a) it
displays solid microeconomic foundations, (h) it
is a highly synthetic theory, which combines in
a unified framework general equilibrium, non-
clearing markets, imperfect competition, growth
theory and rational expectations, and (c) it is
also an empirical success, leading to substantial
progress towards matching real world statistics.
Keywords
Budget constraints; Dynamic models with non-
clearing markets; Dynamic stochastic general
equilibrium (DSGE) models; Efficiency wages;
Fixprice macroeconomic model; General
equilibrium; Growth theory; Implicit contracts;
IS-LM model; Keynesianism; Market clearing;
Menu costs; Microfoundations; Monetary
shocks; Monopolistic competition; Nominal
rigidities; Non-clearing markets in general
equilibrium; Objective
Dynamic Models with Non-clearing Markets
demand curve; Phillips curve; Quantity con- ability of DSGE to reproduce a number of mac-
straint; Quantity signals; Rational expectations; roeconomic facts.
Real business cycles; State dependent price
rigidities; Sticky prices; Technological shocks;
Time-dependent contracts; Walrasian History
equilibrium
JEL Cl assi f i cat i ons
D5
This article studies a new class of models which
synthesize the two traditions of general equilibrium
with non-clearing markets and imperfect
competition on the one hand, and dynamic sto-
chastic general equilibrium (DSGE) models on the
other hand. Although this line of models is still
recent, it has clearly become in a short time a
central paradigm of modem macroeconomics. The
reasons are at least threefold.
The first is that it displays solid microeconomic
foundations. This is quite natural since from the two
constituent fields above this one inherited a strong
general equilibrium framework where all agents
(households or firms) maximize their respective
objectives subject to well defined constraints.
The second is that it is a highly synthetic theory,
which combines in a unified framework general
equilibrium, non-clearing markets, imperfect
competition, growth theory and rational expecta-
tions, so that it can appeal to macro economists with
very different backgrounds.
The third reason is empirical. A key motivation
for DSGE models is to compare the ‘statistics’
generated by these models with the real-world ones.
In that respect the addition of non-clearing markets
and imperfect competition has led to substantial
progress in matching these statistics, and this has
certainly been an important factor in the success of
these models.
Now such a wide synthesis did not come all at
once. So we begin by recalling briefly a little bit of
history and some of the antecedents of the field.
We then present a series of models with explicit
solutions. These will demonstrate analytically how
the introduction of non-clearing markets allows us
to substantially improve the
3125
Early Times
At the time when many of the developments lead-
ing to these models were initiated, there was a
profound split between microeconomics and mac-
roeconomics. On the one hand microeconomics, in
its general equilibrium version, was dominated by
Walras’s ( ) model, as developed by Arrow
and Debreu ( ), Arrow ( ), and Debreu
( ). In these models all adjustments are carri ed
out via frilly flexible prices, and agents never
experience any quantity constraint. On the other
hand in the standard macroeconomic model in the
Keynes ( ) and Hicks ( ) tradition, as
exemplified by the IS-LM model, there are price
and wage rigidities, unemployment is present and
most adjustments are carried out through variations
in real income, a quantity, not a price.
Confronted with this inconsistency, the strate-
gies of macroeconomists turned out to be quite
diverse and they took two different routes.
General Equilibrium with Non-clearing
Markets
On the one hand, a first set of authors aimed at
achieving a synthesis between the then existing
microeconomics and macroeconomics. This was
achieved by generalizing the traditional general
equilibrium model, by introducing non-clearing
markets, introducing quantity signals into demand
and supply functions, and endogenizing prices in a
framework of imperfect competition.
Patinkin ( ) and Clower ( ) showed
that the presence of quantity constraints in non-
clearing markets would drastically modify the
demands for labour and goods, an insight further
emphasized by Leijonhufvud ( ).
Barro and Grossman ( , ) combined
these insights into a fixprice macromodel. Dreze (
) and Benassy ( , ) constructed full
general equilibrium concepts with price rigidities,
where price movements are partially replaced by
endogenous quantity constraints. Benassy ( )
3126 Dynamic Models with Non-clearing Markets

linked these concepts with general equilibrium A first type of rigidities is ‘real’ rigidities,
under imperfect competition a la Negishi ( ). which create an endogenous noncompetitive wedge
This link was furthered with the construction of a between various prices. As an example,
full general equilibrium concept of objective monopolistic competition a la Dixit and Stiglitz (
demand curve based on quantity constraints ) introduces a markup between marginal
(Benassy ; see also Gabszewicz and Vial , for a cost and price. In this class Danthine and
Coumotian view). All these developments are Donaldson ( ) introduce efficiency wages,
reviewed in the dictionary entry ‘non clearing Danthine and Donaldson ( , ) introduce
markets in general equilibrium’. implicit contracts in the vein of Azariadis ( ),
Baily ( ) and Gordon ( ). Rotemberg and
Dynamic Market Clearing Macroeconomics Woodford ( , ) study imperfect
A second set of authors achieved consistency competition.
between microeconomics and macroeconomics by Models with nominal rigidities study situations
importing into macroeconomics the basic where the nominal prices themselves (and not
assumption of the then dominant general equilib- relative prices) are sluggish. Several devices have
rium microeconomic models, market clearing. At been used. The first, following the early works on
the same time they paid strong attention to the wage and price contracts by Gray ( ), Fischer (
issues of dynamics and expectations. A central part ), Phelps and Taylor
of these developments was the use of ‘rational ( ), Taylor ( , ) and Calvo ( ),
expectations’ in the sense of Muth ( ). This assumes that there is a system of contracts expiring
was an important addition, as in the Keynesian at deterministic or stochastic dates. For that reason
system it was sometimes difficult to disentangle the they are called ‘time dependent’. Such contracts
results due to price or wage rigidity from those due have been integrated in DSGE models by Cho (
to incorrect expectations. Rational expectations ), Cho and Cooley ( ), Benassy
allowed the suppression of the second type of ( , , ), Yun ( ), Cho
results. It appeared also that, even with rational et al. ( ), Andersen ( ), Jeanne ( ),
expectations and market clearing, it was possible to Ascari ( ), Chari et al. ( ), Collard and
build rigorous models displaying fluctuations Ertz ( ), Ascari and Rankin ( ), Huang
(Lucas ; Kydland and Prescott ; Long and Plosser ). and Liu ( ), Smets and Wouters ( ) and
Christiano et al. ( ), to name only a few.
Non-Walrasian Cycles Another type of price rigidity, called ‘state
Starting in the mid-1980s authors began combining dependent’, is based on costs of changing prices.
elements of the two paradigms described above, Two specifications are favourite in the literature:
achieving the synthesis that is the subject of this quadratic costs of changing prices (Rotemberg ,
article. Svensson ( ) studies a ), which have been implemented, for
dynamic stochastic general equilibrium monetary example, in Hairault and Portier ( ), and
economy subject to supply and demand shocks. fixed costs of changing prices (Barro ), often
Prices are preset one period in advance by renamed ‘menu costs’. Clearly these costs should
monopolistically competitive firms, so we have be interpreted as surrogates for other unspecified
both imperfect competition and sticky prices. causes, and identifying these causes is a challenge
Because of price presetting the model has multiple that faces this line of research.
regimes. Now most of the contributions of this field are
Various types of rigidities have been then intro- based on numerical evaluations of various models.
duced in dynamic models, leading to different So we present next a number of models with
patterns of cycles. Andersen ( ) reviews vari explicit solutions which will make clear why this
ous causes and consequences of price and wage line of models has been successful in solving
rigidities. problems that were difficult to solve in market-
clearing models.
Dynamic Models with Non-clearing Markets 3127

An Analytical Illustration At the beginning of period t the household faces a

monetary shock a la Lucas ( ), whereby the
We shall now show in this section in a series of quantity of money Mt \ coming from t 1 is
explicitly solved models how the introduction of multiplied by /<,, so that its budget constraint for
nominal rigidities in DSGE models allows to con- period t is:
siderably improve the capacities of these models to
reproduce the dynamic evolutions of actual M, = W,N ^
economies. (3)
Pi P, P t
We first present a basic model and compute as a
reference its Walrasian equilibrium and dynamics. There are thus two shocks in this economy, the
Then we introduce a first nominal rigidity, one- technology shock Zt and the monetary shock /(, =
period wage contracts. This improves some MJMt _ !* As an illustration we shall use below the
correlations, but cannot create strong persistence as following traditional processes (in all that follows
in reality. We next introduce multi-periodic wage lower-case letters represent the logarithm of the
contracts, and show that this allows us to obtain a variable represented by the corresponding
persistent response of output to demand shocks. uppercase letter):
Finally, simultaneous rigidities of wages and prices
are considered, and we show that one can obtain in &mt ^ ___ £-/
this way with fairly realistic values of the m, — m,-\ 1 - pC' ~ 1 -<j)L (4)
parameters a persistent and hump-shaped response
of both output and inflation. where £-, and the innovations in zt and mt, are
uncorrelated white noises with:

The Basic Model var(;:z/) = ^var(;:m() = a2m (5)

We study a dynamic monetary economy a la
Sidrauski ( ) and Brock ( ), where goods
are exchanged against money at the (average) price
P, and work against money at the (average) wage
Wt. There are two types of agents: households and
firms. Firms have a simple technology:
Y, = Z , N « (1)
where N, is the quantity of labour used by firms and
Z, a technological shock common to all firms. Note
that we do not introduce capital in this model.
Because its rate of depreciation is low, it would not
add much to our argument, and would substantially and that employment Nt is constant:
complicate the results and exposition.
The representative household works Nt, con-
sumes C„ and ends period t with a quantity of
money Mt. It maximizes the expectation of its
discounted utility:
Walrasian Dynamics
As a benchmark we shall study here the case where
both labour and goods markets are in Walrasian
equilibrium in each period, as in the first traditional
real business cycle (RBC) models, and we shall see
how this economy reacts to technological and mon-
etary shocks. Solving the model we find that money
holdings are a multiple of consumption:
M, co
(6)
~hC, = T^
N, = N = («/£)1/v. (7)
Using ( ) and ( ) we find (we eliminate some
irrelevant constant terms):

n, = ny, = z, + an w, — p, = yt — n. ( 8 )
U = E0J2P‘ M, Nv'
logC + mlog-i-^ ■ (2) Although we will not do any real calibration in this
Pi v
1=0 article, we can note at this stage a few issues
3128
that posed a problem to researchers in the RBC
domain.
First, real wages are much too pro-cyclical in
this Walrasian model. From ( ) we see that the real
wage-output correlation is equal to 1. Even
though this correlation is lower than 1 in cali-
brated models where Nt varies, it is always quite
above what is observed in real economies.
A second problem concerns the
inflation-output correlation, a problem related to
the literature on the Phillips curve. Whereas it is
generally considered that this correlation is posi-
tive, the above Walrasian model yields a negative
correlation:
cov(Ap„ yt) = - < 0.
Finally, an important and recurrent critique of
RBC-type models has been that they do not gen-
erate any internal propagation mechanism, and
that the only persistence in output movements is
that already present in the exogenous process of
technological shocks zt (see, for example, Cogley
and Nason , ). This appears here in
Eq. ( ), where the dynamics of output^ is exactly
the same as that of the technological shock zt.
We shall now introduce wage contracts, first
lasting one period, and then multiperiod over-
lapping contracts, and we shall see that the above
problems find a natural solution in this framework.
Single-Period Wage Contracts
Let us thus assume (Benassy , and Benassy
, for microfoundations), that the wages are
predetermined at the beginning of each period at
the expected value of the Walrasian wage
(in logarithms), and that at this contractual wage
the households supply the quantity of work
demanded by firms (this type of contract was
introduced by Gray ).
Combining ( ) and Ct = Yt we find that the
Walrasian wage w* is, up to an unimportant
constant, equal to mt, so that the preset wage wt
is given by:
(10)
wt = Et \w* = Et xmt
Dynamic Models with Non-clearing Markets
where E, _imt is the expectation of m, formed at the
beginning of period t, before shocks are known.
The difference with the Walrasian case is that
employment^ is now variable and demand deter-
mined. Equations ( ) become:
y, = z, + xn,w, - p, =y, - nt ( 1 1 )
while nt n is replaced by ( ). So we first obtain
the level of employment in period t:
nt = n + m, — E,-\m, = n + em, ( 1 2 )
since mt — E,_1m, = emt. Contrarily to what happened
in the Walrasian version of the model, unanticipated
monetary shocks now have an impact on the level of
(9) employment, and therefore output. We shall now use
the preceding formulas to show that the hypothesis
of preset wages allows to substantially improve some
correlations relative to the Walrasian model.
Let us start with the real wage which, in the
Walrasian model, has a much too high positive
correlation with output. Let us combine ( ) and ( ), to
obtain the values of output and real wage:
y, = z, + ae„„w, - pt = z, - (1 - ct)smL (13)
We see that supply shocks create a positive corre-
lation between the real wage and output. However,
monetary shocks create a negative correlation. Our
model thus allows us to combine this last
characteristic, typical of traditional Keynesians
models, with the usual results of RBC models. If one
considers the technological and monetary shocks ( ),
one obtains the following correlation:
corr(w,-p,.yt)
=____________<r?-(l-<ft2)a(l-a)ff2_____________
{{a: + (1 -4>2)a-2am)}1/2[of + (1 - - c c f f fl ]'
(14)
We see that the real-wage-output correlation is equal
to 1 if there are only technological shocks. But this
correlation diminishes as soon as there are
Dynamic Models with Non-clearing Markets
monetary shocks, and it can even become negative.
One can thus reproduce the correlations observed in
reality by adequate combinations of technological
and monetary shocks.
Let us now consider the relation between infla-
tion and output, which are generally considered to
be positively correlated, at least in Keynesian
tradition. If we assume again the monetary and
technological shocks (4), we find:
Covariance (Ap,, y,) = a(l — ot)o2m
Formula ( ) shows us that the positive
covariance (and thus correlation) between inflation
and output is linked to the presence of demand
shocks, and that the sign of this correlation may
change if there are sufficiently strong technological
shocks.
So we just saw that one-period contracts allow
us to improve some important correlations. We now
naturally ask a question already posed for the
standard RBC model: is the response to shocks, and
in particular to demand shocks, sufficiently
persistent? Let us recall Eq. ( ):
yt = z, + tt£mt-
(16)
We see that monetary shocks now have an imme-
diate effect on output (and employment), but that,
starting with the second period, the effect of these
shocks is completely dampened. One-period con-
tracts allow us to solve the puzzle raised by some
correlations, but certainly not the persistence
problem. We shall see in the next two sections that
multi-periodic contracts allow us to solve that
problem.
Multi-periodic Wage Contracts
The models that we have examined so far share
with traditional RBC models the defect of having an
extremely weak internal propagation mechanism. In
particular, the response of output to monetary
demand shocks is almost entirely transitory. But We thus see that varying y from 0 to 1 the average
several empirical studies (see, for example, duration of the contract varies from zero to infinity.
Christiano et al. , ) have
3129
pointed out that in reality the response to monetary
shocks not only was persistent but also had a hump-
shaped response function. We shall now introduce
multi-periodic wage contracts in rigorous stochastic
dynamic models, and show that they allow us to
reproduce these features. Models with such multi-
periodic wage or price contracts have been studied
notably by Yun ( ), Ander-
sen ( ), Jeanne
( ), Ascari
( ), Chari
et al. ( ), Collard and Ertz ( ), and
Benassy ( , , ).
In order to make our demonstration analytically,
we use a contract, inspired by Calvo ( ) and
developed in Benassy ( , ),
which has three advantages: (a) the average dura-
tion of contracts can take any value from zero to
infinity, (b) an analytical solution can be found with
both wage and price contracts, and (c) it has explicit
microfoundations.
In this framework in each period s a contract is
made for wages at period t > s. As in the Gray
contract, the contract wage is the expectation of the
market-clearing wage in period t. So if we denote as
xst the contract wage made in 5 for period l:
xst = Es(w*).
(17)
Now, as in Calvo ( ), each wage contract
has a probability y to stay unchanged, and a prob-
ability 1 — y to be broken. If the contract is broken,
a new contract is immediately renegotiated on the
basis of current period information. So for y 0,
wages are totally flexible, for y 1 they are totally
rigid.
It is easy to compute the average duration of
these contracts. The probability for a contract to be
still valid j periods after the date it was concluded is
equal to (1 y)y'. The expected dura
tion of the contract is thus:
00
v
(18)
;=o '
3130 Dynamic Models with Non-clearing Markets

The average wage wy is the mean of past x y/s
weighted by the probability for the corresponding
contract to be still in effect. Because of the law of
large numbers, and since the probability of survival
of wage contracts is y, the proportion of contracts
coming from period s < t is (1 — y)'/ ,v. Therefore,
the average wage in the economy is given by:
i we saw above (formula ) that the average duration
M-, = (1 - y) ^2 (19)
5=—OO
Formula ( ) shows clearly that, contrarily
to the case of one-period contracts, the response to
a monetary shock can be quite persistent. We can
have an idea of the temporal profile of this response
by computing the response function of output and
employment to a monetary shock. The value of p
most often found in the literature is p 0:5. As for y,
of wage contracts is equal to y/ (1 — y). One

considers generally that the average duration of

wage contracts is about one year (see, for example,
If we now solve the model with the shocks ( ) Taylor ),
we find that the dynamics of employment is char- which corresponds to y = 4/5. Figure shows the
acterized by (Benassy , ): response of employment (output is derived via 21)
to a monetary shock for y = 4/5.
'('''ml
n> n
(i-y£)(i-y pl) (20) We see that the response function displays
persistence in the effects of monetary shocks, and
has even a hump-shaped response. If we plot,
where L is the lag operator: (JX, = Xt The response
however, the response function of inflation, we find
of output is deduced from that of employment
that it is steadily decreasing after the initial jump,
through:
whereas it seems to have a delayed hump-shaped
response in reality.
yt = an, + z,. (21)

Employment

Dynamic Models with Non-dearing Markets, Fig. 1

Dynamic Models with Non-clearing Markets
Wage and Price Multi-periodic Contracts
We shall now enlarge our model by considering
simultaneously wage and price multi-periodic
contracts (see Benassy , for such a model with
explicit microfoundations). Numerically solved
models with both wage and price multi- periodic
contracts are found in Christiano et al. ( ),
Huang and Liu ( ), Smets and
Wouters ( ).
Wage contracts are exactly the same as in the
preceding section: each contract is maintained with
probability y, or renegotiated with probability 1 —
y. Symmetrically, price contracts are maintained
with probability tp, or break down and are
renegotiated with probability 1 tp. The average
price p, is given by:
P, = (! - fl)v‘~ st (22)
5=—OO
where qst is the price contract negotiated in period s
for period /. Using again the shock processes ( ),
and taking v = 1, we find the following dynamics
for output and inflation:
^P^Zt . ^7 &mt
Z
‘ 1 — <p<j)L (1 — yL)(l — y pL)
. m,
( 1 -<pL)(l — ippL)
_________
(l - y<pL)(l - ytppL) (23)
7i, = (1 - L)p, = (1 - L) (m, - y,).
As in the preceding section we take as an
illustration a = 2/3, p = 1/2 and y = 4/5 (one-year
wage contracts). As for prices, we want to take a
rather low duration of contracts, so we shall take tp
1/2 (one quarter). Simulations show that in that case
we obtain a persistent and hump-shaped response
for both output and inflation.
So we see that with only reasonable nominal
rigidities we obtain some realistic response func-
tions. Clearly the adjunction of ‘real’ rigidities
would allow to reproduce even better the actual
dynamic macroeconomic patterns.
3131
See Also
Bibliography
Andersen, T. 1994. Price rigidity: Causes and macroeco-
nomic implications. Oxford: Oxford University Press.
Andersen, T. 1998. Persistency in sticky price models.
European Economic Review 42: 593-603.
Arrow, K. 1963. The role of securities in the optimal
allocation of risk-bearing. Review of Economic Studies
31:91-96.
Arrow, K., and G. Debreu. 1954. Existence of an equilibrium
for a competitive economy. Econometrica 22: 265-290.
Ascari, G. 2000. Optimising agents, staggered wages and
persistence in the real effects of money shocks. Economic
Journal 110: 664-686.
Ascari, G., and N. Rankin. 2002. Staggered wages and output
Sq
dynamics under disinflation. Journal of Economic
Dynamics and Control 26: 653-680.
Azariadis, C. 1975. Implicit contracts and underemployment
equilibria. Journal of Political Economv 83: 1183-1202.
Baily, M. 1974. Wages and employment under uncertain
demand. Review of Economic Studies 41: 37-50.
Barro, R. 1972. A theory ofmonopolistic price adjustment.
Review of Economic Studies 39: 17-26.
Barro, R., and H. Grossman. 1971. A general disequilibrium
model of income and employment. American Economic
Review 61: 82-93.
Barro, R., and H. Grossman. 1976. Money, employment and
inflation. Cambridge: Cambridge University Press.
Benassy, J.-P. 1975. Neo-Keynesian disequilibrium theory in
a monetary economy. Review of Economic Studies 42:
503-523.
Benassy, J.-P. 1976. The disequilibrium approach to
monopolistic price setting and general monopolistic
equilibrium. Review of Economic Studies 43: 69-81.
(24) Benassy, J.-P. 1977. A neoKeynesian model of price and
quantity determination in disequilibrium. In Equilibrium
and disequilibrium in economic theory, ed. G.
Schwodiauer. Boston: Reidel Publishing Company.
Benassy, J.-P. 1982. The economics of market disequilibrium.
New York: Academic.
Benassy, J.-P. 1988. The objective demand curve in general
equilibrium with price makers. Economic Journal
98(Suppl): 37^49.
Benassy, J.-P. 1990. Non-Walrasian equilibria, money and
macroeconomics. In Handbook of monetary economics,
ed. B. Friedman and F. Hahn. Amsterdam: North-
Holland.
Benassy, J.-P. 1995. Money and wage contracts in an
optimizing model of the business cycle. Journal of
Monetary Economics 35: 303-315.
3132
Benassy, J.-P. 2002. The macroeconomics of imperfect
competition and nonclearing markets: A dynamic general
equilibrium approach. Cambridge, MA: M.I.T. Press.
Benassy, J.-P. 2003a. Staggered contracts and persistence:
microeconomic foundations and macroeconomic
dynamics. Recherches Economiqu.es de Louvain 69: 125
144.
Benassy, J.-P. 2003b. Output and inflation persistence under
price and wage staggering: Analytical results. Annales d
'Economic et de Statistique, n 69: 1-30.
Brock, W. 1975. A simple perfect foresight monetary model.
Journal of Monetary Economics 1: 133-150.
Calvo, G. 1983. Staggered prices in a utility-maximizing
framework. Journal of Monetatv Economics 12: 383 398.
Chari, V., P. Kehoe, and El McGrattan. 2000. Sticky price
models of the business cycle: Can the contract multiplier
solve the persistence problem? Econometrica 68: 1151
1179.
Cho, .1.-0. 1993. Money and the business cycle with one-
period nominal contracts. (Canadian Journal of
Economics 26: 638 659.
Cho, .1.-0., and T. Cooley. 1995. Business cycles with
nominal contracts. Economic Theory 6: 13-34.
Cho, .1.-0., T. Cooley, and L. Phaneuf. 1997. The welfare
cost of nominal wage contracting. Review of Economic
Studies 64: 465 484.
Christiano, L., M. Eichenbaum, and C. Evans. 1999. Mon-
etary policy shocks: what have we learned and to what
end? In Handbook of macroeconomics, ed. .1. Taylor and
M. Woodford. Vol. 1A. Amsterdam: North-Holland.
Christiano. L.. M Eichenbaum. and C. Evans. 2005. Nominal
rigidities and the dynamic effects of a shock to monetary
policy. Journal of Political Economy 113: 1 46.
Clower, R. 1965. The Keynesian counterrevolution: A
theoretical appraisal. In The theory of interest rates, ed. F.
Hahn and F. Brechling. London: Macmillan.
Cogley. T.. and J. Nason. 1993. Impulse dynamics and
propagation mechanisms in a real business cycle model.
Economics Letters 43: 77-81.
Cogley. T.. and J. Nason. 1995. Output dynamics in real-
business-cycle models. American Economic Review 85:
492-511.
Collat'd. F.. and G. Ertz. 2000. Stochastic nominal wage
contracts in a cash-in-advance model. Recherches
Economiques de Louvain 66: 281-301.
Danthine. J.-P.. and J. Donaldson. 1990. Efficiency wages and
the business cycle puzzle. European Economic Review
34: 1275-1301.
Danthine. J.-P., and J. Donaldson. 1991. Risk sharing, the
minimum wage and the business cycle. In Equilibrium
theory and applications: A conference in honor of
Jacques Dreze. ed. W. Barnett et al. Cambridge: Cam-
bridge University Press.
Danthine. J.-P. and J. Donaldson. 1992. Risk sharing in the
business cycle. European Economic Review 36: 468-475.
Dynamic Models with Non-clearing Markets
Debreu, G. 1959. Theory of value. New York: Wiley.
Dixit, A., and .1. Stiglitz. ll>77. Monopolistic competition and
optimum product diversity. American Economic Review
67:297 308.
Dreze, J. 1975. Existence of an exchange equilibrium
underprice rigidities. International Economic Review 16:
301-320.
Fischer, S. 1977. Long-term contracts, rational expectations,
and the optimal money supply rule. Journal of Political
Economy 85: 191 205.
Gabszewicz, J., and J.-P. Vial. 1072. Oligopoly ‘A la
Cournot’ in a general equilibrium analysis. Journal of
Economic Theory 42: 381 400.
Gordon, D. 1974. A neo-classical theory of Keynesian
unemployment. Economic Inquiry 12: 4.31 459.
Gray, J.-A. 1976. Wage indexation: A macroeconomic
approach. Journal of Monetary Economics 2:221 2.35.
Hairault, J.-O., and F. Fortier. 1993. Money, new-Keynesian
macroeconomics and the business cycle. European
Economic Review .37: 1533 1568.
Hicks, J. 1937. Mr. Keynes and the ‘classics’: A suggested
interpretation. Econometrica 5: 147 159.
Huang, K., andZ. Liu. 2002. Staggered price-setting, stag-
gered wage-setting, and business cycle persistence.
Journal of Monetary Economics 4l): 405 4.33.
Jeanne, O. 1998. Generating real persistent effects ofmon-
etary shocks: How much nominal rigidity do we really
need'.' European Economic Review 42: 1009 1032.
Keynes. J.M. 1936. The general theory of employment,
interest and money. New York: Harcourt Brace.
Kydland, F., and E. Prescott. 1982. Time to build and
aggregate fluctuations. Econometrica 50: 1345 1370.
Leijonhufvud, A. 1968. On Keynesian economics and the
economics of Keynes. Oxford: Oxford University Press.
Long, J., and C. Plosser. 1983. Real business cycles. Journal
of Political Economy 91: 39 69.
Lucas, R. Jr. 1972. Expectations and the neutrality of money.
Journal of Economic Theory A: 103 124.
Muth, J. 1961. Rational expectations and the theory of price
movements. Econometrica 29: 315 335.
Negishi, T. 1961. Monopolistic competition and general
equilibrium. Review of Economic Studies 28: 196 201.
Patinkin, D. 1956. Money, interest and prices. 2nd ed, 1965.
New York: Harper and Row.
Phelps, E.. and J. Taylor. 1977. Stabilizing powers of
monetary policy under rational expectations. Journal of
Political Economy 85: 163 190.
Rotemberg, J. 1982a. Monopolistic price adjustment and
aggregate output. Review of Economic Studies 44: 517-
531.
Rotemberg, J. 1982b. Sticky prices in the United States.
Journal of Political Economy 90:1187 1211.
Rotemberg, J.. andM. Woodford. 1992. Oligopolistic pricing
and the effects of aggregate demand on economic activity.
Journal of Political Economy 100:1153 1207.
Rotemberg, J.. and M Woodford. 1995. Dynamic general
equilibrium models with imperfectly competitive product
markets. In Frontiers of business cycle research, ed. T.
Cooley. Princeton: Princeton University Press.
Dynamic Programming
Sidrauski, M. 1967. Rational choice and patterns of growth in
a monetary economy. American Economic Review
57(Suppl): 534-544.
Smets, F., and R. Wouters. 2003. An estimated dynamic
stochastic general equilibrium model of the euro area.
Journal of the European Economic Association 1: 1123-
1175.
Svensson, L. 1986. Sticky goods prices, flexible asset prices,
monopolistic competition and monetary policy. Review
of Economic Studies 53: 385-405.
Taylor, J. 1979. Staggered wage setting in a macro model.
American Economic Review 69: 108-113.
Taylor, J. 1980. Aggregate dynamics and staggered contracts.
Journal of Political Economy 88: 1-23.
Taylor, J. 1999. Staggered price and wage setting in mac-
roeconomics. In Handbook of macroeconomics, ed. J.
Taylor and M. Woodford, Vol. 1. Amsterdam: North-
Holland.
Walras, L. 1874. Elements d'economic politique pure.
Lausanne: Corbaz. Definitive edition Trans. W. Jaffe as
Elements of Pure Economics. London: Allen and Unwin,
1954.
Yun, T. 1996. Nominal price rigidity, money supply endo-
geneity, and business cycles. Journal of Monetary Eco-
nomics 37: 345-370.
Dynamic Programming
John Rust
Abstract
This article reviews the history and theory of
dynamic programming (DP), a recursive method
of solving sequential decision problems under
uncertainty. It discusses computational
algorithms for the numerical solution of DP
problems, and an important limitation in our
ability to solve realistic large-scale dynamic
programming problems, the ‘curse of dimen-
sionality’. It also summarizes recent research in
complexity theory that delineates situations
where the curse can be broken (allowing us to
solve DPs using fast polynomial time algo-
rithms), and situations where it is insuperable.
The literature on econometric estimation and
testing of DP models is reviewed, as is another
‘scientific limit to knowledge’, namely, the
identification problem.
3133
Keywords
Backward induction; Bellman equation; Com-
putational complexity; Computational experi-
ments; Concavity; Continuous and discrete time
models; Curse of dimensionality; Decision
variables; Discount factor; Dynamic discrete
choice models; Dynamic games; Dynamic
programming; Econometric estimation; Euler
equations; Game tree; Identification;
Independence; Indirect inference; Infinite
horizons; Kalman filtering; Kuhn-Tucker th;
Markov chain Monte Carlo methods; Markovian
decision problems; Maximum likelihood;
Method of simulated moments; Method of
simulated scores; Minimum residual method;
Monotonicity; Monte Carlo integration; Neural
networks; Nonlinear regression; Nonparametric
regression; Optimal decision rules; Policy
iteration; Principle of optimality; Rational
expectations; Sequential decision problems;
Simulated maximum likelihood; Simulated
method of moments; Simulated minimum
distance; Simulation-based estimation; State
variables; Stationarity; Statistical decision
theory; Structural estimation; Subgame
perfection; Uncertainty; Wald, A
JEL Classification
C51; C61
Introduction
Dynamic programming is a recursive method for
solving sequential decision problems (hereafter
abbreviated as SDP). Also known as backward
induction, it is used to find optimal decision rules in
‘games against nature’ and sub game perfect
equilibria of dynamic multi-agent games, and
competitive equilibria in dynamic economic
models. Dynamic programming has enabled
economists to formulate and solve a huge variety of
problems involving sequential decisionmaking
under uncertainty, and as a result it is
3134
now widely regarded as the single most important
tool in economics. Section “ ” provides a
brief history of dynamic programming. Section “
” discusses some of the main
theoretical results underlying dynamic program-
ming, and its relation to game theory and optimal
control theory. Section “
” pro-
vides a brief survey of numerical dynamic
programming. Section “
” sur-
veys the experimental and econometric literature
that uses dynamic programming to construct
empirical models economic behaviour.
History
The earliest reference to the use of the method of
backward induction to solve decision problems
appears to be Arthur Cayley’s solution to the
secretary problem (I am grateful to Arthur F. Veinott
Jr. for alerting me to this). In the mid-1940s a
number of different researchers in economics and
statistics appear to have independently discovered
backward induction as a way to solve SDPs involv-
ing risk or uncertainty. Von Neumann and
Morgenstem, in their seminal work on game theory (
), used backward induction to find what we
now call subgame perfect equilibria of extensive
form games. (‘We proceed to discuss the game T by
starting with the last move Mv and then going
backward from there through the moves M.v-2- ■ ■
, p. 126.) Abraham Wald,
who is credited with the invention of statistical
decision theory, extended this theory to sequential
decision-making in his 1947 book Sequential Anal-
ysis. Wald generalized the problem of gambler’s
ruin from probability theory and introduced the
sequential probability ratio test that minimizes the
expected number of observations in a sequential
generalization of the classical hypothesis test. How-
ever, the role of backward induction is less obvious
in Wald’s work. It was more clearly elucidated in
the 1949 paper by Anow, Blackwell and Girshick.
They studied a generalized version of the statistical
decision problem and formulated and solved it in a
way that is a readily recognizable application of
Dynamic Programming
modem dynamic programming. Following Wald,
they characterized the optimal rule for making a
statistical decision (for example, accept or reject a
hypothesis), accounting for the costs of collecting
additional observations. In the section ‘The Best
Truncated Procedure’ they show how the optimal
rule can be approximated ‘Among all sequential
procedures not requiring more than N
observations ...’ and solve for the optimal truncated
sampling procedure ‘by induction backwards’ (
, p. 217).
Other early applications of backward induction
include the work of Pierre Masse ( , p. 196) on
statistical hydrology and the management of reser-
voirs, and Dvoretzky et al. ( ) analysis of opti
mal inventory policy. Richard Bellman is widely
credited with recognizing the common structure
underlying SDPs, and showing how backward
induction can be applied to solve a huge class of
SDPs under uncertainty. Most of Bellman’s work in
this area was done at the RAND Corporation,
starting in 1949. It was there that he invented the
term ‘dynamic programming’ that is now the gen-
erally accepted synonym for backward induction.
Bellman ( , p. 159) explained that he invented
the name ‘dynamic programming’ to hide the fact
that he was doing mathematical research at RAND
under a Secretary of Defense who ‘had a patholog-
ical fear and hatred of the term, research’. He
settled on ‘dynamic programming’ because it would
be difficult give it a ‘pejorative meaning’ and
because ‘It was something not even a Congressman
could object to’.
Theory
Dynamic programming can be used to solve for
optimal strategies and equilibria of a wide class of
SDPs and multiplayer games. The method can be
applied both in discrete time and continuous time
settings. The value of dynamic programming is that
it is a ‘practical’ (that is, constructive) method for
finding solutions to extremely complicated
problems. However, continuous time problems
involve technicalities that I wish to avoid in this
survey. If a continuous time problem does not admit
a closed-form solution, the most commonly used
numerical approach is to solve an
Dynamic Programming
approximate discrete time version of the problem or
game, since under very general conditions one can
find a sequence of discrete time DP problems whose
solutions converge to the continuous time solution
the time interval between successive decisions tends
to zero (Kushner ). I start by describing how
dynamic programming is used to solve single agent
‘games against nature’. The approach can be
extended to solve multiplayer games, dynamic
contracts, principal-agent problems, and
competitive equilibria of dynamic economic
models. See recursive competitive equilibrium.
Sequential Decision Problems
There are two key variables in any dynamic pro-
gramming problem: a state variable st, and a deci-
sion variable d, (the decision is often called a
‘control variable’ in the engineering literature).
These variables can be vectors in R", but in some
cases they might be infinite-dimensional objects.
For example, in Bayesian decision problems, one of
the state variables might be a posterior distribution
for some unknown quantity 6. In general, this
posterior distribution lives in an infinite dimen-
sional space of all probability distributions on 0. In
heterogeneous agent equilibrium problems state
variables can also be distributions. The state vari-
able evolves randomly over time, but the agent’s
decisions can affect its evolution. The agent has a
utility or payoff function U(s\, d\, ..., sr, df) that
depends on the realized states and decisions from
period t = 1 to the horizon T. In some cases T = oo,
and we say the problem is infinite horizon. In other
cases, such as a life-cycle decision problem, T
might be a random variable, representing a
consumer’s date of death. As we will see, dynamic
programming can be adapted to handle either of
these possibilities. Most economic applications
presume a discounted, time-separable objective
function, that is, U has the form
T Hall and Rust ( ), where a speculator has a discrete
U(si,du- ■ - , S T , d T ) = Yt P u t M (1)
t= l
where j> is known as a discount factor that is
typically presumed to be in the (0, 1) interval,
3135
and ujst, dL) is the agent’s period t utility (payoff)
function. Discounted utility and profits are typical
examples of time separable payoff functions studied
in economics. However, the method of dynamic
programming does not require time separability, and
so I will describe it without imposing this
restriction.
We model the uncertainty underlying the deci-
sion problem via a family of history and decision-
dependent conditional probabilities {pfsfif ()} where
IIt \ = (st, du ..., st ,, dt () denotes the history, that is
the realized states and decisions from the initial date
l 1 to date l T. Note that this includes all
deterministic SDPs as a special case where the
transition probabilities pt are degenerate. In this case
we can represent the Taw of motion’ for the state
variables by deterministic functions st , | f(st, dt).
This implies that in the most general case, {s„ df,
evolves as a history dependent stochastic process.
Continuing the ‘game against nature’ analogy, it
will be helpful to think of {pt(sfif-i)\ as constituting
a ‘mixed strategy’ played by ‘nature’ and the
agent’s optimal strategy as a ‘best response’ to
nature’s strategy.
The final item we need to specify is the timing
of decisions. Assume that the agent selects dt after
observing st, which is ‘drawn’ from the distribution
p,(st\H, i). The alternative case where dt is chosen
before st is realized can also be handled, but requires
a small change in the formulation of the problem.
The agent’s choice of dt is restricted to a state-
dependent constraint (choice) set Dt(H, i, st). We
can think of D, as the generalization of a ‘budget
set’ in standard static consumer theory. The choice
set could be a finite set, in which case we refer to
the problem as discrete choice, or D, could be a
subset of it with non-empty interior, then we have a
continuous choice problem. In many cases, there is
a mixture of types of choices, which we refer to as
discrete- continuous choice problems. An example
is commodity price speculation; see for example
choice of whether or not to order to replenish his
inventory and a continuous decision of how much of
the commodity to order. Another example is
retirement: a person has a discrete decision of
3136 Dynamic Programming

whether to retire and a continuous decision of how that maximizes the decision-maker’s expected
much to consume. utility

Definition A (single agent) sequential decision 3* = aigm?LxE{u({s„d,}s)}, ( 2 )

problem (SDP) consists of (1) a utility function U, beF
(2) a sequence of choice sets [ D, (, and (3) a
where T denotes the class of feasible history-
sequence of transition probabilities {pt(s,\H,_i)}
dependent decision rules, and {st,dt}s denotes the
where we assume that the process is initialized at
stochastic process induced by the decision rule
some given initial state s,.
3 = (r),........ST). Problem ( ) can be regarded as a
In order to solve this problem, we have to make
static, ex ante version of the agent’s problem. In
assumptions about how the decision-maker eval-
game theory, ( ) is referred to as the normal form or
uates alternative risky strategies. The standard
the strategic form of a dynamic game, since the
assumption is that the decision-maker maximizes
dynamics are suppressed and the problem has the
expected utility. Backward induction does not nec-
superficial appearance of a static optimization
essarily result in optimal strategies for non-
problem or game in which an agent’s problem is to
expected utility maximizers, except for certain
choose a best response, either to nature (in the case
classes of recursive preferences.
of single agent decision problems) or to other
As the name implies, an expected utility max-
rational opponents (in the case of games). The
imizer makes decisions that maximize their ex ante
strategic formulation of the agent’s problem is quite
expected utility. However, since information unfolds
difficult to solve since the solution is a sequence of
over time, it is generally not optimal to pre-commit
histoiy-dependent functions S ' = (d j ,.... d f ) for
to any fixed sequence of actions (c/,,
which standard finite
..., df). Instead, the decision-maker can generally
dimensional constrained optimization techniques
obtain higher expected utility by adopting a histoiy-
(for example, the Kuhn-Tucker th) are inapplicable.
dependent strategy or decision rule (di,
(If we consider problems where all states can
..., () ■/ ). This is a sequence of functions such that
assume only a finite number of values, it is possible
for each time t the realized decision is a function of
to apply standard finite dimensional Kuhn-Tucker
all available information. In the engineering
constrained optimization methods, but if the state
literature, a decision rule that does not depend on
variables can assume a continuum of possible
evolving information is referred to as an open- loop
values, the programming problem becomes an
strategy, whereas one that does is referred to as a
infinite dimensional programming problem for
closed-loop strategy. In deterministic control
which optimal control and dynamic programming
problems, the closed-loop and open-loop strategies
methods are more appropriate. See Luenberger , for
are the same since both are simple functions of time.
a more thorough discussion of how Lagrange
However in stochastic control problems, open-loop
multipliers and Kuhn-Tucker methods can be
strategies are a strict subset of closed-loop
extended to problems where decisions are infinite-
strategies. Under our timing assumptions the
dimensional objects. These methods are usually
information available at time t is (I f i, ,v,), so we
applied in deterministic context, and there is a
can write dt = St(If |, st). By convention we set //(,
specialized literature on optimal control for solving
0 so that the available informa
such problems.) See Pontryagin’s principle of
tion for making the initial decision is just 5!. A
optimality.
decision rule is feasible if it also satisfies St(ff i, st) C
D,(H, i, st) for all (sh If ,). Each feasible decision rule
Solving Sequential Decision Problems by
can be regarded as a Tottery’ whose payoffs are
Backward Induction
utilities, the expected value of which corresponds to
To carry out backward induction, we start at the last
expected utility associated with the decision rule.
period, T, and for each possible combination
An optimal decision rule 3* is simply a feasible
(Ht_ i, sj) we calculate the time T value function
decision rule
Dynamic Programming 3137

and decision rule (we will discuss how backward where we have written U(IIr , . sT, dT) instead of
induction can be extended to cases where T is C/(si, du ..sT, dT) since //-, , = (v,, du .. sT~i, dT_ |).
random or where T = oo shortly). Next we move backward one time period to time 7'
1 and compute
Vt(IIt i , ) = max U(HT-\,ST,dr)
8T(HT |,.Vy) dj € I , St)
= argmax
U(HT_\,sT,dT ) ,
DJ € D7 (//r-1, ST )
(3)

Vt-i(Ht-2,st-i) — m a x E{Vj(llr 2,xr-\,dr \,sr)\Hj^2,sj-\,dr i }

dT -1 £D T - I (H T -2,S T - I )

IVj(HT ~2, S T - i,dr \, S T ) P { S T \H T -2, S T - i,dr i)

T

max (4)
dT -1 £D T - I (H T -2,S T -1

8t-\(Ht-2,st-\) — a r g m a x E{VT(HT-2,ST-i,dT-i,ST)\HT-2,ST-i,dT-i}
dT-l

where the integral in Eq. ( ) is the formula for the
conditional expectation of Vr, where the expectation
is taken with respect to the random variable St whose
value is not known as of time T 1. We
continue the backward induction recursively for time
periods T — 2, T — 3,... until we reach time period t
1. The equation for the value function V, in an
arbitrary period t is defined recursively by an
equation that is now commonly called the Bellman
equation
V t(Ht-t> $t)
= max
= max
d,£D (H, i,s ) J
t t
optimal policy, starting in state sq implied by the
recursively constructed sequence of decision rules
d = (<5i, ..., S T ) .
The Principle of Optimality
The key idea underlying why backward induction
produces an optimal decision rule is called
The Principle of Optimality An optimal decision
ride 8* = ( < 5 j , . . . , d ) ) has the property
that given any t e { 1 , . . . , T] and any history llt
\ in the support of the
c o n t r o l l e d p r o c e s s { , s y , dt}s,,8*
remains optimal for the ‘subgame ’starting at time t

(5) and history Ht_\. That is, 8* maximizes the

“continuation payoff’’ given by the conditioned
The decision rule St is defined by the value of d, that expectation of utility from period t to T, given
attains the maximum in the Bellman equation for history H _\.
t
each possible value of st)
8* = argmax£'{f/({i„d;}^)|/7;_1}. (7)
s
= a r g m a x E{V,+i(Ht-i,st,dt,St+\)\Hl-i,st,dt}.
dt €Dt{Ht-\,St)
In game theory, the principle of optimality is
(6)
equivalent to the concept of a subgame perfect
Backward induction ends when we reach the first equilibrium in an extensive form game. When all
period, in which case, as we will now show, the actions and states are discrete, the stochastic deci-
function Vi(si) provides the expected value of an sion problem can be diagrammed as a game tree.
The principle of optimality, which in game theory is
equivalent to the concept of a subgame perfect
3138
equilibrium, guarantees that if 8* is an optimal
strategy (or equilibrium strategy) for the overall
game tree, then it must also be an optimal strategy
for every subgame, or, more precisely, all subgames
that are reached with positive probability from the
initial node.
It should now be evident why there is a need for
the qualification ‘for all IIt \ in the support of
{si,di}s> ’ in the statement of the principle of
optimality. There are some subgames that are never
reached with positive probability under an optimal
strategy. Thus, it is easy to construct alternative
optimal decision rules that do not satisfy the
principle of optimality because they involve taking
suboptimal decisions on ‘zero probability
subgames’. Since these subgames are never reached,
such modifications do not jeopardize ex ante
optimality. However we cannot be sure ex ante
which subgames will be irrelevant ex post unless we
carry out the full backward induction process.
Dynamic programming results in strategies that are
optimal in every possible subgame, even those
which will never be reached when the strategy is
executed. Since backward induction results in a
decision rale 8 that is optimal for all possible
subgames, it is intuitively clear that 8 is optimal for
the game as a whole, that is, it is a solution to the ex
ante strategic form of the optimization problem ( ).
For a formal proof of this result for games
against nature (with appropriate care taken to ensure
measurability and existence of solutions), see
Gihman and Skorohod ( ). If in addition to
‘nature’ we extend the game tree by adding another
rational expected utility maximizing player, then
backward induction can be applied in the same way
to solve this alternating move dynamic game.
Assume that player 1 moves first, then player 2, then
nature, and so on. Dynamic programming results in
a pair of strategies for both players. Nature still
plays a ‘mixed strategy’ that could depend on the
entire previous history of the game, including all the
previous moves of both players. The backward
induction process ensures that each player can
predict the future choices of their opponent, not only
in the succeeding move but in all future stages of the
game. The pair of
Dynamic Programming
strategies (d1, 82) produced by dynamic program-
ming are mutual best responses, as well as being
best responses to nature’s moves. Thus, these
strategies constitute a Nash equilibrium. They
actually satisfy a stronger condition: they are Nash
equilibrium strategies in every possible subgame of
the original game, and thus are subgame- perfect
(Selten ). Subgame-perfect equilibria exclude
‘implausible equilibria’ based on incredible threats.
A standard example is an incumbent’s threat to
engage in a price war if a potential entrant enters the
market. This threat is incredible if the incumbent
would not really find it advantageous to engage in a
price war (resulting in losses for both firms) if the
entrant called its bluff and entered the market. Thus
the set of all Nash equilibria to dynamic multiplayer
games is strictly larger than the subset of subgame-
perfect equilibria, a generalization of the fact that, in
single agent decision problems, the set of optimal
decision rules includes ones which take suboptimal
decisions on subgames that have zero chance of
being reached for a given optimal decision mle.
Dynamic programming ensures that the decision-
maker would never mistakenly reach any such
subgame, similar to the way subgame perfection
ensures that a rational player would not be fooled by
an incredible threat.
Dynamic Programming for Stationary,
Markovian, Infinite-Horizon Problems
The complexity of dynamic programming arises
from the exponential growth in the number of
possible histories as the number of possible values
for the state variables, decision variables, and/or
number of time periods T increases. For example, in
a problem with N possible values for st and D
possible values for dt in each time period t, there are
[AD]1 possible histories, and thus the required
number of calculations to solve a general T period,
history-dependent dynamic programming problem
is 0{[ND]1). Bellman and Dreyfus ( ) referred to
this exponential growth in the
number of calculations as the curse of dimension-
ality. In the next section, I will describe various
strategies for dealing with this problem, but an
immediate solution is to restrict attention to time
Dynamic Programming
separable Markovian decision problems. These are
problems where the payoff function U is addi- tively
separable as in Eq. ( ), and where both the choice
sets {Dt} and the transition probabilities {pt} depend
only on the contemporaneous state variable st and
not on the entire previous history H, |. We say a
conditional distributionp, satisfies the Markov
property if it depends on the previous history only
via the most recent values, that is, if pt{st\Ht_x) =
pt(st\st-1, dt_i). In this case backward induction
becomes substantially easier. For example, in this
case the dynamic programming optimizations have
to be performed only at each of the N possible
values of the state variable at each time t, so only
O(NDT) calculations are required to solve a time T
period time separable Markovian problem instead of
0{[ND]T) calculations when histories matter. This is
part of the reason why, even though time non-
separable utilities and non-Markovian forms of
uncertainty may be more general, most dynamic
programming problems that are solved in practical
applications are both time separable and Markovian.
SDPs with random horizons 7 can be solved by
backward induction provided there is some finite
time T satisfying Pr{f <T l. In this case, backward
induction proceeds from the maximum possible
value T and the survival probability pt = Pr{f > t\f >
t is used as to capture the probability that the
problem will continue for at least one more period.
The Bellman equation for the discounted, time-
separable utility with uncertain lifetime is
V,(s,)= m a x [u,(s„d) + pfiEVt+\(s,,d)\
deD,(s,)
8,(s,) = a r g m a x [u,(s„ d) + pfiEV,+i
{s„ r f ) ] ,
where
EVt+fis,d) Vl+1{s')pt+1{s'\s,d). (9)
In many problems there is no finite upper bound
Ton the horizon. These are called infinite horizon
problems and they occur frequently in
3139
economics. For example, SDPs used to model
decisions by firms are typically treated as infinite
horizon problems. It is also typical in infinite
horizon problems to assume stationarity. That is,
the utility function u(s. d). the constraint set D(s),
the survival probability p, and the transition
probability d) do not explicitly depend on time t. In
such cases, it is not hard to show the value function
and the optimal decision rules are also stationary,
and satisfy the following version of Bellman’s
equation
V ( s ) = m a x \u{s,d) + pfiEV{s,d)\
deD(s)
not
d(s) = a r g m a x | m ( . v , d) + pjlEV(s,
d)\, ’
deD(s)
where
EV(s,d) = | V(s')p(s'\s,d)- (11)
This is a fully recursive definition of V, and as
such there is an issue of existence and uniqueness of
a solution. In addition, it is not obvious how to carry
out backward induction, since there is no ‘last’
period from which to begin the backward induction
process. However, under relatively weak
assumptions one can show there is a unique V
satisfying the Bellman equation, and the implied
decision rule in Eq. ( ) is an optimal decision rule
for the problem. Further, this decision rule can be
approximated by solving an approximate finite
horizon version of the problem by backward
induction.
For example, suppose that «(s, d)isa continuous
function of (s, d), the state space S is compact, the
constraint sets D(s) are compact for each 5 e 5, and
the transition probability /?(.v'|.s', d) is weakly
continuous in (s, d) (that is, EV(s, d)=
W(s')p(s'\s,d) is a continuous function of (5,
J S1
d) for each continuous function W:S —> R). Black-
well ( , ), Denardo ( ) and others have
proved that, under these sorts of assumptions, Kis
the unique fixed point to the Bellman operator F: B
—> B, where B is the Banach space of
3140
continuous functions on S under the supremum
norm, and I is given by
r(w)(s)
= max u(s,d) + pfi IT(,s,)/?(.v'|.v,d) .
d <ED(s)
The existence and uniqueness of V is a consequence
of the contraction mapping th, since T can be shown If W 0 (that is the zero function in B), then VT = 1
to satisfy the contraction property,
|| rw- rv ||< a || w-v ||, (13) programming problem. Thus, this result implies that
where a c_ (0,1) and \\W\\ = sup, s ,v|W(s)\. In this
case, a = pfj, so the Bellman operator will be a
contraction mapping if pfi e (0,1).
The proof of the optimality of the decision rule
8 in Eq. ( ) is somewhat more involved. Using the
Bellman equation ( ), we will show that (see Eq. ( )
in section “
V(s) = «(s, S(s))
+pp V{s'W |,,d(,))
So = s}
7=0
that is, V is the value function implied by the
decision rule 8. Intuitively, the boundedness of the
utility function, combined with discounting of future
utilities, p[> C (0,1), implies that if we truncate the
infinite horizon problem to a 7’period problem, the
error in doing so would be arbitrarily small when T
is sufficiently large. Indeed, this is the key to
understanding how to find approximately optimal
decision rules to infinite horizon SDPs: we
approximate the infinite horizon decision rule 8 by
solving an approximate finite horizon version of the
problem by dynamic programming. The validity of
this approach can be formalized using a well-known
property of contraction mappings, namely, that the
method of successive approximations starting from
any initial guess W converges to the fixed point of
T, that is
Dynamic Programming
lim V, = T'(VT) = V VW c li. (15)
'
where F‘W denotes t successive iterations of the
Bellman operator 1 ,
y0 = r°(wo = w v, = r'(w)
(12)
(1<5)
v, = r'0*0 = i fr 1 iv) = r(v,_i).
'({)) is simply the period / = 1 value function
resulting from the solution of a T period dynamic
the optimal value function VT for a 7:pcriod
approximation to the infinite horizon problem
converges to V as 7’ oo. Moreover, the difference in
the two functions satisfies the bound
[pP\‘ I
||Vr-V|| < (17)
»
1 ~PP
Let 8t = 81 ■/; r>2,7v • •, 8ttbe the optimal decision
mle to the T period problem. It can be shown that, if
(14) we follow this decision rule up to period T and then
use f>, ,7- in every period after T, the resulting
decision mle is approximately optimal in the sense
that the value function for this infinite horizon
problem also satisfies inequality ( ), and
thus can be made arbitrarily small as T increases.
In many cases in economics the state space 5
has no natural upper bound. An example might be
where s, denotes an individual’s wealth at time t, or
the capital stock of the firm. If the unboundedness
of the state space results in unbounded payoffs, the
contraction mapping argument must be modified
since the Banach space structure under the
supremum norm no longer applies to unbounded
functions. Various alternative approaches have been
used to prove existence of optimal decision mles for
unbounded problems. One is to use an alternative
norm (for example, a weighted norm) and demon-
strate that the Banach space/contraction mapping
argument still applies. However, there are cases
where there are no natural weighted norms, and the
contraction mapping property cannot hold
Dynamic Programming 3141

since the Bellman equation can be shown to have actions D, dynamic programming requires O ( [ A 7 J ] ')
multiple solutions. The most general conditions operations to find a solution. Thus it appears that
under which the existence and uniqueness of the the time required to compute a solution via dynamic
solution V to the Bellman equation and the optimal- programming increases exponentially fast with the
ity of the implied stationary decision rule <5 has number of possible decisions or states in a dynamic
been established is in Bhattacharya and Majumdar ( programming problem.
). However, as I discuss in the next section, Fortunately, computer power (for example,
considerable care must be taken in solving operations per second) has also been growing
unbounded problems numerically. exponentially fast, a consequence of Moore’s Law
and other developments in information technology,
such as improved communications and massive
Numerical Dynamic Programming and the parallel processing. Bellman and Dreyfus ( )
Curse of Dimensionality carried out calculations on RAND’s
‘Johnniac’ computer (named in honour of Jon von
The previous section showed that dynamic pro- Neumann, whose work contributed to the
gramming is a powerful tool that has enabled us to development of the first electronic computers) and
formulate and solve a wide range of economic reported that this machine could do 12,500
models involving sequential decision-making under additions per second. Nowadays, in 2007, a typical
uncertainty - at least ‘in theory’. Unfortunately, the laptop computer can do over a billion operations per
cases where dynamic programming results in second and we now have supercomputers that are
analytical, closed-form solutions are rare and often approaching a thousand trillion operations per
rather fragile in the sense that small changes in the second—a level known as a ‘petaflop’. In addition
formulation of a problem can destroy the ability to to faster ‘hardware’, research on numerical methods
obtain an analytical solution. However even though has resulted in significantly better ‘software’ that
most problems do not have analytical solutions, the has had a huge impact on the spread of numerical
theorems in the previous section guarantee the dynamic programming and on the range of
existence of solutions, and these solutions can be problems we can solve. In particular, algorithms
calculated (or approximated) by numerical methods. have been developed that succeed in ‘breaking’ the
Since the 1980s, faster computers and better curse of dimensionality, enabling us to solve in
numerical methods have made dynamic polynomial time classes ofproblems that were
programming a tool of substantial practical value by previously believed to be solvable only in
significantly expanding the range of problems that exponential time. The key to breaking the erase of
can be solved. In particular, it has led to the dimensionality is the ability to recognize and
development of a large and rapidly growing exploit special structure in an SDP problem. We
literature on econometric estimation and testing of have already illustrated an example of this in
‘dynamic structural models’ that I will discuss in section “
the next section.
However, there are still many difficult chal- if the SDP is Markovian and
lenges that prevent us from formulating and solving utility is time separable, a finite horizon, finite state
models that are as detailed and realistic as we mightSDP can be solved by dynamic programming in
like, a problem that is especially acute in empiricalonly O(NDT) operations, compared to the 0{[ND]t)
applications. The principal challenge is what operations that are required in the general history-
Bellman and Dreyfus ( ) called the curse
dependent case. There is only enough space here to
of dimensionality. We have already illustrated this discuss several of the most commonly used and
problem in section “ most effective numerical methods for solving
different types of SDPs by dynamic programming. I
for history-dependent SDPs with a finite horizon T refer the reader to
and a finite number of states N and
3142
Puterman ( ), Rust ( ) and Judd ( ) for
more in-depth surveys on the literature on numer-
ical dynamic programming. See computational
methods in econometrics.
Naturally, the numerical method that is appro-
priate or ‘best’ depends on the type of problem
being solved. Different methods are applicable
depending on whether the problem has (a) finite
versus infinite horizon, (b) finite versus continuous-
valued state and decision variables, and (c) single
versus multiple players. In finite horizon problems,
backward induction is the essentially the only
approach, although as we will see there are many
different choices about how to most implement it
most efficiently - especially in discrete problems
where the number of possible values for the state
variables is huge (for example, chess) or in prob-
lems with continuous state variables. In the latter
case, it is clearly not possible to carry out backward
induction for every possible history (or value of the
state variable at stage t if the problem is Markovian
and time separable), since there are infinitely many
(indeed a continuum) of them. In these cases, it is
necessary to interpolate the value function, whose
values are only explicitly computed at a finite
number of points in the state space. I use the term
‘grid’ to refer to the finite number of points in the
state space where the backward induction
calculations are actually performed. Grids might be
lattices (that is, regularly spaced sets of points
formed as Cartesian products of unidimensional
grids for each of the continuous state variables), or
they may be quasirandom grids formed by
randomly sampling the state space from some
probability distribution, or by generating
deterministic sequences of points such as low
discrepancy sequences. The reason why one might
choose a random or low-discrepancy grid instead of
regularly spaced lattice is to break the curse of
dimensionality, as I discuss shortly. Also, in many
cases it is advantageous to refine the grid over the
course of the backward induction process, starting
out with an initial ‘coarse’ grid with relatively few
points and subsequently increasing the number of
points in the grid as the backward induction
progresses. I will have more to say about such
multigrid and
Dynamic Programming
adaptive grid methods when I discuss solution of
infinite horizon problems below.
Once a particular grid is chosen, the backward
induction process is carried out in the way it would
be normally be done in a finite state problem. On
the assumption that the problem is Markovian and
the utility is time separable and there are n grid
points {.v i, ..., ,v„}, this involves the following
calculation at each grid point sh i = 1,.. .,n
V,(si) = ma x M,(.v„ d) + pPEVt+1 (.V„ d) ,
(18)
where EVt+\(st, d) is a numerical estimate of the
conditional expectation of next period’s value
function. I will be more specific below about
which numerical integration methods are appro-
priate, but at this point it suffices to note that they
are all simple weighted sums of values of the
value function at t + 1, F,+1(s). We can now see
that, even if the actual backward induction calcu-
lations are carried out only at the n grid points
{si,..., s„}, we will still have to do numerical
integration to compute EVt+\ (,v,-, d) and the latter
calculation may require values of V/+i(s) at points
s off the grid, that is at points ,v (j_ { s s „ } .
This
is why some form or interpolation (or in some
cases extrapolation) is typically required. Almost
all methods of interpolation can be represented as
weighted sums of the value function at its known
values {F, 11 (v |),..., Vt, i (s„)} at the n grid points,
which were calculated by backward induction at
the previous stage. Thus, we have
k,+iW = ^H',(r)V(+1(i,), (19)
j= i
where w/.v) is a weight assigned to the ith grid point
that depends on the point s in qst. These weights are
typically positive and sum to 1. For example in
multilinear interpolation or simplicial interpolation
the w,(i) weights are those that allow s to be
represented as a convex combination of the vertices
of the smallest lattice hypercube containing s. Thus,
the weights wt(s) will be zero for all i except the
immediate neighbours of the
Dynamic Programming
point s. In other cases, such as kernel density and
local linear regression, the weights W/(s) are gen-
erally non-zero for all i, but the weights will be
highest for the grid points {.v,,..........v„ j which are
the nearest neighbours of s. An alternative approach
can be described as curve fitting. Instead of
attempting to interpolate the calculated values of the
value function at the grid points, this approach treats
these values as a data-set and estimates parameters
6 of a flexible functional form approximation to
Vt+i(s) by nonlinear regression. Using the estimated
d/+i from this nonlinear regression, we can ‘predict’
the value of Vt+i(s) at any s e S
Vl+1(s)=f(s,6l+iy
(20)
A frequently used example of this approach is to
approximate Vf+1 (s) as a linear combination of K
‘basis functions’ ,/y (.v),..., hK(s)\. This implies that
fis, 6) takes the form of a linear regression function
K
f{s,6) =J2&kbk(s),
(21)
k= 1
and can be estimated by ordinary least
squares. Neural networks are an example where
/depends on 6 in a nonlinear fashion. Partition 6 into
subvectors 9 = (y, 1, a), where y and X are vectors in
RJ. and a - (xb ..., a,). where each otj has the same
dimension as the state vector s. Then the neural
network/is given by
f(s,6) = f(s,y,X, a)
J
= ’^2yM1j + (W)
(22)
j=i
where (s, ~j.fi is the inner product between s and the
conformable vector a,, and y is a ‘squashing func-
tion’ such as the logistic function fix) = expj.r}/ (1 +
exp [A'| ). Neural networks are known to be
‘universal approximators’ and require relatively few
parameters to provide good approximations to
nonlinear functions of many variables. For further
details on how neural networks are applied,
3143
see the book by Bertsekas and Tsitsiklis ( ) on
Neuro-Dynamic Programming.
All these methods require extreme care for
problems with unbounded state spaces. By def-
inition, any finite grid can cover only a small subset
of the state space in this case, and thus any of the
methods discussed above would require
extrapolation of the value function to predict its
values in regions where there are no grid points, and
thus ‘data’ on what its proper values should be. Not
only may mistakes that lead to incorrect
extrapolations in these regions lead to errors in the
regions where there are no grid points, but the
errors can ‘unravel’ and also lead to considerable
errors in approximating the value function in
regions where we do have grid points. Attempts to
‘compactify’ an unbounded problem by arbitrarily
truncating the state space may also lead to
inaccurate solutions, since the truncation is itself an
implicit form of extrapolation (for example, some
assumption needs to be made what to do when state
variables approach the ‘boundary’ of the state
space: do we assume a ‘reflecting boundary’, an
‘absorbing boundary’, and so on?). For example in
life- cycle optimization problems, there is no
natural upper bound on wealth, even if it is true that
there is only a finite amount of wealth in the entire
economy. We can always ask the qst, if a person
had wealth near the ‘upper bound’, what would
happen to next period wealth if he invested some of
it? Here we can see that, if we extrapolate the value
function by assuming that the value function is
bounded in wealth, this means that by definition
there is no incremental return to saving as we
approach the upper bound. This leads to lower
saving, and this generally leads to errors in the
calculated value function and decision rale far
below the assumed upper bound. There is no good
general solution to this problem except to solve the
problem on a much bigger (bounded) state space
than one would expect to encounter in practice, in
the hope that extrapolation-induced errors in
approximating the value function die out the further
one is from the boundary. This property should hold
for problems where the probability that the next
period state will hit or exceed the ‘truncation
3144
boundary’ gets small the farther the current state is
from this boundary.
When a method for interpolating/extrapolating
the value function has been determined, a second
choice must be made about the appropriate method
for numerical integration in order to approximate
the conditional expectation of the value function
EVt+\(s, d) given by
EVl+i(s,d) (23)
There are two main choices here: (T) determin-
istic quadrature rales or (2) (quasi-) Monte Carlo
methods. Both methods can be written as weighted
averages of form
N
EVt+i (s, d) = ^ Wi(s,d)V,+i(ai),
(24)
where {w;(s, d)} are weights, and {«,•} are quad-
rature abscissae. Deterministic quadrature methods
are highly accurate (for example, an /V-point
Gaussian quadrature rale is constructed to exactly
integrate all polynomials of degree 2 N — 1 or less),
but become unwieldy in multivariate integration
problems when product rules (tensor products of
unidimensional quadrature) are used. Any sort of
deterministic quadrature method can be shown to be
subject to the curse of dimensionality in terms of
worst-case computational complexity (see Traub
and Werschulz ). For example, if N = 0( 1/e)
quadrature points are necessary to approximate a
univariate integral within s, then in a d-dimensional
integration problem bf1 = 0(\/ed) quadrature points
would be necessary to approximate the integral with
an error of s, which implies that computational
effort to find an e-approximation increases
exponentially fast in the problem dimension d.
Using the theory of computational complexity, one
can prove that any deterministic integration
procedure is subject to the curse of dimensionality,
at least in terms of a ‘worst case’ measure of
complexity. The curse of dimensionality can dis-
appear if one is willing to adopt a Bayesian per-
spective and place a ‘prior distribution’ over the
space of possible integrands and consider an
Dynamic Programming
‘average case’ instead of a ‘worst case’ notion of
computational complexity.
Since multivariate integration is a ‘subproblem’
that must be solved in order to carry out dynamic
programming when there are continuous state
variables (indeed, dynamic programming in
principle involves infinitely many integrals in order
to calculate EVt+i{s, d), one for each possible value
of (s, d) ), if there is a erase of dimensionality
associated with numerical integration of a single
multivariate integral, then it should also not be
surprising that dynamic programming is also subject
to the same erase. There is also a curse of
dimensionality associated with global optimization
of nonconvex objective functions of continuous
variables. Since optimization is also a sub-problem
of the overall dynamic programming problem, this
constitutes another reason why dynamic
programming is subject to a curse of
dimensionality. Under the standard worst case
definition of computational complexity, Chow and
TsitsiUis ( ) proved that no
deterministic algorithm can succeed in breaking the
curse of dimensionality associated with a suf-
ficiently broad class of dynamic programming
problems with continuous state and decision vari-
ables. This negative result dashes the hopes of
researchers dating back to Bellman and Dreyfus (
), who conjectured that there might be suf
ficiently clever deterministic algorithms that can
overcome the curse of dimensionality.
However, there are examples of random algo-
rithms that can circumvent the curse of dimen-
sionality. Monte Carlo integration is a classic
example. Consider approximating the
(multidimensional) integral in Eq. ( ) by using
random quadrature abscissae {a, | that are N
independent and identically distributed (IID) draws
from the distribution £>,+1(s'|s, d) and uniform
quadrature weights equal to w,(s, d) = UN. Then the
law of large numbers and the central limit theorem
imply that the Monte Carlo integral EV,+\ (s, d)
converges to the true conditional expectation EV,
+i(s, d) at rate 1 /\//Vregardless of the dimension of
the state space d. Thus a random algorithm, Monte
Carlo integration, succeeds in breaking the curse of
dimensionality of multivariate integration.
Unfortunately,
Dynamic Programming
randomization does not succeed in breaking the
curse of dimensionality associated with general
nonconvex optimization problems with continuous
multidimensional decision variables d (see
Nemirovsky and Yudin ).
However, naive application of Monte Carlo
integration will not necessarily break the curse of
dimensionality of the dynamic programming
problem. The reason is that a form of uniform
convergence (as opposed to pointwise) convergence
of the conditional expectations EV,+1 (s, d) to E Vt
11 (.v, d) is required in order to guarantee that the
overall backward induction process converges to the
true solution as the number of Monte Carlo draws,
N, gets large. To get an intuition why, note that if
separate IID sets of quadrature abscissae {df where
drawn for each (s, d) point that we wish to evaluate
the Monte Carlo integral£V,+i (,v, d) at, the
resulting function would be an extremely ‘choppy’
and irregular function of (5, d) as a result of all the
random variation in the various sets of quadrature
abscissae. Extending an idea introduced by Tauchen
and Hussey ( ) to solve rational expectations
models. Rust ( ) proved that it is possible to
break the curse of dimensionality in a class of SDPs
where the choice sets D,(s) are finite, a class he
calls discrete decision processes. The restriction to
finite choice sets is necessary, since, as noted above,
randomization does not succeed in breaking the
curse of dimensionality of nonconvex optimization
problems with continuous decision variables. The
key idea is to choose, as a random grid, the same set
of random points that are used quadrature abscissae
for Monte Carlo integration. That is, suppose /y-
ds'ly d) is a transition density and the state space
(perhaps after translation and normalization) is
identified with the d-dimensional hypercube S =
[0,1 \d. Apply Monte Carlo integration by drawing
N IID points {.?i.... ,sN} from the this hypercube
(this can be accomplished by drawing each com-
ponent of Si from the uniform distribution on the
[0,1] interval). We have
_ lN
EV t+\ ( s , cl) = - ^Vt+fsfp^fsils, d). ( 2 5 )
3145
Applying results from the theory of empirical
processes (Pollard ), Rust showed that this form of
the Monte Carlo integral does result in uniform
convergence (that is, PEVt+i(s,d) EVt 11 (,v, d)P =
Op{ 1 /fN)), and, using this, he showed that this
randomized version of backward induction succeeds
in breaking the curse of dimensionality of the
dynamic programming problem. The intuition of
why this works is, instead of trying to approximate
the conditional expectation in ( ) by
computing many indepen
dent Monte Carlo integrals (that is, drawing sep-
arate sets of random abscissae {df from pt, 1 (v'|.v,
d) for each possible value of (5, dj), the approach in
Eq. ( ) is to compute a single Monte Carlo
integral where the random quadrature points {,v,}
are drawn from the uniform distribution on [0,l] rf,
and the integrand is treated as the function Vt+1(s')
Pn i(.v'|.v, d) instead of Vt+1 (s'). The second impor-
tant feature is that Eq. ( ) has a self- approximating
property: that is, since the quadrature abscissae are
the same as the grid points at which we compute the
value function, no auxiliary interpolation or function
approximation is necessary in order to evaluate EV,
,,
+\(s,d). In particular, if p,+1(s |s', d) is a smooth
function of 5, then EVt+\(s, d) will also be a smooth
function of 5. Thus, backward induction using this
algorithm is extremely simple. Before starting back-
ward induction we choose a value for N and draw N
IID random vectors {sq,...,Ty} from the uniform
distribution on the d-dimensional hypercube. This
constitutes a random grid that remains fixed for the
duration of the backward induction. Then we begin
ordinary backward induction calculations, at each
stage t computing V) (.?,■) at each of the N random
grid points, and using the self- approximating
formula ( ) to calculate the con
ditional expectation of the period t + 1 value
function using only the N stored values (ym(.?i), ■ ■
■ ,V,+i(sN)) from the previous stage of the backward
induction. See Keane and Wolpin ( ) for an
alternative approach, which
combines Monte Carlo integration with the curve-
fitting approaches discussed above. Note that the
Keane and Wolpin approach will not generally
succeed in breaking the curse of dimensionality
since it requires approximation of functions of
3146 Dynamic Programming

d variables which is also subject to a curse of
dimensionality, as is well known from the literature
on nonparametric regression.
There are other subclasses of SDPs for which it
is possible to break the curse of dimensionality. For
example, the family of linear quadratic/ Gaussian
(LQG) can be solved in polynomial time using
highly efficient matrix methods, including efficient
methods for solving the matrix Ricatti equation
which is used to compute the Kalman filter for
Bayesian LQG problems (for example, problems
where agents only receive a noisy signal of a state
variable of interest, and they update their beliefs
about the unknown underlying state variable via
Bayes rule).
Now consider stationary, infinite horizon Mar-
kovian decision problems. As noted in section
where I ' is the Gateaux or directional derivative of
r, that is, it is the linear operator given by
r(v + tw ) - r(v)
r'{v){w) lim t
(28)
Newton’s method converges quadratically inde-
pendent of the value of the discount factor, as long
as it is less than 1 (to guarantee the contraction
property and the existence of a fixed point). In fact,
Newton’s method turns out to be equivalent to the
method ofpolicy iteration introduced by Howard (
). Let 8 be any stationary deci
sion rale, that is, a candidate policy. Define the
policy-specific conditional expectation operator Es
by

”, there is no ‘last’ EsV{s) V(s')p(s'\s,8(s)). (29)

period from which to begin the backward induction
process. However, if the utility function is time
separable and discounted, then, under fairly general
conditions, it will be possible to approximate the
solution arbitrarily closely by solving a finite
horizon version of the problem, where the horizon
Tis chosen sufficiently large. As we noted in section
“ (30)
”, this is
equivalent to solving for V, the fixed point to the
contraction mapping V = T( F) by the method of
successive approximations, where T is the Bellman
operator defined in Eq. ( ) of section
J s'
Given a value function V,. let 8, , be the decision
rule implied by Vt, that is
*(s,d) + p / l j V,(s')p(s/\s.d)
<5t+i(s) = argmax
d€D(s)
It is not hard to see that the value of policy 8t+1
must be at least as high as V„ and for this reason,
Eq. ( ) is called the policy improvement step of
the policy iteration algorithm. It is also not hard to
show that

r'(V t )(W)(s) = pfSEs,+lW(s), ( 3 1 )

and this implies that the Newton iteration, Eq. (

V,+1 = r{ V,).
), is numerically identical to policy
(26) iteration

Since successive approximations converges at a V,+i (s) = [ I - ppE&t+l]-\{s, 8I+1 ( s ) ) ,

geometric rate, with errors satisfying the upper (32)
bound in Eq. ( ), this method can converge at an
unacceptably slow rate when the discount factor is where 8t+1 is given in Eq. ( ). Equation ( ) is called
close to 1. A more effective algorithm in such cases the policy valuation step of the policy iteration
is Newton s method whose iterates are given by algorithm since it calculates the value function
implied by the policy 8t+Note that, since Eg is an
v l + i = v, - [i - r'(v,)} 1 [v, - r(v t )}, (27) expectation operator, it is linear and satisfies
11.Ball < 1, and this implies that the operator
Dynamic Programming 3147

[I — pfiEg] is invertible and has the following infinite horizon problems for which the discount
geometric series expansion factor is sufficiently close to 1. However, if the
discount factor is far enough below 1, then suc-
OO
[7-p^]-1 ='£\pffEi (33) cessive approximations can be faster since policy
j=o iteration requires 0(N3) operations per iteration
whereas successive approximations requires 0(N2)
where E’s is the j step ahead expectations operator. operations per iteration. At most T(s:, j’>)
Thus, we see that successive approximation iterations are required to
OO compute an £-approximation to an infinite horizon
[ / - pfiE&Ylu(s, < 5 ( s ) ) = lpfflEJ8u(s’ S(s)) Markovian decision problem with discount factor fj,
j=0 where T(s, ( j ) = log((l — fl)t:)/\og((l). Roughly
speaking, if T(e, /I) < N, then successive
= £ j x ! [p^]' u (^^( s i ))\ s o = j , approximations are faster than policy iteration.
Successive approximations can be accelerated
(34)
by a number of means discussed in Puterman ( ) and
so that value function Vt from the policy iteration ( Rust ( ). Multigrid algorithms
) corresponds to the expected value implied by are also effective: these methods begin backward
policy (decision rule) dt. induction with a coarse grid with relatively few grid
If there are an infinite number of states, the points N, and then as iterations proceed, the number
expectations operator E$ is an infinite- dimensional of grid points is successively increased, leading to
linear operator, so it is not feasible to compute an finer and finer grids as the backward induction
exact solution to the policy iteration Eq. ( ). starts to converge. Thus, computational time is not
However if there are a finite number of wasted early on in the backward induction iterations
states (or an infinite state space is discretized to a when the value function is far from the true
finite set of points, as per the discussion above), solution. Adaptive grid methods are also highly
then E$ is an N x N transition probability matrix, effective in many problems: these methods can
and policy iteration is feasible using ordinary matrix automatically detect regions in the state space where
algebra, requiring at most 0{N3) operations to solve there is higher curvature in the value function, and
a system of linear equations for Vt at each policy in these regions more grid points are added in order
valuation step. Further, when there are a finite to ensure that the value function is accurately
number of possible actions as well as states, there approximated, whereas in regions where the value
are only a finite number of possible policies |£>| |,s|, function is ‘flatter’ grid points can be removed, so
where \D\ is the number of possible actions and |5| as to direct computational resources to the regions
is the number of states, and policy iteration can be of the state space where there is the highest payoff
shown to converge in a finite number of steps, since in terms of accurately approximating the value
the method produces an improving sequences of function. See Grime and Semmler ( ) for more
decision mles, that is Vt < Vt+1. Thus, since there is details and an interest
an upper bound on the number of possible policies ing application of adaptive grid algorithms.
and policy iteration cannot cycle, it must converge I conclude this section with a discussion of
in a finite number of steps. The number of steps is several other alternative approaches to solving
typically quite small, far fewer than the total stationary infinite horizon problems that can be
number of possible policies. Santos and Rust ( ) extremely effective relative to ‘discretization’
show that the methods when the number of grid points N required
number of iterations can be bounded independent of to obtain a good approximation becomes very large.
the number of elements in the state space |S|. Thus, Recall the curve-fitting approach discussed above in
policy iteration is the method of choice for finite horizon SDPs: we approximate the value
function F by a parametric function as V g ( s ) — /(.v,
9 ) for some
3148
flexible functional form f where 9 are treated as
unknown parameters to be ‘estimated’. For infinite
horizon SDPs, our goal is to find parameter values 9
so that the implied value function satisfies the
Bellman equation as well as possible. One approach
to doing this, known as the minimum residual
method, is a direct analogue of nonlinear least
squares: if 9 is a vector with K components, we
select N > K points in the state space (potentially at
random) and find 9 that minimizes the squared
deviations or residuals in the Bellman equation
r~ 12
9 = argminV'
a
f(F«)(.v,) - F«(.v,
7“T L
(35)
eR il
K
where r denotes an approximation to the Bellman
operator, where some numerical integration and
optimization algorithm are used to approximate the
hue expectation operator and maximization in the
Bellman equation ( ). Another approach, called the
collocation method, finds 9 by choosing K grid
points in the state space and setting the residuals at
those K points to zero:
V-(Sl)=f(v?)(Sl)V^2)
= r(v7)(s2y-vftK)
= r(v 7 )(s K ). (36)
Another approach, called parametric policy iter-
ation, carries out the policy iteration algorithm in
Eq. ( ) above, but, instead of solving the linear
system ( ) for the value function V, at each
policy valuation step, they approximately solve this
system by finding 9, that solves the regression
problem
r iterations tends to infinity. A key assumption
9, = argmin V' V~(s;) - M(S;, <5,(s;))-
e 6 rk L e' (37)
ppEStVgt(Si)
Other than this, policy iteration proceeds exactly as
discussed above. Note that, due to the lineal ity
Dynamic Programming
of the expectations operator, the regression problem
above reduces to an ordinary linear regression
problem when Vg is approximated as a linear
combination of basis functions as in ( ) above.
There are variants of the minimum residual and
collocation methods that involve parameterizing the
decision rule rather than the value function. These
methods are frequently used in problems where the
control variable is continuous, and construct
residuals from the Euler equation - a functional
equation for the decision rule that can in certain
classes of problems be derived from the first-order
necessary condition for the optimal decision rule.
These approaches then try to find 9 so that the Euler
equation (as opposed to the Bellman equation) is
approximately satisfied, in the sense of minimizing
the squared residuals (minimum residual approach)
or setting the residuals to zero at K specified points
in the state space (collocation method). See Judd (
) for further
discussion of these methods and a discussion of
strategies for choosing the grid points necessary to
implement the collocation or minimum residual
method.
There is a variety of other iterative stochastic
algorithms for approximating solutions to dynamic
programming problems that have been developed in
the computer science and ‘artificial intelligence’
literatures on reinforcement learning. These
methods include Q-learning, temporal difference
learning, and reed time dynamic programming. The
general approach in all these methods is to
iteratively update an estimate of the value function,
and recursive versions of Monte Carlo integration
methods are employed in order to avoid doing
numerical integrations to calculate conditional
expectations. Using methods adapted from the
literature on stochastic approximation, it is possible
to prove that these methods converge to the true
value function in the limit as the number of
underlying the convergence proofs is that there is
sufficient stochastic noise to ensure that all possible
decisions and decision nodes are visited ‘infinitely
often’. The intuition of why such an assumption is
necessary follows from the discussion in section “
”: suppose
Dynamic Programming
that at some state s an initial estimate of the value
function for decision that is actually optimal hap-
pens to be so low that the action is deemed to be
‘nonoptimal’ relative to the initial estimate. If the
agent does not ‘experiment’ sufficiently, and thus
fails to choose suboptimal decisions infinitely often,
the agent may fail to learn that the initial estimated
value was an underestimate of the true value, and
therefore the agent might never learn that the
corresponding action really is optimal. There is a
trade-off between learning and experimentation, of
course. The literature on ‘multiarmed bandits’
(Gittins ) shows that a fully rational Bayesian
decision-maker will generally not find it optimal to
experiment infinitely often. As a result such an
agent can fail to discover actions that are optimal in
an ex post sense. However, this does not contradict
the fact that their behaviour is optimal in an ex ante
sense: rather, it is a reflection that learning and
experimentation is a costly activity, and thus it can
be optimal to be incompletely informed, a result
that has been known as early as Wald ( ). A
nice feature
of many of these methods, particularly the real time
dynamic programming developed in Barto et al. (
), is that these methods can be used in
‘real time’, that is, we do not have to ‘precalculate’
the optimal decision rule in ‘offline’ mode. All
these algorithms result in steady improvement in
performance with experience. Methods similar to
these have been used to produce highly effective
strategies in extremely complicated problems. An
example is IBM’s ‘Deep Blue’ computer chess
strategy, which has succeeded in beating the
world’s top human chess player, Garry Kasparov.
However, the level of computation and repetition
necessary to ‘train’ effective strategies is hugely
time consuming, and it is not clear that any of these
methods succeed in breaking the curse of
dimensionality. For further details on this literature,
see Bertsekas and Tsitsiklis ( ). Pakes ( )
applies these
methods to approximate Markov perfect equilibria
in games with many players. All types of stochastic
algorithms have the disadvantage that the
approximate solutions can be ‘jagged’ and there is
always at least a small probability that
3149
the converged solution can be far from the true
solution. However, they may be the only feasible
option in many complex, high-dimensional prob-
lems where deterministic algorithms (for example,
the Pakes and McGuire , algorithm for Markov
perfect equilibrium) quickly become intractable due
to the curse of dimensionality.
Empirical Dynamic Programming and the
Identification Problem
The developments in numerical dynamic pro-
gramming described in the previous section paved
the way for a new, rapidly growing literature on
empirical estimation and testing of SDPs and
dynamic games. This literature began to take shape
in the late 1970s, with contributions by Sargent (
) on estimation of dynamic labour
demand schedules in a linear quadratic framework,
and Hansen and Singleton ( ), who
developed a generalized method of moment esti-
mation strategy for a class of continuous choice
SDPs using the Euler equation as an orthogonality
condition. About the same time, a number of papers
appeared that provided different strategies for
estimation and inference in dynamic discrete choice
models including Gotz and McCall’s ( )
model of retirements of air force pilots,
Wolpin’s ( ) model of a family’s decision
whether or not to have a child, Pakes’s ( )
model of whether or not to renew a patent, and
Rust’s ( ) model of whether or not to replace a
bus engine. Since 1987, hundreds of different
empirical applications of dynamic programming
models have been published. For surveys of this
literature see Eckstein and Wolpin ( ), Rust
( ), and the very readable book by Adda and
Cooper ( ) - which also provides accessible
introductions to the theory and numerical methods
for dynamic programming. The remainder of this
section will provide a brief overview of estimation
methods and a discussion of the identification
problem.
In econometrics, the term structured estimation
refers to a class of methods that tries to go beyond
simply summarizing the behaviour of
3150
economic agents by attempting to infer their
underlying preferences and beliefs. This is closely
related to the distinction between the reduced-form
of an economic model and the underlying structure
that ‘generates’ it. (Structural estimation methods
were first developed at the Cowles Commission at
Yale University, starting with attempts to
structurally estimate the linear simultaneous
equations model, and models of investment by
firms. Frisch, Haavelmo, Koopmans, Marschak, and
Tinbergen were among the earliest contributors to
this literature.) The reason why one would want to
do structural estimation, which is typically far more
difficult (for example, computationally intensive)
than reduced-form estimation, is having knowledge
of underlying structure enables us to conduct
hypothetical/counterfac- tiial policy experiments.
Reduced-form estimation methods can be quite
useful and yield significant insights into behaviour,
but they are limited to summarizing behaviour
under the status quo. However, they are inherently
limited in their ability to forecast how individuals
change their behaviour in response to various
changes in the environment, or in policies (for
example, tax rates, government benefits,
regulations, laws, and so on) that change the
underlying structure of agents’ decision problems.
As long as it is possible to predict how different
policies change the underlying structure, we can use
dynamic programming to re-solve agents’ SDPs
under the alternative structure, resulting in corres-
ponding decision rules that represent predictions of
how their behaviour (and welfare) will change in
response to the policy change.
The rationale for structural estimation was rec-
ognized as early as Marschak ( ); however, his
message appears to have been forgotten until the
issue was revived in Lucas’s ( ) critique of the
limitations of reduced-form methods for policy
evaluation. An alternative way to do policy evalu-
ation is via randomized experiments in which sub-
jects are randomly assigned to the treatment group
(where the ‘treatment’ is some alternative policy of
interest ) and the con trol group (who continue with
the policy under the status quo). By comparing the
outcomes in the treatment and control groups, we
Dynamic Programming
can assess the behavioural and welfare impacts of
the policy change. However, human experiments
can be very time consuming and expensive to carry
out, whereas ‘computational experiments’ using a
structural model are very cheap and can be
conducted extremely rapidly. The drawback of the
structural approach, though, is the issue of
credibility of the structural model. If the structural
model is misspecified, it can generate incorrect
forecasts of the impact of a policy change. There are
numerous examples of how structural models can be
used to make policy predictions: see Todd and
Wolpin ( ) for an example that compares
the prediction of a structural model with the results
of a randomized experiment, where the structural
model is estimated using subjects from the control
group, and out-of-sample predictions are made to
predict the behavioural response by subjects in the
treatment group. They show that the structural
model results in accurate predictions of how the
treatment group subjects responded to the policy
change.
I illustrate the main econometric methods for
structural estimation of SDPs in the case of a sta-
tionary infinite horizon Markovian decision prob-
lem, although all the concepts extend in a
straightforward fashion to finite horizon, non-
stationary and non-Markovian problems. Estimation
requires a specification of the data generating
process. Assume we observe N agents, and we
observe agent i from time period —7’, to 7’, (or via
appropriate re-indexing, from t 1,..., Tf). Assume
observations of each individual are independently
distributed realizations from the controlled process
{st,dt}. However, while we assume that we can
observe the decisions made by each agent, it is more
realistic to assume that we only observe a subset of
the agent’s state st. If we partition^ = (xt, £,), assume
that the econometrician observes xt but not et, so this
latter component of the state vector constitutes an
unobserved state variable. Then the reduced-form of
the SDP is the decision rule 8
d=S(x,e),
(38)
since the decision rule embodies all the behavioural
content of the SDP model. The
Dynamic Programming
structure A consists of the objects A = {fi, p, u(s,
d),p(s'\s, d)\. Equation ( ) specifies the mapping
from the structure A into the reduced form, 8. The
data-set consists of {(x, h dit), t= 1,..Th i = 1, ..N.
The econometric problem is to infer the underlying
structure A from our data on the observed states and
decisions by a set of individuals. Although the
decision rule is potentially a complicated nonlinear
function of unobserved state variables in the
reduced-form Eq. ( ), it is often possible to
consistently estimate the decision mle under weak
assumptions as N ^ oo, or as Tj —> oo if the data
consists only of a single agent or a small number of
agents i who are observed over long intervals. Thus,
the decision rule 8 can be treated as a biown
function for purposes of a theoretical analysis of
identification. The identification problem is the qst,
under what conditions is the mapping from the
underlying structure A to the reduced form 1 to 1
(that is invertible)? If this mapping is 1 to 1, we say
that the structure is identified since in principle it
can be inverted to uniquely determine the
underlying structure A. In practice, we construct an
estimator A based on the available data and show
that A converges to the true underlying structure A
as N —y oo and/or 7} —> oo for each i.
Unfortunately, rather strong a priori assumptions
on the form of agents’ preferences and beliefs are
required in order to guarantee identification of the
structural model. Rust ( ) and
Magnac and Thesmar ( ) have shown that an
important subclass of SDPs, discrete decision pro-
cesses (DDPs), are nonparametrically uniden tified.
That is, if we are unwilling to make any parametric
functional form assumptions about preferences or
beliefs, then in general there are infinitely many
different structures A consistent with any reduced
form 8. In more direct terms, there are many
different ways to rationalize any observed pattern
of behaviour as being ‘optimal’ for different
configurations of preferences and beliefs. It is likely
that these results extend to continuous choice
problems, since it is possible to approximate a
continuous decision process (CDP) by a sequence
of DDPs with expanding numbers of elements in
their choice sets. Further, for dynamic games,
Ledyard ( ) has shown
3151
that any undominated strategy profile can be a
Bayesian equilibrium for some set of preferences
and beliefs. Thus, the hypothesis of optimality or
equilibrium per se does not have testable empirical
content: further a priori assumptions must be
imposed in order for SDPs models to be identified
and result in empirically testable restrictions on
behaviour.
There are two main types of identifying
assumptions that have been made in the literature to
date: (a) parametric functional form assumptions on
preferences u(s,d) and components of agents’ beliefs
/>(.v'|.v, d) that involve unobserved state variables e
and (b) rational expectations. Rational expectations
states that an agent’s subjective beliefs /?(,v'|.v, d)
coincide with objective probabilities that can be
estimated from data. Of course, this restriction is
useful only for those components of 5, x, that the
econometrician can actually observe. In addition,
there are other more general functional restrictions
that can be imposed to help identify the model. One
example is monotonicity and shape restrictions on
preferences (for example, concavity and
monotonicity of the utility function), and another
example is independence or conditional
independence assumptions about variables entering
agents’ beliefs. I will provide specific examples
below; however, it should be immediately clear why
these additional assumptions are necessary.
For example, consider the two parameters p (the
agent’s subjective survival probability) and [1 (the
agent’s subjective discount factor). We have seen in
section “ ” that only the product of p
and ft enter the SDP model, and not p and fj
separately. Thus, at most the product pjl can be
identified, but without further assumptions it is
impossible to separately identify the subjective
survival probability p from the subjective discount
factor [1 since both affect an agent’s behaviour in a
symmetrical fashion. However, we can separately
identify p and [1 if we assume that an individual has
rational survival expectations, that is, that their subj
ective survival probability p coincides with the
‘objective’ survival probability. Then we can
estimate p ‘outside’ the SDP model, using data on
the lifetime distributions of similar types of agents,
and then fj can be
3152
identified if other restrictions are imposed to guar-
antee that the product pfj is identified. However, it
can be very difficult to make precise inferences
about agents’ discount factors in many problems,
and it is easy to think of models where there is
heterogeneity in survival probabilities and discount
factors, and unobserved variables affecting one’s
beliefs about them (for example, family
characteristics such as a predisposition for cancer,
and so on, that are observed by an agent but not by
the econometrician) where identification is
problematic.
There are two main approaches for conducting
inference in SDPs: (a) maximum likelihood and (b)
‘simulation estimation’. The latter category includes
a variety of similar methods such as indirect
inference (Gourieroux and Monfort ), simulated
method of moments (McFadden ; Gallant and
Tauchen ), simulated maximum likelihood and
method of simulated scores (see simulation-based
estimation), and simulated minimum distance (Hall
and Rust ). To simplify the discussion I will define
these initially for single agent SDPs and at the end
discuss how these concepts naturally extend to
dynamic games. I will illustrate maximum
likelihood and show how a likelihood can be
derived for a class of DDPs; however, for CDPs, it
is typically much more difficult to derive a
likelihood function, especially when there are issues
of censoring, or problems involving mixed discrete
and continuous choice. In such cases simulation
estimation is often the only feasible way to do
inference.
For discrete decision processes, assume that the
utility function has the following parametric,
additively separable representation
u(x, £, d) = u(x, d,6i) + e{d) ( A S ) . A particularly tractable special case is where qU..
(39)
where e = {s{d)\d c D(x)}, and rid) is interpreted as
an unobserved component of utility associated with
choice of alternative d e D(x). Further, suppose that
the transition density p(x. e'| x, e, d) satisfies the
following conditional independence assumption
p(xfs'\x,£,d) = p(x'\x,d,62)q(£' ,df)(Cl).
(40)
Dynamic Programming
The Cl assumption implies that is an IID
‘noise’ process that is independent of {xt, dt).
Thus all of the serially correlated dynamics in the
state variables are captured by the observed com-
ponent of the state vector x,. If, in addition, q(ch Of)
is a distribution with unbounded support with finite
absolute first moments, one can show that the
following conditional choice probabilities exist
P(d\x,0) = J/£{d = S(x,e,6)}q(£)d£,
(41)
where 6 = (p, [1,01; 02, Of constitute the vector of
unknown parameters to be estimated. (Identification
of fully parametric models is a ‘generic’ property,
that is, if there are two different parameters 6 that
produce the same conditional choice probability
P(d[x, 0) for all x and d e D(x) - and thus led to
the same limiting expected log-likelihood - small
perturbations in the parameterization will ‘almost
always’ result in a nearby model for which 0 is
uniquely identified.) In general, the parametric func-
tional form assumptions, combined with the
assumption of rational expectations and the AS
and Cl assumptions, are sufficient to identify the
unknown parameter vector 6.6 can be estimated
by maximum likelihood, using the fidl infonnation
likelihood function Lf given by
Xf(°\{xi,t'di,l},t= 1 i=l,...,N)
N Ti
i=1 1=2
(42)
(f)
has a multivariate extreme value distribution where
O x is a common scale parameter (linearly related to
die standard deviation) for each variable in this
distribution (see McFadden, Daniel; logit models
of individual choice for the exact formula for this
density). This specification leads to a dynamic gen-
eralization of the multinomial logit model
P(d\x, 6)
exp{v(x, d, 6)/6i}
Erf' 6 D(v)exP{v(x,d',6)/63},
Dynamic Programming 3153

where v(x, d, 9) is the expected, discounted utility general specification for the way unobservable state
from taking action d in observed state x given by variables to enter the decision mle that result in a
the unique fixed point to the following smoothed nondegenerate likelihood function (that is, where
Bellman equation the likelihood £(9) is non-zero for any data-set and
any value of 9). For this reason, maximum
v(x,d,9) = likelihood estimation of CDPs is rare, outside
?j 03 k)g ^ certain special subclasses, such at linear quadratic
+ PP Y exp{v(y, d', 9)/9f\ CDPs (Hansen and Sargent ; Sargent ). However,
d' 6 D(x>) simulation-based methods of inference can be used
x p(x'\x,d, 02)dx'. in a huge variety of situations where a likelihood is
difficult or impossible to derive. These methods
(44) have a great deal of flexibility, a high degree of
generality, and often permit substantial
Define Tgby computational savings. In particular, generalizations
of McFadden’s ( )
Tfl (W) (x, d) = u(x,d,81) method of simulated moments (MSM) have enabled
estimation of a wide range of CDPs. The MSM
+ PP 03 log | exp{exp {W(x',d',d)/d3}
estimator minimizes a quadratic form between a set
A' e D(A)
of moments constructed from the data, hN and a
xp(x'\x, d, 82)dx'. vector of simulated moments hN S (9), that is
(45)

It is not hard to show that under weak assumptions 1N

h
Tg is a contraction mapping, so that v(x, d, 9) exists N =^Yh^Xit’dit})
and is unique. Maximum likelihood estimation can 'V ;=1
be carried out using a nested fixed point maximum vsw=^E^I>({4(0)X(0)})
likelihood algorithm consisting of an ‘outer’ ° j= i <=i
optimization algorithm to search for a value of 0 (46)
that maximizes ffiff), and an ‘inner’ fixed point
algorithm that computes \’g Vg(vg) each time the where h is a vector of./ > K ‘moments’ (that is,
outer optimization algorithm generates a new trial functionals of the data that the econometrician is
guess for 9. The implicit function theorem trying to ‘match’), where K is the dimension of 9,
guarantees that vn is a smooth function of 9. See {xit, dit} are the data, and |xj;(0), </^(0)|, / = 1,
Aguirregabiria and Mira ( ) for an . . ., S are 5 IID realizations of the controlled
ingenious alternative that ‘swaps’ the order of the process.
inner and outer algorithms of the nested fixed- point
algorithm resulting in significant computational The estimate 9 is given by 9=
speedups. See also Rust ( ) for further
details on the nested fixed-point algorithm and the a r g m i n [hN - hN,s(9)]'wN [hN - hN,s{9)],
properties of the maximum likelihood estimator, e e RK
and Rust ( ) for a survey of alternative less (47)
efficient but computationally simpler estimation
strategies. where Wm is a J x /positive-definite weighting
As noted above, econometric methods for CDPs, matrix. The most efficient choice for WN is W/g
that is, problems where the decision variable is
n -l
continuous (such as firm investment decisions, price DA where D/v is the variance-covariance
settings, or consumption/savings decisions) are matrix formed from the vector of sample moments
harder, since there is no tractable, hN. Simulation estimators require a nested fixed-
3154
point algorithm since each time the outer minimi-
zation algorithm tries a new trial value for 6, the
inner fixed point problem must be called to solve the
CDP problem, using the optimal decision rule =
s(x’i,,E/it,d\ to generate
the simulated decisions, and the fransition density
p{^i,l+1 ’ i+l Kn4,n4,n°2) t o g e n e r a t e
y=l,..., S IID realizations for a simulated panel each
potential value of 6. (It is important to simulate
using ‘common random numbers’ that remain fixed
as 6 varies over the course of the estimation, in
order to satisfy the stochastic equicontinuity
conditions necessary to establish consistency and
asymptotic normality of the simulation estimator.)
Simulation methods are extremely flexible for
dealing with a number of data issues such as
attrition, missing data, censoring and so forth. The
idea is that, if we are willing to build a stochastic
model of the data ‘problem’, we can account for it in
the process of simulating the behavioiual model. For
example, Hall and Rust ( ) develop a dynamic
model of commodity
price speculation in the steel market. An object of
interest is to estimate the stochastic process
governing wholesale steel prices; however, there is
no public commodity market where steel is traded
and prices are recorded on a daily basis. Instead,
Hall and Rust observe only the actual wholesale
prices of a particular steel trader, who records
wholesale prices only on the days he actually buys
steel in the wholesale market. Since the speculator
makes money by ‘buying low and selling high’, the
set of observed wholesale prices are endogenously
sampled, and failure to account for this can lead to
incorrect inferences about wholesale prices - a
dynamic analogue of sample selection bias.
However, in a simulation model it is easy to censor
the simulated data in the same way it is censored in
the actual data, that is, by discarding simulated
wholesale prices on days where no simulated
purchases are made. Hall and Rust show that even
though moments based on the observed (censored)
data are ‘biased’ estimates, the simulated moments
are biased in exactly the same fashion, so
minimizing the distance between actual and
simulated biased moments nevertheless
Dynamic Programming
results in consistent and asymptotically normal
estimates of the parameters of the wholesale price
process and other parameters entering the
speculator’s objective function.
Simulation methods have also enabled the use of
Bayesian methods, resulting in methods of
inference that do not require asymptotic approxi-
mations, although they generally use Markov chain
Monte Carlo methods to generate simulated draws
from a distribution that approximates the exact
finite sample posterior distribution for the
parameters of interest (see for example, Lancaster ;
Imai et al. ; Nourets ).
The most recent literature has extended the
methods for estimation of single-agent SDPs to
multi-agent dynamic games. For example, Rust (
) described applications of dynamic discrete
choice models to multiple-agent discrete dynamic
games. The unobserved state variables s, entering
any particular agent’s payoff function are assumed
to be unobserved both by the econometrician and by
the other players in the game. The Bayesian-Nash
equilibria of this game can be represented as a
vector of conditional choice probabilities (/>1(z/!|-v),
..., P„(d„\x)), one for each player, where P^dfc)
represents the econometrician’s and the other
players’ beliefs about the probability player i will
take action dh ‘integrating out’ over the
unobservable states variable si t affecting player Vs
decision at time t similar to Eq. ( ) for single-agent
problems. If one adapts the numerical methods for
Markov-perfect equilibrium described in section “
”, it
is possible to compute Bayesian-Nash equilibria of
discrete dynamic games using nested fixed- point
algorithms. While it is relatively straightforward to
write down the likelihood function for the game,
actual estimation via a straightforward application
of full information maximum likelihood is
extremely computationally demanding since it
requires a doubly nested fixed point algorithm (that
is, an ‘outer’ algorithm to search over 6 to
maximize the likelihood, and then an inner
algorithm to solve the dynamic game for each value
of 6, but this inner algorithm is itself a nested fixed-
point algorithm). Alternative, less computationally
demanding estimation methods
Dynamic Programming
have been proposed by Aguirregabiria and Mira ( SDPs whose decision rales provide arbitrarily good
), Bajari and Hong ( ), Bajari et al. approximations to individual behaviour.
( ), and Pesendorfer and Schmidt-Dengler
( ). This research is at the current frontier of
development in numerical and empirical applications of
dynamic programming.
Besides econometric methods, which are applied for
structural estimation for actual agents in their ‘natural’
settings, an alternative approach is to try to make
inferences about agents’ preferences and beliefs (and
even their ‘mode of reasoning’) for artificial SDPs in a
laboratory setting. The advantage of a laboratory experi-
ment is experimental control over preferences
and beliefs. The ability to control these aspects of
decision-making can enable much tighter tests of
theories of decision-making. For example, Binmore et al.
( ) structured a laboratory
experiment to determine whether individuals do
backward induction in one- and two-stage alternating
offer games, and ‘find systematic violations of backward
induction that cannot be explained by payoff-
interdependent preferences’ ( , P- 49).
Comments
There has been tremendous growth in research
related to dynamic programming since the 1940s.
The method has evolved into the main tool for
solving sequential decision problems, and research
related to dynamic programming has led to
fundamental advances in theory, numerical methods
and econometrics. As we have seen, while dynamic
programming embodies the notion of rational
decision-making under uncertainty, there is mixed
evidence as to whether it provides a good literal
description of how human beings actually behave in
comparable situations. Although human reasoning
and decision-making is undoubtedly both more
complex and more ‘frail’ and subject to foibles and
limitations than the idealized notion of ‘full
rationality’ that dynamic programming embodies,
the discussion of the identification problem shows
that, if we are given sufficient flexibility about how
to model individual preferences and beliefs, there
exist
3155
Thus, dynamic programming can be seen as a usefiil
‘first approximation’ to human decisionmaking, but it will
undoubtedly be superseded by more descriptively
accurate psychological models. Indeed, in the future one
can imagine behavioural models that are not derived
from some a priori axiomatization of preferences, but will
result from empirical research that will ultimately deduce
human behaviour from yet even deeper ‘structure’, that is
the very underlying neuroanatomy of the human brain.
Even if dynamic programming is unlikely to be
a descriptively accurate model of human decision-
making, it will probably still remain highly relevant
for the foreseeable future as the embodiment of
rational decision-making. There are well-defined
problems, for example, profit maximization or cost
minimization, where there is agreement on the
objective function to be maximized or minimized,
and where there will be a demand for dynamic
programming methods to find the optimal profit- or
cost-minimizing strategies. There are many
examples of this in the operations research litera-
ture. Practical applications include optimal inven-
tory management (Hall and Rust ) and optimal
harvesting of timber (Paarsch and Rust ).
Some observers such as Kurzweil ( ) pre
dict that in the not too distant future (for example,
approximately 2050) a singularity will occur, ‘dur-
ing which the pace of technological change will be
so rapid, its impact so deep, that human life will be
irreversibly transformed’ ( , p. 7). The singu
larity is a complex of accelerating improvements in
computer hardware and software, and a merger of
machine- and biological-based intelligence that will
blur the distinction between ‘artificial intelligence’
and human intelligence, that will overcome many of
current limitations of the human brain and human
reasoning: ‘By the end of this century, the
nonbiological portion of our intelligence will be
trillions and trillions of times more powerful than
unaided human intelligence’ ( , p. 9). Dynamic
programming will undoubtedly continue to be a
critical tool in this brave new world.
Whether this prognosis will ever come to pass,
or come to pass as soon as Kurzweil forecast, is
3156
debatable; but it does suggest that there will be
continued interest in and research on dynamic
programming. However, the fact that reasonably
broad classes of dynamic programming problems
are subject to a curse of dimensionality suggests
that it may be too optimistic to think that human
rationality will soon be superseded by ‘artificial
rationality’. While there are many complicated
problems that we would like to solve by dynamic
programming in order to understand what ‘fully
rational’ behaviour actually looks like in specific
situations, the curse of dimensionality still limits us
to very simple ‘toy models’ that only very partially
and simplistically capture the myriad of details and
complexities we face in the real world. Although we
now have a number of examples where artificial
intelligence based on principles from dynamic
programming outstrips human intelligence, for
example computerized chess, all these cases are for
very specific problems in very narrow domains. I
believe that it will be a long time before
technological progress in computation and
algorithms produce truly general-purpose
‘intelligent behaviour’ that can compete success-
fully with human intelligence in widely varying
domains and in the immensely complicated situa-
tions that we operate in every day. Despite all our
psychological frailties and limitations, there is an
important unanswered question of ‘how do we do
it?’, and more research is required to determine if
human behaviour is simply suboptimal, or whether
the human brain uses some powerful implicit
‘algorithm’ to circumvent the curse of
dimensionality that digital computers appear to be
subject to for solving problems such as SDPs by
dynamic programming. For a provocative theory
that deep principles of quantum mechanics can
enable human intelligence to transcend com-
putational limitations of digital computers, see
Penrose ( ).
See Also
► - ..'ii
► e "Wee::/
Dynamic Programming
Bibliography
This article has benefited from helpful feedback from Kenneth
Arrow, Daniel Benjamin, Larry Blume, Moshe
Buchinsky, Larry Epstein, Chris Phelan and Arthur F.
Veinott, Jr.
Adda, J., and R. Cooper. 2003. Dynamic economics quan-
titative methods and applications. Cambridge, MA: MTT
Press.
Aguirregabiria, V., and P. Mira. 2004. Swapping the nested
fixed point algorithm: A class of estimators for discrete
Markov decision models. Econometrica 70: 1519-1543.
Aguirregabiria, V., and P. Mira. 2007. Sequential estimation
of dynamic discrete games. Econometrica 75: 1-53.
Arrow, K.J., D. Blackwell, and M.A. Girshik. 1949. Bayes
and minimax solutions of sequential decision problems.
Econometrica 17: 213-244.
Bajari, P., and H. Hong. 2006. Semiparametric estimation of
a dynamic game of incomplete information, Technical
Working Paper No. 320. Cambridge, MA: NBER.
Bajari, P, L. Benkard, and J. Levin. 2007. Estimating dynamic
models of imperfect competition. Econometrica 75:
1331-1370.
Barto, A.G., S.J. Bradtke, and S.P. Singh. 1995. Learning to
act using real-time dynamic programming. Artificial
Intelligence 72: 81-138.
Bellman, R. 1957. Dynamic programming. Princeton:
Princeton University Press.
Bellman, R. 1984. Eye of the hurricane. Singapore: World
Scientific.
Bellman, R., and S. Dreyfus. 1962. Applied dynamic pro-
gramming. Princeton: Princeton University Press.
Bertsekas, D.P. 1995. Dynamic programming and optimal
control, vols 1 and 2. Belmont: Athena Scientific.
Bertsekas, D.P., and J. Tsitsiklis. 1996. Neuro-dynamic
programming. Belmont: Athena Scientific.
Bhattacharya, R.N., and M. Majumdar. 1989. Controlled
semi-Markov models - The discounted case. Journal of
Statistical Planning and Inference 21: 365-381.
Binmore, K., J. McCarthy, G. Ponti, L. Samuelson, and A.
Shaked. 2002. A backward induction experiment. Journal
of Economic Theoty 104: 48-88.
Blackwell, D. 1962. Discrete dynamic programming. Annals
of Mathematical Statistics 33: 719-726.
Dynamic Programming
Blackwell, D. 1965a. Positive dynamic programming. Pro-
ceedings of the 5th Berkeley Symposium 3: 415-428.
Blackwell, D. 1965b. Discounted dynamic programming.
Annals of Mathematical Statistics 36: 226-235.
Cayley, A. 1875. Mathematical qsts and their solutions.
Problem No. 4528. Educational Times 27, 237.
Chow, C.S., and J.N. Tsitsiklis. 1989. The complexity of
dynamic programming. Journal of Complexity 5: 466-
488.
Denardo, E. 1967. Contraction mappings underlying the the-
ory of dynamic programming. SIAM Review 9: 165-177.
Dvoretzky, A., J. Kiefer, and J. Wolfowitz. 1952. The
inventory problem: I. Case of known distributions of
demand. Econometrica 20: 187-222.
Eckstein, Z., and K.I. Wolpin. 1989. The specification and
estimation of dynamic stochastic discrete choice models:
A survey. Journal of Human Resources 24: 562-598.
Gallant, A.R., and G.E. Tauchen. 1996. Which moments to
match? Econometric Theory 12: 657-681.
Gihman, I.I., and A.V. Skorohod. 1979. Controlled stochastic
processes. New York: Springer.
Gittins, J.C. 1979. Bandit processes and dynamic allocation
indices. Journal of the Royal Statistical Society B 41:
148-164.
Gotz, G.A., and J.J. McCall. 1980. Estimation in sequential
decision-making models: A methodological note. Eco-
nomics Letters 6: 131-136.
Gourieroux, C., and A. Monfort. 1997. Simulation-based
methods of inference. Oxford: Oxford University Press.
Grime, L., and W. Semmler. 2004. Using dynamic pro-
gramming with adaptive grid scheme for optimal control
problems in economics. Journal of Economic Dynamics
and Control 28: 2427-2456.
Hall, G., and J. Rust. 2006. Econometric methods for
endogenously sampled time series: The case of com-
modity price speculation in the steel market. Manuscript:
Yale University.
Hansen, L.P., and T.J. Sargent. 1980. Formulating and esti-
mating dynamic linear rational expectations models.
Journal of Economic Dynamics and Control 2: 7-46.
Hansen, L.P., and K. Singleton. 1982. Generalized instru-
mental variables estimation of nonlinear rational expec-
tations models. Econometrica 50: 1269-1281.
Howard, R.A. 1960. Dynamic programming and Markov
processes. New York: Wiley.
Imai, S., N. Jain, and A. Ching. 2005. Bayesian estimation of
dynamic discrete choice models. Manuscript: University
of Illinois.
Judd, K. 1998. Numerical methods in economics. Cambridge,
MA: MIT Press.
Keane, M., and K.I. Wolpin. 1994. The solution and esti-
mation of discrete choice dynamic programming models
by simulation: Monte Carlo evidence. Review of
Economics and Statistics 76: 648-672.
Kurzweil, R. 2005. The singularity is near when humans
transcend biology. New York: Viking Press.
3157
Kushner, H.J. 1990. Numerical methods for stochastic control
problems in continuous time. SIAM Journal on Control
and Optimization 28: 999-1048.
Lancaster, A. 1997. Exact structural inference in optimal job
search models. Journal of Business Economics and
Statistics 15: 165-179.
Ledyard, J. 1986. The scope of the hypothesis of Bayesian
equilibrium. Journal of Economic Theory 39: 59-82.
Lucas Jr., R.E. 1976. Econometric policy evaluation: A
critique. In The phillips curve and labour markets,
Camegie-Rochester Conference on Public Policy, ed. K.
Brunner and A.K. Meltzer. Amsterdam: North- Holland.
Lucas Jr., R.E. 1978. Asset prices in an exchange economy.
Econometrica 46: 1426-1445.
Luenberger, D.G. 1969. Optimization by vector space
methods. New York: Wiley.
Magnac, T., and D. Thesmar. 2002. Identifying dynamic
discrete decision processes. Econometrica 70: 801-816.
Marschak, T. 1953. Economic measurements for policy and
prediction. In Studies in econometric method, ed. W.- C.
Hood and T. J. Koopmans. New York: Wiley.
Masse, P. 1945. Application des probabilites en chaine d
I’hydrologie statistique et au jeu des reservoirs. Report
to the Statistical Society of Paris. Paris: Berger-
Levrault.
Masse, P. 1946. Les reserves et la regulation de Vavenir.
Paris: Hermann.
McFadden, D. 1989. A method of simulated moments for
estimation of discrete response models without numerical
integration. Econometrica 57: 995-1026.
Nemirovsky, A.S., and D.B. Yudin. 1983. Problem complexity
and method efficiency in optimization. New York: Wiley.
Nourets, A. 2006. Inference in dynamic discrete choice
models with serially correlated unobserved state vari-
ables. Manuscript, University of Iowa.
Paarsch, H.J., and J. Rust 2007. Stochastic dynamic pro-
gramming in space: An application to British Columbia
forestry. Working paper.
Pakes, A. 1986. Patents as options: Some estimates of the
values of holding European patent stocks.
Econometrica 54: 755-784.
Pakes, A. 2001. Stochastic algorithms, symmetric Markov
perfect equilibria and the ‘curse’ of dimensionality.
Econometrica 69: 1261-1281.
Pakes, A., and P. McGuire. 1994. Computing Markov perfect
Nash equilibrium: Numerical implications of a dynamic
differentiated product model. RAND Journal of
Economics 25: 555-589.
Penrose, R. 1989. The emperor’s new mind. New York:
Penguin.
Pesendorfer, M., and P. Schmidt-Dengler. 2003. Identifica-
tion and estimation of dynamic games. Manuscript,
University College London.
Pollard, D. 1989. Asymptotics via empirical processes.
Statistical Science 4: 341-386.
3158
Putemian. M.L. 1994. Markovian decision problems. New
York: Wiley.
Rust. J. 1985. Stationary equilibrium in a market for durable
goods. Econometrica 53: 783-805.
Rust. J. 1987. Optimal replacement of GMC bus engines: An
empirical model of Harold Zurcher. Econometrica 55:
999-1033.
Rust. J. 1988. Maximum likelihood estimation of discrete
control processes. SIAM Journal on Control and Opti-
mization 26: 1006 1024.
Rust. J. 1994. Structural estimation of Markov decision
processes. In Handbook of econometrics, vol. 4. ed. R.F.
Engle and D.L. McFadden. Amsterdam: North- Holland.
Rust, J. 1996. Numerical dynamic programming in eco-
nomics. In Handbook of computational economics. ed. H.
Amman, D. Kendrick, and J. Rust. Amsterdam: North-I
lolland.
Rust, J. 1997. Using randomization to break the curse of
dimensionality. Econometrica 65: 487-516.
Rust, J., and G.J. Hall. 2007. The (S. s) rule is an optimal
trading strategy in a class of commodity price speculation
problems. Economic Theory 30: 515-538.
Rust, .1., and C. Phelan. 1997. How social security and
medicare affect retirement behavior in a world with
incomplete markets. Econometrica 65: 781 832.
Rust, .1., J.F. Traub. andH. Wozniakowski. 2002. Is there a
curse of dimensionality for contraction fixed points in the
worst case? Econometrica 70: 285 329.
Santos, M., and J. Rust. 2004. Convergence properties of
policy iteration. SIAM Journal on (Control and Optimi-
zation 42: 2094 2115.
Sargent, T.J. 1978. Estimation of dynamic labor demand
schedules under rational expectations. Journal of Polit-
ical Economy 86: 1009-1044.
Sargent, T.J. 1981. Interpreting economic time series. Journal
of Political Economy 89: 213-248.
Selten, R. 1975. Reexamination of the perfectness concept for
equilibrium points in extensive games. International
Journal of Game Theory 4: 25-55.
Tauchen, G., and R. Hussey. 1991. Quadrature-based methods
for obtaining approximate solutions to nonlinear asset
pricing models. Econometrica 59: 371-396.
Todd, P., and K.I. Wolpin. 2005. Ex ante evaluation of social
programs. Manuscript, University of Pennsylvania.
Traub, J.F, and A.G. Werschulz. 1998. Complexity and
information. Cambridge: Cambridge University Press.
Von Neumann, J., and O. Morgenstem. 1944. Theory of
games and economic behavior. Princeton: Princeton
University Press. 3rd edn, 1953.
Wald, A. 1947a. Foundations of a general theory of sequential
decision functions. Econometrica 15: 279-313.
Wald, A. 1947b. Sequential analysis. New York: Dover.
Wald, A., and J. Wollbwitz. 1948. Optimum character of the
sequential probability ratio test. Annals of Mathematical
Statistics 19: 326-339.
Wolpin, K. 1984. An estimable dynamic stochastic model of
fertility and child mortality. Journal of Political Economy
92: 852-874.
Dynamic Programming and Markov Decision Processes
Dynamic Programming and Markov
Decision Processes
Steven A. Lippman
A great many problems in economics can be
reduced to determining the maximum of a given
function. Dynamic programming is one of a number
of mathematical optimization techniques applicable
in such problems. As will be illustrated, the
dynamic programming technique or viewpoint is
particularly useful in complex optimization
problems with many variables in which time plays a
crucial role. Unlike calculus-based techniques it
does not require the function being optimized to be
differentiable in the (decision) variables.
In a nutshell, dynamic programming is a math-
ematical approach designed for analysing decision
processes in which the multi-stage or sequential
character of the process is prominent. In particular,
dynamic programming is likely to be applicable
whenever an economic agent makes a sequence of
decisions (such as how much to con sume in year /')
in a prespecified order (year fs decision is made
prior to year i + l’s decision). In contrast to the
familiar first-order condition which, for example,
balances marginal revenues against marginal costs,
the orientation of the typical dynamic programming
approach entails balancing current profits against
all fuUire profits. In so doing, it transforms a
complex n variable optimization problem into n
simple one variable optimization problems.
While it is important to understand the conditions
under which this approach induces a computationally
attractive technique, economists’ interest in dynamic
programming emanates from its analytic rather than
computational power. As empiricists we are interested in
the numbers (for example, the optimal amount
consumed in period 1), but wearing our theoretical or
policy-making hats we are more interested in the
intrinsic struc- Uire of the solution (for example, the
optimal amount consumed in period I decreases in
Dynamic Programming and Markov Decision Processes
response to an increase in the riskiness associated
with the income stream). With this in mind, we
emphasize the use of dynamic programming as a
conceptual framework enabling us to understand the
nature of the solution to the decision-maker’s
problem.
The basic components describing a multi-stage
decision process are states, stages, actions
(decisions), rewards, state transitions or law of
motion, and constraints. The relevant dynamic
programming concepts are those of policy, return
function, and functional equation. For pedagogical
purposes we define and explain these objects in the
context of consumption under uncertainty.
An Example
At the beginning of each of N periods (of equal
length such as one year) our economic agent must
decide how much of his current wealth to consume
and how much to save. After making his
consumption decision, his remaining wealth is
invested and experiences a (possibly random) return
of R per unit of capital. Assume R > 0 so that losing
his entire investment is the worst that can happen.
The agent is gainfully employed; accordingly, at the
end of each period his wealth is augmented by his
(possibly random) labour income L. To keep
matters simple, assume that the 2N random
variables are independent and that the distributions
of both return on capital and on labour do not
change with time. The agent’s goal is to maximize
the expected discounted utility of his consumption
stream. The literature on this topic usually
postulates that the agent’s utility function is
separable in time and that u(c), the utility of
consuming c units of capital in a given period, does
not change with time and is strictly concave and
strictly increasing. Denoting the one-period
discount factor by [1 > 0, £",r 1 «(<?,-) is the utility
associated with the consumption stream c 1; c2, ..., cN.
Setting the agent’s initial endowment at wu the
problem specification is complete.
The state of the system or process is the agent’s
wealth, and the set of possible states, called the
3159
state space, is the nonnegative numbers. (Although
typically the state is a real number or a vector of
real numbers, occasionally the state is a more
complicated object such as a probability measure.)
The points in time when decisions must be made
divide the process into stages. In this example, each
period is a stage. The action or decision at each
stage is how much to consume. Given a wealth of
w, i.e., the agent finds himself in state w, the agent’s
consumption level c must satisfy the constraint 0 <
c < w. The agent’s action space is [0, w] and
reflects the fact that his consumption cannot be
negative and is not permitted to exceed his current
wealth. (Thus, borrowing against future income is
prohibited. )
In our consumption example the objective
function or overall return (the functional being
maximized, here f [>' 1 «(<;,-)) is additively
separable in the consumption levels c\, c2,..., cN\ the
change in the overall return associated with a
change in c,- to c, is ff 1 [w(cj) — «(<?,-)] and does
not depend upon any of the other consumption
levels. (Of course, altering c, has an impact upon
the future wealth levels.) Consequently, we can
speak of the one-period reward function u. Given
the current state (wealth level) w, the system passes
(the state is transformed) to a new state T(w, c) in
response to the action (consumption decision) c
selected. The new state T(w, c) is simply the state of
the system at the beginning of the next stage or
period. Thus \v]+ x = T(wj, c) where Wj is the state of
the system in period j. In our example, the law of
motion or transition function is
T(w,c) = (w-c)R + L. (1)
It reflects the facts that labour income is unaf-
fected by either consumption or the return on
investment and that the investment has constant
returns to scale.
It is often more convenient to label time back-
wards and to speak of the number of stages
remaining. Accordingly, define V„(w) to be the
maximum expected discounted utility obtainable
when n stages remain and the current wealth is w, n
= 1,2,... and w > 0; V„ is called the -period
3160 Dynamic Programming and Markov Decision Processes

return function. Our immediate goal is to write a f u n c t i o n a l Vi(vr) = n(vr), all w>0 (2)
e q ua t i o n or recursive formula relating

V to V/ [ —i .
n
To obtain an expression for V2 in terms of V\,
In order to obtain our recursive relation, we note that the total return when two stages remain
shall employ implicitly Bellman’s famous Principle consists of the immediate reward u(c) from the first
of Optimality. Bellman stated it thus: stage plus the discounted return from the second
An optimal policy has the property that whatever the stage. At the second stage the agent will have an
initial state and initial decision are, the remaining amount (w — c)R + L of wealth to allocate between
decisions must constitute an optimal policy with consumption and saving; clearly, it must be
regard to the state resulting from the first decision.
(1, p. 83) allocated in the best possible manner - in this
instance all of it is consumed - in order to obtain an
Our use of the Principle of Optimality will
optimal two stage allocation. Thus, given an initial
become clear as the development of the function
consumption of c, an additional return of fl V] [(vv
equation (3 ) unfolds. To facilitate the connection
— c) R + L\ is garnered if consumption in the final
between the functional equation and Bellman’s
stage is chosen optimally. Therefore, the total
Principle of Optimality one last piece of terminol-
expected return for the two stage process when c is
ogy requires introduction. A policy is a rule which
the consumption in the first of the two stages is
specifies the decisions to be made as the system
simply u(c) - flElfKw - c) R + L], where A' denotes
passes through the various states. Of course the
the expectation with respect to the random variables
decisions specified must be feasible in that they
in the state description. Finally, by selecting the best
satisfy the system’s constraints: each action selected
consumption level in the first of the two stages wc
must lie in the action space associated with the
obtain the desired relationship between V\ and V2:
given state. While the action specified by the policy
for state s at time n can be a (random)
function of the history (suaus2, a2..............v„ |, an |,
,v„) of the system up to time n where .v, was the Iffv) = max {;/(<.)-I f$EV\ [(w — c)R I A]},
state at time i and at the action selected at time i, it
is usually the case that attention can be restricted to all w > 0.
policies under which the action specified for time n (3a)
depends upon the history of the system only through
the state of the system at time n; such a policy is Utilizing the same logic, the return function V2
referred to as a Markov policy. It may be helpfi.il to can be employed to compute V3 and, more
think of a (Markov) policy as a contingency plan generally, l’„ i can be used to compute V„ as
specifying the action to be selected if a given state is follows:
reached at a given stage rather than as a schedule of
N actions that will occur. A policy is said to be an Vn(w) max {K(M’) I- fiEVn-\ [(w — c)R + A]}.
optimal policy if the return associated with using it 0<c<w

equals the maximal renim attainable. n = 2,3, . . ., A, all w > 0.

(3b)

For each n and u; define c„(w) to be the largest

value of c for which the maximum in (3) is attained.
Then the policy which consumes the amount c„(w)
Developing the Functional Equation
when in state w with n stages remaining is an
To begin, recall that the agent’s utility function u is optimal policy. When the optimal return exists, it is
nondecreasing, whence lie consumes all his wealth unique, but it is often the case that there is more
in period TV when one stage remains: than one optimal policy.
Dynamic Programming and Markov Decision Processes 3161

Structure of the Optimal Policy using induction it can be shown the return function
is strictly concave and strictly increasing like u.
As noted earlier, the problem is not solved until the Doing so in this instance is a bit more difficult than
structure of an optimal policy is exhibited. Our usual.
analysis of the consumption problem is typical and
illustrative of many analyses of finite planning Lemma 2 The return function V„ is strictly
horizon (N < oo) dynamic programming problems; increasing and strictly concave. Consequently, there
in particular, mathematical induction is critical to is a unique optimal policy: c„(w) is the unique
the analysis. optimal level of consumption.
To ensure a positive level of consumption each
period, assume «'(0) = oo and L > e > 0 (so labour Proof By ( ) V\ is trivially seen to be strictly
income is bounded away from zero whence V„(w) increasing. Noting that c (w + 5) need not equal
n
> — oo for w > 0 as Eu(L) > M(e) > oo). The first c„(w) + (5 for <5 > 0, we have
condition provides the incentive to consume and the
second provides the capital. If R and/or fj is small, V»{w + 8) > u[c (w) + 5 ]
n
the agent’s optimal policy may entail the comer
+/?£V„_i {[w - cn(w)]R + L}
solution of consuming all of his wealth in a given
period. Assume R and [1 are sufficiently large - e.g., > MMh’)] + J8£V„_I{[W - c„(w)]K + L)
= V„{w)
ii'(e) < IlE(R) Eii' (E) - to ensure c„(w) < w for n >
1. so V„ is strictly increasing.
It is our intention to illustrate common analyt- Strict concavity is proved by induction. Clearly
ical approaches as well as lay bare the structure of E, is strictly concave. Assume V„_i is strictly
the optimal policy. In so doing the nature of the concave. It is easy to demonstrate (see Heyman and
return function V„ is also characterized. Sobel , p. 535) thatthe function J(w, c) = Vn_i[(w-
One technique often employed is that of com- c)r — /] is jointly concave in w and c on the convex
puting the return of a suboptimal policy which set C = {(TV, c) : 0 < c < w, w > 0}. (The concavity is
mimics the actions of another (perhaps optimal) strict ifr > 0.) As the sum of concave functions is
policy. Mimicking is the method of proof of the itself concave, EVn_i[(\v-c)R + L\ is strictly concave
following minor result, which we facetiously on C as is J(w, c) = u(c) + fiEVn-1 [(vr — c)R +
interpret as ‘life is worth living’. L] . While the maximum of a set of concave
functions need not be concave, a standard result
Lemma l I f Eu(L) > 0 , t h e n Vn+\ > Vn.
(ibid, p. 525) reveals that V,„ the maximum of the
Proof Let the consumption nfw) dictated by policy jointly strictly concave function J, is strictly
7i when i stages remain and the current wealth is w concave on C. This completes the induction argu-
be specified as follows: nfw) w and re, , (u) for i > 1. ment. Strict concavity of ./(vr, c) ensures unique-
Thus, when i + 1 stages remain, k acts like an ness. Q.E.D.
optimal policy if there were but i stages remaining. The decreasing marginal utility of wealth readily
Consequently, it yields a return of V„(w) + [>" Eli implies that optimal consumption increases with,
(Z.) > (w). The result now follows as the optimal but not as quickly as, wealth.
return V„+1(w) is at least as large as the return from
using n for n + 1 stages. Q.E.D. Lemma 3 The optimal level of consumption cfw)
The return function often inherits properties of satisfies
the oneperiod reward function. For example,
0 < cn{w + A) — cn{w) < A for A > 0. (4)
3162 Dynamic Programming and Markov Decision Processes

Proof For ease in presentation only, assume that u" K(w) = n> \ . (6)
2 /
exists so that V"n_, and d £V „_1 [(vc — c) R +
L]/dc2both exist. Differentiating the first order From ( ) and the guarantee of an interior solu-
condition u'[cn{w)\ — fiE{RV'n_x [(vv — cn(w)) R + tion for n > 1 we have cqfvv) = w > c2(w), whereas
L]} = 0 with respect to w yields ( ) and u strictly decreasing yield Vi '(vv) =
n'[ci(w)] < m'[c2(m’)] = Vfiw). Assume
c'n(w)u"[cn{w)} - [ 1 - c'n(w)]
C«(H’) > C„+1(H>), all w>0 (7a)
x E(R V^_l{[w - cn(w)}R + L})
2

= 0, (5)
V'n(w) <V'n+1(w), all w>0.
Asu" < 0,R2 > 0, andV'" , < 0,cn < 0would
violate ( ). Similarly, c'n> 1 violates ( ). Q.E.D.
In the context of a teenager’s lament and the (7b)
associated parental response, the two intuitive
inequalities in ( ) have the interpretations ‘What is -~pEVn[{w-c)R + L} =
ac
money for if not to spend?’ and, ‘Don’t let it bum a -pE{RV'n[{W-c)R + L]} >
hole in your pocket! ’
The demonstration that the overall return -m ( W - c ) R + L = T(SEV„+l [ ( W - c)R + L}.

increases with the time remaining offered in the

proof of Lemma was straightforward. Verifying that
from which we obtain immediately
the marginal utility of wealth declines with the
agent’s age, whence consumption increases with
Cn+i(w) > C„+2(M>). (8a)
age, entails the application of a frequently
employed induction technique we call Now ( ), ( ), and u strictly decreasing yield V'n+\
bootstrapping induction.

Lemma 4 The marginal utility of wealth decreases
and the optimal level of consumption increases with
age:
C n ( w ) > C„+i(w)
and
v» < K+iM-
Proof For ease in presentation assume only that u"
exists. Comer solutions have been eliminated: the
assumption u(s:) < fiE(R)Eu(L) ensures cn (w) < w
for n > 1 whereas u(0) = oo and L > s ensures
c„(w) >0 for n > 1. Therefore, the marginal benefit
u\cn(wj\ of immediate consumption equals the
marginal benefit fjE(RVf |{[w — cn(wj\R + L)) of
savings for n > 1. Consequently, regardless of the
percentage of marginal increase in wealth the agent
allocates to savings, we find
{W) = U'\cn+X{w)\ < U'[cn+2{w)\
= Vn+2(w).
(8b)
Having established (8), the induction argument
is complete. Q.E.D.
The impact on consumption of increased
uncertainty (in the sense of second order stochastic
dominance) in capital income R or labour income L
has been a focal point of the literature which
models the agent’s allocation between immediate
consumption and saving. Will the prospect of either
uncertainty vis-a-vis certainty or increased
uncertainty induce the agent to increase his
immediate consumption as a hedge against the
(increasingly) uncertain future in which nature
herself may, in effect, consume his wealth, or will
the agent decrease his immediate consumption in an
attempt to provide against an adverse future? The
former strategy adopts a ‘get while the getting is
good’ philosophy while the latter evokes one of
‘save for a rainy day’.
Dynamic Programming and Markov Decision Processes
The best response to an increase in uncertainty
depends on the shape ofthe utility function as well
as the source of the uncertainty. On this account, the
family [«,.} of utility functions with constant
relative risk aversion plays an important role:
UQ(C) = I n c a n d uy(c) =
cVy, f o r y < 1 , y ^ 0 .
When the utility function has a positive third
derivative and there is pure income risk (i.e. R is a
constant), consumption decreases in the face of an
increase in risk (see Miller ). For y < 1, u > 0. When
there is pure capital risk (i.e. L is a constant ) and
the utility function exhibits constant relative risk
aversion, an increase in risk causes consumption to
decrease if y < 0 and increase if y > 0 (see Phelps or
Mirman ).
Of course the dynamic programming approach
can be gainfully employed to address other inter-
esting questions such as the conditions which imply
capital will (on average) accumulate, whether
V„ converges, and the impact of an uncertain
lifetime. The second question is a recurring one in
dynamic programming models; in this instance, Vhr
converges as /V > oo provided (jER < 1. Ifp, is the
probability the agent lives i or more years and PN+i
= 0, then the return function V„ satisfies ( ) and (3)
with «(c) replaced by Pn-„+i W(C) and an increase
in consumption is the agent’s response to an
increase in the risk associated with his own
longevity when there is neither income nor capital
risk and u = uv for u = uy for y < i (see Levhari and
Mirman ).
Markov Decision Processes
The consumption model was considered in unctuous
detail for several reasons. It is intrinsically
interesting to economists, it is relatively simple to
describe, and its risk structure can be ascertained
without undue effort. In addition, the analytical
approach and the techniques employed as well as
the formulation of the functional equation are Bibliography
standard fare in dynamic programming models. The Bellman, R. 1957. Dynamic programming. Princeton:
most important reason, however, emanates Princeton University Press.
3163
from the fact that it is an example of the seemingly
ubiquitous Markov Decision Process (MDP).
A discrete time MDP is a process that is
observed at time points 0, 1, 2, ...; the k th
observation finds the process to be in some states sk
C S. When in state 5 at time n, an action a C As is
chosen. As a result of this action a reward r(.v, a) is
received and the next state of the process is deter-
mined according to the transition probability of a
stationary Markov process. The objective is to
maximize the sum of the expected discounted
rewards. Thus, if S is indexed by the non-negative
integers, the optimal return function V for this infi-
nite stage process can be shown (see Ross for the
standard proof when r is bounded and Lippman for
appropriate conditions on r and P when r is
unbounded) to satisfy the functional equation
O',a) + I^PMVQ) |, (9) j =
r
0,1,2,...
where /’,,(«) is the conditional probability that the
process will be in stage j at time n + 1 given that it
was in state i at time n and action a was selected.
The theory of MDP, including the roles played
by successive approximation and policy iteration, is
rather extensive, though not by comparison with its
host of applications. Excellent modem treatments
emphasizing theory and computation, respectively,
are given in Heyman and Sobel ( )
and Ross ( ) and in Denardo ( ).
Bellman’s original book ( ) on dynamic pro
gramming remains a very worthwhile read as does
Howard’s book ( ) on MDP.
See Also
Blackwell, D. 1962. Discrete dynamic programming. Annals
of Mathematics and Statistics 33: 719-726.
3164
Blackwell, D. 1965. Discounted dynamic programming.
Annals of Mathematics and Statistics 36: 226-235.
Denardo, E.V. 1967. Contraction mappings in the theory
underlying dynamic programming. SIAM Review 9: 165-
177."
Denardo, E.V. 1982. Dynamic programming. Englewood
Cliffs: Prentice-Hali.
Hakansson, N.H. 1970. Optimal investment and consumption
strategies under risk fora class of utility functions.
Econometrica 38: 587-607.
Heyman, It., and M. Sobel. ll>84. Stochastic models in
operations research, vol. II. New York: McGraw- Hill.
Howard, R.A. 1960. Dynamic programming and Markov
processes. Cambridge, MA: MIT Press.
Dynamic Programming and Markov Decision Processes
Levhari, D., and LJ. Mirman. 1977. Savings and uncertainty
with an uncertain horizon. Journal of Political Economy
85: 265-281.
Lippman, S. 1975. On dynamic programming with un-
bounded rewards. Management Science 21: 1225 1233.
Miller, B.L. 1974. Optimal consumption with a stochastic
income stream. Econometrica 42: 253 266.
Miller, B.L. 1976. The effect on optimal consumption of
increased uncertainty in labor income in the multi period
case. Journal of Economic Theory 13: 154 167.
Mirman, LJ. 1971. Uncertainty and optimal consumption
decisions. Econometrica 39: 179-185.
Phelps, E.S. 1962. The accumulation of risky capital: A
sequential utility analysis. Econometrica 30: 729-743.
Ross, S. Il)83. Introduction to stochastic dynamic pro-
gramming. New York: Academic Press.
E

Easterlin Hypothesis JEL Classifications

Diane J. Macunovich and Richard A. Easterlin
Abstract
The economic and social fortunes of a birth
cohort tend to vary as a function of that cohort’s
relative size, approximated by the crude birth
rate surrounding the cohort’s birth.
Effects have been observed on young men’s
earnings and unemployment rates, college
enrolment rates, marriage and divorce, fertility,
crime, and suicide rates. These effects have been
found to be asymmetrical about the peak of a
baby boom, and the original hypothesis has been
extended to suggest a wide range of effects on
the economy as a whole, from GDP growth rate,
through interest rates and stock market
performance, to measures of productivity.
Keywords
Aggregate demand; Cohort size effects;
Crowding; Demographic transition; Easterlin
hypothesis; Female labour force participation; 1. Direct or first-order effects of relative cohort
Fertility; Inflation; Interest rates; Life cycle size on male relative income (the earnings of
models; Marriage and divorce; Saving rates;
Relative income; Relative cohort size; Produc-
tivity; GDP growth; Unemployment rates; Col-
lege enrolment rates; Material aspirations;
Preferences; Crime rates; Suicide rates
Jll
The Easterlin, or ‘relative cohort size’, hypothesis
as originally formulated posits that, other things
constant, the economic and social fortunes of a
cohort (those bom in a given year) tend to vary as a
function of its relative size, approximated by the
crude birth rate surrounding the cohort’s birth
(Easterlin ). This hypothesis has since been
extended to suggest a wider range of effects on the
economy as a whole (Macunovich ).
Although cohort size effects were originally
expected to be symmetrical around the peak of the
baby boom, which in the United States entered the
labour market around 1980, it is now thought that
they are tempered by aggregate demand effects and
by feedback effects from adjustments made by
young adults on the ‘leading edge’ of a baby boom.
As a result, cohorts - and the economy generally -
on the ‘leading edge of a baby boom fare much
better than those on the ‘trailing edge’, when all else
is equal.
The ultimate effects of changing relative cohort
size are hypothesized to fall into these three
categories:
young men relative to their aspirations); male
unemployment and hours worked; men’s and
women’s college wage premium (the extra
earnings of a college graduate relative to

(C> Macmillan Publishers Ltd 2018

Macmillan Publishers Ltd (ed.), The New Palgrave Dictionary of Economics,
1 ■ c //-.. o: g/10.1057/;78-:i.-h' 1 - 1 i S S - l
3166
those of a secondary school graduate); and
levels of income inequality generally.
2. Second-order effects operating through male
relative income, especially the demographic
adjustments people make in response to chang-
ing relative income, such as changes in
women’s labour force participation and their
occupational choices; men’s and women’s col-
lege enrolment rates; marriage and divorce;
fertility; crime, drag use, and suicide rates; out-
of-wedlock childbearing and the incidence of
female-headed families; and living
arrangements.
3. Third-order effects on the economy of changing
relative cohort size and the resulting demo-
graphic adjustments, such as changes in average
wage growth; the overall demand for goods and
services in the economy and hence the growth
rate of the economy; inflation, interest rates, and
savings rates; stock market performance;
industrial structure; measures of gross domestic
product (GDP); and productivity measures.
The three categories of effect are discussed first
in this article, followed by a consideration of
feedback effects and a discussion of empirical
evidence.
First-Order Effects
The linkage between higher birth rates and adverse
social and economic effects arises from ‘crowding
mechanisms’ operating within three major social
institutions, the family, school and the labour
market. Within the family, a sustained upsurge in
the birth rate is likely to entail an increase in the
average number of siblings, higher average birth
order, and a shorter average birth interval, and there
is a substantial literature in psychology, sociology
and economics linking child development
negatively to one or more of these magnitudes
(Ernst and Angst ; Heer ). The negative effects that
have been investigated range over a wide variety of
phenomena. With regard to mental health, for
example, there is evidence that problem behaviours
such as
Easterlin Hypothesis
fighting, breaking rales, and delinquency are asso-
ciated with increased family size. Adverse effects
on morbidity and mortality of children have been
found to be associated with increased family size
and shorter birth spacing. A negative association
between IQ and number of siblings has been found
in a number of studies, and, with IQ controlled for,
between educational attainment and family size.
The principal mechanism underlying such
developments is likely to be the dilution of parental
time and energy per child and family economic
resources per child, associated with increased
family size.
The family mechanisms just discussed imply
that, on average, a larger cohort is likely to perform
less well in school. But even in the absence of any
adverse effects within the family, a large cohort is
likely to experience crowding in schools, which
reduces average educational performance
(Freeman ). At any given time the human and
physical capital stock comprising the school system
tends to be either fixed in amount or to expand at a
fairly constant rate, so that a surge in entrants into
the school system tends to be accompanied by a
reduction in physical facilities and teachers per
student. In the United States, school planning
decisions are divided among numerous local gov-
ernments and private institutions, and expansion has
tended to occur in reaction to, rather than in
anticipation of, a large cohort’s entry. Moreover,
even when expansion occurs it is usually not
accompanied by maintenance of curriculum stan-
dards, partly because of the diminishing pool of
qualified teachers available to supply the needs of
educational expansion.
The experience of a large cohort both in the
family and in school is likely, in turn, to leave the
cohort less well prepared, on reaching adulthood,
for success in the labour market. But even if there
were no prior effects, the entry of a large proportion
of young and relatively inexperienced workers into
the labour market creates a new set of crowding
phenomena, because the expansion of
complementary factor inputs is unlikely to be
commensurate with that of the youth labour force.
Additions to physical capital stock tend to be
dominated by considerations other than the relative
supply of younger workers, and the growth in
Easterlin Hypothesis
older, experienced, workers is largely governed by
prior demographic conditions. Growth in the
relative supply of younger workers results, in
consequence, in a deterioration of their relative
wage rates, unemployment conditions and upward
job mobility (Welch ). The adverse effects of labour
market crowding tend to reinforce those of
crowding within the school and family. For
example, the deterioration in relative wage rates of
the young translates into lower returns to education
and consequent adverse impact on school drop-out
rates and college enrolment (Freeman ). Also,
problems encountered in finding a good job may
reinforce feelings of inadequacy or frustration
already stirred up by some prior experiences at
home or in school, and lead to lower labour force
participation among young men.
Second-Order Effects
The relative economic standing of successive gen-
erations at a given point in time may be altered
systematically by fluctuations in relative cohort
size. If parents’ living levels play an important role
in setting their children’s material aspirations, as
socialization theory leads one to believe, then an
increase in the shortfall of children’s wage rates
relative to parents will cause the children to feel
relatively deprived and under greater pressure to
keep up. The importance of relative status influ-
ences of this type in affecting attitudes or behaviour
has been widely recognized in social science theory
(Duesenberry ).
Confronted with the prospect of a deterioration
in its living level relative to that of its parents, a
large young adult cohort may make a number of
adaptations in an attempt to preserve its compar-
ative standing. Foremost among these are changes
in behaviour related to family formation and family
life (Macunovich and Easterlin ; McNown and
Rajbhandary ). To avoid the financial pressures
associated with family responsibilities, marriage
may be deferred. If marriage occurs, wives are more
likely to work and to put off childbearing. If a wife
bears children, she is more likely to couple labour
force participation
3167
with childrearing, and to have a smaller number of
children more widely spaced (Macunovich ; Jeon
and Shields ).
The process of demographic adjustment to
changing relative income can best be thought of in
terms of ex ante and ex post income; that is, the
disposable per capita income of individuals prior to
and then following the adjustments. Analyses of
baby boom cohorts in the United States have found
that a cohort’s male relative income - individual
earning potential of baby boomers relative to that of
their parents - was significantly lower than the
individual earning potential of pre-boom cohorts
relative to their parents. But after making the type
of demographic adjustments indicated above, the
boomers managed to bring their per capita
disposable income on a par with that of their parents
(Easterlin et al. ).
Other reactions to the psychological stresses
induced by large cohort size may be viewed as
socially dysfunctional. Feelings of inadequacy and
frustration, for example, may lead to dispro-
portionate consumption of alcohol and drugs, to
mental depression, and, at the extreme, to a higher
rate of suicide (Pampel ; Stockard and O’Brien ).
Feelings of bitterness, disappointment and rage may
induce a higher incidence of crime (O’Brien et al. ).
Within marriage, the stresses of conflicting work
and motherhood roles for women, and feelings of
inadequacy as a breadwinner for men, are likely to
result in a higher incidence of divorce
(Macunovich ). In the political sphere, the
disaffection felt by a large cohort because of its lack
of success may make it more responsive to the
appeals of those who are politically alienated
(O’Brien and Gwartney- Gibbs ).
Third-Order Effects
The second-order effects described in the previous
section will, through reduced marriage rates and
increased divorce and female labour force partic-
ipation rates, reduce the proportion of households
with stay-at-home spouses, which increases the
tendency to purchase market replacements for the
goods and services traditionally produced by
3168
women in the home. The result is a ‘commoditi-
zation’ of many goods and services that used to be
produced in the home. They are now exchanged in
the market - and thus counted in official measures
of GDP and productivity - whereas previously they
were part of the excluded ‘non-market’ economy.
This commoditization of goods and services
causes measures of industrial structure to skew
strongly toward services and retail, away from
agriculture and manufacturing, creating low- wage
service jobs. In addition, the influx of inexperienced
young workers as members of a large birth cohort -
both men and women - into the labour market
exacerbates any decline in productivity growth by
changing the composition of the workforce to one
dominated by inexperienced and therefore lower-
productivity workers. This decline in relative wages
of younger workers resulting from their oversupply
would lead employers to substitute cheaper labour
for more expensive capital, thus lowering the young
workers’ productivity still further by providing
those low-wage workers with less productivity-
enhancing machinery and technology.
Although some analysts maintain that the
potential age structure effect of the baby boomers
on personal savings is not large enough to explain
the full drop in US national savings rates since the
1980s, studies of this phenomenon to date have
focused only on the behaviour of the baby boomers
themselves. However, one might argue that the
baby boomers have affected the propensity to save
in age groups other than their own. For example,
because boomers’ earnings were depressed and they
experienced an inflated housing market when they
went to buy homes (both the effects of their own
large cohort size), many parents of baby boomers
drew on their own savings in order to help with
down payments.
When the age structure of children is permitted
to affect consumption and savings, a very strong
age-related pattern of expenditures and saving can
be identified. Children induce savings on the part of
their parents between the ages of five and 16,
possibly in anticipation of later educational
expenses. When the relationships identified in this
way are combined with the changing age
Easterlin Hypothesis
distribution in the US population during the 20th
century, they produce a savings rate that fluctuates
by plus or minus 25 per cent around the mean,
simply as a result of changing age structure
(Macunovich ).
Similarly, a strong effect has been identified of
changing age structure (measured simply as the
proportion of young to old in the population) on real
interest rates and inflation, because of differential
patterns of savings and consumption with age
(McMillan and Baesel ). A higher proportion of
young adults in a population will produce lower
aggregate savings levels - and hence higher interest
rates. In this model, today’s lower interest and
inflation rates are the result of the ageing of the
baby boomers, as they begin to acquire assets for
their retirement years. The converse of this
phenomenon - the potential ‘meltdown’ effect of a
retiring baby boom on financial markets, asset
values and interest rates - has been described as
well (Schieber and Shoven ).
Some research has estimated a strong effect of
age structure on housing prices in the United States,
with the entry of the baby boom into the housing
market causing the severe house price inflation of
the 1970s and 1980s, and the entry of the baby bust
causing house price deflation (Mankiw and Weil ).
Although some have disputed the magnitude of the
effect estimated there, most researchers have
confirmed its existence. A later study, for example,
found significant effects of detailed (single year)
age structure in the adult population on all forms of
consumption, including housing demand, and on
money demand (Fair and Dominguez ).
These potential effects on aggregate demand,
savings rates, interest rates and inflation suggest
that there might have been a connection between
changing age structure and macroeconomic fluc-
tuations in the United States and elsewhere during
the 20th century. When the population of young
adults is expanding, the resultant growth in demand
for durable goods creates confidence in investors,
while an unexpected slowdown in the growth rate of
young adults could cause cutbacks in production
and investment in response to inventory buildups,
with a snowball effect throughout the economy.
There was a close
Easterlin Hypothesis
correspondence in the United States in the 20th
century between ‘turnaround points’ of growth in
the key age group of 15-24, and significant eco-
nomic dislocations in 1908,1929,1938 and 1974.
Similarly, there was a correlation between age
structure and economic performance in industri-
alized nations in the 1930s, and in both industri-
alized and developing nations since the 1980s, with
the ‘Asian Tigers’ some of the most recent
examples (Macunovich ).
Feedback Effects on the Relationship
Between Relative Cohort Size and Relative
Income
Easterlin’s original statements recognized the
potential effects of outside influences on the relative
cohort size mechanism (Easterlin ). Efowever, the
dynamic nature of the mechanism - the fact that
many of these other factors would, in fact, be
secondary and tertiary results of changing relative
cohort size, and thus endogenous in any empirical
application - has not been fully appreciated in most
analyses to date. As a result, it is often concluded
that the hypothesis may have been relevant in the
post-Second World War period up to about 1980,
but that it fails to extend beyond one full cycle to
apply to the period since 1980.
The aggregate demand effect of changing rel-
ative cohort size, discussed in the previous section,
is hypothesized to contribute significantly to the
observed asymmetry in relative cohort size effects
on male relative income. Although cohorts on the
leading edge of a baby boom experience declining
wages relative to those of older workers, they do so
in an economy experiencing strong growth in
aggregate demand resulting from the increasing
relative cohort size among young adults. Cohorts on
the lagging edge of a baby boom, however, enter a
labour market weakened by the economic slump
resulting from a transition from expanding to
contracting relative cohort size.
Similarly, as one of the secondary effects of
changing relative cohort size discussed earlier,
female labour force participation is hypothesized
3169
to have increased in response to declining male
relative income as the leading edge of the baby
boom entered the labour market. If, as hypothe-
sized, these young women also increased their
levels of educational attainment in anticipation of
future labour market participation, they would have
in many cases competed directly with the male
members of their cohort and exacerbated the effects
of relative cohort size on male relative income. This
effect would have been greatest for cohorts on the
lagging edge of the boom - those who should have
benefited from declining relative cohort size. It is
important in empirical analyses to recognize the
potential endogeneity of these other factors, rather
than treat relative cohort size effects as ‘contingent’
on exogenous changes in female labour force
participation, educational attainment and wages.
Wage analyses based on relative cohort size which
control for a cohort’s position in the US baby boom
- and thus allow for aggregate demand and female
labour force changes - can explain most of the
observed change in young men’s entry level wages
and in their returns to experience and education
(Macunovich ).
Empirical Analyses
Empirically, the most important application of the
hypothesis has been to explain the varying expe-
rience of young adults in the United States since the
Second World War. There is, however, some
evidence of its relevance to the experience of
developed countries more generally in this period
(Korenman and Neumark ; Pampel ; Stockard and
O’Brien ; Jeon and Shields ), and perhaps as a
mechanism leading to fertility decline during the
demographic transition in developing countries
(Macunovich ).
Overall, however, empirical analyses testing
various aspects of the Easterlin hypothesis have
produced fairly mixed results. By 2007 there have
been two comprehensive analyses of the literature
on the Easterlin hypothesis, and one meta-analysis
of 19 studies completed between 1976 and 2002.
The meta-analysis (Waldorf and Byun ) focused on
the age structure-fertility link, and concluded that
analytical problems contribute to
3170
an apparent lack of empirical support for the
Easterlin hypothesis. Most significant among these
were the failure to recognize the endogeneity of an
income variable when combined with a relative
cohort size variable, and the use of very broad age
groups in defining relative cohort size.
The first of the literature reviews considered a
broad range of topics, including labour market
experience and education; marriage, fertility and
divorce; and crime, suicide and alienation. It
concluded:
[T]he evidence for the Easterlin effect proves mixed
at best and plain wrong at worst... Aggregate data
support the hypothesis more than individual level
data, period-specific or time-series data support the
hypothesis more than cohort-specific data, experi-
ences from 1945-1980 support the hypothesis more
than the years since 1980, and trends in the United
States support the hypothesis more than trends in
European nations. (Pampel and Peters , p. 189.
The second literature review evaluated 76
published analyses focused solely on fertility, and
concluded:
With an equal number of micro- and macro-level
analyses using North American data (twenty-two),
the ‘track record’ of the hypothesis is the same in
both venues, with fifteen providing significant sup-
port in each case. The literature suggests unequivocal
support for the relativity of the income concept in
fertility but is less clear regarding the source! s) of
differences in material aspirations, and suggests that
the observed relationship between fertility and cohort
size has varied across countries and time periods due
to the effects of additional factors not included in
most models. (Macunovich , p. 53)
This review suggests that, because of data lim-
itations and idiosyncratic interpretations of the
hypothesis by individual researchers, many of the
studies with unfavourable findings have been only
peripherally related to the Easterlin hypothesis.
Conclusion
Since the early 1980s, demographic concepts have
encroached modestly on economic theory, as
evidenced by the appearance of life cycle,
overlapping generations and vintage models. The
cohort size hypothesis might be viewed as
Easterlin Hypothesis
another in this sequence. Its roots, however, extend
beyond economics, reaching out into sociology,
demography and psychology, and it seeks to
encompass a wider range of attitudinal and
behavioural phenomena than is traditionally con-
sidered economic.
See Also
Bibliography
Duesenberry, J.S. 1949. Income, saving, and the theory of
consumer behaviour. Cambridge, MA: Harvard Uni-
versity Press.
Easterlin, R.A. 1980. Birth andfortune, 1st ed. New York:
Basic Books.
Easterlin, R.A. 1987. Birth and fortune, 2nd ed. Chicago:
University of Chicago Press.
Easterlin, R.A., C. Macdonald, and D.J. Macunovich. 1990.
How have the American baby boomers fared? Earnings
and well-being of young adults 1964-1987. Journal of
Population Economics 3: 277-290.
Ernst, C., and J. Angst. 1983. Birth order: Its influence on
personality. Berlin: Springer.
Fair, R.C., and K. Dominguez. 1991. Effects of the changing
U.S. age distribution on macroeconomic equations.
American Economic Review 81: 1276-1294.
Freeman, R.B. 1976. The overeducated American. New York:
Academic Press.
Heer, D.M. 1985. Effects of sibling number on child outcome.
Annual Review of Sociology 11: 27-47.
Jeon, Y., and M.P. Shields. 2005. The Easterlin hypothesis in
the recent experience of higher-income OECD countries:
A panel-data approach. Journal of Population Economics
18: 1-13.
Korenman, S., and D. Neumark. 2000. Cohort crowding and
the youth labour market: A cross-national analysis. In
Youth employment and joblessness in advanced
countries, NBER comparative labour market series, ed.
D.G. Blanchflower and R.B. Freeman. Chicago:
University of Chicago Press.
Macunovich, D.J. 1998. Fertility and the Easterlin hypothesis:
An assessment of the literature. Journal of Population
Economics 11: 1-59.
Macunovich, D.J. 2002. Birth quake: The baby boom and its
after shocks. Chicago: University of Chicago Press.
Macunovich, D.J., and R.A. Easterlin. 1990. How parents
have coped: The effect of life cycle decisions on the
economic status of pre-school age children, 1964-1987.
Population and Development Review 16: 301-325.
East-West Economic Relations
Mankiw, N.G., and N.D. Weil. 1989. The baby boom, the
baby bust and the housing market Regional Science and
Economics 19: 235-258.
McMillan, H.M., and J.B. Baesel. 1990. The macroeconomic
impact of the baby boom generation. Journal of
Macroeconomics 12: 167-195.
McNown, R., and S. Rajbhandary. 2003. Time series analysis
of fertility and female labour market behaviour. Journal
of Population Economics 16: 501-523.
O’Brien, R.M., and RA. Gwartney-Gibbs. 1989. Relative
cohort size and political alienation: Three methodological
issues and a replication supporting the Easterlin
hypothesis. American Sociological Review 54: 476-480.
O’Brien, R.M., J. Stockard, and L. Isaacson. 1999. The
enduring effects of cohort characteristics on age- specific
homicide rates 1960-1995. American Journal of
Sociology’ 104: 1061-1095.
Pampel, F.C. 2001. The institutional context of population
change: Patterns of fertility and mortality across high-
income nations. Chicago: University of Chicago Press.
Pampel, F.C., and H.E. Peters. 1995. The Easterlin effect.
Annual Review of Sociology 21: 163.
Schieber, S.J., and J.B. Shoven. 1994. The consequences of
population aging on private pension fund saving and
asset markets, Working paper no. 4665. Cambridge, MA:
NBER.
Stockard, J., and R.M. O’Brien. 2002. Cohort effects on
suicide rates: International variations. American Socio-
logical Review 67: 854-872.
Waldorf, B., and P. Byun. 2005. Meta-analysis of the impact
of age structure on fertility. Journal of Population
Economics 18: 14-40.
Welch, F. 1979. Effects of cohort size on earnings: The baby
boom babies’ financial bust. Journal of Political
Economy 87: 65-97.
East-West Economic Relations
Marie Lavigne
The decade 1966-1975 is usually considered as the
golden age of East-West economic relations.
Already during the previous decade, i.e. since the
end of the cold war, the USSR and the Eastern
European countries had increased their trade with
the West at an annual rate of growth slightly higher
than their total trade. But after 1966 the expansion
of trade and cooperation was sustained both by a
favourable political climate and by strong economic
complementarities between the
3171
West (here equated to the OECD countries) and the
East (the USSR and the six European countries that
are members of the CMEA, or Council for Mutual
Economic Assistance; hereafter we shall mention
them as CPEs or centrally planned economies, for
the sake of brevity).
These years were marked by detente, initiated in
1966 with the triumphal visit to the USSR of the
French President General de Gaulle. This was not
only a bilateral event, but it set the stage for
diversified and institutionalized links between
Eastern and Western European economies. Later on,
in 1972, US President Nixon’s visit to Moscow
opened the shorter phase of bright USUSSR
economic relations which ended in 1975. At the
beginning of that year, the Soviet Union unilaterally
repudiated the Soviet-American treaty of
commerce, as a retaliation for the deprivation of the
most favoured nation clause; according to the
American legislation just introduced, the clause
could not be granted to a country restricting the
rights for its citizens to emigrate. Before detente
came altogether to its end, it was symbolically
magnified in the final Act of the Conference for
Security and Co-operation in Europe, signed in
Helsinki in August 1975. The economic ‘basket’ of
this text was meant to appear as the Charter of East-
West mutually profitable relations.
From the economic point of view, the 1966-
1975 decade was indeed a time of converging
interests. The USSR and Eastern European
countries had just engaged in economic reforms.
They needed to modernize their industries. The
Western firms found new markets for selling
equipment and turnkey plants. High rates of eco-
nomic growth, both in the West and in the East,
sustained the prospects for increased exports from
the East to the West, once the new capacities
acquired from the West were put into operation. An
era of deepening industrial cooperation, based upon
technology imports and reverse flows of
manufactured goods, seemed to open.
It was then almost forgotten that even in such a
favourable context, East-West trade accounted for
less then 3% of world trade. While in 1975 it
amounted to slightly under 30% of total trade for
the CPEs (slightly more for the USSR and less for
the six smaller CPEs taken together), it never
3172 East-West Economic Relations

exceeded 5% of total trade for the Western coun- important for the East than for the West. However,
tries, except for some non-typical cases (such as dependencies arc to be found on both sides, with an
Austria or Finland). uneven distribution.
The following decade, ending in 1985, has In the West, European countries arc the main
witnessed a general shrinking of East-West trade. group of partners. They account for roughly 75% of
There was a conspicuous deterioration of the sales to the East and 90% of imports from the East
political climate with the invasion of Afghanistan (figures of 1983). This pattern has been stable since
by the Soviet troops in December 1979 and, 2 years the end of the 1970s. In 1970 the share of Western
later, martial law in Poland. The world economic Europe was vciy similar on the import side, but
larger on the export side (about 10 points more).
crisis exerted some adverse effects as well. True, it
benefited the USSR as an oil exporter. But the Since then, two major exporters have emerged
Western recession hampered the export drive of the outside Europe, Japan (for technology) and the
smaller CPEs. The manufactured goods which they United States (for grain, mainly to the USSR).
intended to export so as to repay their imports of In the East, the USSR gained a growing share of
equipment became less saleable in the East. Thus East-West trade after 1970. From twofifths of the
the imbalance between imports and exports, which total trade of the European CPEs with the West, it
had been steadily growing since 1970, could not be reached 50% in the mid-1970s and over 60% in the
corrected through expanded sales. An easier way 1980s. This is mainly due to the increase in oil
out was to borrow on Western financial markets. prices after 1973; it allowed the Soviet Union to
The CPEs were still creditworthy, and the level of secure a higher rate of growth of its trade with the
international liquidity was high as a result of theWest compared with the other CPEs up to 1980, and
inflow of petro-dollars. The total indebtedness of to avoid the decrease in trade which the other CPEs
the CPEs culminated in 1981. The subsequent experienced in the beginning of the 1980s.
adjustments conducted in 1981-1983 (though a This growing concentration of East-West trade
decrease in imports and domestic investment) ended on the Soviet Union is an expression of stronger
up with a marked improvement in the CPEs external interdependences.
financial position and with a decrease in their For Western Europe, especially for the large
foreign debt (except for Poland). But the general industrial corporations, the USSR emerged in the
slowdown of growth in the East, partly due to these1970s as a major purchaser of heavy equipment,
adjustments, does not allow for a steep upward whose orders helped to sustain the level of activities
trend in East-West trade. and jobs during the recession years. The
The outlook for East-West economic relations iscontroversial multi-billion dollars gas pipeline deal
to be evaluated through the combination of two concluded in 1981 is a clear demonstration of such
opposed sets of factors. On the one hand, there areinterests. When in 1982 the US government tried to
strong interests on both sides pressing for the oppose the supply of tubes and other equipment for
expansion of trade and cooperation. On the other, the pipeline, as a retaliation for the Soviet role in
equally strong obstacles are hindering such a the Polish crisis, and also as an attempt to reduce
development. The outcome is probably to be seen in the export capacities of natural gas of the USSR, the
a stabilization of those relations, below the levelEuropean governments backed theh films. Even
reached during the 'golden age' decade. though the Soviet orders for equipment substantially
declined after then, the Soviet Union remains a
huge market.
Economic Interests On the other side, the Soviet Union has become
a significant supplier of energy to Western Europe.
East-West trade is sometimes said to be a one-way Fuels now account for about 80% of its sales to the
street. As the magnitudes of shares in total trade West, from about half that share in the
show, these relations are several times more
East-West Economic Relations
beginning of the 1970s. The major Western Euro-
pean energy importers (Germany, France, Italy) are
now dependent for 6-7% of their total energy
imports on the Soviet Union. For natural gas alone,
their dependence may be above the 30% mark at the
end of the 1980s, from about 15% to 20% a decade
earlier. The Soviet market is a means of achieving a
diversification in energy imports; it is a cheaper
supplier for oil and gas because of the distance
factor, and may be considered as a more reliable
one, than the Third World.
Regarding trade with the United States, the
major link is grain. The Soviet Union began to buy
large quantities of American grain in 1975-1976
and has been the largest single customer of the
United States since then. US sales never again
reached the 70% share of Soviet grain imports
which they formed in 1979. However, the strength
of economic versus political interests is clearly
demonstrated by the failure of the grain embargo,
which had to be lifted under the pressure of
American farmers. The long-term grain sales
agreement linking the two countries, first signed in
1975, has not only been renewed but also
supplemented with an anti-embargo clause (in
1983).
The Western trade of Eastern Europe lacks these
powerful interdependences. The smaller CPEs taken
together are on average less involved in trade with
the West than the USSR. In 1984, the share of
Western trade in their total trade was about 25%
(against 30% for the USSR) and had been declining
since 1980. But while Bulgaria and Czechoslovakia,
much more oriented toward trade within CMEA,
have a very low share of their total trade with the
West (12—15%), Hungary (35%), Poland, GDR
and Romania (30%) are potentially interested in
expanding their trade with the West. However,
opportunities for that are low. Their supply is made
of sensitive goods (steel, chemicals, textiles,
manufactured goods, agricultural products), the
demand for which is sluggish in the West - and they
complain of growing protectionism. For these goods
competition is growing on Western markets from
the new industrializing countries of the Third
World, which in addition are more advanced in
some high
3173
technology fields (electronics). They can hardly
expect concessions from Western countries, for
which they provide less promising markets than the
USSR. The development of compensation deals is
only a marginal way of securing outlets for their
goods.
Obstacles
In the background of these differentiated economic
interests, specific obstacles hinder East- West trade,
in the political, institutional (systemic) and financial
fields, to which must be added the 1986
developments on the world oil market.
Is East-West trade political in essence? In
Western Europe, politics and economic relations are
regarded as distinct by governments and firms. The
lasting failure to find an agreement between EEC
and the CMEA, since the beginning of official talks
in 1976, is mainly due to the lack of institutional
competence of the CMEA in matters of trade as
appraised by the EC Commission (even if on the
side of the Commission there is a political concern
to avoid strengthening the Soviet- dominated
CMEA as an organization). The major involvement
of politics in East-West economic relations is
related to US policy. The ‘linkage’ concept of tying
economic advantages to Soviet concessions in the
political sphere was associated in the 1970s with
commercial policies (the granting of the MFN
clause) or financial conditions (for access to bank
credits). Since the end of that decade it has evolved
into a policy of sanctions, first as a retaliation for
the Soviet invasion of Afghanistan in 1979 (the
grain embargo against the USSR, which was lifted
in April 1981, and a tighter control of high
technology sales); then as a response to the martial
law imposed in December 1981 in Poland. In this
last case the sanctions hit Poland (though credit and
export restrictions, a suspension of the MFN
clause), and the USSR (through attempts to stop the
Eurosiberian pipeline deal by preventing the
Western European countries from selling equipment
to the USSR and from concluding the agreements
for the purchases of gas). They were also
3174
extended to the other CPEs through a very severe
credit squeeze. All these measures culminated in
1982. They proved largely ineffective but generated
conflicts within the Western Alliance. The major
and lasting field of political pressure is to be found
in the embargo on high technology sales to the
CPEs, conducted through the Cocom (Coordinating
Committee), an informal organization set up in
1949 and including the NATO countries plus Japan.
Very active during the years of the cold war, it
seemed to be withering in the late 1970s but
regained momentum from 1980 on. The present
rationale of the Cocom restrictions is threefold: to
impose sanctions; to prevent the Soviet bloc from
acquiring dual-use technologies (for military as well
as civilian ends); to enlarge the scope of controls by
restricting high- technology exports of non-Cocom
members (Sweden, Switzerland, Austria, and even
some Third World countries such as India).
The systemic obstacles in trade are related to the
specific organization of state trading in the CPEs.
The monopoly of foreign trade and the related
planning of trade flows remain very rigid in the
Soviet Union. Increased flexibility has been
introduced in the trade mechanisms of all the other
CPEs, where enterprises are gaining easier access to
foreign trade transactions. Direct interfirm contacts
have been stimulated through industrial
cooperation. In all these countries except for GDR,
it is now possible to create joint enterprises with
foreign equity capital (the experiences remain
limited). The state trading system, however
reformed, still prevents the CPEs from successfully
adjusting to the market requirements in the West.
The financial problems of East-West relations
are less dramatic than in 1980-1981, when the total
indebtedness of the USSR and Eastern Europe
combined exceeded $80 billion, more than four
times its level of the end of 1974. Two countries,
Poland and Romania, entered in 1981 a process of
rescheduling, which is still going on for Poland.
Two others, GDR and Hungary, successfully
managed to restore their external accounts in 1982-
1984. Since then, the Western banks have again
been ready to expand their loans not only to the
Soviet Union, which has always remained a
East-West Economic Relations
good risk, but also to the other CPEs, which by all
accounts seem more creditworthy than the Third
World.
East-West economic relations are finally to be
replaced in the broader context of the CPEs’ foreign
economic relations, including infia-CMEA trade.
The move toward closer integration, advocated by
the Soviet Union at the Summit meeting of the
CMEA in June 1984 and based upon the heavy
requirements of the USSR regarding its imports
from its partners, might well appear as an additional
constraint to the expansion of East- West relations
for the smaller CPEs.
The fall in oil prices, since the end of 1985, may
have strong adverse effects on East-West trade. If
the average price for oil is for some time stabilized
at half its 1985 level, the Soviet Union will lose at
least one third of its export gains in its trade with
the West. These losses may be compensated for, in
the short ran, by cuts in imports and increased
borrowing, together with a stronger pressure on the
smaller CMEA countries. The latter will thus have
to divert to the Soviet market goods exportable to
the West. In addition, they too will lose as sellers of
refined oil products, with the same consequences as
for the USSR. The ‘golden age’ of East-West trade
is definitely not to be renewed.
See Also
► Convergence - /p ott1
► Cycles in So 3-;onomie
► Socialist Economic:
References
Bomstein, M., Z. Gilclman, and W. Zimmerman (eds.). 1981.
East-west relations and the future of eastern Europe:
Politics and economics. London: Allen and Unwin.
Economic Bulletin for Europe. 1949 onwards. Geneva:
Economic Commission for Europe, United Nations. Each
volume contains developments on East-West trade.
Fallenbuchl, Z., and C. McMillan (eds.). 1980. Partners in
east-west relations, the determinants of choice. New
York: Pergamon Press.
Eckstein, Otto (1927-1984)
Holzman, F. 1976. International trade under communism,
politics and economics. New York: Basic Books.
Lavigne, M. 1979. Les relations economiques Est-Ouest.
Paris: Presses Universitaires de France.
Lavigne, M. 1985. Economie Internationale des pays
socialistes. Paris: Armand Colin.
Levcik, F. (ed.). 1978. International economics -
Comparisons and interdependencies. Essays in honour of
F. Nemschak. Vienna: Springer.
Marer, P., and J.M. Montias (eds.). 1980. East European
integration and east-west trade. Bloomington: Indiana
University Press.
Eckstein, Otto (1927-1984)
Lester C. Thurow
Keywords
Eckstein, O.; Forecasting; Macroeconometric
models
JEL Classifications
B31
Eckstein was an entrepreneur who moved a whole
technology from the research community into the
marketplace. Until he founded Data Resources, Inc.,
macroeconometric models were research vehicles
and not vehicles for aiding business decision
making. Under his direction Data Resources came
to dominate the marketplace for this type of
information, but more importantly it changed the
nature of the game. To be taken seriously after his
innovation, all economic forecasts had to be but-
tressed with econometric equations and no large
firm would attempt to begin its decision-making
processes without an understanding of the national
and international economic forecasts emanating
from such models.
Bom in Ulm, Germany, in 1927, Dr. Eckstein
fled to England in 1938 and came to the United
States in 1939. He graduated from Stuyvesant High
School in New York City and served in the United
States Army Signal Corps from 1946 to 1947. He
received an AB degree from Princeton
3175
University in 1951 and a Ph.D. from Harvard
University in 1955.
In 1968, he and Donald B. Marron founded Data
Resources, Inc., which has grown into the largest
economic information company in the world. The
firm became a subsidiary of McGraw-Hill, Inc. in
1979. He directed the development of the Data
Resources Model of the US economy, and was
responsible for its forecasting operations.
As an immigrant to the United States from Nazi
Germany, Otto Eckstein wanted to contribute
something to America’s future success. Better
economic policies that would lead to a higher
American standard of living were not an abstraction
to him. They were the centre of his professional life.
His professional career began with the analysis
of large scale multi-year water resources projects
and how one might better allocate national
resources in such projects. In the late 1950s he was
the principal intellectual director of a Joint
Economic Committee study on how the United
States might break out of what was then seen as the
stagnation of the mid-1950s. His study on growth,
full employment and price stability laid the basis for
the successful economic policies that were followed
in the first two-thirds of the 1960s. But he went on
to implement those intellectual foundations as a
member of the President’s Council of Economic
Advisers under President Johnson.
No one who knew the enthusiasm of Otto
Eckstein for studying, teaching, and practising
economics could thereafter think of economics as
the dismal science.
Selected Works
1958a. Water resources development: The eco-
nomics of project evaluation. Cambridge, MA:
Harvard University Press.
1958b. (With J.V. Krutilla.) Multiple purpose river
development. Baltimore: Johns Hopkins Press.
1964a. Public finance. New York: Prentice-Hall.
4th edn, 1979.
3176
1964b. (With E.S. Kirschen and others.) Economic
policy in our rime. Amsterdam: North- Holland.
1967. (ed.) Studies in the economics of income
maintenance. Washington. DC: Brookings.
1970. (ed.) Die econometrics of price determina-
tion. Washington, DC: Board of Governors of
the Federal Reserve System and Social Science
Research Council.
1976. (ed.) Parameters and policies in the U.S.
economy. Amsterdam: North- Holland.
1978. Die great recession. Amsterdam: North-
Holland.
1981. Core inflation. New York: Prentice-Hall.
1983. Die DR1 model of the U.S. economy. New
York: McGraw-Hill.
1984. (With C. Cat on, R. Brinner and P. Duprey.)
Die DR1 report on U.S. manufacturing indus-
tries. New York: McGraw-Hill.
Ecole Nationale des Ponts et Chaussees
Robert B. Ekelund Jr and Robert F. Hebert
French School of Civil Engineering, located at 28
rue des Saint-Peres, Paris. Established in 1747 by
Daniel Trudaine, Finance Minister to Louis XV, the
Ecole has traditionally produced economists of
exceptional talent and originality. Isnard, Dupuit
and Cheysson were sUidents there and at various
times its faculty included the likes of Hcmi Navier,
Joseph Minard, Joseph Gamier, Henri Baudrillart,
Charles Gide, Clement Colson and Franpois
Divisia.
The idea of an instimtion dedicated to the pro-
fessionalization of French engineers had its roots in
the 17th century. In 1690 Vauban created the Corps
of Military Engineers, which was to serve as a
model for future public bodies of this sort. He even
went so far as to propose a public examination to
test the scientific knowledge of young people
aspiring to become engineers. After an inauspicious
beginning, the Ecole slowly
Ecole Nationale des Ponts et Chaussees
acquired more scholarly aspirations. It was directed
in its formative years by J.R. Perronet, who
established the high standards and pedagogical
technique responsible for the ultimate success of the
school, so much so that French engineers became
the envy of the world. Although a formal course in
economics was not established until 1847 (receiving
impetus from Dupuit’s pioneer researches in 1844),
engineers were ‘officially’ exhorted to study
economics as early as 1792.
The Revolution of 1789 brought sweeping
changes to higher education in France. For a time it
seemed as though the Ecole would be swept away
as a vestige of the ancien regime, but Mirabeau
successfully defended its existence, and by the time
Napoleon came to power, a major expansion of
faculty, students and curriculum was under way.
With the establishment of the Ecole Polytechnique
in 1794, the nature of the Ecole des Ponts ct
Chaussees changed from an undergraduate to a
postgraduate institution, offering admission by
competitive examination and specialized training
for polytechnicians seeking to become civil
engineers. These civil engineers became problem-
solvers in the areas of flood control, municipal
water distribution and sewage disposal, canal
building, railway construction, road building and
myriad other matters of concern to engineers.
The 19th cenUtty was the ‘golden age’ of the
Ecole, a time when the faculty was upgraded and
the curriculum was stretched to include stereo-
tomy (1799), modem languages (1806), mineralogy
and geology (1826), administrative law (1831),
political economy (1847), thermodynamics (1851),
and applied chemistry (1864). The role of the Ecole
was pivotal and international in both engineering
and economics. Henri Navier, for example, was sent
in 1821 and in 1823 by the Director General of the
Corps to study British achievements in suspension
bridge design and construction. Upon his return
Navier, who wrote a number of essays on the
economic value of public works, offered a Memoire
sur les ponts suspendus which brought the French to
the forefront of such technology for much of the
19th century. Jules Dupuit entered the Ecole in
1824, where he reacted to both Navier’s engineering
and
Ecological Economics
economic studies, later producing a theory of
marginal utility and a full scale welfare analysis of
markets and market structure. In 1830 an American
student, Charles Ellet, Jr., entered the Ecole,
returning home as the premier suspension bridge
builder and designer of his age and as one of the
most creative American economic theorists of the
century. In short, the 19th century is the period
when economic inquiry at the Ecole burgeoned,
easily outdistancing the policy squabbles that
occupied French academic economists at the
universities and in academic journals. It was the era
of Dupuit, Cheysson and Colson, the unrecognized
giants of 19th century French economics.
Today the Ecole des Ponts et Chaussees stands
as the oldest of France’s grandes ecoles. Perched at
the top of a rigid and highly centralized educational
system, it persists in admitting the country’s
intellectual elite and in providing them with solid
training in economics.
References
Ekelund Jr., R.B., andR.F. Hebert. 1973. Public economics at species such as animals, plants and microorgan-
the Ecole des Ponts et Chaussees: 1830-1850. Journal of
Public Economics 2(3): 241-256.
Ecological Economics
Anastasios Xepapadeas
Abstract
Ecological economics is the study of the inter-
actions and co-evolution in time and space of
human economies and the ecosystems in which
human economies are embedded. It uncovers the
links and feedbacks between human economies
and ecosystems, and so provides a unified
picture of ecology and economy. The link
between ecology and human economies has
been manifested in the development of resource
management or bio-economic
3177
models, in which the main focus has been on
fishery or forestry management where the
impact of humans on ecosystems is realized
through harvesting. More closed links have
been developed, however, as both disciplines
evolve.
Keywords
Adiabatic parameter; Biodiversity; Bio-
economics; Boulding, K.; Diffusion;
Eecological economics; Ecology; Ecosystems;
Externalities; Fisher-Kolmogorov equation;
Intrinsic growth rate; Kolmogorov model;
Logistic function; Lotka, A.; Nash equilibrium;
Optimality behaviour; Predator-prey models;
Random walk models; Red Queen hypothesis;
Spatial economics; Tragedy of the commons;
Turing mechanism; Volterra, V
JEL Classifications
B5;Q2
Ecology can be regarded as the study of living
isms, and the relations among them and their
natural environment. In this context, an ecosystem
includes these species and their non-living envi-
ronment, their interactions, and their evolution in
time and space (see, for example, Roughgarden et
al. ). Economics, meanwhile, is the study of how
human societies use scarce resources to produce
commodities and to distribute them among their
members.
The need for an interdisciplinary approach -
‘ecological economics’ - stems from the fact that
natural ecosystems and human economies are
closely linked. In the process of production and
consumption, human beings use ecosystems and
their services, influence their evolution, and are the
recipients of feedbacks originating from their
actions upon ecosystems. As Kenneth Boulding (
) notes in his classic
paper ‘Earth as a space ship’, which can be
regarded as a landmark in the emergence of eco-
logical economics, ‘Man is finally going to have to
face the fact that he is a biological system living in
an ecological system, and that his survival
3178
power is going to depend on his developing sym-
biotic relationships of a closed-cycle character with
all the other elements and populations of the world
of ecological systems.’
Thus, ecological economics can be regarded as
the study of the interactions and co-evolution in
time and space of human economies and the eco-
systems in which human economies are embedded.
This implies that the task of ecological economics is
to bridge the gap between economy and ecology by
uncovering the links and the feedbacks between
human economies and ecosystems, and by using
these links and feedbacks to provide a unified
picture of ecology and economy and their
interactions and co-evolution. In a sense, ecological
economics aims at linking ecological models and
economic models in order to provide insights into
complex and interrelated phenomena stemming
from and affecting both ecosystems and human
economies.
The natural link between ecology and human
economies has been manifested in the traditional
development of resource management or bio-
economic models (for example, Clark ), in which
the main focus has been on fishery or forestry
management where the impact of humans on
ecosystems is realized through harvesting. More
close links have been developed, however, as both
disciplines evolve.
Common methodological approaches may also
be encountered in ecology and economics. Opti-
mality behaviour, which is fundamental in eco-
nomics, has also been used to provide insights into
the structure of ecological systems, in the context of
optimal foraging behaviour, species competition, or
net energy maximization by organisms (for
example, Tschirhart ; Tilman
et al. ) with the purpose of founding macro-
behaviours in ecosystems - such as those emerging
from population dynamics - on microfoundations.
In the same context, the classical
phenomenological-descriptive approach to species
competition based on Lotka-Volterra systems has
recently been complemented by mechanistic
resource-based models of species competition for
limiting resources (Tilman , ). This
Ecological Economics
approach has obvious links to competition among
economic agents for limited resources. Furthermore,
by linking the functioning of natural ecosystems
with the provision of useful services to humans, or
by using concepts such as ecosystems productivity,
insurance from the genetic diversity of ecological
systems against catastrophic events, or development
of new products using genetic resources existing in
natural ecosystems (Heal ), new insights into the
fundamental issues of the valuation of ecosystems
or the valuation of biodiversity have been derived.
(Examples of useful services to humans include
provisioning services, such as food, water, fuel,
genetic material; regulation services, such as cli-
mate regulation, disease regulation; and cultural
services and supporting services, such as soil for-
mation, nutrient cycling; see Millennium Ecosystem
Assessment .)
Ecological Models
The traditional bio-economic models (Clark ),
which describe the evolution of the population or
the biomass of species when harvesting takes place,
have formed the building blocks of ecological-
economic modelling. These models can be extended
along various lines to provide a more realistic
picture of ecosystems (for a detailed analysis, see
Murray ) and help build meaningful ecological-
economic models. To start with, let x(t) denote the
biomass of a certain species at time t. Then
evolution of the biomass is described by an ordinary
differential equation
dx(t)
dt = birth naturaldeath + migration
— harvesting. (1)
In the analysis of population models it is com-
mon, unless it is a specific case, to set the migration
rate at zero, and to represent the natural rate of
population growth (birth-natural death) by a func-
tion Fix ). The most common specification of this
function is the famous logistic function, which is
Ecological Economics 3179

F(x) rx(l—x/K). In this function r is a positive Lotka-Volterra predator-prey models are classic. If
constant called intrinsic growth rate and K is the we denote the prey population by x(t) and the
carrying capacity of the environment which predator population by y(t) and ignore harvesting
depends on factors such as resource availability or for the moment to simplify things, the model can be
environmental pollution. If we denote by h(t) the written as
rate of harvesting of the species biomass by humans,
the population model becomes: dx{t) r(l - yR(x) ,x(0) = x0
dt = x (5)

= F(x) - h(t),x(0 ) = x0. dy(t) ym{\ — —^ , y ( 0 ) =y0m,n > 0

dt ( 6)
(2)

If Mi) = F(x), the population remains constant where R(x) is a function called the predation term,
and the harvesting rate corresponds to sustainable which can be specified as yx/(x 2+r>2), y, S > 0. A
yield. Harvesting rate is usually modelled as pop- more general multi-species model with /prey and
ulation dependent or h qEx, where q is a positive /predators can be written, for i = 1,.../, as
constant, referred to as a catchability coefficient in
fishery models, and E is harvesting effort. Human d%i(t)
~ir=Xi ,x,(0) = x,0 (7)
activities can affect the species population, in
addition to harvesting, by affecting parameters such
as the intrinsic growth rates or the carrying capacity. d
yM
For example, if the stock of environmental pollution dt Si > H (°) =H O (8)
of a certain pollutant (such as phosphorus in a lake)
in a natural ecosystem is denoted by P. with
dynamics described by where all parameters are positive constants. An
even more general model of interacting populations
can be obtained by the generalized Kolmogorov
^- = g(s(t),P(t)),P(0)=P0, model where the evolution of each species biomass

(3) is described by:

where s(t) is the rate of emissions (such as phos- dxft)

dt X i F f x i , x 2,x3, . . .) i = 1 , 2 , 3 , . . . ( 9 )
phorus loadings), and the pollutant affects param-

dx(t) x In the mechanistic resource-based models of

dt r( P ) x W ). — qEx,r'(P)
<0,K'{P) <0.
If the catchability coefficient is affected by
technical change, then it can be expressed by a
function of time as q(t). In this case ( ) is not
autonomous. Alternatively q can be a function of
technological variables like R&D evolving in the
economic module.
The population model ( ) can be generalized to
age-structured populations and multi-species
populations. In multi-species populations the
species competition emerging from the work of
Tilman (for example, Tilman , ), species
(4)
compete for limiting resources. (Forthe use ofthis
model in ecological-economic modelling, see Brock
and Xepapadeas ; Tilman
et al. .) In these models the growth of a species
depends on the limiting resource, and interactions
among species take place through the species’
effects on the limiting resource. Let x = (xb.. .,x„) be
the vector of species biomasses and R the amoimt of
the available limiting resource. Then a mechanistic
resource-based model with a single limiting factor
in a given
3180
area and i =1,.. ,,n species can be described by growth function leads to the so-called the following
equations: Fisher-Kolmogorov equation, which can be writ
ten as
x,i
g i ( R ) - d i , X j ( 0) = x i 0 (10)
n
R = S - a R ~ Y / ^ i 8 i ( R ) ( I D where x{z,t) denotes the concentration of the bio-
/-I
where g,(R) is resource-related growth, d, is the
species’ natural death rate, S is the amount of
resource supplied, a is the natural resource removal
rate (leaching rate), and w, is specific resource
consumption by species i. The main result in this
framework relates to an exclusion principle stating
that, in a landscape free of disturbances, the species
with the lowest resource requirement in equilibrium
will competitively displace all other species, driving
the system to a monoculture. Species coexistence
and polycultures in equilibrium can be supported in
a system with more than one limiting resource, or
even in single resource systems if there is
temperaturedependent growth and temperature
variation in the ecosystem, spatial or temporal
variations in resource ratios, differences in local
palatabilities and local abundance of herbivores.
In addition to the temporal variation captured by
the models described above an important char-
acteristic of ecosystems is that of spatial variation.
Biological resources tend to disperse in space under
forces promoting ‘spreading’ or ‘concentrating’
(Okubo ); these processes, along with intra- and
inter-species interactions, induce the formation of
spatial patterns for species. A central concept in
modelling the dispersal of biological resources is
that of diffusion. Diffusion is defined as a process
whereby the microscopic irregular movement of
particles such as cells, bacteria, chemicals, or
animals results in some macroscopic regular motion
of the group. Biological diffusion is based on
random walk models which, when coupled with
population growth equations, lead to general
reaction-diffusion systems (see, for example, Okubo
and Levin ; Murray ). When only one species is
examined, the coupling of classical diffusion with a
logistic
Ecological Economics
dx{z, t) d2x{z, t)
F( x ( z , t ) ) +D x dz2 (12)
dt
mass at spatial point z at time t. The biomass grows
according to a standard growth function F(x) which
determines the resource’s kinetics but also disperses
in space with a constant diffusion coefficient Dx,
(Nonlinear reaction diffusion equations are
associated with propagating wave solutions.) In
general, a diffusion process in an ecosystem tends
to produce a uniform population density, that is,
spatial homogeneity. Thus it might be expected that
diffusion would ‘stabilize’ ecosystems where
species disperse and humans intervene through
harvesting.
There, is however, one exception, known as
‘diffusion induced instability’ or ‘diffusive insta-
bility’. It was Alan Turing ( ) who suggested
that under certain conditions reaction-diffusion
systems can generate spatially heterogeneous
patterns. This is the so-called ‘Turing mechanism’
for generating diffusion instability. With two
interacting species evolving according to
dx{z, t) d2x{z, t)
F ( x , y) I D x 2 (13)
dt dz
dy(z, t) d2y(z, t)
G ( x, y ) + D y dz2 (14)
dt
if in the absence of diffusion (Dx =Dy =0) the
system tends to a spatially uniform stable steady
state, then under certain conditions, depending on
the relationship DJDy, spatially heterogeneous
patterns can emerge due to diffusion-induced
instability.
Spatial variations in ecological systems can also
be analysed in terms of meta-population models. A
meta-population is a set of local populations
occupying isolated patches which are connected by
migrating individuals. Metapopulation dynamics
can be developed for single
-rx-yQ(d,y)) (18)
Ecological Economics 3181

(xQ(d,y)-S).
or many species (19)
(Levin ). For the single species to evolve slowly, in contrast to the population, host-
case the dynamics become parasite, dynamics which are assumed to evolve fast
(see Dieckmann and Law ).
— = F(x)x + Dx (15) A simple co-evolutionary model can be devel-
at oped in a system with one harvested (‘useful’)
species or host species whose biomass is denoted
where x=(xlv . ,,xj) is a column vector of species by x and a parasite denoted by y, where the abun-
densities, F has its rth row depending on the rth row dance of x and y depends on the evolution of two
of x, and D = \dtj\ is a connectivity matrix, where dy characteristics or traits denoted by d and y (see, for
is the rate of movement from patch j to patch i (j example, the Red Queen dynamic models devel-
=/=■ i). Thus dynamics are local with the exception oped by Krakauer and Jansen ), where d affects the
of movements from one patch to the other. fitness of x and y affects the fitness ofv.
A more general model encompassing i = 1,.. ,,n Let the growth rates of x and the pathogeny be
species competing for j = limiting given by
resources, with density-dependent growth and
interactions across patches c = 1 , . C in a given
landscape, can be written as

— = Fic(xc,X-c)gic(Rc,dic),\fi,c ( 1 6 )
Xci

R j C = S j c ( R c , R - c ) — D j c ( x c , x - c , Rc, R_c), V / , c
If we measure fitness by growth rates, then 9 Q!
(17) ff^ < 0, so that an increase in d increases fitness of
x. In the same way, > 0 for an
where R_c, x c are respectively vectors of resources
increase in y to increase fitness ofy. In equilibrium
and species outside patch c.
of the fast population system where x = y = 0, it
(For a detailed analysis, see Brock and
holds that
Xepapadeas .) A more general setup can be obtained
in the context of co-evolutionary models which
x S s — rx
describe the interactions between population (or m7 r y = W y ) , s - r x ’ (20)
biomass) dynamics and mutation (or trait
dynamics). Antagonistic co-evolution of species on
the one hand and pests or parasites or the other can that On the assumption of constant mutation rates i-
be described by the so-called Red Queen hypothesis id and /<T, the evolutionary dynamics for the traits d
(see, for example, Van Valen , and Kawecki ). and y, when population dynamics have reached the
According to this hypothesis, asymptotically stable steady state, are given by
parasites evolve ceaselessly in response to perpetual
evolution of species’ (or hosts’) resistance. The co-
evolution of the parasites’ ability to attack d UQ (d , y
-t*dxy ) (21)
(virulence) and the hosts’ resistance is expected to
dd
indicate persistent fluctuations of resistance and
vimlence. In this context the Red Queen hypothesis dQ(d,y)
y = F y x y dy (22)
generates a continuous need for variation, resulting
in a limit cycle or other non-point attractor in trait
space dynamics, which are called Red Queen races. See Krakauer and Jansen ( ) who, by con
Red Queen cycles are observed in a slow time scale, sidering the slow time scale trait dynamics, show
since trait dynamics are assumed that the equilibrium point (d*,y*) : d y 0 is not
attracting; the dynamics spiral away from this
3182 Ecological Economics

point. This behaviour is the oscillatory, Red Queen of control variables - is defined as a formal optimal
dynamics. control problem. In this problem the objective is the
optimization ofthe criterion function subject to the
constraints imposed by the structure of the
Ecological-Economic Modelling ecosystem. These constraints, which provide the
transition equations of the optimal control problem,
The ecological models developed above are the
are the dynamic equations of the ecological models
cornerstones of the development of meaningful
described in the previous section.
ecological-economic models. The impact of humans
The solution of the ecological-economic model,
on the population of species can be realized through
provided it exists, will determine the paths of the
direct harvesting h as described in ( ) and ( ). This
type of impact can be easily incorporated into the state and the control variables and the steady state
more general population dynamic models by of the system, which will determine the long-run
selecting the harvested species. Human influence equilibrium values of the ecological populations as
can also be realized in an indirect way by having the well as the approach dynamics to the steady state. In
environmental carrying capacity affected by this context, managed ecological systems which are
environmental pollution generated in the non- predominantly nonlinear could exhibit dynamic
harvesting sector of the economy, as in ( ) and ( ), or behaviour characterized by multiple, locally stable
by having technological considerations affecting and unstable steady states, limit cycles, or the
catchability coefficients. It is also possible that emergence of hysteresis, bifurcations or
external environmental conditions which are irreversibilities.
anthropogenic, such as global warming, can make The way in which the objective function is set
some parameters associated with population up and the ecological constraints which are taken
dynamics or mutation dynamics change slowly. into account determine the solution of the
This can be modelled in ( ) and ( ) by consid ecological-economic model. In principle, a socially
ering p,/ and py as slow varying parameters, defined optimal solution can be distinguished from a
as pd(£t) and py(£t), where 0 < e <C 1 is the privately optimal solution. The socially optimal
adiabatic parameter. This slowly varying system solution corresponds to the so-called problem ofthe
could be used to model slow anthropogenic impacts social planner, where the objective function takes
on ecosystem structure. into account not only benefits from harvesting
However, the size and the severity of the impact certain resoiuces of the ecological system, which
of human economies in ecosystems depend on the corresponds to harvesting commercially valuable
way in which variables, such as harvesting or other biomass, but in addition a wide spectrum of flows
variables which can be chosen by humans (such as of services generated by the whole ecosystem.
emissions, investment in harvesting capacity) and These include, as described above, regulation,
which influence the evolution of ecosystems, are cultural or supporting services, existence values, or
actually chosen. These variables can be regarded as benefits associated with productivity or insiuance
control variables, and the way in which they are gains. If V (h(/)) denotes harvesting benefits at time
chosen affects the evolution of ecological variables, t associated with harvesting vector h, and U(x(t))
such as species biomasses or traits, which can be denotes the flow of benefits associated with
considered as the state variables of the problem. ecosystem service generated by species biomasses
The typical approach in economics is to asso- existing in the ecosystem and not removed by
ciate the choice of the control variables with opti- harvesting, then the total flow of benefit is V (h(/)) )
mizing behaviour. Thus, the control variables are U(x(t)). In this formulation, the V ( ) and U( )
chosen so that a criterion function is optimized, and functions are usually assumed to be monotonically
the economic problem of ecosystem management - increasing and concave. In a more general setup, the
where management means choice total benefit function can be non-separable, defined
as w(h(t);
Ecological Economics
The objective can then be written as
•OO
ma x e-p,[V(h(f)) + U (x ( t )) ] d t (2 3)
{h(')}Jo
where p > 0 is a discount rate. It should be noted
that in principle benefits associated with V (h(t))
can be estimated using market data, while benefits
associated with U(\U)) are hard to estimate because
markets for the larger part of the spectrum of
ecosystem services are missing. (Valuation of
ecosystem services is an open question. For details,
see, for example, Bingham et al. .) The social
optimum corresponds to the maximization of ( ),
subject to the constraints imposed
by the ecological system. For example, if we use the
generalized model of resource competition, the
constraints arc:
— = F i c ( x c , x-c)gic{R f , di c ) - V/, c (24)
%ci
R j c = S J C ( R C , R - C ) — D j c ( x c , X - c , Rc, R_c), V/, c .
(25)
A solution (h (/), x (/)) is regarded as the socially
optimal solution.
The privately optimal solution is distinguished
from the socially optimal by the fact that only
harvesting benefits enter the objective function. The
assumption is that management is carried out by a
‘small’ profit-maximizing private agent that ignores
the general flows of ecosystem services. In this
case, the private agents do not take into account
externalities associated with their management
practices on ecosystem service flow and U(x(t)) =
0. Market externalities associated with the
definition of F(h) could relate to imperfections in
the markets for the harvested commodities, or to
property rights-related externalities, as the well
known ‘tragedy of the commons’ emerging in the
harvesting of open access resources.
In general the privately optimal solution
(h°(0,xo(0) will deviate from the socially optimal
solution. Another type of externality can be See Also
associated with strategic behaviour in resource
harvesting if more than one private agent harvests
the resource. If l = 1,..., L harvesters
3183
are present, then the biomass equation ( ) for patch c
becomes
L
— = F i c ( x c , X- c )g i c ( R c , d i c ) - ^ ~yic,\Ji,c.
Xci
i=i
In this case the privately optimal solution can be
obtained as an open loop or feedback Nash
equilibrium.
Privately optimal solutions can also be distin-
guished from the socially optimal by the extent to
which the ecological constraints are taken into
account For example, if resource dynamics or trait
dynamics are not taken into account in the optimi-
zation problem, the management rule will deviate
from the social optimum. Furthermore, since all the
ecological constraints are operating, there will be
discrepancies between the perceived evolution of
ecosystems under management that ignores certain
constraints, and the actual evolution of the ecosys-
tem. Brock and Xepapadeas ( ), show that, by
ignoring genetic constraints associated with the
development of resistance to genetically modified
organisms, the actual system loses any productivity
advantage because of resistance development
These discrepancies might be a cause for sur-
prises in ecosystem management. For example, with
reference to the co-evolutionary model ( ), ( ), ( ),
( ), and ( ), profit-maximizing decisions which
ignore evolution might steer the system to a certain
steady state on a fast time scale, but then the
underlying trait dynamics might move the system in
slow time to another attractor.
The deviations between the private solution and
the social optimum provide a basis for regulation
which is similar to the rationale behind the
regulation of environmental externalities. Regula-
tion could take the form, in general spatial models
of ecosystem management, of species-specific and
site-specific taxes on harvesting, or equivalent quota
and zoning systems.
3184 Ecological Inference

► Consumption Externalities Roughgarden, J., R. May, and S. Levin. 1989. Perspectives in

► Dynamic Programming
► Environmental Economics
► atial Econor
► Spatial Econometrics
Bibliography
Bingham, G., et al. 1995. Issues inecosystem valuation:
Improving information for decision making. Ecological
Economics 14(2): 73-90.
Boulding, K. 1965. Earth as a space ship. Washington State
University Committee on Space Sciences. Kenneth E.
Boulding Papers, Archives (Box # 38), University of
Colorado at Boulder Libraries. Online. Available at
ecological theory. Princeton: Princeton University Press.
Tilman, D. 1982. Resource competition and community
structure. Princeton: Princeton University Press.
Tilman, D. 1988. Plant strategies and the dynamics and
structure of plant communities. Princeton: Princeton
University Press.
Tilman, D., S. Polasky, and C. Lehman. 2005. Diversity,
productivity and temporal stability in the economies of
humans and nature. Journal of Environmental Economics
and Management 49: 405-426.
Tschirhart, J. 2000. General equilibrium of an ecosystem.
Journal of Theoretical Biology 203: 13-32.
Turing, A. 1952. The chemical basis of morphogenesis.
Philosophical Transactions of the Royal Society of
London. Series B: Biological Sciences 237(641): 37-72.
Van Valen, L. 1973. A new evolutionary law. Evolutionary
Theory 1: 1-30.

. Accessed 13 July 2005.

Brock, W., and A. Xepapadeas. 2002. Optimal ecosystem
management when species compete for limiting
resources. Journal of Environmental Economics and
Management 44: 189-230.
Brock, W., and A. Xepapadeas. 2003. Valuing biodiversity
from an economic perspective: A unified economic,
ecological and genetic approach. American Economic- Ecological Inference
Review 93: 1597-1614.
Clark, C. 1990. Mathematical bioeconomics: The optimal Gary King, Ori Rosen and Martin Tanner
management of renewable resources. 2nd ed. New York:
Wiley.
Dieckmann, U., and R. Law. 1996. The dynamical theory of
coevolution: A derivation from stochastic ecological
processes. Journal of Mathematical Biology 34: 579-612.
Abstract
Kawecki, T. 1998. Red queen meets Santa Rosalia: Arms
races and the evolution of host specialization in organ- Ecological inference is a general statistical
isms with parasitic lifestyles. American Naturalist 152(4): problem where a response variable is not avail-
635-651. able at the subject level because summary sta-
Krakauer, D., and V. Jansen. 2002. Red queen dynamics in
protein translation. Journal of Theoretical Biology 218:
tistics are reported for groups only. It consists of
97-109. merging information from different databases
Heal, G. 2000. Nature and the marketplace: Capturing the which are not linked to each other at the record
value of ecosystem services. Washington, DC: Island level. We consider an election scenario where in
Press.
Levin, S. 1974. Dispersion and population interactions.
each electoral precinct the fraction of voting-age
American Naturalist 108: 207-228. people who turn out to vote, the fraction of black
Millennium Ecosystem Assessment. 2005. Ecosystems and population and the number of voting-age people
human well-being, Policy responses. Vol. 3. Washington, are observed. The proportions of blacks and of
DC: Island Press.
whites who vote are unobserved because
Murray, J. 2003. Mathematical biology. 3rd ed. Springer:
Berlin. electoral results and census data are not linked.
Okubo, A. 2001. Introduction: The mathematics of ecological
diffusion. In Diffusion and ecological problems: Modern
perspectives, 2nd ed., ed. A. Okubo and S. Levin. Berlin:
Springer.
Okubo, A., and S. Levin. 2001. The basics of diffusion. In Keywords
Diffusion and ecological problems: Modern perspectives, Aggregation; Ecological inference; Likelihood;
2nd ed., ed. A. Okubo and S. Levin. Berlin: Springer. Markov chain Monte Carlo methods; Method of
bounds; Nonparametric models; Statistical
approaches
Pbi i-Pb
Ecological Inference
P7 i - ft
JEL Classifications
CIO
The Ecological Inference Problem
For expository purposes, we discuss only an
important but simple special case of ecological
inference, and adopt the running example and
notation from King ( : eh. 2). The basic prob
lem has two observed variables (T, and Xt) and two
unobserved quantities of interest (/> ■' and />“') for
each of p observations. Observations represent
aggregate units, such as geographic areas, and each
individual-level variable within these units is
dichotomous.
To be more specific, in Fig. we observe for each
electoral precinct i(i = 1, ... , p) the fraction of
voting age people who turnout to vote (7-) and who
are black (JQ, along with the number of voting age
people (N,). The quantities of interest, which remain
unobserved because of the secret ballot, are the
proportions of blacks who vote (/if) and whites who
vote (/?/). The proportions fib and (fJ>J) are not
observed because 7} and Xj are from different data
sources (electoral results and census data,
respectively) and record linkage is impossible (and
illegal), and so the crosstabulation cannot be
computed.
Also of interest are the district-wide fractions of
blacks and whites who vote, which are respectively
Race of Voting decision voting
age
person Vote No Vote
T l _ r.
1
1 L1
i Method of Bounds
Ecological Inference, Fig. 1 Notation for Precinct i. Note:
The goal is to estimate the quantities of interest, j8f (the
fraction of blacks who vote) and /if (the fraction of whites
who vote), from the aggregate variables X, (the fraction of
voting age people who are black) and T, (the fraction of
people who vote), along with N, (the known number of voting
age people)
3185
Y.N.XJI1’
-----------------. and
Bb =—p (1)
1>*
i=\
!>«( i - w
Bw = ^-p--------------------- - ( 2 )
i1
These are weighted averages of the
corresponding precinct-level quantities. Some
methods aim to estimate only Bb and Bw without
giving estimates of /if and for all i.
Deterministic and Statistical Approaches
The ecological inference literature before King (
) was bifurcated between supporters of the
method of bounds, originally proposed by Duncan
and Davis ( ), and supporters of statistical
approaches, proposed even before Ogbum and
Goltra ( ) but first formalized into a coherent
statistical model by Goodman ( , ). (For
the historians of science among us: although these
two monumental articles were written by two col-
leagues and friends in the same year and in the same
department and university - the Department of
Sociology at the University of Chicago - the
principal did not discuss their work prior to com-
pletion. Even by today’s standards, nearly a half
century after their publication, the articles are
models of clarity and creativity.) Although Good-
man and Duncan and Davis moved on to other
interests following their seminal contributions, most
of the ecological inference literature in the five
decades since 1953 was an ongoing war between
supporters of these two key approaches, and often
without the usual academic decorum.
Extracting Deterministic Information: The
The purpose of the method of bounds and its
generalizations is to extract deterministic informa-
tion, known with certainty, about the quantities of
interest.
3186
The intuition behind these quantities is simple.
For example, if a precinct contained 150 African-
Americans and 87 people in the precinct voted, then
how many of the 150 African-American actually
cast their ballot? We do not know exactly, but
bounds on the answer are easy to obtain: in this
case, the answer must lie between 0 and 87. Indeed,
conditional only on the data being correct, [0,87] is
a 100 per cent confidence interval. Intervals like this
are sometimes narrow enough to draw mean- ingftd
inferences, and sometimes they are too wide, but the
ability to provide (non-trivial) 100 per cent
confidence intervals in even some situations is quite
rare in any statistical field.
In general, before any data are seen, the
unknown parameters fjb and /i“' are each bounded on
the unit interval. Once we observe Tt and Xt they are
bounded more narrowly, as:
Tj-il-Xj) • iTi , min
fie max ^0, Ti-X A . —, 1
Xi‘
fie T^Ti) ’min Tt , 1
max I 0.
1 -xt
(3)
Deterministic bounds on the district-level
quantities Bb and B" are weighted averages of these
precinct-level bounds.
The bounds then indicate that the parameters in
each case fall within these deterministic bounds
with certainty, and in practice they are almost
always narrower than [0,1]. Whether they are
narrow enough in any one application depends on
the nature of the data.
Extracting Statistical Information: Goodman's
Regression
Leo Goodman’s ( , ) approach is very
different from, but just as important as, Duncan and
Davis’s. He looked at the same data and focused on
the statistical information. His approach examines
variation in the marginals (X, and T) over the
precincts to attempt to reason back to the district-
wide fractions of blacks and whites who vote, Bhand
Bw. The outlines of this approach and the problems
with it have been known at least since Ogbum and
Goltra ( ). For example, if in pre
cincts with large proportions of black citizens we
Ecological Inference
observe that many people do not vote, then it may
seem reasonable to infer that blacks turn out at
lower rates than whites. Indeed, it often is reason-
able, but not always. The problem is that it could
instead be the case that the whites who happen to
live in heavily black precincts are the ones who vote
less frequently, yielding the opposite ecological
inference to the individual-level truth.
What Goodman accomplished was to formalize
the logic of the approach in a simple regression
model, and to give the conditions under which
estimates from such a model are unbiased. To see
this, note first that the accounting identity
Ti = Xifi + {\-Xi)fi
(4)
holds exactly. Then he showed that a regression of
Tt on X j and (1 — X , ) with no constant term could
be used to estimate Bb and B"\ respectively. The key
assumption necessary for unbiasedness that
Goodman identified is that the parameters and X , be
uncorrelated: Cov(fi,Xi) = Cov(fi,Xj) = 0. In the
example, the assumption is that blacks vote in the
same proportions in homogeneously black areas as
in more integrated areas. Obviously, this is true
sometimes and it is false other times. (King : ch. 3,
showed that Goodman’s assumption was necessary
but not sufficient. To have unbiasedness, it must
also be true that the parameters and Ni are
uncorrelated.)
As Goodman recognized, when this key
assumption does not hold, estimates from the model
will be biased. Indeed, they can be very biased,
outside the deterministic bounds, and even outside
the unit interval. This technique has been used
extensively since the 1950s, and impossible
estimates occur with considerable frequency (some
estimates range to a majority of real applications;
Achen and Shively ).
Extracting Both Deterministic
and Statistical Information: King's El
Approach
From 1953 until 1997, the only two approaches
used widely in practice were the method of bounds
and Goodman’s regression. King’s
Ecological Inference
0.75 -
Ecological Inference, Fig. 2 Two views of fhe same data.
Note: The left graph is a scatterplot of the observables, X, by
Tj. The right graph displays this same information as a
tomography plot of the quantities of interest, Pf by pj. Each
precinct i that appears as a point in the left
( ) idea was that the insights from these two
conflicting literatures in fact do not conflict with
each other; the sources of information are largely
distinct and can be combined to improve inference
overall and synergistically. The idea is to combine
the information from the bounds, applied to both
quantities of interest for each and every precinct,
with a statistical approach for extracting informa-
tion within the bounds. The amount of information
in the bounds depends on the data-set, but for many
data-sets it can be considerable. For example, if
precincts are spread uniformly over a scatterplot
ofXt by Th the average bounds on ft1’ and ftj are
narrowed from [0,1] to less than half of that range -
hence eliminating half of the ecological inference
problem with certainty. This additional information
also helps make the statistical portion of the model
far less sensitive to assumptions than previous
statistical methods which exclude the information
from the bounds.
To illustrate these points, we first present all the
information available without making any
assumptions, thus extending the bounds approach as
far as possible. As a starting point, the left graph in
Fig. provides a scatterplot of a sample data set as
observed, Xt horizontally by I vertically. Each
3187
0.0 0.2 5 0.5 0.75 1
P?
graph is a line (rather than a point because of information
lost due to aggregation) in the right graph. For example,
precinct 52 appears as the dot with a little square around it
in the left graph and the dark line in the right graph (Source:
The data are from King ( : Figs. 5.1 and 5.5))
point in this figure corresponds to one precinct, for
which we would like to estimate the two unknowns.
We display the unknowns in the right graph of the
same figure; any point in the right graph portrays
values of the two unknowns, ft' which is plotted
horizontally, and ft'ft which is plotted vertically.
Ecological inference involves locating, for each
precinct, the one point in this unit square
corresponding to the true values of ft1’ and ft", since
values outside the square are logically impossible.
To map the knowns onto the unknowns, King
begins Goodman’s accounting identity from Eq. ).
From this equation, which holds exactly, King
solves for one unknown in terms of the other:
(5)
which shows that ft" is a linear function of ft1’,
where the intercept and slope are known (since they
are functions of the data, Xt and 'i).
King then maps the knowns from the left graph
onto the right graph by using the linear relationship
in Eq. ). A key point is that each dot on the
3188
left graph can be expressed, without assumptions or
loss of information, as what King called a
‘tomography’ line within the unit square in the right
graph. It is precisely the information lost due to
aggregation that causes us to have to plot an entire
line (on which the true point must fall) rather than
the goal of one point for each precinct on the right
graph. In fact, the information lost is equivalent to
having a graph of the by (3J points
but having the ink smear, making the points into
lines and partly but not entirely obscuring the
correct positions of the (/3f, fff j points. (King also
showed that the ecological inference problem is
mathematically equivalent to the ill-posed
‘tomography’ problem of many medical imaging
procedures, such as CAT and PET scans, where one
attempts to reconstruct the inside of an object by
passing X-rays through it and gathering information
only from the outside. Because the line sketched out
by an X-ray is closely analogous to Eq. ), King
labels the latter a tomography line and the
corresponding graph a tomography graph.)
What does a tomography line tell us? Before we
know anything, we know that the true (/?f, fff) point
must lie somewhere within the unit square. After A)
and 7] are observed for a precinct, we also know
that the true point must fall on a specific line
represented by Eq. ) and appearing in the tomog-
raphy plot in Fig. . In many cases narrowing the
region to be searched for the true point from the
entire square to the one line in the square can
provide a significant amount of information. To see
this, consider the point enclosed in a box in the left
graph, and the corresponding dark line in the right
graph. This precinct, number 52, has observed
values of X52 = 0.88 and T52 = 0.19. As a result,
substituting into Eq. ) gives />“ = 1.58
7.33/if, which when plotted appears as the
dark line on the right graph. This particular line
tells us that, in our search for the true point
on the right graph, we can eliminate with certainty
all area in the unit square except that on the line,
which is clearly an advance over not having the
data. Translated into the quantities of interest, this
line tells us (by projecting the line downward to
Ecological Inference
the horizontal axis) that, wherever the hue point
falls on the line, /?| 2 must fall in the relatively
narrow bounds of [0.07,0.21], Unfortunately, in this
case, [1]' can only be bounded (by projecting to the
left) to somewhere within the entire unit interval.
More generally, lines that are relatively steep, like
this one, tell us a great deal about and little about
[if. Tomography lines that are relatively flat give
narrow bounds on (3W, and wide bounds on Lines
that cut off the bottom left (or top right) of the
figure give narrow bounds on both quantities of
interest.
If the only information available to learn about
the unknowns in precinct i is Xt and 7], a tomog-
raphy line like that in Fig. exhausts all this available
information. This line immediately tells us the
known bounds on each of the parameters, along
with the precise relationship between the two
unknowns, but it is not sufficient to narrow in on
the right answer any further. Fortunately, additional
information exists in the other observations in the
same data set (Xj and 7} for all i =/=■ j) which,
under the right assumptions, can be used to learn
more about and in our precinct of interest.
In order to borrow statistical strength from all
the precincts to learn about fl1’ and [1™ in precinct
i, some assumptions are necessary. The simplest
version of King’s model (that is, the one most
useful for expository purposes) requires three
assumptions, each of which can be relaxed in
different ways.
First, the set of (/!*, /i‘ v) points must fall in a
single cluster within the unit square. The cluster can
fall anywhere within the square; it can be widely or
narrowly dispersed or highly variable in one
unknown and narrow in the other; and the two
unknowns can be positively, negatively, or not at all
correlated over i. An example that would violate
this assumption would be two or more distinct
clusters of (/!*, [if) points, as might result from
subsets of observations with fundamentally
different data generation processes (such as from
markedly different regions). The specific
mathematical version of this one-cluster assumption
is that fl1’ and [!'■ follow a truncated bivariate
normal density
Ecological Inference 3189

Kft-ft
three assumptions together produce what has come
TN(#./?r i®.E) = iv(#./r®.£ to be known as King’s ‘basic’ El model. (The use of
El to name this method comes from the name of his
(6) software, available at
) King also generalizes this
where the kernel is the untruncated bivariate assumption, in what has come to be known as the
normal. ‘extended’ El model, by allowing the truncated
normal parameters to vary as functions of measured
covariates, Zf and Zf, giving:
N
, -1/2
u 6
= (27I)-1 E exp
<8 <Piol +0.25) + 0.5] + (zf - Z ) ab

(oP +0.25) +0.5 (ZJ -zw) a"

and 1 (//'. I>)') is an indicator function that equals 1 (10)
if [>b € [0,1] and E [0,1] and zero otherwise. The
normalization factor in the denominator, R ('23J, E ') where a6 and xw are parameter vectors to be esti-
- is the volume under the untruncated normal mated along with the original model parameters and
distribution above the unit square: that have as many elements as Z ) and Z " ' have
columns. This relaxes the mean independence
1 1
r r u u assumptions to:
Jo Jo ^
E ( # | M ) = E(pb\Zj)E(PJ\Xj,Zj)
(8) =m\zt)

When divided into the untruncated normal, this

factor keeps the volume under the truncated dis- Note that this extended model also relaxes the
tribution equal to 1. The parameters of the truncated assumptions of truncated bivariate normality, since
density, which we summarize as there is now a separate density being assumed for
each observation. Because the bounds, which differ
in width and information content for each i,
generally provide substantial information, even X,
can be used as a covariate in Zh (The recommended
default setting in El includes X, as a covariate with a
are on the scale of the untruncated normal (and so, prior on its coefficient.) In contrast, under
for example, <B*U and Q5wJ need not be constrained Goodman’s regression, which does not include
to the unit interval even though and are constrained information in the bounds, including X, leads to an
by this density). unidentified model (King : sec. 3.2).
The second assumption, which is necessary to These three assumptions - one cluster, no spatial
form the likelihood function, is the absence of autocorrelation, and mean independence between
spatial autocorrelation: conditional on Xh T, and Tj the regressor and the unknowns conditional on Xt
are mean independent. Violations of this assumption and Z;- - enable one to compute a posterior (or
in empirically reasonable (and even some sampling) distribution of the two unknowns in each
unreasonable) ways do not seem to induce much if precinct. A fundamentally important component of
any bias. El is that the quantities of interest are not the
The final, and by far the most critical, assumption is parameters of the likelihood
that X\ is independent of fl1’ and fSJ. The
3190
but instead come from conditioning on Tt and
producing a posterior for /l* and in each precinct.
Failing to condition on Tt and examining the
parameters of the truncated bivariate normal only
makes sense if the model holds exactly and so is
much more model-dependent than King’s approach.
Since the most important problem in ecological
inference modelling is precisely model
misspecification, failing to condition on Tassumes
away the problem without justification. This point
is widely regarded as a critical step in applying the
El model (Adolph et al. ).
When bounds are narrow, El model assumptions
do not matter much. But, for precincts with wide
bounds on a quantity of interest, inferences can
become model dependent. This is especially the
case with ecological inference problems precisely
because of the loss of information due to
aggregation. In fact, this loss of information can be
expressed by noting that the joint distribution of /J*
and [>)' cannot be frilly identified from the data
without some untestable assumptions. To be
precise, distributions with positive mass over any
curve or combination of curves that connects the
bottom left point (/?* = 0, [3J = 0) to the top right
point ([l1’ = l,[f- = l) of a tomography plot cannot
be rejected by the data (King : 191). Other features
of the distribution are estimable. This fundamental
indeterminacy is, of course, a problem because it
prevents pinning down the quantities of interest
with certainty, but it can also be something of an
opportunity since different distributional
assumptions can lead to the same estimates,
especially since only those pieces of the
distributions above the tomography lines are used in
the final analysis.
Alternative Approaches to Ecological
Inference
In the continuing search for more information to
bring to bear on ecological inferences, King et al. (
) extend King’s ( ) model another
step. They incorporate King’s main advance of
combining deterministic and statistical information
but begin modelling a step earlier at the individuals
who make up the counts. They also
Ecological Inference
build a hierarchical Bayesian model, using easily
generalizable Markov chain Monte Carlo (MCMC)
technology (Tanner ).
To define the model formally, let T\ denote the
number of voting age people who turn out to vote.
At the top level of the hierarchy they assume that
T0' follows a binomial distribution with probability
equal to 0; Aj/J* I (1 Xt) ft? and count Nt. Note that
at this level it is assumed that the expectation of T'i,
rather than 7’', is equal to + (1 Xi)f?i ■ In other
words, King ( ) models
Tt as a continuous proportion, whereas King et al. (
) recognize the inherently discrete
nature of the counts of voters that go into comput-
ing this proportion. The two models are connected,
of course, since TJN] approaches 7} as N] gets
large.
The connection to King’s tomography tine can
be seen in the contribution of the data from precinct
i to the likelihood, which is.
+ (in
By taking the logarithm of this contribution to
the likelihood and differentiating with respect to
and (3J , King, Rosen and Tanner show that the
maximum of Eq. ( ) is not a unique point, but
rather a line whose equation is given by the
tomography line in Eq. ). Thus, the log-1 ike I i
hood for precinct i looks like two playing cards
leaning against each other. As long as 7} is fixed
and bounded away from 0.5 (and A) is a fixed
known value between 0 and 1), the derivative at this
point is seen to increase with Nt, that is, the pitch of
the playing cards increases with the sample size. In
other words, for large N], the log-likelihood for
precinct i degenerates from a surface defined over
the unit square into a single playing card standing
perpendicular to the unit square and oriented along
the corresponding tomography line.
At the second level of the hierarchical model,
jSf is distributed as a beta density with parameters
cb and db and [>'■ follows an independent beta with
parameters cw and c/„. While (j- and (?■’ are
assumed a priori independent, they are a
Econometric Issues in the Presence of Multiple Equilibria
posteriori dependent. At the third and final level of
the hierarchical model, the unknown parameters cb,
db, cw and dw follow an exponential distribution with
a large mean.
A key advantage of this model is that it gener-
alizes immediately to arbitrarily large R x C tables.
This approach was pursued by Rosen et al. ( ),
who also provided a much faster
method of moment-based estimator. For an appli-
cation, see King et al. ( ).
Wakefield ( ) presents an alternative
approach based on the Bayesian paradigm using a
Markov chain Monte Carlo inference scheme. King
et al. ( ) survey the latest strategies for
solving ecological inference problems in various
fields, many of which do not fit the textbook case of
a 2 x 2 table with known marginals and unknown
cell entries. Staniswalis ( ) proposes
a nonparametric model for ecological inference with
an application to renal failure data.
Bibliography
Achen, C., and W. Shively. 1995. Cross-level inference.
Chicago: University of Chicago Press.
Adolph, C., G. King, M. Herron, and K. Shotts. 2003. A
consensus position on second stage ecological inference
models. Political Analysis 11: 86-94.
Duncan, O., and B. Davis. 1953. An alternative to ecological
correlation. American Sociological Review 18: 665-666.
Goodman, L. 1953. Ecological regressions and the behavior
of individuals. American Sociological Review 18: 663-
666.
Goodman, L. 1959. Some alternatives to ecological corre-
lation. American Journal of Sociology’ 64: 610-624.
King, G. 1997. A solution to the ecological inference
problem: Reconstructing individual behavior from
aggregate data. Princeton: Princeton University Press.
King, G., O. Rosen, and M. Tanner. 1999. Binomial-beta
hierarchical models for ecological inference. Sociological
Methods and Research 28: 61-90.
King, G., O. Rosen, and M. Tanner, eds. 2004. Ecological
inference: New methodological strategies. Cambridge:
Cambridge University Press.
BCing, G., O. Rosen, M. Tanner, and A. Wagner. 2003. The
ordinary election of Adolf Hitler: A modern voting
behavior approach. Online. Available at
Accessed 16 Aug 2006.
Ogbum, W., and I. Goltra. 1919. How women vote: A study
of an election in Portland, Oregon. Political Science
Quarterly 34: 413—433.
3191
Rosen, O., W. Jiang, G. King, and M. Tanner. 2001. Bayesian
and frequentist inference for ecological inference: The R
x C case. Statistica Neerlandica 55(2): 134-156.
Staniswalis, J. 2005. On fitting generalized non-linear models
with varying coefficients. Computational Statistics and
Data Analysis 50: 893-902.
Tanner, M. 1996. Tools for statistical inference: Methods for
the exploration of posterior distributions and likelihood
functions. 3rd ed. New York: Springer- Verlag.
Wakefield, J. 2004. Prior and likelihood choices in the
analysis of ecological data. In Ecological inference: New
methodological strategies, ed. G. King, O. Rosen, and
M. Tanner. Cambridge: Cambridge University Press.
Econometric Issues in the Presence of
Multiple Equilibria
Francesca Molinari
Abstract
Multiplicity of equilibria implies that the rela-
tionship between the outcome variable and the
exogenous variables characterising a model is a
correspondence rather than a function. This
results in an incomplete econometric model.
Incompleteness complicates identification and
statistical inference on functionals of the prob-
ability distribution of the population of interest.
This is because it implies that the sampling
process and the maintained assumptions may be
consistent with a set of values for these
functionals, rather than with a single one. As a
result, the econometric analysis of models with
multiple equilibria needs to either:
(1) rely on simplifying assumptions that shift
focus to outcome features that are common
across equilibria; or (2) augment the model with
a “selection mechanism” that chooses the
equilibrium played in the regions of multiplicity;
or (3) maintain only minimal assumptions that
partially identify the functionals of interest. Each
of these approaches is reviewed, focusing on
static game theoretic models.
3192
Keywords
Aumann expectation; Multiple equilibria; Nor-
mal form games; Partial identification; Point
identification; Qualitative choice models; Ran-
dom sets
JEL Classifications
C01; C14; C15; C35
The Basic Problem
Finite game theoretic models have been employed
to study a wide range of economic decisions, where
each agent’s utility is allowed to depend on the
choice made by each of the other agents. Examples
include the analysis of social interaction models
(Brock and Durlauf ), labour force participation
(Bjom and Vuong ), market entry (Bresnahan and
Reiss , ,1991;
Berry ; Bajari et al. ; Jia ; Ciliberto and Tamer ),
product differentiation (Mazzeo ; Borzekowski and
Cohen ), advertising (Sweeting ), and many others.
From the econometric perspective, a finite game is
a generalisation of a standard discrete choice
model. For example, a bivariate simultaneous
response model may give a stochastic
representation of equilibria in a two-player, two-
action game.
Generically, given a set of payoffs for the
agents, finite games may admit multiple equilibria.
Multiplicity of equilibria implies that the mapping
from the model’s exogenous variables to outcomes
is a correspondence rather than a function. This
violates the classical “principal assumptions” or
“coherency conditions” for simultaneous discrete
response models discussed extensively in the
econometrics literature (e.g. Heckman ; Gourieroux
et al. ; Schmidt ; Blundell and Smith ; Maddala ).
Such coherency conditions require the existence of
a unique reduced form, mapping the model’s
exogenous variables and parameters to a unique
realisation of the endogenous variable; hence they
constrain the model to be recursive or triangular in
nature. Tamer ( ), however, clarifies the
distinction
Econometric Issues in the Presence of Multiple Equilibria
between a model which is incoherent, so that no
reduced form exists, and a model which is incom-
plete, so that multiple reduced forms may exist.
Models with multiple equilibria belong to the latter
category. Jovanovic ( ) discusses the
observable implications of these models. Berry and
Tamer ( ) review and extend a number
of results on the identification of entry models
extensively used in the empirical literature. The
insights in their analysis extend to models where
the discrete outcome has larger support.
This article reviews the challenges posed by the
presence of multiple equilibria for the econometric
analysis of static game theoretic models. These
models do not specify how an equilibrium is
selected in the regions of the exogenous variables
which admit multiple equilibria, and therefore they
are “incomplete”. Incompleteness complicates
identification and statistical inference on functional
of the probability distribution of the population of
interest, because it implies that the sampling
process and the maintained assumptions may be
consistent with a set of values for these functionals,
rather than with a single one. The literature has
provided various approaches to dealing with this
basic problem: (1) imposing simplifying
assumptions that shift focus to outcome features
that are common across equilibria (e.g. Bresnahan
and Reiss ( , ,1991)
and Berry ( ), who study entry games where
the number, though not the identities, of entrants is
uniquely predicted by the model in equilibrium);
(2) explicitly modelling a selection mechanism
which specifies the equilibrium played in the
regions of multiplicity (e.g. Bjom and Vuong ( ),
who choose a constant; and Bajari
et al. ( ), who allow for a more flexible, covar
iate dependent parametrisation); (3) partially iden-
tifying and setestimating the parameters, without
imposing assumptions on the selection mechanism
or on the extent of heterogeneity in payoffs (e.g.
Tamer ( ), who also provides large sup
port conditions and exclusion restrictions that
guarantee point identification of the payoff param-
eters, Ciliberto and Tamer ( ), Andrews
et al. ( ), Beresteanu et al. ( )).
Each of these approaches is reviewed in turn,
using the simple example of a complete
 1 S9 = { (1,1)}
1
a
1
£2 So = 1
1
{(0 ,1 )} S« ={(1 ,0 ),
-X2P2-02 III:
(-(C2+X2p2)/82,-(r.l+X|[l|)/8,),
-X2P2 1 (0,1)}
-X|p| -X|p|-i>il
1 fit
S,, = {(0,0)} 1s,, = {
1 (LO)}
X1

b Y0={ (1,1)}
1
1
1
Ye ={(0 ,1 )} 1
1

-X2P2-
Yo ={(0,0),(1 ,0 ),(0 ,1 ),(!,l)} !
1
82 ill
-X1P1 -X|P,-5 |I C|
1 Ye =
Ye = {((),())} 1{
1 (1,0)}
1
3194 Econometric Issues in the Presence of Multiple Equilibria

simplify the notation, in what follows S‘s x denotes
the region of values for e where t e {(0,0), (1,0),
(0,1), (1,1)} is the unique MSNE of the game; for
example, the grey region in Fig. is the region where
(0,1) is the unique equilibrium of the game, &,'/ v,
denotes the region of values for e where multiple
equilibria occur. In the example, this region is the
centre box of Fig. , where
(1,1) , , (0,1) are the MSNE of
the game.
For realisations of e ^ <#)'/ x, the model admits a
unique equilibrium which is in pure strategies, and
therefore predicts a unique equilibrium outcome.
However, for realisations of se £)! x the model
predicts that any of (0,0), (1,0), (0,1), (1,1) might
result as an equilibrium outcome of the game.
One way to reconcile multiplicity of equilibria
with the fact that only one equilibrium outcome is
realised in each market is to augment die model
with an ‘‘admissible selection mechanism” i//(-; x,
e) : Sg(x, e) —> A'1 with A'1 1 the unit simplex in
MK and k the cardinality of Sg(x, e). For each a €
Sg(x, e), if/(a; x, e) specifies the probability with
which that equilibrium is played. (For £ ^
Sg(x, e) is a singleton and therefore ij/ is a scalar
identically equal to 1.) For a selection mechanism
to be admissible, it is required that i//(cr; x, s) > 0
for all a C ^(x.fifand^^^ ; y//(<x; y,£) = 1.
Hence, when no restrictions are placed on it, x, e)
can depend on market unobservables (s) even after
conditioning on market observables (x). It then
follows that given x, 9, and an admissible if/, the
model predicts that the probability of observing an
equilibrium outcome t e {(0,0), (1,0), (0,1),
(1,2) } is
P(y = t|x; 6, i/O
\
dFg( s ) .
For example, for t = (0,0),

P(y = (o,o)\M) =
S
-------v-------'
P((0,0) is the unique equilibrium |x)
E\ + Xl/?A / £2 +^2^2^ I f fE2 + x2^2 £
1 +Xl/^1N\
-£i -S2 J f { { -82 ’ -5, J
,x, e)dFg(e

P((0,0) is observed when multiple equilibria are possible |x)

The identification problem arises because one may

occurs both in the identity and in the number of
find many values for the parameter vector 6 which,
players that enter the market in equilibrium. How-
when coupled with different admissible selection
ever, if one restricts players to play only pure
mechanisms ij/, generate the same distribution of
strategies, and if one assumes 8j < 0 ,j= 1,2, the
outcomes and payoff shifters as the one observed in
model uniquely predicts the equilibrium number of
the data (i.e. P(y t\x; 6 , i//) = P (y = %) x — a.s.).
entrants. In other words, there is an outcome feature
which is common across equilibria. In this case,
under certain restrictions, the “incompleteness” of
Point Identification Based on Outcome the model can be circumvented, without the need to
Features that are Common Across introduce a selection mechanism.
Equilibria Consider first the case that potential entrants are
identical in both their observable and unobservable
Even in the simple two-player entry game with
characteristics, so that each firm operating in
multiple MSNE described above, multiplicity
equilibrium makes the same profit (Bresnahan and
Reiss ). With mixed
Econometric Issues in the Presence of Multiple Equilibria
strategies explicitly ruled out, the equilibrium
number of entrants is uniquely predicted by a
simple zero profit condition: in equilibrium, no firm
will enter if <5(1) + xj8 + e < 0, one firm will enter
if <5(1) + x/3 + e > 0 and S- (2) + xfi + £ < 0, and
both firms will enter if 5- (2) + xfi + e > 0. Here
d(m) denotes the effect on payoffs of m firms
entering the market, m C {1,2}. Hence point
identification of the model’s parameters can be
achieved, and estimation can be conducted, using
familiar techniques for ordered response models.
These considerations can be extended to the case
where the number of potential entrants is larger,
provided that each entrant makes the same profit;
see Bresnahan and Reiss ( )•
Consider now the case that profits differ among
firms, as in the example in the previous section
(Bresnahan and Reiss ; Berry ; Tamer ). With
heterogeneity in payoffs, but mixed strategies
explicitly ruled out, with two players there is still a
unique prediction for the number of entrants. Hence
the choice probabilities for having an equilibrium
with no firms entering the market, and for having
an equilibrium with both firms entering the market,
are uniquely predicted by the model (this is because
(0,0) and (1,1) can only occur as unique equilibrium
outcomes of the game), and given by
P(y=(O,O)|x;0) = P( £ e<f ( 4 O ) ), P(y=(l,l)|
x;0) = P( £ e4 1 ,k 1 ) )-
In this case, Tamer ( ) shows that under suit
able exclusion restrictions and large support con-
ditions on elements of x, one can use the
information in P(y = (0, 0)|x) to identify and that in
P(y = (1, l)|x) to identify cij, d2. A sufficient set of
restrictions is as follows. Assume that the matrices
X| and x2 have full column rank, and that for either
j = 1 or j = 2, xj contains an element which is not
part of x3 _ j, has full support, and has a
corresponding coefficient in fy that is nonzero. Then
there exist sufficiently large and sufficiently small
values of Xj such that player j will always be in or
out of the market regardless of what her rival
chooses. For these values of xj, the game simplifies
to a single
3195
decision problem for player 3 j, and one can identify
/(; , by using those observations with no entrants
and sufficiently small/large values of x7. One can
then leam and y. Similar reasoning allows one to
leam d1; S2 from P(y (1, l)|x). Tamer ( ) shows that
while using the informa
tion contained in the outcomes uniquely predicted
by the model suffices for point identification ofthe
parameters of interest, one can obtain efficiency
gains by exploiting restrictions on the outcomes of
the game resulting from multiple equilibria ((0, 1)
and (1, 0)).
Under restrictions on the payoff functions (e.g.
homogeneous competition effects) but allowing for
a large number of players, Berry ( )
shows that a pure strategy Nash equilibrium for the
model exists, and is such that the equilibrium
number of entrants is uniquely determined. In this
case, inference can be conducted as in nonlinear
parametric method of moments problems. Point
identification of the model parameters is likely to
hold if there is variation in the number of potential
entrants across markets; see Berry and Tamer ( ).
Importantly, Tamer ( ) shows that large
support and exclusion restrictions can be used to
point identify the parameters even when one allows
for mixed strategies, and no outcome feature is
common across equilibria. Generalisation of this
result to games with a larger number of players
under related assumptions is discussed in Bajari et
al. ( ); see the following section.
Point Identification Based on Specifying a
Selection Mechanism
Early on, Bjom and Vuong ( ) suggested solv
ing the identification problem caused by the pres-
ence of multiple equilibria by specifying a selection
mechanism that assigns the probability mass of the
region of multiplicity among the possible
equilibrium outcomes of the game. They considered
a two-player, two-action game such as the simple
example discussed above, and assumed that players
play only pure strategies. (Bjom and Vuong ( )
were interested in learn
ing the determinants of a husband and wife’s
3196 Econometric Issues in the Presence of Multiple Equilibria

decision to join the labour force. Hence they did not suitable large support conditions on the covariates
constrain the sign of S to be known a priori. Here or with exclusion restrictions, and with scale
the exposition is simplified by assuming Sj < 0, / invariance conditions on i//( -) (the equilibrium
1,2). For simplicity, they assumed selection probabilities are required to depend only
that i/r((l, 0); x, e) = i/t((0,1); x, e) =5 for all x, e on the relative but not absolute scales of payoffs),
such that s C S’f x. Under this restriction, Bjorn and both 9 and a can be point identified. They then
Vuong ( ) provided a necessary and suf propose a method of simulated moments estimator
ficient condition for the parameters of the model to to estimate these parameters, which embeds a
be point identified, based on the classic criterion of computationally feasible procedure to calculate all
nonsingularity of the information matrix. In their the MSNE of the game. Importantly, under the
model, estimation can be carried out maintained assumptions, this also yields an
straightforwardly using maximum likelihood. estimator of the selection mechanism.
Bajari et al. ( ) suggest a more flexible
specification of the selection mechanism, while
accounting for the possibility that players Partial Identification of Model Parameters
randomise across their actions. In their model, i//( -)
cannot depend on (x, s) directly, but only through Given knowledge of P(y, x), model parameters can
be partially identified even in the absence of
players’ payoffs. Using this restriction, Bajari et al.
( ) provide a parametrisation of assumptions on the nature of competition, hetero-
geneity of firms, availability of covariates with
i//( -) which explicitly accounts for criteria of equi-
librium selection often discussed in economic the- sufficiently large support and exclusion restrictions,
ory. In particular, they assume: and restrictions on the selection mechanism i//( •).
In particular, the sharp identification region of 9,
ij/(cT;x, e ) = i j / ( c r ; denoted 0/, is given by the set of parameter vectors
which are consistent with the sampling process and
Sg(x, e ) , a) exp (a ■ z(a, the maintained modelling assumptions, and
71)) therefore may have generated the distribution of
Y.o’eSe{x,z)eXP^-<a'^y observables. Berry and Tamer ( )
provide the following definition of 0/ in the two-
where z is a vector of covariates including, for player entry model described in ‘A simple example’:
example, dummy variables for whether the equi-
librium a C Sg(x, e) is in piue strategies, is Pareto
dominated, maximises industry profits, and is risk
dominant. Bajari et al. ( ) show that under

3ij/ such that Vt € {(0,0), (1,0), (0,1), (1,1)},

(0 . 1 ) 0 7
9 e 0 : P(y = t\x) = Y iA(cr;ue)o'i(0)o'2(t2) I dFe{e)x - a . s .
o (E Sj? (A , g)

where \]/ is an admissible equilibrium selection A computationally simple procedure to estimate

mechanism as described in ‘A simple example’.
This formulation is theoretically attractive, but
computationally challenging to implement. This is
because when no assumptions are placed on it, the
selection mechanism \j/ may represent an infinite-
dimensional nuisance parameter.
an outer region for the model parameters is
provided by Ciliberto and Tamer ( ). An
outer region includes all the parameter values in the
parameter space that may have generated the
observables, but may include other (infeasible)
parameter values as well. They
Econometric Issues in the Presence of Multiple Equilibria 3197

observe that for a given t e {(0,0), (1,0), (0,1),
(2.1) }, the model implies that P(y /|v) cannot
be larger than the probability that t is a possible
equilibrium outcome of the game, and cannot be
smaller than the probability that t is the unique
equilibrium outcome of the game. This is because
for a given 9 C 0 and any realisation of (v, e) such
that t is a possible equilibrium outcome of the
game, there can be another
outcome f e {(0,0), (1,0), (0,1), (1,1)} which is also
a possible equilibrium outcome of the game, and
when both are possible t is selected only part of the
time. Similarly, t is certainly realised whenever it is
the only possible equilibrium outcome, but it can
additionally be realised when it belongs to a set of
multiple equilibrium outcomes. In the twoplayer
entry game, these considerations yield the
following outer region:

P(£ € #§£) < P(y = (0,0)\x) < P(e e 4!k0))

£l +X I J 8 J S2 + X2[l2
+ Si dFe(e)

p (fie 4‘f) < P(y = (1,0) |x) < p(£e 4!i0))

6e0■ +P(ce<x)
f>CT P(£ e 4°k1)) < P( y = (0, i)|x) < P(e e 4!k1))
u
o
+P(£ € 4y*)
P(e € 4‘iO < P(y = (1, l)\x) < P(e € 41k11)
f £ i + e2 T x2fi2

-dFg{e)
Si
7
e,x

Andrews et al. ( ) suggest using only the
information provided by the model implication that
P(y = i\x) cannot be larger than the probability that
t is a possible equilibrium outcome of the game
(hence, using only the upper bounds in the above
expression), thereby obtaining an outer region that
is simpler to compute than but wider.
Exploiting results in Random Set Theory
Molchanov ( ), Beresteanu et al. (
) propose a formulation of the sharp identifi-
cation region 0/which is computationally tractable.
While their formulation is computationally more
intensive than Ciliberto and Tamer’s, the benefits in
terms of identification yielded by their
methodology can be substantial; see Beresteanu et
al. ( , ) for examples.
Their approach can be summarised as follows.
Given a 6 e © and a realisation of (x, e), one obtains
a realisation of the random set of MSNE Sg(x, e);
see Fig. . Each of the equilibria in
Sg(x, e) determines a probability distribution over
the game’s outcomes conditional on the realisation
of x and £. Denote by Q(Sg(x, s)) the random set of
such probability distributions; see Fig. . Beresteanu
et al. ( , ) establish
that the collection of probability distributions over
outcomes of the game conditional on x which are
consistent with the model (i.e. with all its
implications) is given by the Aumann expectation
of Q(Sg(x, e)) conditional on x, denoted E(Q(Sg(x,
£))\x), which is a closed convex set. (Formally,
E(g(50)|x) = {E(^|.r) : q <5 Q(Sg) a.s.}, see
Molchanov ( , Definition 2.1.13).)
Hence if the model is correctly specified, a
candidate value for 6 belongs to 0/ if and only if
Pty'ly) C E (Q(Sg(x, £))|.r), x — a.s. In other words,
if this condition is satisfied, the candidate 6 may
have generated the observed conditional
distribution F{y\x). Exploiting the notion of support
function of a closed convex set (recall that the
support function of a non-empty compact
a 1
1
! Q(Se) = {(0,0,0,1)}
Q(S0) = 1(0,0,1,0)}
——
-X2P2-82 Q(S0)={(0,1,0,0), —
q(-(t:2+x2 P 2)/$2*"(C | +X | Pi )/61), ‘
-X2P2 (0,0,1,0)} 1 III
-X1P1 -XiPl-fi|l El
». ........ ,: 1: „„ 1 rm s 1 «j ................................... § .
Q(S0) = {(1,0,0,0)} ! «s») = 1(0,1,0,0)}
.......... ■ • ...... ................... 1.......... 11
b e2 1
h(Q(S0),u) = u.,
11 h(Q(S0),u) = u4
1
-X2P2-82 h(Q(So),u) = ii|l:
nuix{u:,u'q(-(e2+X2p2)/82,-
-X2P2 (ci+Xi(3i)/8|), ui} '
-X||l 1 -X|P|-8|!
h(Q(Se),u) = m 1 h(Q(S„>,u) = u2
1
1

Econometric Issues in the Presence of Multiple Equi-
libria, Fig. 2 The random set of probability distributions
over outcome profiles implied by MSNE, Q(Sg), in panel (a),
and the support function in direction u of the random set of
probability distributions over outcome profiles
implied by MSNE, h{Q{Sf),u), in panel (b), in a static,
complete information, simultaneous move, two player entry
game with <5, < 0, j = 1,2 and q(a) = [(1 — af (1 - <r2)
crfil - rr2) (1 - 01)02 o-joJ

convex set B e RK>, denoted h(B,-), is given by h(B, regions with a prespecified probability, can be
u) = maxfc e su'b, u € RK>), Beresteanu et al. ( , carried out using the methodology proposed by
) show that one can verify this Chemozhukov et al. ( ).
condition by checking if the minimum of a sub-
linear (hence convex) function over a convex set is
equal to zero. Specifically, they show that See Also

< 0 € © : min (E[/z(Q(Se), «)Lyl

©i l ;,:|M|<T
—n'P(y|x)) = Ox — a.s.},

with h(Q(Sg), u) the support function of Q(Sg) in

direction u. This minimisation problem can be
solved efficiently using algorithms in convex pro-
gramming. For certain special cases (e.g., games
where players use only pure strategies), Galichon Andrews, D.W.K., S.T. Berry, and P. Jia. 2004. Confidence
and Henry ( ) provide alternative computa
tional methods based on optimal transportation
theory.
Estimation of 0/ and ©\>r, and construction of
confidence sets that asymptotically cover these
Bibliography
regions for parameters in discrete games with multiple
equilibria, with an application to discount chain store
location, mimeo.
Aradillas-Lopez, A., and E. Tamer. 2008. The identification
power of equilibrium in simple games. Journal of
Business and Economic Statistics 26(3): 261-310.
Econometrics
Bajari, P., H. Hong, and S. Ryan. 2009. Identification and
estimation of a discrete game of complete information.
Econometrica (forthcoming).
Beresteanu, A., I.S. Molchanov, and F. Molinari. 2008. Sharp
identification regions in games. CeMMAP working paper
CWP15/08.
Beresteanu, A., I.S.Molchanov, and F. Molinari. 2009. Sharp
identification regions in models with convex predictions:
Games, individual choice, and incomplete data.
CeMMAP working paper CWP27/09.
Berry, S.T. 1992. Estimation of a model of entry in the airline
industry. Econometrica 60(4): 889-917.
Berry, S.T., and E. Tamer. 2007. Identification in models of
oligopoly entry, Chap. 2. In Advances in economics and
econometrics: Theory and application, Ninth world
congress, vol. II, 46-85. Cambridge: Cambridge Uni-
versity Press.
Bjorn, P.A., and Q.H. Vuong. 1985. Simultaneous equations
models for dummy endogenous variables: A game
theoretic formulation with an application to labor force
participation. CalTech DHSS working paper number 557.
Blundell, R., and J.R. Smith. 1994. Coherency and estimation
in simultaneous models with censored or qualitative
dependent variables. Journal of Econometrics 64: 355-
373.
Borzekowski, R., and A.M. Cohen. 2005. Estimating strategic
complementarities in credit unions’ outsourcing decisions.
Working paper, Federal Reserve Board of Governors.
Bresnahan, T.F., and P.C. Reiss. 1988. Do entry conditions
vary across markets. Brookings Papers on Economic
Activity 3: 833-871.
Bresnahan, T.F., and P.C. Reiss. 1990. Entry in monopoly
markets. Review of Economic Studies 57: 531-553.
Bresnahan, T.F., and P.C. Reiss. 1991a. Empirical models of
discrete games. Journal of Econometrics
48: 57-82.
Bresnahan, T.F., and P.C. Reiss. 1991b. Entry and compe-
tition in concentrated markets. Journal of Political
Economy 99(5): 977-1009.
Brock, W., and S. Durlauf. 2001. Discrete choice with social
interactions. Review of Economic Studies 68: 235-260.
Brock, W., and S. Durlauf. 2007. Identification of binary
choice models with social interactions. Journal of
Econometrics 140: 52-75.
Chemozhukov, V., H. Hong, and E. Tamer. 2007. Estimation
and confidence regions for parameter sets in econometric
models. Econometrica 75: 1243-1284.
Ciliberto, F., and E. Tamer. 2009. Market structure and
multiple equilibria in airline markets. Econometrica
(forthcoming).
Galichon, A., and M. Henry. 2008. Inference in models with
multiple equilibria, mimeo.
Gourieroux, C., J.J. Laffont, and A. Monfort. 1980. Coher-
ency conditions in simultaneous linear equation models
with endogenous switching regimes. Econometrica 48:
675-695.
3199
Heckman, J. 1978. Dummy endogenous variables in a
simultaneous equation system. Econometrica 46: 931-
959.
Jia, P. 2008. What happens when Wal-Mart comes to town:
An empirical analysis of the discount retailing industry.
Econometrica 76: 1263-1316.
Jovanovic, B. 1989. Observable implications of models with
multiple equilibria. Econometrica 57: 1431-1437.
Maddala, G.S. 1983. Limited-dependent and qualitative
variables in econometrics. New York: Cambridge Uni-
versity Press.
Mazzeo, M. 2002. Product choice and oligopoly market
structure. RAND Journal of Economics 33(2): 221-242.
Molchanov, I.S. 2005. Theory of random sets. London:
Springer.
Schmidt, P. 1981. Constraints on the parameters in simul-
taneous tobit and probit models, Chap. 12. In Structural
analysis of discrete data and econometric applications,
ed. C. Manski and D. McFadden, 422-434. Cambridge,
MA: MIT Press.
Sweeting, A. 2008. The strategic timing of radio commer-
cials: An empirical analysis using multiple equilibria.
RAND Journal of Economics (forthcoming).
Tamer, E. 2003. Incomplete simultaneous discrete response
model with multiple equilibria. Review of Economic
Studies 70: 147-165.
Econometrics
John Geweke, Joel Horowitz and Hashem Pesaran
Abstract
As a unified discipline, econometrics is still
relatively young and has been transforming and
expanding very rapidly. Major advances have
taken place in the analysis of cross- sectional
data by means of semiparametric and
nonparametric techniques. Heterogeneity of
economic relations across individuals, firms and
industries is increasingly acknowledged and
attempts have been made to take it into account
either by integrating out its effects or by
modelling the sources of heterogeneity when
suitable panel data exist. The counterfac- tual
considerations that underlie policy analysis and
treatment valuation have been given a more
satisfactory foundation. New time-series
3200
econometric techniques have been developed
and employed extensively in the areas of
macroeconometrics and finance. Nonlinear
econometric techniques are used increasingly in
the analysis of cross-section and time-series
observations. Applications of Bayesian tech-
niques to econometric problems have been
promoted largely by advances in computer
power and computational techniques. The use of
Bayesian techniques has in turn provided the
investigators with a unifying framework where
the tasks of forecasting, decision making, model
evaluation and learning can be considered as
parts of the same interactive and iterative
process, thus providing a basis for ‘real time
econometrics’.
Keywords
Acceptance sampling; Adaptive expectations
hypothesis; ARM A processes; Asset pricing
models; Asset return volatility; Auctions;
Baehelier, L.; Bayesian computation; Bayesian
econometrics; Bayesian inference; Benini, R.;
Binary logit and probit models; Bootstrap;
Building cycle; Bunch maps; Causality in eco-
nomics and econometrics; Censored regression
models; Central limit theorems; Cointegration;
Common factors; Conditional hazard functions;
Conditional mean functions; Conditional median
functions; Confluence analysis; Convexity;
Correlation analysis; Cowles Commission; Curse
of dimensionality; Davenant, C.; Diagnostic
tests; Discrete choice models; Discrete response
models; Distributed lags; Douglas, P.H.; Duhem-
Quine thesis; Duration models; Dynamic
decision models; Dynamic specification;
Dynamic stochastic general equilibrium models;
Econometric Society; Econometrics; Economic
distance; Economic laws; Edgeworth
expansions; Edgeworth, F. Y.; Efficient market
hypothesis; Engel curve; Error correction
models; Euler equations; Experimental
economics; Financial econometrics; Fisher, I.;
Fisher, R. A.; Fixed effects and random effects;
Forecast error variances; Forecast evaluation;
Forecasting; Frisch, R. A. K.; Full information
maximum likelihood; Galton, F.; Gaussian
quadrature; Generalized method
Econometrics
of moments; Geometric distributed lag model;
Gibbs sampling; Haavelmo, T.; Habit persis-
tence; Hastings-Metropolis algorithm; Hedonic
prices; Homogeneity; Hooker, R.H.;
Identification; Impulse response analysis; Indi-
rect utility function; Inference; Instrumental
variables; Integration; Inventory cycle; Joint
hypotheses; Juglar cycle; Juglar, C.; K-class
estimators; Kernel estimators; King, G.; Kitchin,
.1.; Kondratieff, N.; Koopmans, T. C.; Kuznets,
S.; Labour market search; Lagrange multiplier;
Latent variables; Least absolute deviations
estimators; Likelihood ratio; Limited
information maximum likelihood; Linear
models; Local linear estimation; Logit models;
Long waves; Longitudinal data; Lucas critique;
Macroeconometric models; Markov chain Monte
Carlo methods; Maximum likelihood;
Measurement; Measurement errors; Method of
simulated moments; Microeconometrics;
Microfoundations; Mis- specification; Mitchell,
W. C.; Model evaluation; Model selection;
Model testing; Model uncertainty; Monotonicity;
Monte Carlo simulation; Moore, H.L.; Multi-
collinearity; Multinomial probit model; National
Bureau of Economic Research; Nonlinear
simultaneous equation models; Nonnested tests;
Nonparametric models; Observed variables;
Ordinary least squares; Parameter uncertainty;
Pearson K.; Petty, W.; Phillips curve; Policy
evaluation; Political arithmeticians; Probability;
Probability calculus; Probability distribution;
Purchasing power parity; Quantile functions;
Random assignment; Random utility models;
Random variables; Random walk theory;
Rational expectations; Real time econometrics;
Regional migration; Regression analysis;
Revealed preference theory; Saddlepoint
expansions; Sampling theory; Schultz, H.;
Semiparametric estimation; Sensitivity analysis;
Series estimators; Significance tests; Simulated
method of moments; Simulation methods;
Simultaneous equations models; Simultaneous
linear equations; Social experimentation in
economics; Spatial econometrics; Specification
tests; Splines; Spurious correlation; State
dependence; State space models;
Econometrics
Statistical inference; Statistics and economics;
Stochastic models; Stock return predictability;
Structural change; Structural estimation; Struc-
tural VAR; Survival models; Three-stage least
squares; Time-series analysis; Tinbergen, J.;
Tobit models; Treatment effect; Truncated
regression models; Uncovered interest parity;
Unit roots; Value distribution; Vector auto-
regressions (VAR); Vining, R.; Waugh, F.;
Weibull hazard model; Working, H.
JEL Classifications
Cl
What Is Econometrics?
Broadly speaking, econometrics aims to give
empirical content to economic relations for testing
economic theories, forecasting, decision making,
and for ex post decision/policy evaluation. The term
‘econometrics’ appears to have been first used by
Pawel Ciompa as early as 1910, although it is
Ragnar Frisch who takes the credit for coining the
term, and for establishing it as a subject in the sense
in which it is known today (see Frisch p. 95;
Bjerkholt ). By emphasizing the quantitative aspects
of economic relationships, econometrics calls for a
‘unification’ of measurement and theory in
economics. Theory without measurement can have
only limited relevance for the analysis of actual
economic problems; while measurement without
theory, being devoid of a framework necessary for
the interpretation of the statistical observations, is
unlikely to result in a satisfactory explanation of the
way economic forces interact with each other.
Neither ‘theory’ nor ‘measurement’ on its own is
sufficient to further our understanding of economic
phenomena.
As a unified discipline, econometrics is still
relatively young and has been transforming and
expanding very rapidly since an earlier version of
this article was published in the first edition of The
New Palgrave: A Dictionary of Economics in 1987
(Pesaran ). Major advances have
taken place in the analysis of crosssectional data by
means of semiparametric and nonparametric
3201
techniques. Heterogeneity of economic relations
across individuals, firms and industries is increas-
ingly acknowledged, and attempts have been made
to take them into account either by integrating out
their effects or by modelling the sources of
heterogeneity when suitable panel data exists. The
counterfactual considerations that underlie policy
analysis and treatment evaluation have been given a
more satisfactory foundation. New time series
econometric techniques have been developed and
employed extensively in the areas of macro-
econometrics and finance. Nonlinear econometric
techniques are used increasingly in the analysis of
cross-section and time-series observations. Appli-
cations of Bayesian techniques to econometric
problems have been given new impetus largely
thanks to advances in computer power and com-
putational techniques. The use of Bayesian tech-
niques has in turn provided the investigators with a
unifying framework where the tasks of forecasting,
decision making, model evaluation and learning can
be considered as parts of the same interactive and
iterative process; thus paving the way for
establishing the foundation of ‘real time
econometrics’. See Pesaran and Timmermann ( )•
This article attempts to provide an overview of
some of these developments. But to give an idea of
the extent to which econometrics has been
transformed over the past decades we begin with a
brief account of the literature that pre-dates
econometrics, and discuss the birth of econometrics
and its subsequent developments to the present.
Inevitably, our accounts will be brief and non-
technical. Readers interested in more details are
advised to consultant the specific entries provided
in the New Palgrave and the excellent general texts
by Maddala ( ), Greene ( ),
Davidson and MacKinnon ( ), and
Wooldridge ( ), as well as texts on specific
topics such as Cameron and Trivedi ( ) on
microeconometrics, Maddala ( ) on econo
metric models involving limited-dependent and
qualitative variables, Arellano ( ), Baltagi
( ), Hsiao ( ), and Wooldridge ( ) on
panel data econometrics, Johansen ( ) on
cointegration analysis, Hall ( ) on generalized
method of moments, Bauwens et al. ( ), Koop
3202
( ), Lancaster ( ), and Geweke ( ) on
Bayesian econometrics, Bosq ( ), Fan and
Gijbels ( ), Horowitz ( ), Hardle ( ),
Hardle and Linton ( ), and Pagan and Ullah
( ) on nonparametric and semiparametric
econometrics, Campbell et al. ( ) and
Gourieroux and Jasiak ( ) on financial econo-
metrics, Granger and Newbold ( ), Lutkepohl
( ), and Hamilton ( ) on time series
analysis.
Quantitative Research in Economics:
Historical Backgrounds
Empirical analysis in economics has had a long and
fertile history, the origins of which can be traced at
least as far back as the work of the 16th-century
political arithmeticians such as William Petty,
Gregory King and Charles Davenant. The political
arithmeticians, led by Sir William Petty, were the
first group to make systematic use of facts and
figures in their studies. They were primarily
interested in the practical issues of their time,
ranging from problems of taxation and money to
those of international trade and finance. The
hallmark of their approach was undoubtedly
quantitative, and it was this which distinguished
them from their contemporaries. Although the
political arithmeticians were primarily and under-
standably preoccupied with statistical measurement
of economic phenomena, the work of Petty, and that
of King in particular, represented perhaps the first
examples of a unified quantitative-theoretical
approach to economics. Indeed Schumpeter in his
History of Economic Analysis ( , p. 209) goes
as far as to say that the works
of the political arithmeticians ‘illustrate to perfec-
tion, what Econometrics is and what Econometri-
cians are trying to do’.
The first attempt at quantitative economic
analysis is attributed to Gregory King, who was the
first to fit a linear function of changes in com prices
on deficiencies in the com harvest, as reported in
Charles Davenant ( ). One
important consideration in the empirical work of
King and others in this early period seems to have
been the discovery of Taws’ in economics, very
Econometrics
much like those in physics and other natural
sciences.
This quest for economic laws was, and to a
lesser extent still is, rooted in the desire to give
economics the status that Newton had achieved for
physics. This was in turn reflected in the conscious
adoption of the method of the physical sciences as
the dominant mode of empirical enquiry in
economics. The Newtonian revolution in physics,
and the philosophy of ‘physical determinism’ that
came to be generally accepted in its aftermath, had
far-reaching consequences for the method as well as
the objectives of research in economics. The
uncertain nature of economic relations began to be
fully appreciated only with the birth of modem
statistics in the late 19th century and as more
statistical observations on economic variables
started to become available.
The development of statistical theory in the
hands of Galton, Edgeworth and Pearson was taken
up in economics with speed and diligence. The
earliest applications of simple correlation analysis in
economics appear to have been carried out by Yule (
, 1896) on the relationship
between pauperism and the method of providing
relief, and by Hooker ( ) on the relationship
between the marriage rate and the general level of
prosperity in the United Kingdom, measured by a
variety of economic indicators such as imports,
exports, and the movement in com prices.
Benini ( ), the Italian statistician was the
first to make use of the method of multiple regres-
sion in economics. But Henry Moore (
) was the first to place the statistical estima-
tion of economic relations at the centre of quanti-
tative analysis in economics. Through his relentless
efforts, and those of his disciples and followers Paul
Douglas, Henry Schultz, Holbrook Working, Fred
Waugh and others, Moore in effect laid the
foundations of ‘statistical economics’, the precursor
of econometrics. The monumental work of Schultz,
The Theory and the Measurement of Demand ( ), in
the United States and that of
Allen and Bowley, Family Expenditure ( ), in
the United Kingdom, and the pioneering works of
Lenoir ( ), Wright ( , ), Working
( ), Tinbergen ( -1930), and Frisch
( ) on the problem of ‘identification’
Econometrics
represented major steps towards this objective. The
work of Schultz was exemplary in the way it
attempted a unification of theory and measurement
in demand analysis; while the work on identification
highlighted the importance of ‘structural estimation’
in econometrics and was a crucial factor in the
subsequent developments of econometric methods
under the auspices of the Cowles Commission for
Research in Economics.
Early empirical research in economics was by
no means confined to demand analysis. Louis
Bachelier ( ), using time-series data on
French equity prices, recognized the random walk
character of equity prices, which proved to be the
precursor to the vast empirical literature on market
efficiency hypothesis that has evolved since the
early 1960s. Another important area was research
on business cycles, which provided the basis of the
later development in time-series analysis and macro
econometric model building and forecasting.
Although, through the work of Sir William Petty
and other early writers, economists had been aware
of the existence of cycles in economic time series, it
was not until the early 19th century that the
phenomenon of business cycles began to attract the
attention that it deserved. Clement Juglar (1819-
1905), the French physician turned economist, was
the first to make systematic use of time-series data
to study business cycles, and is credited with the
discovery of an investment cycle of about 7-11
years duration, commonly known as the Juglar
cycle. Other economists such as Kitchin, Kuznets
and Kondratieff followed Juglar’s lead and
discovered the inventory cycle (3-5 years duration),
the building cycle (15-25 years duration) and the
long wave (45-60 years duration), respectively. The
emphasis of this early research was on the
morphology of cycles and the identification of
periodicities. Little attention was paid to the quan-
tification of the relationships that may have under-
lain the cycles. Indeed, economists working in the
National Bureau of Economic Research under the
direction of Wesley Mitchell regarded each business
cycle as a unique phenomenon and were therefore
reluctant to use statistical methods except in a
nonparametric manner and for purely descriptive
purposes (see, for example, Mitchell
3203
; Bums and Mitchell ). This view of business
cycle research stood in sharp contrast to the
econometric approach of Frisch and Tinbergen and
culminated in the famous methodological
interchange between I jailing Koopmans and Rut-
ledge Vining about the roles of theory and mea-
surement in applied economics in general and
business cycle research in particular. (This inter-
change appeared in the August 1947 and May 1949
issues of the Review of Economics and Statistics.)
The Birth of Econometrics
Although, quantitative economic analysis is a good
three centuries old, econometrics as a recognized
branch of economics began to emerge only in the
1930s and the 1940s with the foundation of the
Econometric Society, the Cowles Commission in
the United States, and the Department of Applied
Economics (DAE) in Cambridge, England. (An
account of the founding of the first two
organizations can be found in Christ ,
, while the history of the DAE is covered in
Stone .) This was largely due to the multi-
disciplinary nature of econometrics, comprising of
economic theory, data, econometric methods and
computing techniques. Progress in empirical
economic analysis often requires synchronous
developments in all these four components.
Initially, the emphasis was on the development
of econometric methods. The first major debate
over econometric method concerned the applica-
bility of the probability calculus and the newly
developed sampling theory of R.A. Fisher to the
analysis of economic data. Frisch ( ) was
highly sceptical of the value of sampling theory and
significance tests in econometrics. His objection
was not, however, based on the epistemological
reasons that lay behind Robbins’s and Keynes’s
criticisms of econometrics. He was more concerned
with the problems of multi- collinearity and
measurement errors which he believed were
pervasive in economics; and to deal with the
measurement error problem he developed his
confluence analysis and the method of ‘bunch
maps’. Although used by some
3204
econometricians, notably Tinbergen ( ) and
Stone ( ), the bunch map analysis did not
find much favour with the profession at large.
Instead, it was the probabilistic rationalizations of
regression analysis, advanced by Koopmans ( ) and
Haavelmo ( ), that formed the
basis of modern econometrics.
Koopmans did not, however, emphasize the
wider issue of the use of stochastic models in
econometrics. It was Haavelmo who exploited the
idea to the full, and argued for an explicit probabil-
ity approach to the estimation and testing of eco-
nomic relations. In his classic paper published as a
supplement to Econometrica in 1944, Haavelmo
defended the probability approach on two grounds.
First, he argued that the use of statistical measures
such as means, standard errors and correlation
coefficients for inferential purposes is justified only
if the process generating the data can be cast in
terms of a probability model. Second, he argued that
the probability approach, far from being limited in
its application to economic data, because of its
generality is in fact particularly suited for the
analysis of ‘dependent’ and ‘nonhomogeneous’
observations often encountered in economic
research.
The probability model is seen by Haavelmo as a
convenient abstraction for the purpose of under-
standing, or explaining or predicting, events in the
real world. But it is not claimed that the model
represents reality in all its details. To proceed with
quantitative research in any subject, economics
included, some degree of formalization is inevita-
ble, and the probability model is one such formal-
ization. The attraction of the probability model as a
method of abstraction derives from its generality
and flexibility, and the fact that no viable alternative
seems to be available. Haavelmo’s contribution was
also important as it constituted the first systematic
defence against Keynes’s ( ) influ
ential criticisms of Tinbergen’s pioneering research
on business cycles and macro- econometric
modelling. The objective of Tinbergen’s research
was twofold: first, to show how a macro
econometric model may be constructed and then
used for simulation and policy analysis
(Tinbergen ); second, ‘to submit to statistical test
some of the theories which have
Econometrics
been put forward regarding the character and causes
of cyclical fluctuations in business activity’
(Tinbergen , p. 11). Tinbergen assumed a rather
limited role for the econometrician in the process of
testing economic theories, and argued that it was the
responsibility of the ‘economist’ to specify the
theories to be tested. He saw the role of the
econometrician as a passive one of estimating the
parameters of an economic relation already
specified on a priori grounds by an economist. As
far as statistical methods were concerned, he
employed the regression method and Frisch’s
method of confluence analysis in a complementary
fashion. Although Tinbergen discussed the
problems of the determination of time lags, trends,
structural stability and the choice of functional
forms, he did not propose any systematic method-
ology for dealing with them. In short, Tinbergen
approached the problem of testing theories from a
rather weak methodological position. Keynes saw
these weaknesses and attacked them with charac-
teristic insight (Keynes ). A large part of Keynes’s
review was in fact concerned with technical
difficulties associated with the application of
statistical methods to economic data. Apart from the
problems of the ‘dependent’ and ‘nonhomogeneous’
observations mentioned above, Keynes also
emphasized the problems of mis- specification,
multicollinearity, functional form, dynamic
specification, structural stability, and the difficulties
associated with the measurement of theoretical
variables. By focusing his attack on Tinbergen’s
attempt at testing economic theories of business
cycles, Keynes almost totally ignored the practical
significance of Tinbergen’s work for econometric
model building and policy analysis (for more
details, see Pesaran and Smith ).
In his own review of Tinbergen’s work, Haa-
velmo ( ) recognized the main burden of the
criticisms of Tinbergen’s work by Keynes and
others, and argued the need for a general statistical
framework to deal with these criticisms. As we have
seen, Haavelmo’s response, despite the views
expressed by Keynes and others, was to rely more,
rather than less, on the probability model as the
basis of econometric methodology. The technical
problems raised by Keynes and
Econometrics
others could now be dealt with in a systematic
manner by means of formal probabilistic models.
Once the probability model was specified, a solu-
tion to the problems of estimation and inference
could be obtained by means of either classical or of
Bayesian methods. There was little that could now
stand in the way of a rapid development of
econometric methods.
Early Advances in Econometric Methods
Haavelmo’s contribution marked the beginning of a
new era in econometrics, and paved the way for the
rapid development of econometrics, with the
likelihood method gaining importance as a tool for
identification, estimation and inference in
econometrics.
Identification of Structural Parameters
The first important breakthrough came with a
formal solution to the identification problem which
had been formulated earlier by Working ( ). By
defining the concept of ‘structure’ in
terms of the joint probability distribution of obser-
vations, Haavelmo ( ) presented a very gen
eral concept of identification and derived the
necessary and sufficient conditions for identifica-
tion of the entire system of equations, including the
parameters of the probability distribution of the
disturbances. His solution, although general, was
rather difficult to apply in practice. Koopmans et al.
( ) used the term ‘identifica
tion’ for the first time in econometrics, and gave the
now familiar rank and order conditions for the
identification of a single equation in a system of
simultaneous linear equations. The solution of the
identification problem by Koopmans ( ) and
Koopmans et al. ( ) was obtained in the case
where there are a priori linear restrictions on the
structural parameters. They derived rank and order
conditions for identifiability of a single equation
from a complete system of equations without
reference to how the variables of the model are
classified as endogenous or exogenous. Other
solutions to the identification problem, also
allowing for restrictions on the elements of the
variance-covariance matrix of the structural
3205
disturbances, were later offered by Wegge ( ) and
Fisher ( ).
Broadly speaking, a model is said to be identi-
fied if all its structural parameters can be obtained
from the knowledge of its implied joint probability
distribution for the observed variables. In the case of
simultaneous equations models prevalent in
econometrics, the solution to the identification
problem depends on whether there exists a sufficient
number of a priori restrictions for the derivation of
the structural parameters from the reduced- form
parameters. Although the purpose of the model and
the focus of the analysis on explaining the variations
of some variables in terms of the unexplained
variations of other variables is an important
consideration, in the final analysis the specification
of a minimum number of identifying restrictions
was seen by researchers at the Cowles Commission
to be the function and the responsibility of
‘economic theory’. This attitude was very much
reminiscent of the approach adopted earlier by
Tinbergen in his business cycle research: the
function of economic theory was to provide the
specification of the econometric model, and that of
econometrics to furnish statistically optimal
methods of estimation and inference. More specif-
ically, at the Cowles Commission the primary task
of econometrics was seen to be the development of
statistically efficient methods for the estimation of
structural parameters of an a priori specified system
of simultaneous stochastic equations.
More recent developments in identification of
structural parameters in context of semiparametric
models is discussed below in section “Nonpara-
metric and Semiparametric Estimation”. See also
Manski ( ).
Estimation and Inference in Simultaneous
Equation Models
Initially, under the influence of Haavelmo’s
contribution, the maximum likelihood (ML)
estimation method was emphasized as it yielded
consistent estimates. Anderson and Rubin ( )
developed the limited information maximum like-
lihood (LIML) method, and Koopmans et al. ( )
proposed the frill information maximum
likelihood (FIML). Both methods are based on the
joint probability distribution of the endogenous
3206
variables conditional on the exogenous variables
and yield consistent estimates, with the former
utilizing all the available a priori restrictions and the
latter only those which related to the equation being
estimated. Soon, other computationally less
demanding estimation methods followed, both for a
fully efficient estimation of an entire system of
equations and for a consistent estimation of a single
equation from a system of equations.
The two-stage least squares (2SLS) procedure
was independently proposed by Theil (
) and Basmann ( ). At about the same
time the instrumental variable (TV) method, which
had been developed over a decade earlier by
Reiersol ( , ) and Geary ( ) for the
estimation of errors-in-variables models, was gen-
eralized and applied by Sargan ( ) to the esti
mation of simultaneous equation models. Sargan’s
generalized IV estimator (GIVE) provided an
asymptotically efficient technique for using surplus
instruments in the application of the IV method to
econometric problems, and formed the basis of
subsequent developments of the generalized method
of moments (GMM) estimators introduced
subsequently by Hansen ( ). A related class of
estimators, known as
k-class estimators, was also proposed by Theil (
). Methods of estimating the entire system
of equations which were computationally less
demanding than the FIML method were also
advanced. These methods also had the advantage
that, unlike the FIML, they did not require the full
specification of the entire system. These included
the three-stage least squares method due to Zellner
and Theil ( ), the iterated instrumental vari
ables method based on the work of Lyttkens ( ),
Brandy and Jorgenson ( ), and
Dhrymes ( ) and the system k-class estimators
due to Srivastava ( ) and Savin ( ). Impor
tant contributions have also been made in the areas
of estimation of simultaneous nonlinear equations
(Amemiya ), the seemingly unrelated regression
equations (SURE) approach proposed by Zellner (
), and the simultaneous
rational expectations models (see section “
” below).
Interest in estimation of simultaneous equation
models coincided with the rise of Keynesian
Econometrics
economics in early 1960s, and started to wane with
the advent of the rational expectations revolution
and its emphasis on the GMM estimation of the
structural parameters from the Euler equations
(first-order optimization conditions). See section
44 : ??
below. But, with the rise of the dynamic stochastic
general equilibrium models in macro
econometrics, a revival of interest in identification
and estimation of nonlinear simultaneous equation
, models seems quite likely. The recent contribution
of Fernandez-Villaverde and Rubio- Ramirez ( )
represents a start in this direction.
Developments in Time Series Econometrics
While the initiative taken at the Cowles Commis-
sion led to a rapid expansion of econometric tech-
niques, the application of these techniques to
economic problems was rather slow. This was
partly due to a lack of adequate computing facilities
at the time. A more fundamental reason was the
emphasis of the research at the Cowles Commission
on the simultaneity problem almost to the exclusion
of other econometric problems. Since the early
applications of the correlation analysis to economic
data by Yule and Hooker, the serial dependence of
economic time series and the problem of nonsense
or spurious correlation that it could give rise to had
been the single most important factor explaining the
profession’s scepticism concerning the value of
regression analysis in economics. A satisfactory
solution to the spurious correlation problem was
therefore needed before regression analysis of
economic time series could be taken seriously.
Research on this topic began in the mid-1940s at the
Department of Applied Economics (DAE) in
Cambridge, England, as apart of a major
investigation into the measurement and analysis of
consumers’ expenditure in the United Kingdom (see
Stone et al. ). Although the first steps towards the
resolution of the spurious correlation problem had
been taken by Aitken ( -1935)
and Champemowne ( ), the
research in the DAE introduced the problem and its
possible solution to the attention of applied
economists. Orcutt ( ) studied the autocorre
lation pattern of economic time series and showed
that most economic time series can be represented
Econometrics
by simple autoregressive processes with similar
autoregressive coefficients. Subsequently, Cochrane
and Orcutt ( ) made the important
point that the major consideration in the analysis of
stationary time series was the autocorrelation of the
error term in the regression equation and not the
autocorrelation of the economic time series
themselves. In this way they shifted the focus of
attention to the autocorrelation of disturbances as
the main source of concern. Although, as it turns
out, this is a valid conclusion in the case of regres-
sion equations with strictly exogenous regressors, in
more realistic set-ups where the regressors are
weakly exogenous the serial correlation of the
regressors is also likely to be of concern in practice.
See, for example, Stambaugh ( ).
Another important and related development was
the work of Durbin and Watson (
) on the method of testing for residual auto-
correlation in the classical regression model. The
inferential breakthrough for testing serial correla-
tion in the case of observed time-series data had
already been achieved by von Neumann (
), and by Hart and von Neumann ( ). The
contribution of Durbin and Watson was, however,
important from a practical viewpoint as it led to a
bounds test for residual autocorrelation which could
be applied irrespective of the actual values of the
regressors. The independence of the critical bounds
of the Durbin-Watson statistic from the matrix of
the regressors allowed the application of the statistic
as a general diagnostic test, the first of its type in
econometrics. The contributions of Cochrane and
Orcutt and of Durbin and Watson marked the
beginning of a new era in the analysis of economic
time-series data and laid down the basis of what is
now known as the ‘time-series econometrics’
approach.
Consolidation and Applications
The work at the Cowles Commission on identifi-
cation and estimation of the simultaneous equation
model and the development of time series
techniques paved the way for widespread applica-
tion of econometric methods to economic and
financial problems. This was helped significantly
3207
by the rapid expansion of computing facilities,
advances in financial and macroeconomic model-
ling, and the increased availability of economic
data-sets, cross section as well as time series.
Macroeconometric Modelling
Inspired by the pioneering work of Tinbergen, Klein
( , ) was the first to construct a
macroeconometric model in the tradition of the
Cowles Commission. Soon others followed Klein’s
lead. Over a short space of time macroeconometric
models were built for almost every industrialized
country, and even for some developing and
centrally planned economies. Macroeconometric
models became an important tool of ex ante
forecasting and economic policy analysis, and
started to grow in both size and sophistication. The
relatively stable economic environment of the 1950s
and 1960s was an important factor in the initial
success enjoyed by macroeconometric models. The
construction and use of large-scale models
presented a number of important computational
problems, the solution of which was of fundamental
significance, not only for the development of
macroeconometric modelling but also for
econometric practice in general. In this respect
advances in computer technology were clearly
instrumental, and without them it is difficult to
imagine how the complicated computational
problems involved in the estimation and simulation
of large-scale models could have been solved. The
increasing availability of better and faster
computers was also instrumental as far as the types
of problems studied and the types of solutions
offered in the literature were concerned. For
example, recent developments in the area of
microeconometrics (see section “
” below) could hardly
have been possible if it were not for the very
important recent advances in computing facilities.
Dynamic Specification
Other areas where econometrics witnessed signif-
icant developments included dynamic specification,
latent variables, expectations formation, limited
dependent variables, discrete choice models,
random coefficient models, disequilibrium models,
nonlinear estimation, and the
3208
analysis of panel data models. Important advances
were also made in the area of Bayesian economet-
rics, largely thanks to the publication of Zellner’s
textbook ( ), which built on his earlier work
including important papers with George Tiao. The
Seminar on Bayesian Inference in Econometrics and
Statistics (SBIES) was founded shortly after the
publication of the book, and was key in the
development and diffusion of Bayesian ideas in
econometrics. It was, however, the problem of
dynamic specification that initially received the
greatest attention. In an important paper, Brown (
) modelled the hypothesis of habit persis
tence in consumer behaviour by introducing lagged
values of consumption expenditures into an
otherwise static Keynesian consumption function.
This was a significant step towards the incor-
poration of dynamics in applied econometric
research, and allowed the important distinction to be
made between the short-run and the long- run
impacts of changes in income on consumption.
Soon other researchers followed Brown’s lead and
employed his autoregressive specification in their
empirical work.
The next notable development in the area of
dynamic specification was the distributed lag
model. Although the idea of distributed lags had
been familiar to economists through the pioneering
work of Irving Fisher ( ) on the
relationship between the nominal interest rate and
the expected inflation rate, its application in
econometrics was not seriously considered until the
mid-1950s. The geometric distributed lag model
was used for the first time by Koyck ( ) in a
study of investment. Koyck arrived
at the geometric distributed lag model via the
adaptive expectations hypothesis. This same
hypothesis was employed later by Cagan (
in a study of demand for money in conditions of
hyperinflation, by Friedman ( ) in a study of
consumption behavioin and by Nerlove (
in a study of the cobweb phenomenon. The geo-
metric distributed lag model was subsequently
generalized by Solow ( ), Jorgenson (
and others, and was extensively applied in empirical
studies of investment and consumption behaviour.
At about the same time Almon (
provided a polynomial generalization of
Econometrics
I. Fisher’s ( ) arithmetic lag distribution
which was later extended further by Shiller ( ).
Other forms of dynamic specification con
sidered in the literature included the partial adjust-
ment model (Nerlove ; Eisner and Strotz ) and the
multivariate flexible accelerator model (Treadway )
and Sargan’s ( )
work on econometric time series analysis which
formed the basis of error correction and
cointegration analysis that followed next. Following
the contributions of Champemowne ( ),
Granger and Newbold ( ), and Phillips ( )
the spurious regression problem was better under-
stood, and paved the way for the development of the
theory of cointegration. For further details see
section “ ” below.
Techniques for Short-Term Forecasting
Concurrent with the development of dynamic
modelling in econometrics there was also a resur-
gence of interest in time-series methods, used
primarily in short-term business forecasting. The
dominant work in this field was that of Box and
Jenkins ( ), who, building on the pioneering
works of Yule ( , ), Slutsky ( ), Wold
( ), Whittle ( ) and others, proposed com
putationally manageable and asymptotically effi-
cient methods for the estimation and forecasting of
univariate autoregressive-moving average (ARMA)
processes. Time-series models provided an
important and relatively simple benchmark for the
evaluation of the forecasting accuracy of
econometric models, and further highlighted the
significance of dynamic specification in the con-
struction of time-series econometric models. Ini-
tially univariate time-series models were viewed as
mechanical ‘black box’ models with little or no
) basis in economic theory. Their use was seen
primarily to be in short-term forecasting. The
potential value of modem time-series methods in
) econometric research was, however, underlined in
the work of Cooper ( ) and Nelson ( )
who demonstrated the good forecasting perfor-
) mance of univariate Box-Jenkins models relative to
that of large econometric models. These results
raised an important question about the adequacy of
) large econometric models for forecasting as well as
for policy analysis. It was argued that a
Econometrics
properly specified structural econometric model
should, at least in theory, yield more accurate
forecasts than a univariate time-series model. The-
oretical justification for this view was provided by
Zellner and Palm ( ), followed by Trivedi
( ), Prothero and Wallis ( ), Wallis
( ) and others. These studies showed that
Box-Jenkins models could in fact be derived as
univariate final form solutions of linear structural
econometric models. In theory, the pure time-series
model could always be embodied within the struc-
ture of an econometric model and in this sense it did
not present a ‘rival’ alternative to econometric
modelling. This literature further highlighted the
importance of dynamic specification in econometric
models and in particular showed that econometric
models that are outperformed by simple univariate
time-series models most probably suffer from
specification errors.
The papers in Elliott et al. ( ) provide
excellent reviews of recent developments in eco-
nomic forecasting techniques.
A New Phase in the Development of
Econometrics
With the significant changes taking place in the
world economic environment in the 1970s, arising
largely from the breakdown of the Bretton Woods
system and the quadnipling of oil prices, econo-
metrics entered a new phase of its development.
Mainstream macroeconometric models built during
the 1950s and 1960s, in an era of relative economic
stability with stable energy prices and fixed
exchange rates, were no longer capable of
adequately capturing the economic realities of the
1970s. As a result, not surprisingly, macro-
econometric models and the Keynesian theory that
underlay them came under severe attack from
theoretical as well as from practical viewpoints.
While criticisms of Tinbergen’s pioneering attempt
at macroeconometric modelling were received with
great optimism and led to the development of new
and sophisticated estimation techniques and larger
and more complicated models, the disenchantment
with macroeconometric models in 1970s prompted
a much more fundamental reappraisal of
3209
quantitative modelling as a tool of forecasting and
policy analysis.
At a theoretical level it was argued that econo-
metric relations invariably lack the necessary
‘microfoundations’, in the sense that they cannot be
consistently derived from the optimizing behaviour
of economic agents. At a practical level the Cowles
Commission approach to the identification and
estimation of simultaneous macroeconometric
models was questioned by Lucas and Sargent and
by Sims, although from different viewpoints (Lucas
; Lucas and Sargent ; Sims ). There was also a move
away from macroeconometric models and towards
microeconometric research with greater emphasis
on matching of econometrics with individual
decisions.
It also became increasingly clear that
Tinbergen’s paradigm where economic relations
were taken as given and provided by ‘economic
theorist’ was not adequate. It was rarely the case
that economic theory could be relied on for a full
specification of the econometric model (Learner ).
The emphasis gradually shifted from estimation and
inference based on a given tightly parameterized
specification to diagnostic testing, specification
searches, model uncertainty, model validation,
parameter variations, structural breaks, and
semiparametric and nonparametric estimation. The
choice of approach often governed by the purpose
of the investigation, the nature of the economic
application, data availability, computing and
software technology.
What follows is a brief overview of some of the
important developments. Given space limitations
there are inevitably significant gaps. These include
the important contributions of Granger ( ), Sims
( ), and Engle et al. ( ) on different concepts
of ‘causality’ and ‘exogeneity’, the literature on
disequilibrium models (Quandt ; Maddala , ),
random coefficient models (Swamy
; Hsiao and Pesaran 2008, unobserved time
series models (Harvey ), count regression models
(Cameron and Trivedi , ), the
weak instrument problem (Stock et al. ), small
sample theory (Phillips ; Rothenbeig ), econometric
models of auction pricing (Hendricks and Porter ;
Laffont et al. ).
3210
Rational Expectations and the Lucas
Critique
Although the rational expectations hypothesis
(REH) was advanced by Muth in 1961, it was not
until the early 1970s that it started to have a
significant impact on time-series econometrics and
on dynamic economic theory in general. What
brought the REH into prominence was the work of
Lucas ( , ), Sargent ( ),
Sargent and Wallace ( ) and others on the
new classical explanation of the apparent break-
down of the Phillips curve. The message of the REH
for econometrics was clear. By postulating that
economic agents form their expectations
endogenously on the basis of the true model of the
economy, and a correct understanding of the
processes generating exogenous variables of the
model, including government policy, the REH
raised serious doubts about the invariance of the
structural parameters of the mainstream
macroeconometric models in the face of changes in
government policy. This was highlighted in Lucas’s
critique of macroeconometric policy evaluation. By
means of simple examples Lucas ( ) showed that
in models with rational
expectations the parameters of the decision rales of
economic agents, such as consumption or
investment functions, are usually a mixture of the
parameters of the agents’ objective functions and of
the stochastic processes they face as historically
given. Therefore, Lucas argued, there is no reason to
believe that the ‘structure’ of the decision rales (or
economic relations) would remain invariant under a
policy intervention. The implication of the Lucas
critique for econometric research was not, however,
that policy evaluation could not be done, but rather
than the traditional econometric models and
methods were not suitable for this purpose. What
was required was a separation of the parameters of
the policy rale from those of the economic model.
Only when these parameters could be identified
separately given the knowledge of the joint
probability distribution of the variables (both policy
and non-policy variables) would it be possible to
carry out an econometric analysis of alternative
policy options.
Econometrics
There have been a number of reactions to the
advent of the rational expectations hypothesis and
the Lucas critique that accompanied it.
Model Consistent Expectations
The least controversial reaction has been the adop-
tion of the REH as one of several possible expec-
tations formation hypotheses in an otherwise
conventional macroeconometric model containing
expectational variables. In this context the REH, by
imposing the appropriate cross-equation parametric
restrictions, ensures that ‘expectations’ and
‘forecasts’ generated by the model are consistent. In
this approach the REH is regarded as a convenient
and effective method of imposing crossequation
parametric restrictions on time series econometric
models, and is best viewed as the ‘model-
consistent’ expectations hypothesis. There is now a
sizeable literature on solution, identification, and
estimation of linear RE models. The canonical form
of RE models with forward and backward
components is given by
y, = Ay;_! + BE'(y;|l | F t ) + w,,
where yt is a vector of endogenous variables, E (.|
Ft) is the expectations operator, F, the publicly
available information at time t, and w, is a vector of
forcing variables. For example, log-linearized
version of dynamic general equilibrium models (to
be discussed) can all be written as a special case of
this equation with plenty of restrictions on the
coefficient matrices A and B. In the typical case
where w, are serially uncorrelated and the solution
of the RE model can be assumed to be unique, the
RE solution reduces to the vector autoregression
(VAR)
y, = Oy,_! + Gw,
where <I> and G are given in terms of the structural
parameters:
B®2-O + A = 0, andG = (I-BO)-1.
The solution of the RE model can, therefore, be
viewed as a restricted form of VAR popularized in
econometrics by Sims ( ) as a response in
Econometrics
macroeconometric modelling to the rational
expectations revolution. The nature of restrictions is
determined by the particular dependence of A and B
on a few ‘deep’ or structural parameters. For
general discussion of solution of RE models see, for
example, Broze et al. ( ) and Binder
and Pesaran ( ). For studies of identification
and estimation of linear RE models see, for exam-
ple, Hansen and Sargent ( ), Wallis (
Wickens ( ), and Pesaran ( , ).
These studies show how the standard econometric
methods can in principle be adapted to the econo-
metric analysis of rational expectations models.
Detection and Modelling of Structural Breaks
Another reaction to the Lucas critique has been to
treat the problem of ‘structural change’ emphasized
by Lucas as one more potential econometric
‘problem’. Clements and Hendry ( , )
provide a taxonomy of factors behind structural
breaks and forecast failures. Stock and Watson (
) provide extensive evidence of structural
break in macroeconomic time series. It is argued
that structural change can result from many factors
and need not be associated solely with intended or
expected changes in policy. The econometric lesson
has been to pay attention to possible breaks in
economic relations. There now exists a large body
of work on testing for structural change, detection
of breaks (single as well as multiple), and modelling
of break processes by means of piece-wise linear or
non-linear dynamic models (Chow ; Brown et al. ;
Nyblom ; Andrews ; Andrews and Ploberger ; Bai
and Perron ; Pesaran and Timmermann , . See One prominent example was the identification
also the surveys
by Stock ; Clements and Hendry ). The implications
of breaks for short-term and longterm forecasting
have also begun to be addressed (McCulloch and
Tsay ; Koop and Potter , ; Pesaran et al. ).
VAR Macroeconometrics
Unrestricted VARs
The Lucas critique of mainstream macro- extensively in the literature. It continues to assume
econometric modelling also led some
3211
econometricians, notably Sims ( , ), to
doubt the validity of the Cowles Commission style
of achieving identification in econometric models.
Sims focused his critique on macro- econometric
models with a vector autoregressive (VAR)
specification, which was relatively simple to
estimate; and its use soon became prevalent in
macroeconometric analysis. The view that eco-
), nomic theory cannot be relied on to yield identifi-
cation of structural models was not new and had
been emphasized in the past, for example, by Liu (
). Sims took this viewpoint a step further
and argued that in presence of rational expectations
a priori knowledge of lag lengths is indispensable
for identification, even when we have distinct
strictly exogenous variables shifting supply and
demand schedules (Sims , p. 7). While it is true that
the REH complicates the necessary conditions for
the identification of structural models, the basic
issue in the debate over identification still centres
on the validity of the classical dichotomy between
exogenous and endogenous variables (Pesaran ). In
the context of closed-economy macroeconometric
models where all variables are treated as endoge-
nous, other forms of identification of the structure
will be required. Initially, Sims suggested a recur-
sive identification approach where the matrix of
contemporaneous effects was assumed to be lower
(upper) triangular and the structural shocks
orthogonal. Other non-recursive identification
schemes soon followed.
Structural VARs
scheme developed in Blanchard and Quah ( ),
who distinguished between permanent
and transitory shocks and attempted to identify the
structural models through long-run restrictions. For
example, Blanchard and Quah argued that the effect
of a demand shock on real output should be
temporary (that is, it should have a zero long-run
impact), while a supply shock should have a
permanent effect. This approach is known as
‘structural VAR’ (SVAR) and has been used
that structural shocks are orthogonal, but uses a
mixture of short-run and long-mn
3212
restrictions to identify the structural model. In their
work Blanchard and Quah considered a bivariate
VAR model in real output and unemployment. They
assumed real output to be integrated of order 1, or
7(1), and viewed unemployment as an 7(0), or a
stationary variable. This allowed them to associate
the shock to one of the equations as permanent, and
the shock to the other equation as transitory. In
more general settings, such as the one analysed by
Gali ( ) and
Wickens and Motto ( ), where there are
m endogenous variables and r long-run or
cointegrating relations, the SVAR approach pro-
vides m(m — r) restrictions which are not sufficient
to fully identify the model, unless m = 2 and r = 1
which is the simple bivariate model considered by
Blanchard and Quah (Pagan and Pesaran
). In most applications additional short-term
restrictions are required. More recently, attempts
have also been made to identify structural shocks by
means of qualitative restrictions, such as sign
restrictions. Notable examples include Canova and
de Nicolo ( ), Uhlig ( ), and
Peersman ( ).
The focus of the SVAR literature has been on
impulse response analysis and forecast error var-
iance decomposition, with the aim of estimating the
time profile of the effects of monetary policy, oil
price or technology shocks on output and inflation,
and deriving the relative importance of these shocks
as possible explanations of forecast error variances
at different horizons. Typically such analysis is
carried out with respect to a single model
specification, and at most only parameter
uncertainty is taken into account (Kilian ). More
recently the problem of model uncertainty and its
implications for impulse response analysis and
forecasting have been recognized. Bayesian and
classical approaches to model and parameter
uncertainty have been considered. Initially,
Bayesian VAR models were developed for use in
forecasting as an effective shrinkage procedure in
the case of high-dimensional VAR models (Doan et
al.
; Litterman ). The problem of model
uncertainty in cointegrating VARs has been
addressed in Garratt et al. ( , 2006), and
Strachan and van Dijk ( ).
Econometrics
Structural Cointegrating VARs
This approach provides the SVAR with the
decomposition of shocks into permanent and tran-
sitory and gives economic content to the long-run or
cointegrating relations that underlie the transitory
components. In the simple example of Blanchard
and Quah this task is trivially achieved by assuming
real output to be 1(1) and the unemployment rate to
be an 1(0) variable. To have shocks with permanent
effects some of the variables in the VAR must be
non-stationary. This provides a natural link between
the SVAR and the unit root and cointegration
literature. Identification of the cointegrating
relations can be achieved by recourse to economic
theory, solvency or arbitrage conditions (Garratt et
al.
). Also there are often long-mn over-
identifying restrictions that can be tested. Once
identified and empirically validated, the long-run
relations can be embodied within a VAR structure,
and the resultant structural vector error correction
model identified using theory-based short-run
restrictions. The structural shocks can be
decomposed into permanent and temporary com-
ponents using either the multivariate version of the
Beveridge and Nelson ( ) decompositions,
or the one more recently proposed by Garratt et al.
( )•
Two or more variables are said to be
cointegrated if they are individually integrated (or
have a random walk component), but there exists a
linear combination of them which is stationary. The
concept of cointegration was first introduced by
Granger ( ) and more formally
developed in Engle and Granger ( ). Rigorous
statistical treatments followed in the papers by
Johansen ( , ) and Phillips ( ).
Many further developments and extensions have
taken place with reviews provided in Johansen (
), Juselius ( ), and Garratt et al.
( ). The related unit root literature is
reviewed by Stock ( ) and Phillips and
Xiao ( ).
Macroeconometric Models with
Microeconomic Foundations
For policy analysis macroeconometric models need
to be based on decisions by individual
Econometrics 3213

households, firms and governments. This is a the dissatisfaction with econometrics during the
daunting undertaking and can be achieved only by 1970s caused a shift of focus from problems of
gross simplification of the complex economic estimation to those of model evaluation and testing.
interconnections that exists across millions of This shift has been part of a concerted effort to
decision-makers worldwide. The dynamic sto- restore confidence in econometrics, and has
chastic general equilibrium (DSGE) modelling received attention from Bayesian as well as classical
approach attempts to implement this task by viewpoints. Both these views reject the ‘axiom of
focusing on optimal decisions of a few represen- correct specification’ which lies at the basis of most
traditional econometric practices, but they differ
tative agents operating with rational expectations
under complete learning. Initially, DSGE models markedly as how best to proceed.
were small and assumed complete markets with It is generally agreed, by Bayesians as well as by
non-Bayesians, that model evaluation involves
instantaneous price adjustments, and as a result did
not fit the macroeconomic time series (Kim and considerations other than the examination of the
Pagan ). More recently, Smets and Wouters ( ) statistical properties of the models, and personal
have shown that DSGE models judgements inevitably enter the evaluation process.
Models must meet multiple criteria which are often
with sticky prices and wages along the lines devel-
oped by Christiano et al. ( in conflict. They should be relevant in the sense that
) are sufficiently
they ought to be capable of answering the questions
rich to match most of the statistical features of the
main macroeconomic time series. Moreover, by for which they are constructed. They should be
applying Bayesian estimation techniques, these consistent with the accounting and/or theoretical
authors have shown that even relatively large structure within which they operate. Finally, they
models can be estimated as a system. Bayesian should provide adequate representations of the
DSGE models have also shown to perform rea- aspects of reality with which they are concerned.
sonably well in forecasting as compared with These criteria and their interaction are discussed in
Pesaran and Smith (
standard and Bayesian vector autoregressions. It is ). More detailed
breakdowns of the
also possible to incorporate long-run cointegrating
relations within Bayesian DSGE models. The criteria of model evaluation can be found in Hendry
problems of parameter and model uncertainty can and Richard ( ) and McAleer et al.
also be readily accommodated using data-coherent ( ). In econometrics it is, however, the crite
DSGE models. Other extensions of the DSGE rion of ‘adequacy’ which is emphasized, often at the
expense of relevance and consistency.
models to allow for learning, regime switches, time
variations in shock variances, asset prices, and The issue of model adequacy in mainstream
multi-country interactions are likely to enhance econometrics is approached either as a model
selection problem or as a problem in statistical
their policy relevance (Del Negro and Schorfheide ;
Del Negro et al. inference whereby the hypothesis of interest is
tested against general or specific alternatives. The
; An and Schorfheide ; Pesaran and Smith ).
use of absolute criteria such as measures of
Further progress will also be welcome in the area of
macroeconomic policy analysis under model fit/parsimony or formal Bayesian analysis based on
uncertainty, and robust policymaking (Brock and posterior odds are notable examples of model
Durlauf ; Hansen and Sargent ). selection procedures, while likelihood ratio, Wald
and Lagrange multiplier tests of nested hypotheses
and Cox’s centred log-likelihood ratio tests of non-
Model and Forecast Evaluation nested hypotheses are examples of the latter
approach. The distinction between these two general
While in the 1950s and 1960s research in econo- approaches basically stems from the way alternative
metrics was primarily concerned with the identi- models are treated. In the case of model selection
fication and estimation of econometric models, (or model discrimination) all the models under
consideration enjoy the same status
3214
and the investigator is not committed a priori to any
one of the alternatives. The aim is to choose the
model which is likely to perform best with respect
to a particular loss function. By contrast, in the
hypothesis-testing framework the null hypothesis
(or the maintained model) is treated differently from
the remaining hypotheses (or models). One
important feature of the model- selection strategy is
that its application always leads to one model being
chosen in preference to other models. But, in the
case of hypothesis testing, rejection of all the
models under consideration is not ruled out when
the models are nonnested. A more detailed
discussion of this point is given in Pesaran and
Deaton ( ).
Broadly speaking, classical approaches to the
problem of model adequacy can be classified
depending on how specific the alternative hypoth-
eses are. These are the general specification tests,
the diagnostic tests, and the non-nested tests. The
first of these, pioneered by Durbin ( ) and
introduced in econometrics by Ramsey ( ),
Wu ( ), Hausman ( ), and subsequently
developed further by White ( , ) and
Hansen ( ), are designed for circumstances
where the nature of the alternative hypothesis is
kept (sometimes intentionally) rather vague, the
purpose being to test the null against a broad class
of alternatives. (The pioneering contribution of
Durbin , in this area has been documented by
Nakamura and Nakamura .) Important examples of
general specification tests are Ramsey’s regression
specification error test (RESET) for omitted
variables and/or mis- specified functional forms,
and the Durbin-Hausman-Wu test of
misspecification in the context of measurement
error models and/or simultaneous equation models.
Such general specification tests are particularly
useful in the preliminary stages of the modelling
exercise.
In the case of diagnostic tests, the model under
consideration (viewed as the null hypothesis) is
tested against more specific alternatives by
embedding it within a general model. Diagnostic
tests can then be constructed using the likelihood
ratio, Wald or Lagrange multiplier (LM) principles
to test for parametric restrictions imposed on the
general model. The application of the LM
Econometrics
principle to econometric problems is reviewed in
the papers by Breusch and Pagan ( ), Godfrey
and Wickens ( ), and Engle ( ). An excel-
lent review is provided in Godfrey ( ). Exam
ples of the restrictions that may be of interest as
diagnostic checks of model adequacy include zero
restrictions, parameter stability, serial correlation,
heteroskedasticity, functional forms, and normality
of errors. The distinction made here between
diagnostic tests and general specification tests is
more apparent than real. In practice some diagnostic
tests such as tests for serial correlation can also be
viewed as a general test of specification.
Nevertheless, the distinction helps to focus attention
on the purpose behind the tests and the direction
along which high power is sought.
The need for non-nested tests arises when the
models under consideration belong to separate
parametric families in the sense that no single
model can be obtained from the others by means of
a suitable limiting process. This situation, which is
particularly prevalent in econometric research, may
arise when models differ with respect to their
theoretical underpinnings and/or their auxiliary
assumptions. Unlike the general specification tests
and diagnostic tests, the application of non-nested
tests is appropriate when specific but rival
hypotheses for the explanation of the same
economic phenomenon have been advanced.
Although non-nested tests can also be used as
general specification tests, they are designed
primarily to have high power against specific
models that are seriously entertained in the
literature. Building on the pioneering work of Cox (
, ), a number of such tests for
single equation models and systems of simultaneous
equations have been proposed (Pesaran and
Weeks ).
The use of statistical tests in econometrics,
however, is not a straightforward matter and in most
applications does not admit of a clear-cut
interpretation. This is especially so in circumstances
where test statistics are used not only for checking
the adequacy of a given model but also as guides to
model construction. Such a process of model
construction involves specification searches of the
type emphasized by Learner ( ) and presents
insurmountable pre-test
Econometrics
problems which in general tend to produce econo-
metric models whose ‘adequacy’ is more apparent
than real. As a result, in evaluating econometric
models less reliance should be placed on those
indices of model adequacy that are used as guides to
model construction, and more emphasis should be
given to the performance of models over other data-
sets and against rival models.
A closer link between model evaluation and the
underlying decision problem is also needed.
Granger and Pesaran ( , ) discuss this prob
lem in the context of forecast evaluation. A recent
survey of forecast evaluation literature can be found
in West ( ). Pesaran and Skouras
( ) provide a review from a decision-theoretic
perspective.
The subjective Bayesian approach to the treat-
ment of several models begins by assigning a prior
probability to each model, with the prior proba-
bilities summing to 1. Since each model is already
endowed with a prior probability distribution for its
parameters and for the probability distribution of
observable data conditional on its parameters, there
is then a complete probability distribution over the
space of models, parameters, and observable data.
(No particular problems arise from nonnesting of
models in this framework.) This probability space
can then be augmented with the distribution of an
object or vector of objects of interest. For example,
in a macroeconomic policy setting the models could
include VARs, DSGEs and traditional large-scale
macroeconomic models, and the vector of interest
might include future output growth, interest rates,
inflation and unemployment, whose distribution is
implied by each of the models considered. Implicit
in this formulation is the conditional distribution of
the vector of interest conditional on the observed
data. Technically, this requires the integration (or
marginalization) of parameters in each model as
well as the models themselves. As a practical matter
this usually proceeds by first computing the
probability of each model conditional on the data,
and then using these probabilities as weights in
averaging the posterior distribution of the vector of
interest in each model. It is not necessary to choose
one particular model, and indeed to do so would be
suboptimal. The ability
3215
to actually carry out this simultaneous consideration
of multiple models has been enhanced greatly by
recent developments in simulation methods,
surveyed in section “
” below;
recent texts by Koop ( ), Lancaster ( ),
and Geweke ( ) provide technical details.
Geweke and Whiteman ( ) specifically out
line these methods in the context of economic
forecasting.
Microeconometrics: An Overview
Partly as a response to the dissatisfaction with
macro econometric time-series research and partly
in view of the increasing availability of micro data
and computing facilities, since the mid-1980s sig-
nificant advances have been made in the analysis of
micro data. Important micro data-sets have become
available on households and firms especially in the
United States in such areas as housing,
transportation, labour markets and energy. These
data sets include various longitudinal surveys (for
example, University of Michigan Panel Study of
Income Dynamics, and Ohio State National
Longitudinal Study Surveys), cross- sectional
surveys of family expenditures, population and
labour force surveys. This increasing availability of
micro-data, while opening up new possibilities for
analysis, has also raised a number of new and
interesting econometric issues primarily originating
from the nature of the data. The errors of
measurement are likely to be important in the case
of some micro data-sets. The problem of the
heterogeneity of economic agents at the micro level
cannot be assumed away as readily as is usually
done in the case of macro data by appealing to the
idea of a ‘representative’ firm or a ‘representative’
household.
The nature of micro data, often being qualitative
or limited to a particular range of variations, has
also called for new econometric models and
techniques. Examples include categorical survey
responses (‘up’, ‘same’ or ‘down’), and censored or
tmncated observations. The models and issues
considered in the microeconometric literature are
wide ranging and include fixed and random effect
3216
panel data models (for example, Mundlak , ), logit
and probit models and their multi- nominal
extensions, discrete choice or quantal response
models (Manski and McFadden ), continuous time
duration models (Heckman and Singer ), and
microeconometric models of count data (Hausman
et al. ; Cameron and Trivedi ).
The fixed or random effect models provide the
basic statistical framework and will be discussed in
more detailed below. Discrete choice models are
based on an explicit characterization of the choice
process and arise when individual decision makers
are faced with a finite number of alternatives to
choose from. Examples of discrete choice models
include transportation mode choice (Domenich and
McFadden ), labour force participation (Heckman
and Willis ), occupation choice (Boskin ), job or
firm location (Duncan ), and models with
neighbourhood effects (Brock and Durlauf ).
Limited dependent variables models are commonly
encountered in the analysis of survey data and are
usually categorized into truncated regression
models and censored regression models. If all
observations on the dependent as well as on the
exogenous variables are lost when the dependent
variable falls outside a specified range, the model is
called truncated, and, if only observations on the
dependent variable are lost, it is called censored.
The literature on censored and truncated regression
models is vast and overlaps with developments in
other disciplines, particularly in biometrics and
engineering. Maddala ( , eh. 6) provides a
survey.
The censored regression model was first intro-
duced into economics by Tobin ( ) in his
pioneering study of household expenditure on
durable goods, where he explicitly allowed for the
fact that the dependent variable, namely, the
expenditure on durables, cannot be negative. The
model suggested by Tobin and its various gener-
alizations are known in economics as Tobit models
and are surveyed in detail by Amemiya ( ),
and more recently in Cameron and Trivedi
( , eh. 14). Continuous time duration models,
also known as survival models, have been used in
analysis of unemployment duration, the period of
Econometrics
time spent between jobs, durability of marriage, and
so on. Application of survival models to analyse
economic data raises a number of important issues
resulting primarily from the non-controlled
experimental nature of economic observations,
limited sample sizes (that is, time periods), and the
heterogeneous nature of the economic environment
within which agents operate. These issues are
clearly not confined to duration models and are also
present in the case of other micro- econometric
investigations that are based on time series or cross-
section or panel data.
Partly in response to the uncertainties inherent
in econometric results based on non-experimental
data, there has also been a significant move towards
social experimentation, and experimental
economics in general. A social experiment aims at
isolating the effects of a policy change (or a treat-
ment effect ) by comparing the consequences of an
exogenous variation in the economic environment
of a set of experimental subjects known as the
‘treatment’ group with those of a ‘control’ group
that have not been subject to the change. The basic
idea goes back to the early work of R.A. Fisher (
) on randomized trials, and has been applied
extensively in agricultural and biomedical research.
The case for social experimentation in economics is
discussed in Burtless ( ).
Hausman and Wise ( ) and Heckman and
Smith ( ) consider a number of actual social
experiments carried out in the United States, and
discuss their scope and limitations.
Experimental economics tries to avoid some of
the limitations of working with observations
obtained from natural or social experiments by
using data from laboratory experiments to test
economic theories by fixing some of the factors and
identifying the effects of other factors in a way that
allows ceteris paribus comparisons. A wide range
of topics and issues are covered in this literature,
such as individual choice behaviour, bargaining,
provision of public goods, theories of learning,
auction markets, and behavioural finance. A
comprehensive review of major areas of
experimental research in economics is provided in
Kagel and Roth ( ).
These developments have posed new problems
and challenges in the areas of experimental
Econometrics
design, statistical methods and policy analysis.
Another important aspect of recent developments in
micro econometric literature relates to the use of
microanalytic simulation models for policy analysis
and evaluation to reform packages in areas such as
health care, taxation, social security systems, and
transportation networks. Cameron and Trivedi (
) review the recent developments in
methods and application of microeconometrics.
Some of these topics will be discussed in more
detail below.
Econometrics of Panel Data
Panel data models are used in many areas of
econometrics, although initially they were devel-
oped primarily for the analysis of micro behaviour,
and focused on panels formed from cross- section of
N individual households or firms surveyed for T
successive time periods. These types of panels are
often referred to as ‘micropanels’. In social and
behavioural sciences they are also known as
longitudinal data or panels. The literature on micro-
panels typically takes N to be quite large (in
hundreds) and T rather small, often less than ten.
But more recently, with the increasing availability
of financial and macroeconomic data, analyses of
panels where both N and T are relatively large have
also been considered. Examples of such data-sets
include time series of company data from
Datastream, country data from International
Financial Statistics or the Penn World Table, and
county and state data from national statistical
offices. There are also pseudo panels of firms and
consumers composed of repeated cross sections that
cover cross-section units that are not necessarily
identical but are observed over relatively long time
periods. Since the available cross-section
observations do not (necessarily) relate to the same
individual unit, some form of grouping of the cross-
section units is needed. Once the grouping criteria
are set, the estimation can proceed using fixed
effects estimation applied to group averages if the
number of observations per group is sufficiently
large; otherwise possible measurement errors of the
group averages also need to be taken into account.
Deaton ( ) are needed for nonlinear
3217
pioneered the econometric analysis of pseudo
panels. Verbeek (2008) provides a recent review.
Use of panels can enhance the power of empir-
ical analysis and allows estimation of parameters
that might not have been identified using the time
or the cross-section dimensions alone. These ben-
efits come at a cost. In the case of linear panel data
models with a short time span the increased power
is usually achieved under assumptions of parameter
homogeneity and error cross-section independence.
Short panels with autocorrelated disturbances also
pose a new identification problem, namely, how to
distinguished between dynamics and state
dependence (Arellano ch. 5). In panels with fixed
effects the homogeneity assumption is relaxed
somewhat by allowing the intercepts in the panel
regressions to vary freely over the cross-section
units, but continues to maintain the error cross-
section independence assumption. The random
coefficient specification of Swamy ( ) further
relaxes the slope homo
geneity assumption, and represents an important
generalization of the random effects model (Hsiao
and Pesaran ). In micro-panels where T is small
cross-section dependence can be dealt with if it can
be attributed to spatial (economic or geographic)
effects. Anselin ( ) and Anselin
et al. ( ) provide surveys of the literature on
spatial econometrics. A number of studies have also
used measures such as trade or capital flows to
capture economic distance, as in Conley and Topa (
), Conley and Dupor ( ), and
Pesaran et al. ( ).
Allowing for dynamics in panels with fixed
effects also presents additional difficulties; for
example, the standard within-group estimator will
be inconsistent unless T —» oo (Nickell ). In linear
dynamic panels the incidental parameter problem
(the unobserved heterogeneity) can be resolved by
first differencing the model and then estimating the
resultant first-differenced specification by
instrumental variables or by the method of
transformed likelihood (Anderson and Hsiao , ;
Holtz-Eakin et al. ;
Arellano and Bond ; Hsiao et al. ). A similar
procedure can also be followed in the case of short
T panel VARs (Binder et al. ). But other approaches
3218
panel data models. See, for example, Honore and
Kyriazidou ( ) and review of the literature on
nonlinear panels in Arellano and Honore ( ).
Relaxing the assumption of slope homogeneity in
dynamic panels is also problematic, and neglecting
to take account of slope heterogeneity will lead to
inconsistent estimators. In the presence of slope
heterogeneity Pesaran and Smith ( ) show that
the within-group estimator
remains inconsistent even if both N and T oo. A
Bayesian approach to estimation of micro dynamic
panels with random slope coefficients is proposed
in Hsiao et al. ( ).
To deal with general dynamic specifications,
possible slope heterogeneity and error cross- section
dependence, large T and N panels are required. In
the case of such large panels it is possible to allow
for richer dynamics and parameter heterogeneity.
Cross-section dependence of errors can also be dealt
with using residual common factor structures. These
extensions are particularly relevant to the analysis
of purchasing power parity hypothesis (O’Connell ;
Imbs et al. ; Pedroni ; Smith et al. ), output
convergence (Durlauf et al. ; Pesaran ), the Fisher
effect (Westerlund ),
house price convergence (Holly et al. ),
regional migration (Fachin ), and uncovered interest
parity (Moon and Perron ). The econometric
methods developed for large panels has to take into
account the relationship between the increasing
number of time periods and cross- section units
(Phillips and Moon ). The relative expansion rates
of A and T could have important consequences for
the asymptotic and small sample properties of the
panel estimators and tests. This is because fixed
Testimation bias tend to magnify with increases in
the cross-section dimension, and it is important that
any bias in the T dimension is corrected in such a
way that its overall impact disappears as both N and
/' * oc, jointly.
The first generation panel unit root tests pro-
posed, for example, by Levin et al. ( ) and Im
et al. ( ) allowed for parameter heterogeneity
but assumed errors were cross-sectionally inde-
pendent. More recently, panel unit root tests that
Econometrics
allow for error cross-section dependence have been
proposed by Bai and Ng ( ), Moon and
Perron ( ), and Pesaran ( ). As compared
with panel unit root tests, the analysis of
cointegration in panels is still at an early stage of its
development. So far the focus of the panel
cointegration literature has been on residual- based
approaches, although there has been a number of
attempts at the development of system approaches
as well (Pedroni ). But once cointegration is
established the long-run parameters can be
estimated efficiently using techniques similar to the
ones proposed in the case of single time-series
models. These estimation techniques can also be
modified to allow for error cross- section
dependence (Pesaran ). Surveys of the panel unit
root and cointegration literature are provided by
Baneijee ( ), Baltagi and Kao
( ), Choi ( ), and Breitung and
Pesaran ( ).
The micro and macro panel literature is vast and
growing. For the analysis of many economic prob-
lems, further progress is needed in the analysis of
nonlinear panels, testing and modelling of error
cross-section dependence, dynamics, and neglected
heterogeneity. For general reviews of panel data
econometrics, see Arellano ( ), Baltagi ( ),
Hsiao ( ), and Wooldridge ( ).
Nonparametric and Semiparametric
Estimation
Much empirical research is concerned with esti-
mating conditional mean, median, or hazard func-
tions. For example, a wage equation gives the mean,
median or, possibly, some other quantile of wages
of employed individuals conditional on
characteristics such as years of work experience and
education. A hedonic price function gives the mean
price of a good conditional on its characteristics.
The function of interest is rarely known a priori and
must be estimated from data on the relevant
variables. For example, a wage equation is
estimated from data on the wages, experience,
education and, possibly, other characteristics of
individuals. Economic theory rarely gives useful
Econometrics
guidance on the form (or shape) of a conditional
mean, median, or hazard function. Consequently,
the form of the function must either be assumed or
inferred through the estimation procedure.
The most frequently used estimation methods
assume that the function of interest is known up to a
set of constant parameters that can be estimated
from data. Models in which the only unknown
quantities are a finite set of constant parameters are
called ‘parametric’. A linear model that is estimated
by ordinary least squares is a familiar and
frequently used example of a parametric model.
Indeed, linear models and ordinary least squares
have been the workhorses of applied econometrics
since its inception. It is not difficult to see why.
Linear models and ordinary least squares are easy to
work with both analytically and computationally,
and the estimation results are easy to interpret.
Other examples of widely used parametric models
are binary logit and probit models if the dependent
variable is binary (for example, an indicator of
whether an individual is employed or whether a
commuter uses automobile or public transit for a
trip to work) and the Weibull hazard model if the
dependent variable is a duration (for example, the
duration of a spell of employment or
unemployment).
Although parametric models are easy to work
with, they are rarely justified by theoretical or other
a priori considerations and often fit the available
data badly. Horowitz ( ), Horowitz
and Savin ( ), Horowitz and Lee ( ), and
Pagan and Ullah ( ) provide examples. The
examples also show that conclusions drawn from a
convenient but incorrectly specified model can be
very misleading. Of course, applied econome-
tricians are aware of the problem of specification
error. Many investigators attempt to deal with it by
carrying out a specification search in which several
different models are estimated and conclusions are
based on the one that appears to fit the data best.
Specification searches may be unavoidable in some
applications, but they have many undesirable
properties. There is no guarantee that a specification
search will include the correct model or a good
approximation to it. If the search includes the
correct model, there is no guarantee
3219
that it will be selected by the investigator’s model
selection criteria. Moreover, the search process
invalidates the statistical theory on which inference
is based.
Given this situation, it is reasonable to ask
whether conditional mean and other functions of
interest in applications can be estimated non-
parametrically, that is, without making a priori
assumptions about their functional forms. The
answer is clearly ‘yes’ in a model whose explan-
atory variables are all discrete. If the explanatory
variables are discrete, then each set of values of
these variables defines a data cell. One can estimate
the conditional mean of the dependent variable by
averaging its values within each cell. Similarly, one
can estimate the conditional median cell by cell.
If the explanatory variables are continuous, they
cannot be grouped into cells. Nonetheless, it is
possible to estimate conditional mean and median
functions that satisfy mild smoothness conditions
without making a priori assumptions about their
shapes. Techniques for doing this have been
developed mainly in statistics, beginning with
Nadaraya’s ( ) and Watson’s ( ) non-
parametric estimator of a conditional mean func-
tion. The Nadaraya-Watson estimator, which is also
called a kernel estimator, is a weighted average of
the observed values of the dependent variable. More
specifically, suppose that the dependent variable is
Y, the explanatory variable is X, and the data consist
of observations {Yh Xt :
; = 1, Then the Nadaraya-Watson
estimator of the mean of Tat X = .r is a weighted
average of the >',’s. L’s corresponding to A/s that
are close to .r get more weight than do >',’s
corresponding to A,’s that are far from .r. The
statistical properties of the Nadaraya-Watson esti-
mator have been extensively investigated for both
crosssectional and time-series data, and the esti-
mator has been widely used in applications. For
example, Blundell et al. ( ) used kernel esti
mates of Engel curves in an investigation of the
consistency of household-level data and revealed
preference theory. Hausman and Newey ( )
used kernel estimates of demand functions to esti-
mate the equivalent variation for changes in
3220
gasoline prices and the deadweight losses associated
with increases in gasoline taxes. Kernel- based
methods have also been developed for estimating
conditional quantile and hazard functions.
There are other important nonparametric
methods for estimating conditional mean functions.
Local linear estimation and series or sieve
estimation are especially useful in applications.
Local linear estimation consists of estimating the
mean of Y at X x by using a form of weighted least
squares to fit a linear model to the data. The weights
are such that observations ( A z ) for which X, is
close to x receive more weight than do observations
for which X, is far from x. In comparison with the
Nadaraya-Watson estimator, local linear estimation
has important advantages relating to bias and
behaviour near the boundaries of the data. These are
discussed in the book by Fan and Gijbels ( ),
among other places.
A series estimator begins by expressing the true
conditional mean (or quantile) function as an
infinite series expansion using basis functions such
as sines and cosines, orthogonal polynomials, or
splines. The coefficients of a truncated version of
the series are then estimated by ordinary least
squares. The statistical properties of series
estimators are described by Newey ( ).
Hausman and Newey ( ) give an example of
their use in an economic application.
Nonparametric models and estimates essentially
eliminate the possibility of misspecification of a
conditional mean or quantile function (that is, they
consistently estimate the true function), but they
have important disadvantages that limit their
usefulness in applied econometrics. One important
problem is that the precision of a nonparametric
estimator decreases rapidly as the dimension of the
explanatory variable X increases. This phenomenon
is called the ‘curse of dimensionality’. It can be
understood most easily by considering the case in
which the explanatory variables are all discrete.
Suppose the data contain 500 observations of Y and
X. Suppose, further, that A is a K-component vector
and that each component can take five different
values. Then the values of Agenerate 5k cells. If K =
4, which is not unusual in applied econometrics,
then there are 625 cells, or more cells than
observations. Thus, estimates
Econometrics
of the conditional mean function are likely to be
very imprecise for most cells because they will
contain few observations. Moreover, there will be at
least 125 cells that contain no data and, conse-
quently, for which the conditional mean function
cannot be estimated at all. It has been proved that
the curse of dimensionality is unavoidable in non-
parametric estimation. As a result of it, impracti-
cably large samples are usually needed to obtain
acceptable estimation precision if A is
multidimensional.
Another problem is that nonparametric estimates
can be difficult to display, communicate, and
interpret when A is multidimensional. Non-
parametric estimates do not have simple analytic
forms. If A is one-or two-dimensional, then the
estimate of the function of interest can be displayed
graphically, but only reduced- dimension
projections can be displayed when A has three or
more components. Many such displays and much
skill in interpreting them can be needed to fully
convey and comprehend the shape of an estimate.
A further problem with nonparametric estima-
tion is that it does not permit extrapolation. For
example, in the case of a conditional mean function
it does not provide predictions of the mean of Fat
values of x that are outside of the range of the data
on A. This is a serious drawback in policy analysis
and forecasting, where it is often important to
predict what might happen under conditions that do
not exist in the available data. Finally, in
nonparametric estimation it can be difficult to
impose restrictions suggested by economic or other
theory. Matzkin ( ) discusses this issue.
The problems of nonparametric estimation have
led to the development of so-called semi- parametric
methods that offer a compromise between
parametric and nonparametric estimation.
Semiparametric methods make assumptions about
functional form that are stronger than those of a
nonparametric model but less restrictive than the
assumptions of a parametric model, thereby
reducing (though not eliminating) the possibility of
specification error. Semiparametric methods permit
greater estimation precision than do nonparametric
methods when A is multidimensional.
Semiparametric estimation results are usually
Econometrics
easier to display and interpret than are nonpara-
metric ones, and provide limited capabilities for
extrapolation.
In econometrics, semiparametric estimation
began with Manski’s ( , ) and Cosslett’s
( ) work on estimating discrete-choice
random-utility models. McFadden had introduced
multinomial logit random utility models. These
models assume that the random components of the
utility function are independently and identically
distributed with the Type I extreme value
distribution. (The Type I extreme value distribution
and density functions are defined, for example, in
Eqs. (3.1) and (3.2) Maddala , p. 60.) The resulting
choice model is analytically simple but has
properties that are undesirable in many applications
(for example, the well-known independence-of-
irrelevant-altematives property). Moreover,
estimators based on logit models are inconsistent if
the distribution of the random components of utility
is not Type I extreme value. Manski ( , ) and
Cosslett ( ) pro
posed estimators that do not require a priori
knowledge of this distribution. Powell’s ( , nomic theory only loosely or through essentially
) least absolute deviations estimator for cen-
sored regression models is another early contribu-
tion to econometric research on semiparametric
estimation. This estimator was motivated by the
observation that estimators of (parametric) Tobit
models are inconsistent if the underlying normality
assumption is incorrect. Powell’s estimator is
consistent under very weak distributional
assumptions.
Semiparametric estimation has continued to be
an active area of econometric research. Semi-
parametric estimators have been developed for a
wide variety of additive, index, partially linear, and
hazard models, among others. These estimators all
reduce the effective dimension of the estimation
problem and overcome the curse of dimensionality
by making assumptions that are stronger than those
of fully nonparametric estimation but weaker than
those of a parametric model. The stronger
assumptions also give the models limited
extrapolation capabilities. Of course, these benefits
come at the price of increased risk of specification
error, but the risk is smaller than with simple
parametric models. This is because
3221
semiparametric models make weaker assumptions
than do parametric models, and contain simple
parametric models as special cases.
Semiparametric estimation is also an important
research field in statistics, and it has led to much
interaction between statisticians and econometri-
cians. The early statistics and biostatistics research
that is relevant to econometrics was focused on
survival (duration) models. Cox’s ( ) proportional
hazards model and the Buck-
ley and James ( ) estimator for censored
regression models are two early examples of this
line of research. Somewhat later, Stone ( )
showed that a nonparametric additive model can
overcome the curse of dimensionality. Since then,
statisticians have contributed actively to research on
the same classes of semiparametric models that
econometricians have worked on.
Theory-Based Empirical Models
Many econometric models are connected to eco-
arbitrary parametric assumptions about, say, the
shapes of utility functions. For example, a logit
model of discrete choice assumes that the random
components of utility are independently and iden-
tically distributed with the Type I extreme value
distribution. In addition, it is frequently assumed
that the indirect utility function is linear in prices
and other characteristics of the alternatives.
Because economic theory rarely, if ever, yields a
parametric specification of a probability model, it is
worth asking whether theory provides useful
restrictions on the specification of econometric
models, and whether models that are consistent
with economic theory can be estimated without
making non-theoretical parametric assumptions.
The answers to these questions depend on the
details of the setting being modelled.
In the case of discrete-choice, random-utility
models, the inferential problem is to estimate the
distribution of (direct or indirect) utility conditional
on observed characteristics of individuals and the
alternatives among which they choose. More
specifically, in applied research one usually is
interested in estimating the systematic
3222
component of utility (that is, the function that gives
the mean of utility conditional on the explanatory
variables) and the distribution of the random
component of utility. Discrete choice is present in a
wide range of applications, so it is important to
know whether the systematic component ofutility
and the distribution of the random component can
be estimated nonparametrically, thereby avoiding
the non-theoretical distributional and functional
form assumptions that are required by parametric
models. The systematic component and distribution
of the random component cannot be estimated
unless they are identified. However, economic
theory places only weak restrictions on utility
functions (for example, shape restrictions such as
monotonicity, convexity, and homogeneity), so the
classes of conditional mean and utility functions
that satisfy the restrictions are large. Indeed, it is not
difficult to show that observations of individuals’
choices and the values of the explanatory variables,
by themselves, do not identify the systematic com-
ponent ofutility and the distribution of the random
component without making assumptions that shrink
the class of allowed functions.
This issue has been addressed in a series of
papers by Matzkin that are summarized in Matzkin (
). Matzkin gives conditions under
which the systematic component ofutility and the
distribution of the random component are identified
without restricting either to a finitedimensional
parametric family. Matzkin also shows how these
functions can be estimated consistently when they
are identified. Some of the assumptions required for
identification may be undesirable in applications.
Moreover, Manski ( ) and Horowitz ( )
have given examples
in which infinitely many combinations of the sys-
tematic component of utility and distribution of the
random component are consistent with a binary
logit specification of choice probabilities. Thus,
discrete-choice, random-utility models can be
estimated under assumptions that are considerably
weaker than those of, say, logit and probit models,
but the systematic component of utility and the
distribution of the random component cannot be
identified using the restrictions of economic theory
alone. It is necessary to make
Econometrics
additional assumptions that are not required by
economic theory and, because they are required for
identification, cannot be tested empirically.
Models of market-entry decisions by oligopo-
listic firms present identification issues that are
closely related to those in discrete-choice, random
utility models. Berry and Tamer ( ) explain
the identification problems and approaches to
resolving them.
The situation is different when the economic
setting provides more information about the relation
between observables and preferences than is the
case in discrete-choice models. This happens in
models of certain kinds of auctions, thereby
permitting nonparametric estimation of the distri-
bution of values for the auctioned object. An
example is a first-price, sealed bid auction within
the independent private values paradigm. Here, the
problem is to infer the distribution of bidders’
values for the auctioned object from observed bids.
A game-theory model of bidders’ behaviour
provides a characterization of the relation between
bids and the distribution of private values. Guerre et
al. ( ) show that this relation non
parametrically identifies the distribution of values if
the analyst observes all bids and certain other mild
conditions are satisfied. Guerre et al. ( )
also show how to carry out nonparametric estima-
tion of the value distribution.
Dynamic decision models and equilibrium job-
search models are other examples of empirical
models that are closely connected to economic
theory, though they also rely on non-theoretical
parametric assumptions. In a dynamic decision
model, an agent makes a certain decision repeatedly
over time. For example, an individual may decide
each year whether to retire or not. The optimal
decision depends on uncertain future events (for
example, the state of one’s future health) whose
probabilities may change over time (for example,
the probability of poor health increases as one ages)
and depend on the decision. In each period, the
decision of an agent who maximizes expected
utility is the solution to a stochastic, dynamic
programming problem. A large body of research,
much of which is reviewed by Rust ( ), shows how
to specify
and estimate econometric models of the utility
Econometrics
function (or, depending on the application, cost
function), probabilities of relevant future events,
and the decision process.
An equilibrium search model determines the
distributions of job durations and wages endoge-
nously. In such a model, a stochastic process
generates wage offers. An unemployed worker
accepts an offer if it exceeds his reservation wage.
An employed worker accepts an offer if it exceeds
his current wage. Employers choose offers to
maximize expected profits. Among other things, an
equilibrium search model provides an explanation
for why seemingly identical workers receive
different wages. The theory of equilibrium search
models is described in Albrecht and Axell ( ),
Mortensen ( ),
and Burdett and Mortensen ( ). There is a
large body of literature on the estimation of these
models. Bowlus et al. ( ) provide a recent
example with many references.
The Bootstrap
The exact, finite-sample distributions of econo-
metric estimators and test statistics can rarely be
calculated in applications. This is because, except in
special cases and under restrictive assumptions (for
example, the normal linear model), finite sample
distributions depend on the unknown distribution of
the population from which the data were sampled.
This problem is usually dealt with by making use of
large-sample (asymptotic) approximations. A wide
variety of econometric estimators and test statistics
have distributions that are approximately normal or
chi-square when the sample size is large, regardless
of the population distribution of the data. The
approximation error decreases to zero as the sample
size increases. Thus, asymptotic approximations can
to be used to obtain confidence intervals for
parameters and critical values for tests when the
sample size is large.
It has long been known, however, that the
asymptotic normal and chi-square approximations
can be very inaccurate with the sample sizes
encountered in applications. Consequently, there
can be large differences between the ttue and
3223
nominal coverage probabilities of confidence
intervals and between the true and nominal prob-
abilities with which a test rejects a correct null
hypothesis. One approach to dealing with this
problem is to use higher-order asymptotic approx-
imations such as Edgeworth or saddlepoint expan-
sions. These received much research attention
during 1970s and 1980s, but analytic higher- order
expansions are rarely used in applications because
of their algebraic complexity.
The bootstrap, which is due to Efron ( ),
provides a way to obtain sometimes spectacular
improvements in the accuracy of asymptotic
approximations while avoiding algebraic com-
plexity. The bootstrap amounts to treating the data
as if they were the population. In other words, it
creates a pseudo-population whose distribution is
the empirical distribution of the data. Under
sampling from the pseudo-population, the exact
finite sample distribution of any statistic can be
estimated with arbitrary accuracy by carrying out a
Monte Carlo simulation in which samples are
drawn repeatedly from the empirical distribution of
the data. That is, the data are repeatedly sampled
randomly with replacement. Since the empirical
distribution is close to the population distribution
when the sample size is large, the bootstrap
consistently estimates the asymptotic distribution of
a wide range of important statistics. Thus, the
bootstrap provides a way to replace analytic
calculations with computation. This is usefiil when
the asymptotic distribution is difficult to work with
analytically.
More importantly, the bootstrap provides a low-
order Edgeworth approximation to the distribution
of a wide variety of asymptotically standard normal
and chi-square statistics that are used in applied
research. Consequently, the bootstrap provides an
approximation to the finite-sample distributions of
such statistics that is more accurate than the
asymptotic normal or chi-square approximation.
The theoretical research leading to this conclusion
was carried out by statisticians, but the bootstrap’s
importance has been recognized in econometrics
and there is now an important body of econometric
research on the topic. In many settings that are
important in applications, the bootstrap essentially
eliminates errors in the
3224
coverage probabilities of confidence intervals and
the rejection probabilities of tests. Thus, the boot-
strap is a very important tool for applied
econometricians.
There are, however, situations in which the
bootstrap does not estimate a statistic’s asymptotic
distribution consistently. Manski’s ( , )
maximum score estimator of the parameters of a
binary response model is an example. All known
cases of bootstrap inconsistency can be overcome
through the use of subsampling methods. In sub-
sampling, the distribution of a statistic is estimated
by carrying out a Monte Carlo simulation in which
the subsamples of the data are drawn repeatedly.
The subsamples are smaller than the original data-
set, and they can be drawn randomly with or
without replacement. Subsampling provides
estimates of asymptotic distributions that are
consistent under very weak assumptions, though it
is usually less accurate than the bootstrap when the
bootstrap is consistent.
Programme Evaluation and Treatment
Effects
Programme evaluation is concerned with estimating
the causal effect of a treatment or policy inter-
vention on some population. The problem arises in
many disciplines, including biomedical research
(for example, the effects of a new medical
treatment) and economics (for example, the effects
of job training or education on earnings). The most
obvious way to leam the effects of treatment on a
group of individuals by observing each individual’s
outcome in both the treated and the untreated states.
This is not possible in practice, however, because
one virtually always observes any given individual
in either the treated state or the untreated state but
not both. This does not matter if the individuals
who receive treatment are identical to those who do
not, but that rarely happens. For example,
individuals who choose to take a certain drug or
whose physicians prescribe it for them may be
sicker than individuals who do not receive the drug.
Similarly, people who choose to obtain high levels
of education may be
Econometrics
different from others in ways that affect future
earnings.
This problem has been recognized since at least
the time of R.A. Fisher. In principle, it can be
overcome by assigning individuals randomly to
treatment and control groups. One can then estimate
the average effect of treatment by the difference
between the average outcomes of treated and
untreated individuals. This random assignment
procedure has become something of a gold standard
in the treatment effects literature. Clinical trials use
random assignment, and there have been important
economic and social experiments based on this
procedure. But there are also serious practical
problems. First, random assignment may not be
possible. For example, one cannot assign high-
school students randomly to receive a university
education or not. Second, even if random
assignment is possible, postrandomization events
may dismpt the effects of randomization. For
example, individuals may drop out of the
experiment or take treatments other than the one to
which they are assigned. Both of these things may
happen for reasons that are related to the outcome
of interest. For example, very ill members of a
control group may figure out that they are not
receiving treatment and find a way to obtain the
drug being tested. In addition, real-world
programmes may not operate the way that
experimental ones do, so real-world outcomes may
not mimic those found in an experiment, even if
nothing has disrupted the randomization.
Much research in econometrics, statistics, and
biostatistics has been aimed at developing methods
for inferring treatment effects when randomization
is not possible or is disrupted by postrandomization
events. In econometrics, this research dates back at
least to Gronau ( )
and Heckman ( ). The fundamental problem
is to identify the effects of treatment or, in less
formal terms, to separate the effects of treatment
from those of other sources of differences between
the treated and untreated groups. Manski ( ),
among many others, discusses this problem. Large
literatures in statistics, biostatistics, and
econometrics are concerned with developing
Econometrics
identifying assumptions that are reasonable in
applied settings. However, identifying assumptions
are not testable empirically and can be con-
troversial. One widely accepted way of dealing with
this problem is to conduct a sensitivity analysis in
which the sensitivity of the estimated treatment
effect to alternative identifying assumptions is
assessed. Another possibility is to forgo contro-
versial identifying assumptions and to find the
entire set of outcomes that are consistent with the
joint distribution of the observed variables. This
approach, which has been pioneered by Manski and
several co-investigators, is discussed in Manski (
, ), among other places. Hotz
et al. ( ) provide an interesting application of
bounding methods to measuring the effects of
teenage pregnancy on the labour market outcomes
of young women.
Integration and Simulation Methods in
Econometrics
The integration problem is endemic in economic
modelling, arising whenever economic agents do
not observe random variables and the behaviour
paradigm is the maximization of expected utility.
The econometrician inherits this problem in the
expression of the corresponding econometric model,
even before taking up inference and estimation. The
issue is most familiar in dynamic optimization
contexts, where it can be addressed by a variety of
methods. Taylor and Uhlig ( )
present a comprehensive review of these methods;
for later innovations see Keane and Wolpin ( ),
Rust (), and Santos and Vigo-
Aguiar ( ).
The problem is more pervasive in econometrics
than in economic modelling, because it arises, in
addition, whenever economic agents observe
random variables that the econometrician does not.
For example, the economic agent may form
expectations conditional on an information set not
entirely accessible to the econometrician, such as
personal characteristics or confidential information.
Another example arises in discrete choice settings,
where utilities of alternatives are never
3225
observed and the prices of alternatives often are not.
In these situations the economic model provides a
probability distribution of outcomes conditional on
three classes of objects: observed variables,
available to the econometrician; latent variables,
unobserved by the econometrician; and parameters
or functions describing the preferences and
decision-making environment of the economic
agent. The econometrician typically seeks to learn
about the parameters or functions given the
observed variables.
There are several ways of dealing with this task.
Two approaches that are closely related and widely
used in the econometrics literature generate
integration problems. The first is to maintain a
distribution of the latent variables conditional on
observed variables, the parameters in the model,
and additional parameters required for completing
this distribution. (This is the approach taken in
maximum likelihood and Bayesian inference.)
Combined with the model, this leads to the joint
distribution of outcomes and latent variables con-
ditional on observed variables and parameters.
Since the marginal distribution of outcomes is the
one relevant for the econometrician in this
conditional distribution, there is an integration
problem for the latent variables. The second
approach is weaker: it restricts to zero the values of
certain population moments involving the latent and
observable variables. (This is the approach taken in
generalized method of moments, which can be
implemented with both parametric and
nonparametric methods.) These moments depend
upon the parameters (which is why the method
works) and the econometrician must therefore be
able to evaluate the moments for any given set of
parameter values. This again requires integration
over the latent variables.
Ideally, this integral would be evaluated ana-
lytically. Often - indeed, typically - this is not
possible. The alternative is to use numerical
methods. Some of these are deterministic, but the
rapid growth in the solution of these problems since
(roughly) 1990 has been driven more by simulation
methods employing pseudo-random numbers
generated by computer hardware and software. This
section reviews the most important
3226
these methods and describes their most significant
use in non-Bayesian econometrics, namely, simu-
lated method of moments. In Bayesian economet-
rics the integration problem is inescapable, the
structure of the economic model notwithstanding,
because parameters are treated explicitly as
unobservable random variables. Consequently
simulation methods have been central to Bayesian
inference in econometrics.
Deterministic Approximation of Integrals
The evaluation of an integral is a problem as old as
the calculus itself. In well-catalogued but limited
instances analytical solutions are available:
Gradshteyn and Ryzhik ( ) is a useful classic
reference. For integration in one dimension there
are several methods of deterministic approximation,
including Newton-Coates (Press et al. ch. 4; Davis
and Rabinowitz , ch. 2), and Gaussian quadrature
(Golub and Welsch ; Judd , s. 7.2). Gaussian
quadrature approximates a smooth function as the
product a polynomial of modest order and a smooth
basis function, and then uses iterative refinements to
compute the approximation. It is incorporated in
most mathematical applications software and is
used routinely to approximate integrals in one
dimension to many significant figures of accuracy.
Integration in several dimensions by means of
deterministic approximation is more difficult.
Practical generic adaptations of Gaussian quadra-
ture are limited to situations in which the integrand
is approximately the product of functions of single
variables (Davis and Rabinowitz , pp. 354-9). Even
here the logarithm of computation time is
approximately linear in the number of variables, a
phenomenon sometimes dubbed ‘the curse of
dimensionality.’ Successful extensions of
quadrature beyond dimensions of four or five are
rare, and these extensions typically require sub-
stantial analytical work before they can be applied
successfully.
Low discrepancy methods provide an alternative
generic approach to deterministic approximation of
integrals in higher dimensions. The approximation
is the average value of the integrand computed over
a well-chosen sequence of points whose
configuration amounts to a
Econometrics
sophisticated lattice. Different sequences lead to
variants on the approach, the best known being the
Halton ( ) sequence and the Hammersley
( ) sequence. Niederreiter ( ) reviews
these and other variants.
A key property of any method of integral
approximation, deterministic or nondeterministic, is
that it should provide as a by-product some
indicator of the accinacy of the approximation.
Deterministic methods typically provide upper
bounds on the approximation error, based on worst-
case situations. In many situations the actual error is
orders of magnitude less than the upper bound, and
as a consequence attaining desired error tolerances
may appear to be impractical, whereas in fact these
tolerances can easily be attained. Geweke ( , s.
2.3) provides an
example.
Simulation Approximation of Integrals
The structure of integration problems encountered
in econometrics makes them often more amenable
to attack by simulation methods than by non-
deterministic methods. Two characteristics are key.
First, integrals in many dimensions are required. In
some situations the number is proportional to the
size of the sample, and, while the structure of the
problem may lead to decomposition in terms of
many integrals of smaller dimension, the resulting
structure and dimension are still unsuitable for
deterministic methods. The second characteristic is
that the integration problem usually arises as the
need to compute the expected value of a function of
a random vector with a given probability
distribution P:
1
= [ £(x)E(x)dx, (1)
Js
where p is the density corresponding to P, g is the
function, x is the random vector, and I is the number
to be approximated. The probability distribution P
is then the point of departure for the simulation.
For many distributions there are reliable algo-
rithms, implemented in widely available mathe-
matical applications software, for simulation of
random vectors x. This yields a sample {glx1'" ’)}
Econometrics
(m 1, ... , M) whose arithmetic mean provides an
approximation of I, and for which a central limit
theorem provides an assessment of the accuracy of
the approximation in the usual way. (This requires
the existence of the first two moments of#, which
must be shown analytically.) This approach is most
useful when p is simple (so that direct simulation of
x is possible) but the structure of# precludes
analytical evaluation of I.
This simple approach does not suffice for the
integration problem as it typically arises in econo-
metrics. A leading example is the multinomial
probit (MNP) model with / discrete choices. For
each individual i the utility of the last choice Uy is
normalized to be zero, and the utilities of the first /
— 1 choices are given by the vector
U ; ~ N(XifS, £), (2)
where X is a matrix of characteristics of individual
i, including the prices and other properties of the
choices presented to that individual, and [1 and X
are structural parameters of the model. If the y’th
element of u, is positive and larger than all the other
elements of u, the individual makes choice j, and if
all elements of u are negative the individual makes
choice /. The probability that individual i makes
choice j is the integral of the (n 1)- variate normal
distribution ( ) taken over the subspace {u,- : ujk
<Ujj\/k= 1, ... , n). This computation is essential in
evaluating the likelihood function, and it has no
analytical solution. (For discussion and review, see
Sandor and Andras .)
Several generic simulation methods have been
used for the problem ( ) in econometrics. One of the
oldest is acceptance sampling, a simple variant of
which is described in von Neumann ( ) and
Hammersley and Handscomb ( ). Suppose it
is possible to draw from the distribution Q with
density q, and the ratio p(\)/c/(\) is bounded above
by the known constant a. If x is simulated succes-
sively from Q but accepted and taken into the
sample with probability #(x)/[«(/(x)], then the
resulting sample is independently distributed with
the identical distribution P. Proofs and further
discussion are widely available; for example, Press
et al. ( , s. 7.4), Bratley et al. (
3227
s. 5.2.5), and Geweke ( , s. 4.2.1). The uncon
ditional probability of accepting draws from Q is 1
la. If a is too large the method is impractical, but
when acceptance sampling is practical it provides
draws directly from P. This is an important com-
ponent of many of the algorithms underlying the
‘black box’ generation of random variables in
mathematical applications software.
Alternatively, in the same situation all of the
draws from Q are retained and taken into a stratified
sample in which the weight w(x<m>) = p(x (m))/?(x(m))
is associated with the m’th draw. The
approximation of / in ( ) is then the weighted
average of the terms #(x <m)). This approach dates at
least to Hammersley and Handscomb ( , s.
5.4), and was introduced to
econometrics by Kloek and van Dijk ( ). The
procedure is more general than acceptance sampling
in that a known upper bound of w is not required,
but if in fact a is large then the weights will display
large variation and the approximation will be poor.
This is clear in the central limit theorem for the
accuracy of approximation provided in Geweke (
), which as a practical
matter requires that a finite upper bound on w be
established analytically. This is a key limitation of
acceptance sampling and importance sampling.
Markov chain Monte Carlo (MCMC) methods
provide an entirely different approach to the solu-
tion of the integration problem ( ). These procedures
construct a Markov process of the form
xM ~/?(x/x("'-1')) (3)
in such a way that
M-x
m= 1
converges (almost surely) to I. These methods have
a history in mathematical physics dating back to the
algorithm of Metropolis et al. ( ).
Hastings ( ) focused on statistical problems
and extended the method to its present form known
as the Hastings-Metropolis (HM) algorithm. HM
draws a candidate x* from a convenient distribution
indexed by x1'" M. It sets
3228
x(m) = x with probability z(x‘"' ' x1"'1) and sets x1'"' =
x1"'1 1 otherwise, the function a being chosen so that
the process ( ) defined in this way has the desired
convergence property. Chib and Greenberg ( )
provide a detailed introduction
to HM and its application in econometrics. Tierney
( ) provides a succinct summary of
the relevant continuous state space Markov chain
theory bearing on the convergence of MCMC.
A version of the HM algorithm particularly
suited to image reconstruction and problems in
spatial statistics, known as the Gibbs sampling (GS)
algorithm, was introduced by Geman and Geman (
). This was subsequently shown to
have great potential for Bayesian computation by
Gelfand and Smith ( ). In GS the vector
x is subdivided into component vectors, x' = (x'j, . .. ,
x^), in such a way that simulation from the
conditional distribution of each x, implied by p(\) in
( ) is feasible. This method has proven very
advantageous in econometrics generally, and it
revolutionized Bayesian approaches in particular
beginning about 1990.
By the turn of the century HM and GS algo-
rithms were standard tools for likelihood-based
econometrics. Their structure and strategic impor-
tance for Bayesian econometrics were conveyed in
surveys by Geweke ( ) and Chib ( ), as
well as in a number of textbooks, including Koop (
), Lancaster ( ), Geweke ( ), and
Rossi et al. ( ). Central limit theorems can be
used to assess the quality of approximations as
described in Tierney ( ) and Geweke ( ).
Simulation Methods in Non-Bayesian
Econometrics
Generalized method of moments estimation has
been a staple of non-Bayesian econometrics since
its introduction by Hansen ( ). In an econo
metric model with k x 1 parameter vector 6 eco-
nomic theory provides the set of sample moment
restrictions
h(0) g(x)p(x| 0,y)d\ = 0, (4)
where g(x) is ap x 1 vector and y denotes the data
including instrumental variables. An example is
Econometrics
the MNP model ( ). If the observed choices are
coded by the variables dy I if individual i makes
choice j and dy 0 otherwise, then the expected value
of dy is the probability that individual i makes
choice j, leading to restrictions of the form ( ).
The generalized method of moments estimator
minimizes the criterion function h(0)'Wh(0) given a
suitably chosen weighting matrix W. If the requisite
integrals can be evaluated analytically, p > k, and
other conditions provided in Hansen () are satisfied,
then there is a well-
developed asymptotic theory of inference for the
parameters that by 1990 was a staple of graduate
econometrics textbooks. If for one or more elements
of h the integral cannot be evaluated analytically,
then for alternative values of it is often possible to
approximate the integral appearing in ( ) by
simulation. This is the situation in the MNP model.
The substitution of a simulation approximation
for the integral in ( ) defines the method of simu-
lated moments (MSM) introduced by McFadden (
) and Pakes and Pollard ( ), who were
concerned with the MNP model ( ) in particular and
the estimation of discrete response models using
cross-section data in general. Later the method was
extended to time series models by Lee and Ingram (
) and Duffie and Singleton
( ). The asymptotic distribution theory
established in this literature requires that the num-
ber of simulations M increase at least as rapidly as
the square of the number of observations. The
practical import of this apparently severe require-
ment is that applied econometric work must estab-
lish that changes in M must have little impact on the
results; Geweke et al. ( , ) provide
examples for MNP. This literature also shows that
in general the impact of using direct simulation, as
opposed to analytical evaluation of the integral, is to
increase the asymptotic variance of the GMM
estimator of 0 by the factor M 1, typically trivial in
view of the number of simulations required.
Econometrics
Substantial surveys of the details of MSM and
leading applications of the method can be found in
Gourieroux and Monfort ( , ), Stem
( ), and Liesenfeld and Breitung ( ).
The simulation approximation, unlike the
(unavailable) analytical evaluation of the integral in
( ), can lead to a criterion function that is
discontinuous in 0. This happens in the MNP model
using the obvious simulation scheme in which the
choice probabilities are replaced by their
proportions in the M simulations, as proposed by
Lerman and Manski ( ). The asymp-
totic theory developed by McFadden ( ) and
Pakes and Pollard ( ) copes with this possibil-
ity, and led McFadden ( ) to used kernel
weighting to smooth the probabilities. The most
widely used method for smoothing probabilities in
the MNP model is the Geweke-Hajivassi- liou-
Keane (GHK) simulator of Geweke ( ),
Hajivassiliou et al. ( ), and Keane
( ); a full description is provided in Geweke
and Keane ( ), and comparisons of alternative
methods are given in Hajivassiliou et al. (
and Sandor and Andras ( ).
Maximum likelihood estimation of 0 can lead to
first-order conditions of the form ( ), and thus
becomes a special case of MSM. This context
highlights some of the complications introduced by
simulation. While the simulation approximation of (
) is unbiased, the corresponding expression enters
the log likelihood function and its derivatives
nonlinearly. Thus for any finite number of
simulations M, the evaluation of the first- order
conditions is biased in general. Increasing M at a
rate faster than the square of the number of
observations eliminates the squared bias relative to
the variance of the estimator; Lee ( ) pro
vides further details.
Simulation Methods in Bayesian Econometrics
Bayesian econometrics places a common proba-
bility distribution on random variables that can be
observed (data) and unobservable parameters and
latent variables. Inference proceeds using the dis-
tribution of these unobservable entities conditional
on the data - the posterior distribution. Results are
typically expressed in terms of the expectations of
parameters or functions of
3229
parameters, expectations taken with respect to the
posterior distribution. Thus, whereas integration
problems are application-specific in non- Bayesian
econometrics, they are endemic in Bayesian
econometrics.
The development of modem simulation methods
had a correspondingly greater impact in Bayesian
than in non-Bayesian econometrics. Since 1990
simulation-based Bayesian methods have become
practical in the context of most econometric models.
The availability of this tool has been influential in
the modelling approach taken in addressing applied
econometric problems.
The MNP model ( ) illustrates the interaction in
latent variable models. Given a sample of n
individuals, the (J — 1) x 1 latent utility vectors ui,
. . ., u„ are regarded explicitly as n(J — 1)
unknowns to be inferred along with the unknown
parameters [1 and 2. Conditional on these param-
eters and the data, the vectors u b . . ., u„ are
independently distributed. The distribution of u, is
( ) truncated to an orthant that depends on the
) observed choice j: ifj < /then uik < utJ for all k // and
whereas for choice ■ /,, 0 for all k.
The distribution of each ulk, conditional on all of the
other elements of u„ is truncated univariate normal,
and it is relatively straightforward to simulate from
this distribution. (Geweke , provides details on
sampling from a multivariate normal distribution
subject to linear restrictions.) Consequently GS
provides a practical algorithm for drawing from the
distribution of the latent utility vectors conditional
on the parameters.
Conditional on the latent utility vectors - that is,
regarding them as observed - the MNP model is a
seemingly unrelated regressions model, and the
approach taken by Percy ( ) applies.
Given conjugate priors the posterior distribution
of /?, conditional on 2 and utilities, is Gaussian, and
the conditional distribution of 2, conditional on [1
and utilities, is inverted Wishart. Since GS provides
the joint distribution of parameters and latent
utilities, the posterior mean of any function of these
can be approximated as the sample mean. This
approach and the suitability of GS for latent
variable models were first recognized by Chib (
). Similar approaches in other latent variable
3230
models in include McCulloch and Tsay ( ),
Chib and Greenberg ( ), McCulloch et al.
( ), and Geweke and Keane ( ).
The Bayesian approach with GS sidesteps the
evaluation of the likelihood function and, of any
moments in which the approximation is biased
given a finite number of simulations, two technical
issues that are prominent in MSM. On the other
hand, as in all MCMC algorithms, there may be
sensitivity to the initial values of parameters and
latent variables in the Markov chain, and substantial
serial correlation in the chain will reduce the
accuracy of the simulation approximation. Geweke
( , ) and Tierney ( ) variables exhibiting clear business cycle variations.
discuss these issues.
Financial Econometrics
Attempts at testing of the efficient market hypoth-
esis (EMH) provided the impetus for the application
of time series econometric methods in finance. The
EMH was built on the pioneering work of Bachelier
( ) and evolved in the
1960s from the random walk theory of asset prices
advanced by Samuelson ( ). By the early
1970s a consensus had emerged among financial
economists suggesting that stock prices could be
well approximated by a random walk model and
that changes in stock returns were basically
unpredictable. Fama ( ) provides an early,
definitive statement of this position. He distin-
guished between different forms of the EMH: the
‘weak’ form that asserts all price information is
fully reflected in asset prices; the ‘semi-strong’
form that requires asset price changes to frilly
reflect all publicly available information and not
only past prices; and the ‘strong’ form that postu-
lates that prices fully reflect information even if
some investor or group of investors have monop-
olistic access to some information. Fama regarded
the strong form version of the EMH as a benchmark
against which the other forms of market efficiencies
are to be judged. With respect to the weak form
version he concluded that the test results strongly
support the hypothesis, and considered the various
departures documented as economically
unimportant. He reached a similar
Econometrics
conclusion with respect to the semi-strong version
of the hypothesis. Evidence on the semi-strong form
of the EMH was revisited by Fama ( ).
By then it was clear that the distinction between the
weak and the semi-strong forms of the EMH was
redundant. The random walk model could not be
maintained either, in view of more recent studies, in
particular that of Lo and MacKinlay ( ).
This observation led to a series of empirical
studies of stock return predictability over different
horizons. It was shown that stock returns can be
predicted to some degree by means of interest rates,
dividend yields and a variety of macroeconomic
See, for example, Fama and French ( ),
Kandel and Stambaugh ( ), and
Pesaran and Timmermann ( ) on predictabil
ity of equity returns in the United States; and Clare
etal. ( ) and Pesaran and Timmermann ( )
on equity return predictability in the UK.
Although it is now generally acknowledged that
stock returns could be predictable, there are serious
difficulties in interpreting the outcomes of market
efficiency tests. Predictability could be due to a
number of different factors such as incomplete
learning, expectations heterogeneity, time variations
in risk premia, transaction costs, or specification
searches often carried out in pursuit of
predictability. In general, it is not possible to
distinguish between the different factors that might
lie behind observed predictability of asset returns.
As noted by Fama ( ) the test of the
EMH involves a joint hypothesis, and can be tested
only jointly with an assumed model of market
equilibrium. This is not, however, a problem that is
unique to financial econometrics; almost all areas of
empirical economics are subject to the joint
hypotheses problem. The concept of market
efficiency is still deemed to be usefiil as it provides
a benchmark and its use in finance has led to
significant insights.
Important advances have been made in the
development of equilibrium asset pricing models,
econometric modelling of asset return volatility
(Engle ; Bollerslev ), analysis of high frequency
intraday data, and market microstructures. Some of
these developments are reviewed in Campbell et al.
( ), Cochrane ( ),
Econometrics
Shephard ( ), and McAleer and Medeiros
( ). Future advances in financial econometrics
are likely to focus on heterogeneity, learning and
model uncertainty, real time analysis, and further
integration with macroeconometrics. Finance is
particularly suited to the application of techniques
developed for real time econometrics (Pesaran and
Timmermann ).
Appraisals and Future Prospects
Econometrics has come a long way over a relatively
short period. Important advances have been made in
the compilation of economic data and in the
development of concepts, theories and tools for the
construction and evaluation of a wide variety of
econometric models. Applications of econometric
methods can be found in almost every field of
economics. Econometric models have been used
extensively by government agencies, international
organizations and commercial enterprises. Macro
econometric models of differing complexity and
size have been constructed for almost every country
in the world. In both theory and practice,
econometrics has already gone well beyond what its
founders envisaged. Time and experience, however,
have brought out a number of difficulties that were
not apparent at the start.
Econometrics emerged in the 1930s and 1940s
in a climate of optimism, in the belief that economic
theory could be relied on to identify most, if not all,
of the important factors involved in modelling
economic reality, and that methods of classical
statistical inference could be adapted readily for the
purpose of giving empi rical content to the received
economic theory. This early view of the interaction
of theory and measurement in econometrics,
however, proved rather illusory. Economic theory is
invariably formulated with ceteris paribus clauses,
and involves unobservable latent variables and
general functional forms; it has little to say about
adjustment processes, lag lengths and other factors
mediating the relationship between the theoretical
specification (even if correct ) and observables.
Even in the choice of variables to be included in
econometric relations, the role of economic theory
is far more
3231
limited than was at first recognized. In a Walrasian
general equilibrium model, for example, where
everything depends on everything else, there is very
little scope for a priori exclusion of variables from
equations in an econometric model. There are also
institutional features and accounting conventions
that have to be allowed for in econometric models
but which are either ignored or are only partially
dealt with at the theoretical level. All this means
that the specification of econometric models
inevitably involves important auxiliary assumptions
about functional forms, dynamic specifications,
latent variables, and so on, with respect to which
economic theory is silent or gives only an
incomplete guide.
The recognition that economic theory on its own
cannot be expected to provide a complete model
specification has important consequences for testing
and evaluation of economic theories, for forecasting
and real time decision making. The incompleteness
of economic theories makes the task of testing them
a formidable undertaking. In general it will not be
possible to say whether the results of the statistical
tests have a bearing on the economic theory or the
auxiliary assumptions. This ambiguity in testing
theories, known as the Duhem-Quine thesis, is not
confined to econometrics and arises whenever
theories are conjunctions of hypotheses (on this, see
for example Cross ). The problem is, however,
especially serious in econometrics because theory is
far less developed in economics than it is in the
natural sciences. There are, of course, other
difficulties that surround the use of econometric
methods for the purpose of testing economic
theories. As a mle economic statistics are not the
results of designed experiments, but are obtained as
by-products of business and government activities
often with legal rather than economic considerations
in mind. The statistical methods available are gener-
ally suitable for large samples while the economic
data typically have a rather limited coverage. There
are also problems of aggregation over time,
commodities and individuals that further complicate
the testing of economic theories that are
microbased.
Econometric theory and practice seek to provide
information required for informed decision
3232
making in public and private economic policy. This
process is limited not only by the adequacy of
econometrics but also by the development of
economic theory and the adequacy of data and other
information. Effective progress, in the future as in
the past, w i l l come from simultaneous
improvements in econometrics, economic theory
and data. Research that specifically addresses the
effectiveness of the interface between any two of
these three in improving policy - to say nothing of
all of them - necessarily transcends traditional sub-
disciplinary boundaries within economics. But it is
precisely these combinations that hold the greatest
promise for the social contribution of academic
economics.
Bibliography
Aitken, A.C. 1934 5. On least squares and linear combinations
of observations. Proceedings of the Royal Society of
Edinburgh 55: 42 48.
Albrecht, J.W., and B. Axell. 1084. An equilibrium model of
search unemployment. Journal of Political Economy 92:
824 84(1.
Allen, R.G.D., and A.L. Bowley. 1935. Family expenditure.
London: P.S. King.
Almon, S. 1965. The distributed lag between capital appro-
priations and net expenditures. Econometrica 33: 178-
196.
Amemiya, T. 1983. Nonlinear regression models. In Hand-
book of econometrics, ed. Z. Griliches and M.D. tntriligator,
vol. 1. Amsterdam: North-1 lolland. Amemiya, T. 1984. Tobit
models: A survey. Journal of
Econometrics 24: 3-61.
An, S., andF. Schorfheide. 2007. Bayesian analysis ofDSGE
models. Econometric Reviews 26 (2 4): 113 172. Anderson,
T.W., and C. Hsiao. 1981. Estimation of dynamic models with
error components. Journal of the American Statistical Society
76: 548 606. Anderson, T.W., and C. Hsiao. 1482.
Formulation and estimation of dynamic models using panel
data. Journal of Econometrics 18: 47 82.
Anderson, T.W., and H. Rubin. 1949. Estimation of the
parameters of a single equation in a complete system of
stochastic equations. Annals of Mathematical Statistics
20: 46-63.
Andrews, D.W.K. 1993. Tests for parameter instability and
structural change with unknown change point.
Econometrica 61: 821 856.
Andrews, D.W.K., and W. Ploberger. 1994. Optimal tests
when a nuisance parameter is present only under the
alternative. Econometrica 62: 1383 1414.
Anselin, L. 1988. Spatial econometrics: Methods and models.
Boston: Kluwer Academic Publishers.
Econometrics
Anselin, L., J. Le Gallo, and 11. Jayet. 2007. Spatial panel
econometrics. In The econometrics of panel data: Fun-
damentals and recent developments in theory and prat -
tice. 3rd ed, ed. L. Matyas and P. Sevestre. Dordrecht:
Kluwer (forthcoming).
Arellano, M. 2003. Panel data econometrics. Oxford: Oxford
University Press.
Arellano, M., and S.R. Bond. 1991. Some tests of specifi-
cation for panel data: Monte Carlo evidence and an
application to employment equations. Review of Eco-
nomic Studies 58: 277-297.
Arellano, M., and B. Honore. 2001. Panel data models: Some
recent developments. In Handbook of econometrics, ed.
J.J. Heckman and E. Learner, vol. 5. Amsterdam: North-
Holland.
Bachelier, L.J.B.A. 1900. Theorie de la Speculation. Paris:
Gauthier-Villars. Reprinted in The random character of
stock market prices, ed. PH. Cootner. Cambridge, MA:
MIT Press, 1964.
Bai, .1., and S. Ng. 2004. A panic attack on unit roots and
cointegration. Econometrica 72: 1127 1177.
Bai, .1., and P. Perron. 1998. Estimating and testing linear
models with multiple structural changes. Econometrica
66: 47-78.
Baltagi, B. 2005. Econometric analysis of panel data. 2nd ed.
New York: Wiley.
Baltagi, B.H.. and C. Kao. 2000. Nonstationary panels,
cointegration in panels and dynamic panels: A survey. In
yonstationary panels, panel cointegration, and dynamic
panels. Advances in Econometrics, col. 15. ed. B.
Baltagi. Amsterdam: JAI Press.
Baneijee, A. 1999. Panel data unit roots and cointegration: An
overview. Oxford Bulletin of Economics and Statistics
61: 607 629.
Basmann, R.L. 1957. A generalized classical method of linear
estimation of coefficients in a structural equation.
Econometrica 25: 77 83.
Bauwens, L., M. Lubrano, and J.F. Richard. 2001. Bayesian
inference in dynamic econometric models. Oxford:
Oxford University Press.
Benini. R. 1907. Sull’uso delle formole empiriche a
nelTeconomia applicata. Giornale degli economisti, 2nd
series 35: 1053 1063.
Berry, S., and E. Tamer. 2006. Identification in models of
oligopoly entry. In Advances in economics and econo-
metrics: Theory and applications, ninth world
congress, ed. R. Blundell, W.K. Newey, and T. Persson.
Cambridge: Cambridge University Press.
Beveridge, S., and C.R. Nelson. 1981. Anew approach to the
decomposition of economic time series into permanent
and transitory components with particular attention to
measurement of the ‘business cycle'. Journal of
Monetary Economics 7: 151 174.
Binder, M., and M.H. Pesaran. 1995. Multivariate rational
expectations models and macroeconometric modelling: A
review and some new results. In Handbook of applied
econometrics, vol l Macroeconomics, ed. M.H. Pesaran
andM.R. Wickens. Oxford: Basil Blackwell.
Econometrics
Binder, M., C. Hsiao, and M.H. Pesaran. 2005. Estimation
and inference in short panel vector autoregressions with
unit roots and cointegration. Econometric Theory 21:
795-837.
Bjerkholt, O. 1995. Ragnar Frisch, editor of Econometrica.
Econometrica 63: 755-765.
Blanchard, O.J., and D. Quah. 1989. The dynamic effects of
aggregate demand and supply disturbances. American
Economic Review 79: 1146-1164.
Blundell, R.W., M. Browning, and I.A. Crawford. 2003.
Nonparanretric Engel curves and revealed preference.
Econometrica 71: 205-240.
Bollerslev, T. 1986. Generalised autoregressive conditional
heteroskedasticity. Journal of Econometrics 51: 307-327.
Boskin, M.J. 1974. A conditional logit model of occupational
choice. Journal of Political Economy 82: 389-398.
Bosq, D. 1996. Nonparametric statistics for stochastic
processes. New York: Springer.
Bowlus, A.J., N.M. Kiefer, and G.R. Neumann. 2001.
Equilibrium search models and the transition from school
to work. International Economic Review 42: 317-343.
Box, G.E.P., andG.M. Jenkins. 1970. Time series analysis:
Forecasting and control. San Francisco: Holden-Day.
Bratley, P., B.L. Fox, and L.E. Schrage. 1987. A guide to
simulation. New York: Springer-Verlag.
Breitung, J., and M.H. Pesaran 2008. Unit roots and
cointegration in panels. In The econometrics of panel
data: Fundamentals and recent developments in theory
and practice. 3rd ed. L. Matyas and P. Sevestre. Dor-
drecht: Kluwer.
Breusch, T.S., and A.R. Pagan. 1980. The Lagrange mul-
tiplier test and its applications to model specification in
econometrics. Review of Economic Studies 47: 239-253.
Brock, W., and S. Durlauf. 2002. A multinomial choice
model with neighborhood effects. American Economic
Review 92: 298-303.
Brock, W., and S. Durlauf. 2006. Macroeconomics and
model uncertainty. In Post-Walrasian macroeconomics:
Beyond the dynamic stochastic general equilibrium
model, ed. D. Colander. New York: Cambridge Univer-
sity Press.
Brown, T.M. 1952. Habit persistence and lags in consumer
behaviour. Econometrica 20: 355-371.
Brown, R.L., J. Durbin, and J.M. Evans. 1975. Techniques
for testing the constancy of regression relationships over
time (with discussion). Journal of the Royal Statistical
Society. Series B 37: 149-192.
Broze, L., C. Gourieroux, and A. Szafarz. 1985. Solutions of
dynamic linear rational expectations models. Econo-
metric Theory 1: 341-368.
Brundy, J.M., and D.N. Jorgenson. 1971. Efficient estimation
of simultaneous equations by instrumental variables. The
Review of Economics and Statistics 53: 207-224.
Buckley, J., and I. James. 1979. Linear regression with
censored data. Biometrika 66: 429-436.
3233
Burdett, K, and D.T. Mortensen. 1998. Wage differentials,
employer size, and unemployment. International Eco-
nomic Review 39: 257-273.
Bums, A.F., and W.C. Mitchell. 1947. Measuring business
cycles. New York: Columbia University Press for the
NBER.
Burtless, G. 1995. The case for randomized field trials in
economic and policy research. Journal of Economic
Perspectives 9 (2): 63-84.
Cagan, P. 1956. The monetary dynamics of hyperinflation. In
Studies in the quantity theory of money, ed. M. Friedman.
Chicago: University of Chicago Press.
Cameron, A.C., and PK. Trivedi. 1986. Econometric models
based on count data: Comparisons and applications of
some estimators and tests. Journal of Applied
Econometrics 1: 29-53.
Cameron, A.C., and PK. Trivedi. 1998. Regression analysis
for count data, Econometric Society Monograph No. 30.
Cambridge: Cambridge University Press.
Cameron, A.C., and PK. Trivedi. 2005. Microeconometrics:
Methods and applications. Cambridge: Cambridge
University Press.
Campbell, J.Y., A.W. Lo, and A.C. MacKinlay. 1997. The
econometrics of financial markets. Princeton: Princeton
University Press.
Canova, F., and G. de Nicolo. 2002. Monetary disturbances
matter for business fluctuations in the G7. Journal of
Monetary Economics 49: 1131-1159.
Champemowne, D.G. 1948. Sampling theory applied to
autoregressive sequences. Journal of the Royal Statistical
Society, Series B 10: 204—231.
Champemowne, D.G. 1960. An experimental investigation of
the robustness of certain procedures for estimating means
and regressions coefficients. Journal of the Royal
Statistical Society 123: 398-412.
Chib, S. 1992. Bayes inference in the Tobit censored
regression model. Journal of Econometrics 51: 79-99.
Chib, S. 2001. Markov chain Monte Carlo methods:
Computation and inference. In Handbook of
econometrics, ed. J.J. Heckman and E. Learner, vol. 5.
Amsterdam: North-Holland.
Chib, S., and E. Greenberg. 1995. Understanding the
Metropolis-Hastings algorithm. The American Statistician
49: 327-335.
Chib, S., and E. Greenberg. 1998. Analysis of multivariate
probit models. Biometrika 85: 347-361.
Choi, I. 2006. Nonstationary panels. In Palgrave handbooks
of econometrics, ed. T.C. Mills and K. Patterson, vol. 1.
Basingstoke: Palgrave MacMillan.
Chow, G.C. 1960. Tests of equality between sets of coef-
ficients in two linear regressions. Econometrica 28: 591-
605.
Christ, C.F. 1952. Economic theory and measurement: A
twenty-year research report, 1932-52. Chicago: Cowles
Commission for Research in Economics.
Christ, C.F. 1983. The founding of the Econometric Society
and Econometrica. Econometrica 51: 3-6.
Christiano, L.J., M. Eichenbaum, and C. Evans. 2005.
Nominal rigidities and the dynamic effects of a shock
3234
to monetary policy. Journal of Political Economy 113: 1-
45.
Clare, A.D., S.H. Thomas, and M.R. Wickens. 1994. Is the
gilt equity yield ratio useful for predicting UK stock
return? Economic Journal 104: 303-315.
Clements, M.P., and D.F. Hendry. 1998. Forecasting eco-
nomic time scries. Cambridge: Cambridge University
Press.
Clements, M.P., and D.F. Hendry. 1999. Forecasting non-
stationary economic time scries. Cambridge, MA: MIT
Press.
Clements, M.P., and D.F. Hendry. 2006. Forecasting with
breaks. In Handbook of economic forecasting. ed. G.
Elliott, C.W.J. Granger, and A. Timmermann, vol. 1.
Amsterdam: Northri lolland.
Cochrane, .1. 2005. Asset pricing, rev. ed. Princeton:
Princeton University Press.
Cochrane, P, and G.H. Orcutt. 1949. Application of least
squares regression to relationships containing auto-
correlated error terms. Journal of the American Statistical
Association 44: 32-61.
Conley. T.G., and B. Dupor. 2003. A spatial analysis of
sectoral complementarity. Journal of Political Economy
111: 311 352.
Conley. T.G., and G. Topa. 2002. Socio-economic distance
and spatial patterns in unemployment. Journal of Applied
Econometrics 17:303 327.
Cooper, R.L. 1972. The predictive performance of quarterly
econometric models of the United States. In Econometric
models of cyclical behavior, ed.
B. G. Ilickman. New York: NBER.
Cosslett, S.R. 1983. Distribution free maximum likelihood
estimation of the binary choice model. Econometrica
51:765 782.
Cox, D.R. 1961. Tests of separate families of hypotheses. In
Proceedings of the fourth Berkeley symposium on
mathematical statistics and probability, vol. 1. Berkeley:
University of California Press.
Cox, D.R. 1962. Further results of tests of separate families of
hypotheses. Journal of the Royal Statistical Society,
Scries II 24: 406-424.
Cox, D.R. ll>72. Regression models and life tables. Journal of
the Roval Statistical Society Series B 34: 187 220.
Cross, R. 1782. The Duhem-Quine thesis, Lakatos and the
appraisal of theories in macroeconomics. Economic
Journal 72: 320-340.
Davenant, C. 1698. Discourses on the public revenues and on
the trade of England, vol. 1. London.
Davidson, R., and J.G. MacKinnon. 2004. Econometric
theory and methods. Oxford: Oxford University Press.
Davis, P.J., and P. Rabinowitz. 1984.Methods of numerical
integration. Orlando: Academic.
Deaton, A. 1985. Panel data from time series of cross-
sections. Journal of Econometrics 30: 109-126.
Del Negro, M., and F. Schorfheide. 2004. Priors from
equilibrium models for YAR's. International Economic
Review 45: 643-673.
Econometrics
Del Negro, M., F. Schorfheide, F. Smets, and R. Wouters.
2005. On the fit and forecasting performance of new
Keynesian models. Working Paper No. 491. Frankfurt:
European Central Bank.
Dhrymes. P. 1971. A simplified estimator for large-scale
econometric models. Australian Journal of Statistics 13:
168-175.
Doan, T.. R. Litteiman. and C.A. Sims. 1984. Forecasting and
conditional projections using realistic prior distributions.
Econometric Reviews 3: 1 100.
Domenich. T.. and D. McFadden. 1975. Urban travel
demand: A behavioral analysis. Amsterdam: North-
Holland.
Duffie. D.. and K. Singleton. 1993. Simulated moments
estimation of Markov models of asset prices.
Econometrica 61: 929 952.
Duncan. G. 1980. Formulation and statistical analysis of the
mixed continuous/discrete variable model in classical
production theory. Econometrica 48: 839 852.
Durbin. J. 1954. Errors in variables. Review of the Interna-
tional Statistical Institute 22: 23 32.
Durbin. J.. and G.S. Watson. 1950. Testing for serial cor-
relation in least squares regression I. Biomctrika 37: 409-
428.
Durbin. .1., and G.S. Watson. 1951. Testing for serial cor-
relation in least squares regression II. Biomctrika 38: 159
178.
Durlauf S.N., PA. Johnson, and J.R.W. Temple. 2005.
Growth econometrics. In Handbook of economic growth,
ed. P. Aghion and S.N. Durlauf, vol. 1A. Amsterdam:
North-Holland.
Efron. B. 1979. Bootstrap methods: Another look at the
jackknife. Annals of Statistics 7: 1 —26.
Eisner. R.. andR.H. Strotz 1963. Determinants of business
investment. In Impacts of monetary policy Englewood
Cliffs: Prentice-Hall, for the Commission on Money and
Credit.
Elliott. G.. C.W.J. Granger, and A. Timmermann. 2006.
Handbook of economic forecasting. Vol. 1. Amsterdam:
North-Holland.
Engle. R.F. 1982. Autoregressive conditional hetero-
scedasticity. with estimates of the variance of United
Kingdom inflation. Econometrica 50: 987-1007.
Engle. R.F. 1984. Wald likelihood ratio and Lagrange
multiplier tests in econometrics. In Handbook of
econometrics, ed. Z. Grihches and M.D. Intriligator, vol.
2. Amsterdam: North-Holland.
Engle, R.F., and G. Granger. 1987. Cointegration and error-
correction: Representation, estimation and testing.
Econometrica 55: 251-276.
Engle, R.F., D.F. Hendry, and J.-F. Richard. 1983. Exo-
geneity. Econometrica 51: 277- 304.
Fachin, S. 2006. Long-run trends in internal migrations in
Italy: A study in panel cointegration with dependent
units. Journal of Applied Econometrics (forthcoming).
Fama, E.F. 1970. Efficient capital markets: A review of
theory and empirical work. Journal of Finance 25: 383
417.
Econometrics
Farna, E.F. 1991. Efficient capital markets: II. Journal of
Finance 46: 1575-1617.
Fama, E.F., and K.R. French. 1989. Business conditions and
expected returns on stocks and bonds. Journal of
Financial Economics 25: 23^49.
Fan, J., and I. Gijbels. 1996. Local polynomial modelling and
its applications. London: Chapman & Hall.
Fernandez-Villaverde, J., and S. Rubio-Ramirez. 2005.
Estimating dynamic equilibrium economies: Linear
versus nonlinear likelihood. Journal of Applied Econo-
metrics 20: 891-910.
Fisher, R.A. 1928. Statistical methods for research workers.
2nd ed. London: Oliver and Boyd.
Fisher, I. 1930. The theory of interest. New York: Macmillan.
Reprinted, Philadelphia: Porcupine Press, 1977.
Fisher, I. 1937. Note on a short-cut method for calculating
distributed lags. Bulletin de I’Institut International de
Statistique 29: 323- 327.
Fisher, F.M. 1966. The identification problem in economet-
rics. New York: McGraw-Hill.
Friedman, M. 1957. A theory of the consumption function.
Princeton: Princeton University Press.
Frisch, R. 1933. Pitfalb in the statistical construction of
demand and supply curves. Leipzig: Hans Buske Verlag.
Frisch, R. 1934. Statbtical confluence analysis by means of
complete regression systems. Oslo: University Institute of
Economics.
Frisch, R. 1936. A note on the term ‘econometrics’.
Econometrica 4: 95.
Gali, J. 1992. How well does the IS-LM model fit postwar US
data? Quarterly Journal of Economics 107: 709-738.
Garratt, A., K. Lee, M.H. Pesaran, and Y. Shin. 2003a. A long
run structural macroeconometric model of the UK.
Economic Journal 113 (487): 412-455.
Garratt, A., K. Lee, M.H. Pesaran, and Y. Shin. 2003b.
Forecast uncertainty in macroeconometric modelling: An
application to the UK economy. Journal of the American
Statistical Association 98 (464): 829-838.
Garratt, A., D. Robertson, and S. Wright. 2006a. Permanent
vs transitory components and economic fundamentals.
Journal of Applied Econometrics 21: 521-542.
Garratt, A., K. Lee, M.H. Pesaran, and Y. Shin. 2006b.
Global and national macroeconometric modelling: A
long-run structural approach. Oxford: Oxford University
Press.
Geary, R.C. 1949. Studies in relations between economic time
series. Journal of the Royal Statistical Society, Series B
10: 140-158.
Gelfand, A.E., and A.F.M. Smith. 1990. Sampling based
approaches to calculating marginal densities. Journal of
the American Statistical Association 85: 398^109.
Geman, S., and D. Geman. 1984. Stochastic relaxation, Gibbs
distributions and the Bayesian restoration of images.
IEEE Transactions on Pattern Analysis and Machine
Intelligence 6: 721-741.
3235
Geweke, J. 1989a. Bayesian inference in econometric models
using Monte Carlo integration. Econometrica 57: 1317-
1340.
Geweke, J. 1989b. Efficient simulation from the multivariate
normal dbtribution subject to linear inequality
constrainb and the evaluation of constraint probabilities.
Discussion paper, Duke University.
Geweke, J. 1991. Efficient simulation from the multivariate
normal and student-t distributions subject to linear
constraints. In Computing science and statistics: Pro-
ceedings of the twenty-third symposium on the interface,
ed. E.M. Keramidas. Fairfax: Interface Foundation of
North America.
Geweke, J. 1992. Evaluating the accuracy of sampling- based
approaches to the calculation of posterior moments. In
Bayesian statistics 4, ed. J.M. Bernardo et al. Oxford:
Clarendon Press.
Geweke, J. 1996. Monte Carlo simulation and numerical
integration. In Handbook of computational economics,
ed. H.M. Amman, D.A. Kendrick, and J. Rust. Amster-
dam: North-Holland.
Geweke, J. 1999. Using simulation methods for Bayesian
econometric models: Inference, development and com-
munication (with discussion and rejoinder). Econometric
Reviews 18: 1-126.
Geweke, J. 2005. Contemporary Bayesian econometrics and
statistics. New York: Wiley.
Geweke, J., and M. Keane. 2001. Computationally intensive
methods for integration in econometrics. In Handbook of
econometrics, ed. J. Heckman and E.E. Learner, vol. 5.
Amsterdam: North-Holland.
Geweke, J., and C. Whiteman. 2006. Bayesian forecasting. In
Handbook of economic forecasting, ed. G. Elliott,
C. W.J. Granger, and A. Timmermann. Amsterdam:
North-Holland.
Geweke, J., M. Keane, and D. Runkle. 1994. Alternative
computational approaches to statistical inference in the
multinomial probit model. The Review of Economics
and Statistics 76: 609-632.
Geweke, J., M. Keane, and D. Runkle. 1997. Statistical
inference in the multinomial multiperiod probit model.
Journal of Econometrics 80: 125-165.
Godfrey, L.G. 1988. Misspecification tests in econometrics:
The LM principle and other approaches. Cambridge:
Cambridge University Press.
Godfrey, L.G., and M.R. Wickens. 1982. Tests of mis-
specification using locally equivalent alternative models.
In Evaluation and reliability of macro-economic models,
ed. G.C. Chow and P. Corsi. New York: Wiley.
Golub, G.H., and J.H. Welsch. 1969. Calculation of Gaussian
quadrature rules. Mathematics of Computation 23: 221-
230.
Gourieroux, C., and J. Jasiak. 2001. Financial econometrics:
Problems, models, and methods. Oxford: Oxford
University Press.
Gourieroux, C., and A. Monfort. 1993. Simulation based
inference: A survey with special reference to panel data
models. Journal of Econometrics 59: 5-33.
3236
Gourieroux, C., and A. Monfort. 1996. Simulation-based
econometric methods. New York: Oxford University
Press.
Gradshteyn, I S., and I.M. Ryzhik. 1965. Tables of integrals,
series and products. New York: Academic.
Granger, C.W.J. 1969. Investigating causal relations by
econometric models and cross-spectral methods.
Econometrica 37: 424—438.
Granger, C.W.J. 1986. Developments in the study of co-
integrated economic variables. Oxford Bulletin of
Economies and Statistics 48: 213-228.
Granger, C.W.J., and P. Newbold. 1974. Spurious regressions
in econometrics. Journal of Econometrics 2: 111-120.
Granger, C.W.J., and P. Newbold. 1986. Forecasting eco-
nomic time series. 2nd ed. San Diego: Academic.
Granger, C.W.J., and M.H. Pesaran. 2000a. A decision
theoretic approach to forecast evaluation. In Statistics and
finance:. In interface, ed. W.S. Chan, W.K. Li, and H.
Tong. London: Imperial College Press.
Granger, C.W.J., and M.H. Pesaran. 2000b. Economic and
statistical measures of forecast accuracy. Journal of
Forecasting 19: 537-560.
Greene, W.H. 2003. Econometric analysis. 5th ed. New
Jersey: Prentice Hall.
Gronau, R. 1974. Wage comparisons - A selectivity bias.
Journal of Political Economy 82: 1119-1143.
Guerre, E., I. Perrigne, and Q. Vuong. 2000. Optimal
nonparametric estimation of first-price auctions.
Econometrica 68: 525-574.
Haavelmo, T. 1943. Statistical testing of business cycle
theories. The Review of Economics and Statistics 25: 13
18.
Haavelmo, T. 1944. The probability approach in econo-
metrics. Econometrica 12 (Supplement): 1 118.
Hajivassiliou, V., D. McFadden, andP. Ruud. 1991. Simu-
lation of multivariate normal rectangle probabilities.
Methods and programs mimeo. Berkeley: University of C
alifomia.
Hajivassiliou, V., D. McFadden, and P. Ruud. 1996. Sim-
ulation of multivariate normal rectangle probabilities and
their derivatives: Theoretical and computational results.
Journal of Econometrics 72: 85 134.
Hall. A.R. 2005. Generalized method of moments. Oxford:
Oxford University Press.
Halton, J.M. 1960. On the efficiency of evaluating certain
quasi-random sequences of points in evaluating multi-
dimensional integrals. Numerische Mathematik 2:84-90.
Hamilton, J.D. 1994. Time series analysis. Princeton:
Princeton University Press.
1 lammersley, J.M. 1960. Monte Carlo methods for solving
multivariate problems. Annals of the New York Academy
of Sciences 86: 844-874.
1 lammersley. J.M., and D.C. Handscomb. 1964. Monte
Carlo methods. London: Methuen.
Hansen. L.P 1982. Large sample properties of generalized
method of moments. Econometrica 50. 1029 1054.
Hansen. L.P. and T.J. Sargent. 1980. Formulating and
estimating dynamic linear rational expectations
Econometrics
models. Journal of Economic Dynamics and Control 2: 7
46.
Hansen, L.P, and T.J. Sargent. 2007. Robustness. Princeton:
Princeton University Press.
I lardle, W. 1990. Applied nonparametric estimation. Cant-
bridge: Cambridge University Press.
Hardle, W., and O. Linton. 1994. Applied nonparametric
methods. In Handbook of econometrics, ed. R.F. Engle
and D. McFadden, vol. 4. Amsterdam: North-Holland.
Hart, B.S., and .1. von Neumann. 1942. Tabulation of the
probabilities for the ratio of mean square successive
difference to the variance. Annals of Mathematical
Statistics 13: 207 214.
Harvey, A. 1989. Forecasting, structural time series models
and Kalman Filter. Cambridge: Cambridge University
Press.
Hastings, W.K. 1970. Monte Carlo sampling methods rising
Markov chains and their applications. Biometrika 57: 97
109.
I laiisman, J.A. 1978. Specification tests in econometrics.
Econometrica 46: 1251 1272.
I kinsman, J.A., and W.K. Newey. 1995. Nonparametric
estimation of exact consumers surplus and deadweight
loss. Econometrica 63: 1445 1476.
Hausman, J.A., and D.A. Wise, eds. 1985. Social experi-
mentation, NBHR Conference Report. Chicago: Uni-
versity of Chicago Press.
I lausman, J.A., B.I I. I kill, and Z. Griliches. 1984.
Econometric models for count data with application to
the patents R&D relationship. Econometrica 52: 909
1038.
Heckman. J.J. 1974. Shadow prices, market wages, and labor
supply. Econometrica 42: 679-694.
Heckman. J.J.. andB. Singer. 1984. Econometric duration
analysis. Journal of Econometrics 24: 63-132.
Heckman. J.J.. and A.J. Smith. 1995. Assessing the case for
social experimentation. Journal of Economic Per-
spectives 9(2): 85-110.
Heckman. J.J.. and R. Willis. 1977. A beta-logistic model for
the analysis of sequential labour force participation by
married women. Journal of Political Economy 85: 27-58.
Hendricks, K.. and R.H. Porter. 1988. An empirical study of
an auction with asymmetric information. American
Economic Review 78: 865-883.
Hendry, D.F.. and J.-F. Richard. 1982. On the formulation of
empirical models in dynamic econometrics. Journal of
Econometrics 20: 3—33.
Holly. S.. M.H. Pesaran. and T. Yanragata. 2006. A spatio-
temporal model of house prices in the US. Mimeo:
University of Cambridge.
Holtz-Eakin, D., W.K. Newey and I I S. Rosen. 1988.
Estimating vector autoregressions with panel data.
Econometrica 56: 1371-1395.
Honore. B.. and E. Kyriazidou. 2000. Panel data discrete
choice models with lagged dependent variables.
Econometrica 68: 839-874.
Hooker. R.H. 1901. Correlation of the marriage rate with
trade. Journal of the Royal Statistical Society 44: 485
492.
Econometrics
Horowitz, J.L. 1998. Semiparametric methods in econo-
metrics. New York: Springer-Verlag.
Horowitz, J.L. 2001. Semiparametric models. In International
encyclopedia of behavioral and social sciences, ed. N.J.
Smelser and P.B. Baltes. Amsterdam: Elsevier.
Horowitz, J.L., and S. Lee. 2002. Semiparametric methods in
applied econometrics: Do the models fit the data?
Statistical Modelling 2: 3-22.
Horowitz, J.L., and N.E. Savin. 2001. Binary response
models: Logits, probits, and semiparametrics. Journal of
Economic Perspectives 15 (4): 43-56.
Hotz, V.J., C.H. Mullin, and S.G. Sanders. 1997. Bounding
causal effects using data from a contaminated natural
experiment: Analyzing the effects of teenage childbear-
ing. Review of Economic Studies 64: 575-603.
Hsiao, C. 2003. Analysis of panel data. 2nd ed. Cambridge:
Cambridge University Press.
Hsiao, C., and M.H. Pesaran. 2007. Random coefficient panel
data models. In The econometrics of panel data:
Fundamentals and recent developments in theory and
practice. 3rd ed., ed. L. Matyas and P. Sevestre. Dor-
drecht: Kluwer.
Hsiao, C., M.H. Pesaran, and A.K. Tahmiscioglu. 1999.
Bayes estimation of short-run coefficients in dynamic
panel data models. In Analysis of panels and limited
dependent variables models, ed. C. Hsiao et al. Cam-
bridge: Cambridge University press.
Hsiao, C., M.H. Pesaran, and A.K. Tahmiscioglu. 2002.
Maximum likelihood estimation of fixed effects dynamic
panel data models covering short time periods. Journal of
Econometrics 109: 107-150.
Irn, K.S., M.H. Pesaran, and Y. Shin. 2003. Testing for unit
roots in heterogenous panels. Journal of Econometrics
115:53-74.
Imbs, J., H. Murntaz, M.O. Ravn, and H. Rey. 2005. PPP
strikes back, aggregation and the real exchange rate.
Quarterly Journal of Economics 120: 1—43.
Johansen, S. 1988. Statistical analysis of cointegration
vectors. Journal of Economic Dynamics and Control 12:
231-254. Reprinted in Long-run economic relationships,
ed. R.F. Engle and C.W.J. Granger. Oxford: Oxford
University Press, 1991.
Johansen, S. 1991. Estimation and hypothesis testing of
cointegrating vectors in Gaussian vector autoregressive
models. Econometrica 59: 1551-1580.
Johansen, S. 1995. Likelihood-based inference in
cointegrated vector autoregressive models. Oxford:
Oxford University Press.
Jorgenson, D.W. 1966. Rational distributed lag functions.
Econometrica 34: 135-149.
Judd, K.L. 1998. Numerical methods in economics.
Cambridge, MA: MIT Press.
Juselius, K. 2006. The cointegrated VAR model: Econometric
methodology and macroeconomic applications. Oxford:
Oxford University Press.
Kagel, J., and A.E. Roth, eds. 1995. The handbook of
experimental economics. Princeton: Princeton University
Press.
3237
Kandel, S., andR.F. Stambaugh. 1996. On the predictability of
stock returns: An asset-allocation perspective. Journal of
Finance 51: 385-424.
Keane, M.P. 1990. A computationally practical simulation
estimator for panel data, with applications to estimating
temporal dependence in employment and wages.
Discussion paper, University of Minnesota.
Keane, M., and K.I. Wolpin. 1994. The solution and esti-
mation of discrete choice dynamic programming models
by simulation: Monte Carlo evidence. The Review of
Economics and Statistics 76: 648-672.
Keynes, J.M. 1939. The statistical testing of business cycle
theories. Economic Journal 49: 558-568.
Kilian, L. 1998. Small-sample confidence intervals for
impulse response functions. The Review of Economics
and Statistics 80: 218-229.
Kim, K, and A.R. Pagan. 1995. The econometric analysis of
calibrated macroeconomic models. In Handbook of
applied econometrics: Macroeconomics, ed. M.H.
Pesaran and M. Wickens. Oxford: Basil Blackwell.
Klein, L.R. 1947. The use of econometric models as a guide
to economic policy. Econometrica 15: 111-151.
Klein, L.R. 1950. Economic fluctuations in the United States
1921-1941, Cowles Commission Monograph No. 11.
New York: Wiley.
Kloek, T., and H.K. van Dijk. 1978. Bayesian estimates of
equation system parameters: An application of integration
by Monte Carlo. Econometrica 46: 1-20.
Koop, G. 2003. Bayesian econometrics. Chichester: Wiley.
Koop, G., and S. Potter. 2004a. Forecasting and estimating
multiple change-point models with an unknown number
of change-points. Mimeo: University of Leicester and
Federal Reserve Bank of New York.
Koop, G., and S. Potter. 2004b. Prior elicitation in multiple
change-point models. Mimeo: University of Leicester and
Federal Reserve Bank of New York.
Koop, G., M.H. Pesaran, and S.M. Potter. 1996. Impulse
response analysis in nonlinear multivariate models.
Journal of Econometrics 74: 119-147.
Koopmans, T.C. 1937. Linear regression analysis of eco-
nomic time series. Haarlem: De Erven F. Bohn for the
Netherlands Economic Institute.
Koopmans, T.C. 1949. Identification problems in economic
model construction. Econometrica 17: 125-144.
Koopmans, T.C., H. Rubin, andR.B. Leipnik. 1950. Mea-
suring the equation systems of dynamic economics. In
Statistical inference in dynamic economic models, Cowles
Commission Monograph No. 10, ed. T.C. Koopmans.
New York: Wiley.
Koyck, L.M. 1954. Distributed lags and investment analysis.
Amsterdam: North-Holland.
Laffont, J.-J., H. Ossard, and Q. Vuong. 1995. Econometrics
of first-price auctions. Econometrica 63: 953-980.
Lancaster, T. 2004. An introduction to modern Bayesian
econometrics. Malden: Blackwell.
Learner, E.E. 1978. Specification searches: Ad Hoc inference
with non-experimental data. New York: Wiley.
3238
Lee, L.F. 1995. Asymptotic bias in simulated maximum
likelihood estimation of discrete choice models. Eco-
nomic Theory 11: 437-483.
Lee, B.S.. and B. Ingram. 1991. Simulation estimation of
time-series models. Journal of Econometrics 47: 197-205.
Lenoir, M. 1913. Etudes sur la formation et le mouvement
desprix. Paris: Giard et Briere.
Lerman, S.. and C.S. Manski. 1981. On the use of simulated
frequencies to approximate choice probabilities. In
Structural analysis of discrete data with econometric
applications, ed. C.F. Manski and D. McFadden. Cam-
bridge. MA: MIT Press.
Levin, A., C. Lin. and C.J. Chu. 2002. Unit root tests in panel
data: Asymptotic and finite-sample properties. Journal of
Econometrics 108: 1-24.
Liesenfeld. R.. and J. Breitling. 1999. Simulation based
method of moments. In Generalized method of moment
estimation, ed. L. Tatyas. Cambridge: Cambridge Uni-
versity Press.
Litterman. R.B. 1985. Forecasting with Bayesian vector
autoregressions: Fir e years of experience. Journal of
Business and Economic Statistics 4: 25 38.
Liu, T.C. 1960. Underidentification, structural estimation and
forecasting. Econometrica 28: 855 865.
Lo, A., and C. MacKinlay. 1988. Stock market prices do not
follow random walks: Evidence from a simple
specification test. Review of Financial Studies 1: 41-66.
Lucas, R.li. ll)72. Expectations and the neutrality of money.
Journal of Economic Theory 4: 103- 124.
Lucas, R.E. 1973. Some international evidence on output-
inflation tradeoffs. American Economic Review 63: 326-
334.
Lucas, R.E. 1976. Econometric policy evaluation: A critique.
In The Phillips curve and labor markets, ed. K. Brunner
and A.M. Meltzer. Amsterdam: North-Holland.
Lucas, R.E., and T. Sargent 1681. Rational expectations and
econometric practice. Introduction to Rational
expectations and econometric practice. Minneapolis:
University of Minnesota Press.
Lutkepohl, H. 1991. Introduction to multiple time series
analysis. New York: Springer-Verlag.
Lyttkens, E. 1970. Symmetric and asymmetric estimation
methods. In Interdependent systems, ed. E. Mosback and
H. Wold. Amsterdam: North-Holland.
Maddala, G.S. 1983. Limited dependent and qualitative
variables in econometrics. Cambridge: Cambridge
University Press.
Maddala, G.S. 1986. Disequilibrium, self-selection, and
switching models. In Handbook of econometrics, ed. Z.
Griliches and M.D. Intriligator, vol. 3. Amsterdam:
North-Holland.
Maddala, G.S. 2001. Introduction to econometrics. 3rd ed.
New York: Wiley.
Manski, C.F. 1975. Maximum score estimation of the
stochastic utility model of choice. Journal of Econo-
metrics 3: 205-228.
Econometrics
Manski. C.F. 1985. Semiparametric analysis of discrete
response: Asymptotic properties of the maximum score
estimator. Journal of Econometrics 27: 3 13 334.
Manski. C.F. 1988. Identification of binary response models.
Journal of the American Statistical Association 83: 729-
738.
Manski. C.F. 1995. Identification problems in the social
sciences. Cambridge. MA: Harvard University Press.
Manski. C.F. 2003. Partial identification of probability
distributions. New York: Springer-Verlag.
Manski. C.F.. and D. McFadden. 1981. Structural analysis of
discrete data with econometric applications. Cambridge.
MA: MIT Press.
Matzkin. R.L. 1994. Restrictions of economic theory in
nonparametric methods. In Handbook of econometrics,
ed. R.F. Engle and D.L. McFadden. vol. 4. Amsterdam:
North-Holland.
McAleer. M.. and M.C. Medeiros. 2007. Realized volatility:
A review. Econometric Reviews.
McAleer. M.. A.R. Pagan, and PA. Volker. 1985. What will
take the con out of econometrics? American Economic
Review 75: 293-307.
McCulloch. R.E.. and PE. Rossi. 1994. An exact likelihood
analysis of the multinomial probit model. Journal of
Econometrics 64: 207-240.
McCulloch. R.E.. and R.S. Tsay. 1993. Bayesian inference
and prediction for mean and variance shifts in auto-
regressive time series. Journal of the American Statistical
Association 88: 965-978.
McCulloch. R.E.. and R.S. Tsay. 1994. Bayesian analysis of
autoregressive time series via the Gibbs sampler. Journal
of Time Series Analysis 15: 235 250.
McCulloch, R.E., N.G. Poison, and P.E. Rossi. 2000. A
Bayesian analysis of the multinomial probit model with
fully identified parameters. Journal of Econometrics 99:
173 193.
McFadden, D. 1989. A method of simulated moments for
estimation of multinomial probits without numerical
integration. Econometrica 57: 995 1026.
Metropolis, N., A.W. Rosenbluth, M.N. Rosenbluth, ATI.
Teller, and E. Teller. 1953. Equation of state calculations
by fast computing machines. Journal of Chemical
Physics 21: 1087 1092.
Mitchell, W.C. 1928. Business cycles: The problem in its
setting. New York: NBER.
Moon, R., and B. Perron. 2004. Testing for unit root in panels
with dynamic factors. Journal of Econometrics 122:81
126.
Moon, R., and B. Perron. 2007. An empirical analysis of
nonstationarity in a panel of interest rates with factors.
Journal of Applied Econometrics 22: 383 400.
Moore, H.L. 1914. Economic cycles: Their law and cause.
New York: Macmillan.
Moore, H.L. 1917. Forecasting the yield and the price of
cotton. New York: Macmillan Press.
Mortensen, D.T. 1990. Equilibrium wage distributions: A
synthesis. In Panel data and labor market studies, ed. .1.
Hartog, G. Kidder, and .1. Theeuwes. New York: North
Holland.
Econometrics
Mundlak, Y. 1961. Empirical production function free of
management bias. Journal of Farm Economics 43: 44-56.
Mundlak, Y. 1978. On the pooling of time series and cross
section data. Econometrica 46: 69-85.
Muth, J.F. 1961. Rational expectations and the theory of price
movements. Econometrica 29: 315-335.
Nadaraya, E.A. 1964. On estimating regression. Theory of
Probability and Its Applications 10: 141-142.
Nakamura, A., and M. Nakamura. 1981. On the relationships
among several specification error tests presented by
Durbin, Wu, and Hausman. Econometrica 49:1583-1588.
Nelson, C.R. 1972. The prediction performance of the FRB-
MIT-Penn model of the US economy. American
Economic Review 62: 902-917.
Nerlove, M. 1958a. Adaptive expectations and the cobweb
phenomena. Quarterly Journal of Economics 72:227-240.
Nerlove, M. 1958b. Distributed lags and demand analysis.
Washington, DC: USDA.
Newey, W.K. 1997. Convergence rates and asymptotic
normality for series estimators. Journal of Econometrics
79: 147-168.
Nickell, S. 1981. Biases in dynamic models with fixed effects.
Econometrica 49: 1399-1416.
Niederreiter, H. 1992. Random number generation and
Quasi-Monte Carol methods. Philadelphia: SIAM.
Nyblom, J. 1989. Testing for the constancy of parameters
over time. Journal of the American Statistical Association
84: 223-230.
O’Connell, P. 1998. The overvaluation of purchasing power
parity. Journal of International Economics 44: 1-19.
Orcutt, G.H. 1948. A study of the autoregressive nature of the
time series used for Tinbergen’s model of the economic
system of the United States, 1919-1932 (with discussion).
Journal of the Royal Statistical Society, Series B 10: 1-53.
Pagan, A.R., and M.H. Pesaran. 2007. On econometric
analysis of structural systems with permanent and tran-
sitory shocks and exogenous variables. Unpublished
manuscript.
Pagan, A., and A. Ullah. 1999. Nonparametric econometrics.
Cambridge: Cambridge University Press.
Pakes, A., and D. Pollard. 1989. Simulation and the asymp-
totics of optimization estimators. Econometrica 57: 1027-
1058.
Pedroni, P. 2001. Purchasing power parity tests in
cointegrated panels. The Review of Economics and
Statistics 83: 727-731.
Pedroni, P. 2004. Panel cointegration: Asymptotic and finite
sample properties of pooled time series tests with an
application to the PPP hypothesis. Econometric Theory
20: 597-625.
Peersman, G. 2005. What caused the early millennium
slowdown? Evidence based on autoregressions. Journal
of Applied Econometrics 20: 185-207.
Percy, D.F. 1992. Prediction for seemingly unrelated
regressions. Journal of the Royal Statistical Society,
Series B 54: 243-252.
3239
Pesaran, M.H. 1981. Identification of rational expectations
models. Journal of Econometrics 16: 375-398.
Pesaran, M.H. 1987a. Econometrics. In The new Palgrave: A
dictionary of economics, ed. J. Eatwell, M. Milgate, and
P. Newman, vol. 2. London: Macmillan.
Pesaran, M.H. 1987b. The limits to rational expectations.
Oxford: Basil Blackwell.
Pesaran, M.H. 2006. Estimation and inference in large
heterogeneous panels with cross section dependence.
Econometrica 74: 967-1012.
Pesaran, M.H. 2007a. A simple panel unit root test in the
presence of cross section dependence. Journal of Applied
Econometrics 22: 265-312.
Pesaran, M.H. 2007b. A pair-wise approach to testing for
output and growth convergence. Journal of Econometrics
138: 312-355.
Pesaran, M.H., and A.S. Deaton. 1978. Testing non-nested
nonlinear regression models. Econometrica 46: 677-694.
Pesaran, M.H., and S. Skouras. 2002. Decision-based
methods for forecast evaluation. In A companion to
economic forecasting, ed. M.P. Cements and
D. F. Hendry. Oxford: Blackwell Publishing.
Pesaran, M.H., and R.P. Smith. 1985a. Keynes on econo-
metrics. In Keynes' economics: Methodological issues, ed.
T. Lawson and M.H. Pesaran. London: Croom Helm.
Pesaran, M.H., and R.P. Smith. 1985b. Evaluation of
macroeconometric models. Economic Modelling 2: 125-
134.
Pesaran, M.H., and R. Smith. 1995. Estimating long-run
relationships from dynamic heterogeneous panels.
Journal of Econometrics 68: 79-113.
Pesaran, M.H., and R.P. Smith. 2006. Macroeconometric
modelling with a global perspective. The Manchester
School 74: 24-49.
Pesaran, M.H., and A. Timmermann. 1995. The robustness
and economic significance of predictability of stock
returns. Journal of Finance 50: 1201-1228.
Pesaran, M.H., and A. Timmermann. 2000. A recursive
modelling approach to predicting UK stock returns.
Economic Journal 110: 159-191.
Pesaran, M.H., and A. Timmermann. 2005a. Real time
econometrics. Econometric Theory 21: 212-231.
Pesaran, M.H., and A. Timmermann. 2005b. Small sample
properties of forecasts from autoregressive models under
structural breaks. Journal of Econometrics 129: 183-217.
Pesaran, M.H., and A. Timmermann. 2007. Selection of
estimation window in the presence of breaks. Journal of
Econometrics 137: 134-161.
Pesaran, M.H., and M. Weeks. 2001. Non-nested hypothesis
testing: An overview. In Companion to theoretical
econometrics, ed. B.H. Baltagi. Oxford: Basil Blackwell.
Pesaran, M.H., T. Schuermann, and S.M. Weiner. 2004.
Modelling regional interdependencies using a global
error-correcting macroeconometric model (with discus-
sion). Journal of Business and Economic Statistics 22
(129-62): 175-181.
3240
Pesaran, M.H., D. Pettenuzzo, and A. Timmermann. 2006.
Forecasting time series subject to multiple structural
breaks. Review of Economic Studies 73: 1057 1084.
Phillips, P.C.B. ll>83. Exact small sample theory in the
simultaneous equations model. In Handbook of
econometrics, ed. /,. Griliches and M.D. Intrilgator, vol. 1.
Amsterdam: North-Holland.
Phillips, P.C.B. ll>86. Understanding spurious regressions in
econometrics. Journal of Econometrics 33:311 340.
Phillips, P.C.B. 1991. Optimal inference in cointegrated
syslcms. Econometrica 59: 283 306.
Phillips, P.C.B., and H.R. Moon. 1999. Linear regression
limit theory for nonstationary panel data. Econometrica
67:1057-1111.
Phillips, P.C.B., and /,. Xiao. 1998. A primer on unit root
testing. Journal of Economic Surveys 12: 423 46 ).
l
Powell, J.L. ll>84. Least absolute deviations estimation for the
censored regression model. Journal of Econometrics 25:
303 325.
Powell, J.L. 1986. Censored regression quantiles. Journal of
Econometrics 32: 143 155.
Press, W.H., B.P Flannery, S.A. Teukolsky, and
W.T. Vetterling. ll>86. Numerical recipes: The art of
scientific computing, lsl ed. Cambridge: Cambridge
University Press.
Press, W.H., B.P. Flannery, S.A. Teukolsky, and
W.T. Vetterling. 1992. Numerical recipes: The art of
scientific computing. 2nd ed. Cambridge: Cambridge
University Press.
Prothero, D.L., and K.F. Wallis. ll>76. Modelling macro-
economic time series. Journal of the Royal Statistical
Society, Series A 139: 468-486.
Quandt, R.E. 1982. Econometric disequilibrium models.
Econometric Reviews 1: 1-63.
Ramsey, J.B. 1969. Tests for specification errors in classical
linear least squares regression analysis. Journal of the
Royal Statistical Society, Series B 31: 350-371.
Reiersol, O. 1941. Confluence analysis by means of lag
moments and other methods of confluence analysis.
Econometrica 9: 1-24.
Reiersol, O. 1945. Confluence analysis by means of instru-
mental sets of variables. Arkiv for Mathematik .istronomi
och Fysik 32: 1-119.
Rossi, P.E., G.M. Allenby, and R. McCulloch. 2005.
Bayesian statistics and marketing. Chichester: Wiley.
Rothenberg, T.J. 1984. Approximating the distributions of
econometric estimators and test statistics. In Handbook of
econometrics, ed. /,. Griliches and M.D. Intriligator, vol.
2. Amsterdam: North-Holland.
Rust, J. 1994. Structural estimation of Markov decision
processes. InHandbook of econometrics, ed. R.F. Engle
and D.L. MeFadden, vol. 4. Amsterdam: North- Holland.
Rust, J. 1997. Using randomization to break the curse of
dimensionality. Econometrica 65: 487-516.
Samuelson, P. 1965. Proof that properly anticipated prices
fluctuate randomly. Industrial Management Review 6: 41
49.
Econometrics
Sandor, Z., and P. Andras. 2004. Alternative sampling
methods for estimating multivariate normal probabilities.
Journal of Econometrics 120:207 234.
Santos, M.S., and J. Vigo-Aguiar. 1998. Analysis of a
numerical dynamic programming algorithm applied to
economic models. Econometrica 66: 40l) 426.
Sargan, J.D. 1958. The estimation of economic relationships
using instrumental variables. Econometrica 26: 393^415.
Sargan, J.D. 1964. Wages and prices in the United Kingdom:
A study in econometric methodology. In Econometric
analysis for national economic planning, ed. PE. Hart, G.
Mills, and J.K. Whitaker. London: Butterworths.
Sargent, T.J. 1973. Rational expectations, the real rate of
interest and the natural rate of unemployment.
Brookings Papers on Economic Activity 1973 (2):
429M72.
Sargent, T.J., andN. Wallace. 1975. Rational expectations and
the theory of economic policy. Journal of Monetary
Economics 2: 169 184.
Savin, N.E. 1973. Systems k-class estimators. Econometrica
41: 1125 1136.
Schultz, M. 1938. The theory and measurement of demand.
Chicago: University of Chicago Press.
Schumpeter, J.A. 1954. History of economic analysis.
London: George Allen & Unwin.
Shephard, N.. ed. 2005. Stochastic volatility: Selected
readings. Oxford: Oxford University Press.
Shiller, R.J. 1973. A distributed lag estimator derived from
smoothness priors. Econometrica 41: 775-788.
Sims, C.A. 1972. Money, income and causality. American
Economic Review 62: 540-552.
Sims, C.A. 1980. Macroeconomics and reality. Econometrica
48: 1M8.
Sims, C.A. 1982. Policy analysis with econometric models.
Brookings Papers on Economic Activity 1982 (1): 107-
164.
Slutsky, E. 1927. The summation of random causes as the
source of cyclic processes. In Problems of economic
conditions, vol. 3. Moscow. English trans. in
Econometrica 5 (1937): 105-146.
Smets, F., and R. Wouters. 2003. .An estimated stochastic
dynamic general equilibrium model of the euro area.
Journal of the European Economic Association 1: 1123-
1175.
Smith, V., S. Leyboume. T.-H. Kim. and P. Newbold. 2004.
More powerful panel data unit root tests with an appli-
cation to mean reversion in real exchange rates. Journal
of Applied Econometrics 19: 147-170.
Solow, R.M. 1960. On a family of lag distributions.
Econometrica 28: 393—406.
Srivastava, V.K. 1971. Three-stage least-squares and gener-
alized double k-class estimator's: A mathematical rela-
tionship. International Economic Review' 12: 312 316.
Stambaugh, R.F. 1999. Predictive regressions. Journal of
Financial Economics 54: 375-421.
Stem, S. 1997. Simulation-based estimation. Journal of
Economic Literature 35: 2006-2039.
Econometrics
Stock, J.H. 1994. Unit roots, structural breaks and trends. In
Handbook of econometrics, ed. R.F. Engle and D.L.
McFadden. Amsterdam: North-Holland.
Stock, J.H., and M.W. Watson. 1996. Evidence on structural
instability in macroeconomic time series relations.
Journal of Business and Economic Statistics 14: 11-30.
Stock, J.H., J.H. Wright, and M. Yogo. 2002. A survey of
weak instruments and weak identification in generalized
method of moments. Journal of Business and Economic
Statistics 20: 518-529.
Stone, J.R.N. 1945. The analysis of market demand. Journal
of the Royal Statistical Society Series A 108: 286-382.
Stone, J.R.N. 1978. Keynes, political arithmetic and
econometrics. Cambridge: Seventh Keynes Lecture in
Economics, British Academy.
Stone, C.J. 1985. Additive regression and other nonpara-
metric models. Annals of Statistics 13: 689-705.
Stone, J.R.N. et al. 1954. Measurement of consumers'
expenditures and behavior in the United Kingdom, 1920-
38,2 vols. London: Cambridge University Press.
Strachan, R.W., and H.K. van Dijk 2006. Model uncertainty
and Bayesian model averaging in vector autoregressive
processes. Discussion Papers in Economics 06/5,
Department of Economics, University of Leicester.
Swamy, P.A.V.B. 1970. Efficient inference in a random
coefficient regression model. Econometrica 38: 311-323.
Taylor, J.B., and H. Uhlig. 1990. Solving nonlinear stochastic
growth models: A comparison of alternative solution
methods. Journal of Business and Economic Statistics 8:
1-18.
Theil, H. 1954. Estimation of parameters of econometric
models. Bulletin of International Statistics Institute 34:
122-128.
Theil, H. 1958. Economic forecasts and policy. 2nd ed.
Amsterdam: North-Holland. 1961.
Tierney, L. 1994. Markov chains for exploring posterior
distributions with discussion and rejoinder. Annals of
Statistics 22: 1701-1762.
Tinbergen, J. 1929-30. Bestimmung und Deutung von
Angebotskurven: ein Beispiel. Zeitschrift fur
Nationalokonomie 1: 669-679.
Tinbergen, J. 1937. An econometric approach to business
cycle problems. Paris: Herman & Cie Editeurs.
Tinbergen, J. 1939. Statistical testing of business cycle
theories, vol. 1: A method and its application to invest-
ment activity; vol. 2: Business cycles in the United States
of America, 1919-1932. Geneva: League of Nations.
Tobin, J. 1958. Estimation of relationships for limited
dependent variables. Econometrica 26: 24—36.
Treadway, A.B.. 1971. On the multivariate flexible accel-
erator. Econometrica 39: 845-855.
Trivedi, P.K. 1975. Time series analysis versus structural
models: A case study of Canadian manufacturing
behaviour. International Economic Review 16: 587-608.
3241
Uhlig, H. 2005. What are the effects of monetary pohcy:
Results from an agnostic identification approach. Journal
of Monetary Economics 52: 381-419.
Verbeek, M. 2007. Pseudo panels and repeated cross-
sections. In The econometrics of panel data: Funda-
mentals and recent developments in theory and practice.
3rd ed., ed. L. Matyas and P. Sevestre. Dordrecht:
Kluwer.
von Neumann, J. 1941. Distribution of the ratio of the mean
square successive difference to the variance. Annals of
Mathematical Statistics 12: 367-395.
von Neumann, J. 1942. A further remark on the distribution of
the ratio of the mean square successive difference to the
variance. Annals of Mathematical Statistics 13: 86-88.
von Neumann, J. 1951. Various techniques used in connection
with random digits. Applied Mathematics Series 12: 36-
38. US National Bureau of Standards.
Wallis, K.F. 1977. Multiple time series analysis and the final
form of econometric models. Econometrica 45: 1481-
1497.
Wallis, K. 1980. Econometric implications of the rational
expectations hypothesis. Econometrica 48: 49-73.
Watson, G.M. 1964. Smooth regression analysis. Sankhyd,
Series A 26: 359-372.
Watson, M.W. 1994. Vector autoregressions and
cointegration. In Handbook of econometrics, ed. R.F.
Engle and D.L. McFadden, vol. 4. Amsterdam: North-
Holland.
Wegge, L.L. 1965. Identifiability criteria for a system of
equations as a whole. Australian Journal of Statistics 7:
67-77.
West, K.D. 2006. Forecast evaluation. In Handbook of
economic forecasting, ed. G. Elliott, C. Granger, and A.
Timmermann, vol. 1. Amsterdam: North-Holland.
Westerlund, J. 2005. Panel cointegration tests of the Fisher
effect, Working Papers 2005:10. Department of Eco-
nomics, Lund University.
White, H. 1981. Consequences and detection of mis- specified
nonlinear regression models. Journal of the American
Statistical Association 76: 419-433.
White, H. 1982. Maximum likelihood estimation of mis-
specified models. Econometrica 50: 1-26.
Whittle, P. 1963. Prediction and regulation by linear least-
squares methods. London: English Universities Press.
Wickens, M. 1982. The efficient estimation of econometric
models with rational expectations. Review of Economic
Studies 49: 55-68.
Wickens, M.R., and R. Motto. 2001. Estimating shocks and
impulse response functions. Journal of Applied
Econometrics 16: 371-387.
Wold, H. 1938. A study in the analysis of stationary time
series. Stockholm: Almqvist and Wiksell.
Wooldridge, J.M. 2002. Econometric analysis of cross section
and panel data. Cambridge, MA: MIT Press.
Wooldridge, J.M. 2006. Introductory econometrics: A modern
approach. 3rd ed. Stamford: Thomson- South-Westem.
3242
Working, E.J. 1927. What do statistical ‘demand curves’
show? Quarterly Journal of Economics 41: 212-235.
Wright, P.G. 1915. Review of economic cycles by Henry
Moore. Quarterlv Journal of Economics 29: 631-641.
Wright, P.G. 1928. The tariff on animal and vegetable oils.
London: Macmillan for the Institute of Economics.
Wu, D. 1973. Alternatives tests of independence between
stochastic regressor and disturbances. Econometrica 41:
733-750.
Yule, G.U. 1895, 1896. On the correlation of total pauperism
with proportion of outrelief. Economic Journal 5: 603-
611; 6: 613-623.
Yule, G.U. 1921. On the time-correlation problem, with
special reference to the variate-difference correlation
method. Journal of the Royal Statistical Society 84: 497-
526.
Yule, G.U. 1926. Why do we sometimes get nonsense
correlations between time-series? A study in sampling
and the nature of time-series. Journal of the Royal
Statistical Society 89: 1-64.
Zellner, A. 1962. An efficient method of estimating seem-
ingly unrelated regressions and tests for aggregation bias.
Journal of the American Statistical Association 57:348-
368.
Zellner, A. 1971. An introduction to Bayesian inference in
econometrics. New York: Wiley.
Zellner, A. 1984. Basic issues in econometrics. Chicago:
University of Chicago Press.
Zellner, A. 1985. Bayesian econometrics. Econometrica 53:
253-270.
Zellner, A., and F. Palm. 1974. Time series analysis and
simultaneous equation econometric models. Journal of
Econometrics 2: 17-54.
Zellner, A., and H. Theil. 1962. Three-stage least squares:
Simultaneous estimation of simultaneous equations.
Econometrica 30: 54-78.
Economic Anthropology
Timothy Earle
Abstract
Economic anthropology is an empirical science
that describes production, exchange and
consumption cross-culturally. All societies have
economies, but they are variable. Anthro-
pologists evaluate the operations of individual
economies and the applicability of Western
theories to these cases. Some economic pro-
cesses work broadly; for example, strategic
Economic Anthropology
decision-making, the law of competitive
advantage, and calculations of transaction costs
help explain many observed patterns. Ehiman
economies, however, are often structured as
intertwined sectors with distinctive processes.
Differences observed in productivity,
specialization, institutional structure and social
motivations across history and across modem
societies are of theoretical significance when
constructing the limits of general theory.
Keywords
Behavioural economics; Capitalism; Cognitive
ability; Commodity chains; Division of labour;
Domestic mode of production; Dual economies;
Economic anthropology; Formalism; Gift
exchange; Household production; Hunting and
gathering economies; Identity; Institutional
economics; Markets; Polanyi, K.; Political
sector; Postmodernism; Prestige economy;
Rationality; Reciprocity; Social networks;
Social sector; Specialization; Staple finance and
wealth finance; Subsistence sector;
Substantivism; Surplus; Technological progress;
Transactions costs; Weber, M
JEL Classifications
Z1
Economic anthropology is an empirical science that
seeks to describe how production, exchange and
consumption operate outside the West (compare
Hunt ). The second edition (1952) of Herskovits’s (
) text, titled Eco
nomic Anthropology’, labelled this sub-discipline in
anthropology. The broader mission of anthropology
has been to make sense of the diversity in the
human experience, which became apparent to
Europeans during progressive stages of exploration,
colonialization and globalization. Underlying
anthropological research is the premise that human
societies have developed parallel institutions of
aesthetics, religion, kinship, politics, and of course
economics. All societies have economies, and the
economic patterns observed in non-Westem
economies both comfort and confront theories
developed by Western scholars.
Economic Anthropology
Common economic processes, such as rational
decision-making, law of competitive advantage, and
institutional economics help explain many patterns
across human economies based on variable
conditions of cost, demand and availability.
Additionally, however, human economies appear
often to be structured quite differently from West-
ern models, and these differences in institutional
structure and motivation are of theoretical signif-
icance. From the beginning, economic anthropology
has contained, and more or less successfully
resolved, a tension between the desire to find cross-
cultural ly general theories and to recognize the
uniqueness of each individual case. In economic
anthropology this tension has been represented in
the formalist-substantivist debate.
Few anthropologists identify themselves pri-
marily as economic anthropologists, but study eco-
nomic matters as part of a broadly integrative
approach to human societies. Founded in 1980, the
Society of Economic Anthropology is the primary
organization for anthropologists with such interests.
Members include ethnographers, applied
development anthropologists, archaeologists and
ethnohistorians, suggesting that economic studies
bridge the diversity of the discipline. The society
sponsors annual meetings on themes that range
across topics including key institutions of labour,
property, markets and consumption, and special
topics from the gift to slow foods. Research Series
in Economic Anthropology and Society for Eco-
nomic Anthropology Monographs offer edited vol-
umes on the sub-discipline.
History of Economic Anthropology
From early in the 20th century, anthropologists
have questioned whether theories developed to
understand Western market economies apply only
to those Western societies for which they were
generated. To answer this question, anthropologists
have described traditional economies, which
survived into the 20th century, which existed in the
past, and which have been transformed by
engagement with the West. Largely empirical, the
work is of substantial theoretical significance for
understanding economies cross-
3243
culturally. Gudeman ( ) has compiled many
of the most highly referenced articles.
Economic anthropology’s beginning traces to
the landmark ethnography Argonauts of the West-
ern Pacific, in which Malinowski ( )
described the circulation of shell valuables among
the islands of the Kula Ring. Malinowski used the
Trobriand Islanders’ obsession with certain shell
valuables to challenge simplistic notions of
‘economic man’, and he argued that a non- Westem
economy could be fundamentally different from
modem market economies in values and socialized
exchange relationships. Anthropological studies of
traditional economies thrived during the first half of
the 20th century. As part of British functionalism,
Malinowski and his students developed the
approach; in French structuralism, Mauss ( )
focused on the gift as a social phe
nomenon; and, within American anthropology,
Herskovits ( ) defined the sub-field. Much of
the work was descriptive, emphasizing how tradi-
tional people meet basic needs and how the
exchange of primitive valuables fashioned and
maintained social relationships.
By mid-century, however, studies of traditional
economy were increasingly adopting the terms and
concepts of Western economic theory. Both
Herskovits ( ) and Firth ( ) revised their
original books on traditional economies so as to
clarify underlying similarities across world econ-
omies. They each took concepts, like scarcity and
specialization, and generalized them to show that
they apply well to societies in which market pen-
etration is not great. They were making the essential
point that traditional economies were not simply
driven by the food quest. Although most
anthropologists took pains to emphasize the dif-
ferences between traditional economies and market-
integrated systems, some seemed to homogenize the
human experience, and a sharp reaction followed.
In the tradition of Max Weber, economic his-
torian Karl Polanyi ( ) wrote his famous trea
tise The Great Transformation to argue that the
integrating structure of modem markets, for which
prices are set by supply and demand, are a very
recent creation of industrialism and capitalism.
Theories based on scarcity, rationality,
3244
equilibrating price mechanisms operated, he argued,
only in the special case of Western capitalism.
Modem market conditions should not be taken as
inherent in the human experience, but as a recent
social artifact malleable in future societies.
Polanyi’s impact on economic anthropology was
profound and created the debate between
substantivists and formalists that raged in the sub-
discipline for a generation. Trade and Markets in
the Early Empires (Polanyi et al. ), the seminal
edited book, came out of a discussion group which
Polanyi led at Columbia University and which
included anthropologists who would be influential
in the field. Polanyi’s chapter ‘The Economy as
Instituted Process’ characterized the substantivist
approach. He defined three forms of distribution
found in societies with different structured
relationships: reciprocity in egalitarian rela-
tionships; redistribution in hierarchical
relationships; and market exchange in the anony-
mous relationships of the market. Because eco-
nomic relations were so deeply embedded in social
structure, variation in social organization was
thought to explain the differences in the economies.
Substantivists recognized that markets were found
widely in traditional economies, but argued that
those markets were peripheral to most economic
activities, which were deeply embedded in social
relationships (Bohannan and Dalton ). A
compendium collected by Dalton ( ) represented the inherent tension within anthropol-
provided empirical cases that illustrate the embed-
ded nature of traditional economies.
In his critique of those using economic theory in
non-Western contexts, Polanyi labelled them as
‘formalist’, meaning that they focused on ‘formal’
(mathematical) maximizing models to predict how
individuals choose among alternative possibilities to
allocate limited time, money and other resources.
The substantivists, in contrast, focused on how
economies were embedded within cultural
institutions to meet the material desires that par-
ticular culture might have. The debate raged
between the two factions through the 1960s and
1970s. Much of the argument became focused on
how extensive markets were in traditional societies.
In a classic cross-cultural study, Pryor ( ) institutions, Marxism, ecology, and economic
Economic Anthropology
showed that markets were very broadly distributed,
sometimes moving primitive valuable, tools and
food. They certainly did not originate with modem
capitalism. In his famously acerbic article, Cook (
) criticized substantivists for
being romantic and naive; after all, even if they had
useful points to make, the penetration of market
economies, he argued, was so pervasive that
formalist theories were now effectively universal.
Articles representing the two sides were col-
lected in a reader by LeClair and Schneider ( )
that has been used to teach the debate ever since.
Articulating the substantivist position, Sahlins (
) then argued that many concepts of Western
economic theory were inapplicable to traditional
economy. He discussed the affluence of hunter-
gatherers, underproduction in household economies,
and the social determinants of reciprocal exchanges.
Schneider ( ) countered with the
fully articulated formalist position, summarizing
how Western economics can be applied cautiously
to a wide range of non-Western transactions and
decisions, including marriage payments, primitive
money, the prestige economy and household pro-
duction. The debate came to focus on definitions of
rationality, scarcity and institutional constraints, but
those reading the papers increasingly saw that the
participants were talking past each other.
The formalist and substantivist factions
ogy: on the one hand, to seek cross-cultural regu-
larities that reflect shared social process; on the
other, to recognize the cultural relativity and
uniqueness of each culture. The two sides of the
debate fought to exhaustion, as both presented
compelling approaches that could be seen as more
complementary than alternative. In 1980, Schneider
helped organize the Society for Economic
Anthropology in order to resolve the debate by
bringing the full spectrum of economic anthro-
pologists together. The first meeting, published as
Economic Anthropology. Topics and Theories
(Ortiz ) gathered an eclectic group of scholars to
bridge the theoretical divides within the sub-
discipline, with broad interests in marketing,
Economic Anthropology
development. An edited text. Economic Anthro-
pology! (Plattner ), provided a new generation of
students with the breadth of economies and
economic conditions that anthropologists were
trying to make sense of.
Important to the new harmony has been respect
for the different objectives of economic anthro-
pologists, including ethnographic work on tradi-
tional economies, applied work on developing
economies, and archaeological and historical studies
of economies. The field has recognized diversity in
both the theoretical and historical nature of human
economies. To maintain a proper balance between
substantivists and formalists (relativists and
universalists) in economic anthropology, the role of
archaeological and historical studies has been
especially important. As ethnographers increasingly
study variants of a single modem system, historical
and archaeological studies continue to study the tme
variation in how human economies are organized
and operate. Earle ( ), for example, looks at
the alternative
means by which political economies have emerged
to finance the evolution of chiefdoms and states,
showing that the development of market systems is
quite rare and specific in that process. Although no
careful comparative study exists, the extent of
exchange in prehistory appears to have been highly
variable.
During the 1980s and 1990s, as economic
anthropology matured as a sub-discipline, it became
marginalized within anthropology. As in many of
the social sciences and humanities, postmodernism
became popular, and its antimaterialist, anti-
scientific critiques were antithetical to much of what
the sub-discipline advocated. As the excesses of
postmodernism have receded, however, economic
anthropology has regained some of its former
popularity, and its potential significance for
anthropology and economics seems promising.
Perhaps the greatest challenge now is that
economics and economic anthropology have
remained far apart because of the strongly formal
(theoretical) basis of the former and empirical basis
of the later. The two approaches would, however,
seem complementary.
3245
Economic Anthropology and Its Perspective
on World Economies
Economists should consider the empirical value of
economic anthropology, and a good place to begin
is the compendium Theory in Economic
Anthropology (Ensminger ). Economic
anthropologists are committed to models of reality.
The empirical observations and theoretical
inferences of anthropology should help recognize
the specific frames of applicability for grand the-
ories. In essence, anthropology makes clear that all
things are never equal. In this section, I summarize
a few conclusions derived from economic
anthropology that make a difference to studies of
economies. These involve human rationality,
consumer behaviour, commodity chains, and the
multi-sectored quality of human economies. This
list is not meant to be exhaustive, but only to
illustrate the importance of cross-cultural
evaluations for the models that economists develop.
As economics begins to look at such concepts as
behavioural economics and personalized networks,
the relevance of anthropology’s research on these
topics becomes particularly significant.
Human decision-making is to a degree rational,
and empirical anthropological work significantly
improves an understanding of decision-making
processes from a cross-cultural and evolutionary
perspective. Although rationality underpins much
economic theorizing, human cognitive abilities and
goals have been under theorized. Recent trends to
rectify this within behavioural economics
emphasize that individuals do not always act
rationally with primary economic objectives and it
would appear that economic anthropology could
provide valuable cross-cultural validation of these
new ideas. Humans prove to be fairly poor decision
makers; they appear rather to use simplified
proximate measures to estimate such considerations
as value and cost (Henrich ). Anthropologists have
experimented with various economic games given
under controlled conditions in non-Western
societies, and their results are often counter-
intuitive (Ensminger ). In
3246
a sample of societies representing different levels of
economic development, for example, as market
integration increases cooperation can be shown in
such game-playing experiments to become more
highly prized.
To understand the evolutionary roots of human
rationality, anthropological research has looked at
decision-making in small-scale hunting and gath-
ering societies (see for example, Cashdan ). As seen
by the rapid expansion in brain size deep in history,
humans must have been under strong selective
pressure for expanded cognitive abilities, and this
selective pressure took place when humans were
low-density hunter-gatherers. Such hunter-gatherers
make daily a wide range of decisions about what
foods to eat, where to camp, what groups to join,
and the like, and the relative scarcity and abundance
of food and their different nutritional qualities
appear to be considered. Human cognitive skill
determines the ability of hunter-gatherers to adjust
rapidly to changing conditions of food availability,
to occupy diverse habitats from the Arctic to the
tropical forests, and to intensify food procurement
as required by population growth. In short,
cognitive abilities in the food quest, in movement
through the landscape, and in deciding which
groups to join must have provided a strong selective
advantage that resulted in the moulding of human
rationality.
As illustrated by economic anthropology, human
decisions often have little direct relationship to
economic factors of cost and financial gain.
Although of more interest recently to economists,
with the notable exception of Thorstein Veblen,
economic theory has not attempted systematically
to explain how potential consumer outcomes are
ranked. Rather, within the West, consumer
behaviour has been studied with a rather eclectic
and under-theorized set of assumptions.
Anthropologists, however, have tried to understand
consumption cross-culturally as a social process
involving issues of identity and association (Rutz
and Qrlove ). From the anthropological literature,
we know how valued objects signify social
relationships. The giving and receiving of gifts
impart form and meaning to social relationships,
and materialize the social distance between actors
(Sahlins ).
Economic Anthropology
Economic anthropologists frequently study the
movements of objects around the globe. These
commodify chains describe how goods are pro-
duced, distributed and transformed as they move
through a sequence of markets (Hansen ; Obukhova
and Guyer ). Commodity chains illustrate how
goods, like used clothing, are transformed in value,
form and meaning as they pass through a sequence
of social worlds and economic sectors. Social
considerations of prestige and personal worth are
always of great concern in this highly creative
process of economic decisionmaking.
Economic anthropologists have emphasized that
economies are multilayered and that the specific
character of an economy has historical routes.
Although economists often refer to ‘dual economy’,
implying a vestigial survival of traditional practices,
they have been reluctant to accept that economies
are always multilayered mosaics with spheres of
exchange that only partially articulate the different
sectors. Economic theory thus radically simplifies
reality by focusing on decision-making and
outcomes under market conditions, and this
simplification makes very different economies
appear superficial similar. In the emergence and
development of capitalism, since wealth was made
in the markets, the primary concern of economists
became directed there. As anthropologists seek to
understand the different motives and dynamics of
economies as articulated in specific social contexts,
they have, however, realized that human economies
are highly variable, combining subsistence, social,
political, and market sectors, each with distinct
logics and historical traditions.
The subsistence sector is family-based and
involves the daily struggles to meet basic needs. It
is universal and represents the economic world of
survival in which humans evolved as a species. The
primary motivation of humans has probably always
been the satisfaction of a family’s basic needs. The
construction of a general theory of human
economies should thus start with how households
and communities make a living. Until recently,
household requirements were handled largely by
family production. Although markets have a long
history in human societies, they
Economic Anthropology
were typically quite marginal to subsistence needs.
Theorized as the domestic mode of production
(DMP; Sahlins ), households were oriented to
meeting their subsistence needs, and distribution
involved sharing between family members with
different tasks appropriate to an elementary division
of labour by age and gender. In the model, the
household is economically self- sufficient, and the
economy is not inherently growth-oriented. The
amazing conclusion of considerable anthropological
research is that the DMP is often at least the model
of what the economy should be, and the amounts of
goods consumed by households that are produced
outside the family have often been but a fraction of
the households’ overall consumption budget. Prior
to the development of full-scale markets,
households probably produced 75 per cent or more
of everything that they consumed.
The social sector is community-based and
involves the lifetime strategies of individuals to
define identity and relationships within a broader
social group. The social sector is probably univer-
sal, finding its roots among early hunter-gatherers
and their need to form networks of support, coop-
eration and exclusion. In cross-cultural perspective,
much of the social sector involves reciprocal
exchanges within highly social worlds that can be
manipulated to emphasize personal prestige. In
traditional societies, such competitive exchanges
commonly produce social ranking in what has been
called a ‘prestige economy’. The social sector was
elaborated following the Neolithic revolution, as the
creation of local corporate groups must have placed
a premium on group identity and status. With deep
and enduring roots in human history, the social
sector would seem to provide a cross-cultural
understanding of consumer behaviour as part of
processes much broader than capitalism.
Economics now questions assumptions about
anonymous markets organized independently of
other social institutions. Goods and services are
seen as flowing through personalized networks that
create the institutions for expanding economic
transactions. Greif ( ), for example, argues that
the social networks of medieval Europe provided
the frameworks for an emergent modem economy.
3247
Almost self-evident to anthropologists, such con-
clusions suggest how economic theory can gain
from insights from comparative empirical studies of
non-industrial political forms.
Political sectors mobilize and allocate goods to
finance regional and interregional institutions of
domination and stratification (Earle ). Importantly,
political economies are not universal. From the
fourth millennium be, the political sector of the
economy developed along with chiefdoms and then
states. Goods became mobilized as a tax or tribute
and then ‘redistributed’ by dominant political
organizations as means to finance their activities.
Recent archaeology has studied how political
sectors were developed and functioned. An inherent
contrast is between staple finance and wealth
finance. In staple finance, food goods are mobilized
and stored centrally as a means to support
craftsmen, warriors and labourers working for the
state. Many of these systems, especially in
chiefdoms but also some states, functioned with few
or no markets. Subsistence and social sectors
continued largely unchanged, but new patterns of
land ownership and domination required the pro-
duction of a surplus for ruling institutions. Wealth
finance worked similarly, but the local surplus was
used to support the production of wealth for tribute
payments.
And what about the market sector, so funda-
mental to most economic theorizing? Archaeolog-
ical evidence documents that exchange and markets
were not universal. From case to case, the amount
and types of goods exchanged varied greatly
according to specific conditions of availability and
production costs and to specific objects of value.
Based on ethnographic analogies, until quite
recently most of the goods traded were probably
handled by down-the-line exchanges between social
partners. Goods moving any distance were primarily
primitive valuables, items of display and tribute.
The extent of exchange in Neolithic and later
Bronze Age communities, for example, has been
discussed for Europe, where the comparative
advantage of one region over another would have
been based on the availability of special materials
(Sherratt ). Subsistence and technological items
were rarely exchanged over long distances until the
end of the medieval
3248
age. Earlier, some market exchange certainly
existed, but their extent and elaboration were
apparently quite small.
This empirical record from economic anthro-
pology contests economic theories based on
asserted long-term trends in the emergence of
marketing. A common assumption among econo-
mists from Adam Smith onwards has been that the
creation of wealth is an outcome of the development
of efficiencies associated with specialization and
trade. For example, in his analysis of institutional
economics, North ( ) argues that states
developed to lower transaction costs between
locally specialized but political independent
regions. To simplify the logic, technological
development and specialization should have created
increasing productive economies that, with the
emergence of integrating political systems to
guarantee the peace of the market, would generate
the surplus used to support the growth of
civilizations.
The development of markets, however, was
quite late and episodic. Following North, economics
might suggest that such failure of markets to
develop was an outcome of high transaction costs
that made exchange unprofitable. Empirically such
a conclusion, however, can be shown to be wrong.
As political superstructures were developed and
imposed broad regional peace that would have
radically lowered transactions costs, markets
surprisingly did not emerge. The reason appears to
be linked to the nature of finance. When finance
was based on staples, markets were only
rudimentary and peripheral. The complex Hawaiian
chiefdoms, for example, conquered and integrated
several islands with local specialties in food, stone
and other materials, but trade remained very small-
scale and local despite the regional peace.
Archaeology has documented only minor trade in
basaltic adzes and obsidian in Hawaiian prehistory,
and these exchanges did not increase with the
formation of the large-scale chiefdoms. As a
dramatic example, the Inka empire conquered a
massive territory that extended 3000 km up the
spine of the Andes, imposed an effective regional
peace across that territory, and constructed nearly
30,000 km of roads to integrate it. Although these
actions
Economic Anthropology
would certainly have lowered transaction costs, the
regional and distant movements of goods, like
metal, ceramics, and foods, remained very limited
and completely unchanged from the pre-imperial
period (Earle ).
Both markets and currencies seem to have
expanded in other circumstances where they were
linked with wealth finance of states. In the Aztec
empire, tribute to the state was in wealth objects
like textiles that could be easily transported long
distances, centrally stored, and then used as pay-
ment to those working for the state. But the use of
wealth objects in payment required that the objects
be convertible into the staple goods and other con-
sumables desired by state personnel. The Aztec
market system provided the mechanism for con-
version and was apparently developed by the state
(Brumfiel ). Afterwards, markets appear to have
escaped from state sponsorship and control to take
on many of the characterizations commonly
associated with market systems.
What are the possibilities for a grand theory of
economies? The relatively low status of historical
and comparative studies within economics is not
promising, but economics would do well to test
theories claimed for generally applicability by
looking closely at the anthropological literature. To
the degree that economic models are used to design
economic development in non-Western societies,
the general relevance of the economic models must
demonstrated. Using a uniform method of analysis,
the economist Pryor ( )
has compared industrial economies and traditional
(hunter-gatherer and agricultural) economies. His
primary conclusions are startling, suggesting the
advantages of such comparative analyses. All
economies appear to consist of a small number of
component parts, probably reflecting the processes
and constraints involved in the production and
movement of material goods. Economies are thus
comparable. Furthermore, the factors that affect
such variables as gross productivity or volume of
exchange appear not to be determined by social
structure but by the particular internal char-
acteristics of the economy. Thus, Polanyi would
appear to be wrong; economies are rather inde-
pendent engines of essential processes. As recent
work in economics has relaxed simplifying
Economic Calculation in Socialist Countries 3249

assumptions about information, frictionless trade, LeClair, E.E., and H.K. Schneider, eds. 1968. Economic
and anonymity of markets, the potential links anthropology,’. New York: Holt, Rinehart.
Malinowski, B. 1922. Argonauts of the Western Pacific.
between economics and economic anthropology London: Routledge.
take on reciprocal value. Mauss, M. 1925. The gift: Forms and functions of exchange
in archaic societies, 1969. London: Cohen and West.
North, D. 1990. Institutions, institutional change, and eco-
nomic performance. Cambridge: Cambridge University
See Also Press.
Obukhova, E., and Guyer, J.I. 2002. Transcending the
► Behavioural Economics and Game Theory formal/informal distinction: Commercial relations in
Africa and Russia in the post-1989 world. In Ensminger
► H ng and Gathering Economies (2002a).
► ‘Political Economy’ and ‘Economics’ Ortiz, S., ed. 1983. Economic anthropology: Topics and
► Topcity R theories. Lanham: University Press of America.
Plattner, S., ed. 1989. Economic anthropology. Stanford:
► :J-; iti .
Stanford University Press.
Polanyi, K. 1944. The great transformation. New York:
Bibliography Rinehart.
Polanyi, K., C. Arensberg, and H. Pearson, eds. 1957. Trade
Bohannan, P., and G. Dalton, eds. 1962. Markets in Africa. and market in the early empires. New York: Free Press.
Evanston: Northwestern University Press. Pryor, F.L. 1977. The origins of the economy. New York:
Brumfiel, E. 1980. Specialization, market exchange, and the Academic.
Aztec state: A view from Huexotla. Current Anthro- Pryor, F.L. 2005. Economic systems of foraging, agricultural,
pology 21: 459-478. and industrial societies. Cambridge: Cambridge
Cashdan, E. 1989. Hunters and gatherers: Economic behavior University Press.
in bands. In Economic Anthropology, ed. S. Plattner.Rutz, H., and B. Orlove, eds. 1989. The social economy of
Stanford: Stanford University Press. consumption. Lanham: University Press of America.
Sahlins, M. 1972. Stone age economics. Chicago: Aldine.
Cook, S. 1969. The ‘anti-market’ mentality: A critique of the
substantive approach to economic anthropology. Schneider, H.K. 1974. Economic man: The anthropology of
Southwestern Journal of Anthropology 25: 378—406. economics. New York: Free Press.
Sherratt, A. 1997. Economy and society in prehistoric
Dalton, G., ed. 1967. Tribal and peasant economies. Garden
City: Natural History Press. Europe. Edinburgh: Edinburgh University Press.
Earle, T. 2002. Bronze age economics. Boulder: Westview.
Ensminger, J., ed. 2002a. Theory in economic anthropology.
Walnut Creek: AltaMira.
Ensminger, J. 2002b. Experimental economics: A powerful
new method for theory testing in anthropology. In
Ensminger (2002a).
Firth, R. 1939. Primitive Polynesian economy, 1965. London: Economic Calculation in Socialist Countries
Routledge & Regan Paul.
Greif, A. 2006. Institutions and the path to the modern Michael Ellman
economy: Lessons from medieval trade. Cambridge:
Cambridge University Press.
Gudeman, S. 1998. Economic anthropology. Cheltenham:
Edward Elgar.
Hansen, K.T. 2002. Commodity chains and the international
secondhand clothing trade: Salaula and the work of
consumption in Zambia. In Ensminger (2002a). Abstract
Henrich, J. 2002. Decision-making, cultural transmission and In the 1930s, when the classical socialist system
adaptation in economic anthropology. In Ensminger emerged, economic decisions were based not on
(2002a).
Herskovits, M. 1940. Economic anthropology, 1952. New detailed and precise economic methods of
York: Knopf. calculation but on rough and ready political
Hunt, R. 1997. Economic anthropology. In The dictionary of methods. An important method of economic
anthropology, ed. T. Barfield. Oxford: Blackwell. calculation - particularly in the post-Stalin
Johnson, A., and T. Earle. 2000. The evolution of human
societies. 2nd ed. Stanford: Stanford University Press.
period - was that of incrementalism. Input norms
were a very important method of both inter-
industry and consumption planning.
3250
Material balances, and later input-output,
were also widely used. Project evaluation,
linear programming, comparisons with the
West, and economic intuition were other
methods used. The influence of methods of
economic calculation on economic outcomes
should not be exaggerated.
Keywords
Coefficient of absolute economic effective-
ness; Coefficient of relative economic effec-
tiveness; Consumption norms method;
Council for mutual economic assistance
(Comecon); Economic calculation in social
economies; Five year plans (USSR); Gosplan;
Gossnab; Incrementalist method of economic
calculation; Input norms method of economic
calculation; Input-Output method of
economic calculation; Kantorovich, L.;
Lemeshev, M.; Linear programming;
Material balances method; Planned economy;
Planning and counter-planning; Stalinism,
political economy of; Standard Methodology
of economic calculation
JEL Classifications
P21
Economic Calculation and Political
Decisions
An important result of the archival revolution of
the 1990s (that is, the access to former Soviet
archives made possible by the collapse of the
USSR) was the additional knowledge it provided
about economic decision-making in the USSR in
the Stalin period. This made it clear that in the
1930s, when the socialist economic system
emerged, economic decisions were based not on
detailed and precise economic methods of calcu-
lation but on rough and ready political methods.
Interesting light has been thrown on the signifi-
cance of this for macroeconomic, mesoeconomic
and microeconomic decision making.
Economic Calculation in Socialist Countries
Economic Calculation in Socialist Countries, Fig. 1
Maximizing the investible surplus (Source: Adapted from
Gregory and Harrison , p. 732)
Macroeconomic policy in the Stalin era
aimed to maximize investment subject to the
need to provide sufficient consumer goods
(mainly food) to maintain labour productivity.
The consumer goods were obtained from
agriculture by force and allocated by the state in
a way which it was hoped would enable
investment to be maximized. A schematic
representation of short-term macro- economic
calculation under these circumstances is set out
in Fig. .
Figure shows an output curve OQ which
depends on the effort the workers provide, and
an effort curve E E which depends on the real
(J meLX
wage and the level of coercion. If the state
chooses too low a level of wages, output will
decline and the intended investment level will be
impossible to meet. If wages are set at the fair
wage level, output will be maximized but
investment less than desired. At the wage level W*,
investment will be maximized. Hence,
macroeconomic calculation involved gathering
information about worker attitudes (via the state
security organizations), allocating the available
food to crucial groups of workers, and using
coercive or ideological methods to reduce the
food-output ratio.
Mesopolicy aimed at developing heavy indus-
try and the defence sector. An important result
was what has been termed the ‘structural milita-
rization’ of the Soviet economy. This resulted
from the Soviet view of international relations,
Economic Calculation in Socialist Countries 3251

of 1941, and the use by the general staff of population derived maximum satisfaction from
absurdly inflated estimates of the mobilization the resources available.
capacity of the USA and other countries. An
example is the USSR’s capacity at the end of the
1980s to produce about four million tons of Planning and Counter-Planning
aluminium annually. This was greatly in excess
A widely used method of economic calculation
of the peacetime economy’s need for aluminium.
was that of planning and counter-planning. If the
However, in the event of mobilization it would
plan were simply handed down to the enterprises
have enabled the country to produce huge num-
from above, in accordance with the planners’
bers of military airplanes. This situation arose as
view of national economic requirements but in
a result of using as a method of economic calcu-
ignorance of the real possibilities of each
lation the attainment of Western levels and of
enterprise, then it would be unfeasible (if it was
these levels in the military sphere being system-
too high) or wasteful (if it was too low) or both at
atically exaggerated.
the same time (that is, unfeasible for some
On the microeconomic level, Lazarev and
products and wasteful for others). Conversely, if
Gregory ( ) have studied the allocation of
plans were simply drawn up by each enterprise,
motor vehicles (cars/autos and lorries/trucks)
they might have failed to use resources in
from the central reserve fund in 1932 and 1933.
accordance with national economic
This showed that an economic planning model
requirements. The process of planning and
was unable to explain their allocation (in the
counter-planning involved a mutual submission
regressions the economic variables were insignif-
and discussion of planning suggestions, designed
icant and frequently had the wrong signs). But a
to lead to the adoption of a plan which was
political model, in which their allocation was
feasible for the enterprise and ensured that the
explained as part of a gift-exchange process,
resources of each enterprise were used in
explained the data quite well.
accordance with national requirements.
Unfortunately, the bureaucratic complexity
Incrementalism of this procedure militated against both
efficiency and consistency.
A basic method of economic calculation used in
the state socialist countries - particularly in the
post-Stalin period - was that of incrementalism, Input Norms
or, as it was known in the USSR, ‘planning from
The main method of economic calculation used
the achieved level’. The starting point of all eco-
to ensure efficiency was that of input norms. An
nomic plans was the actual or expected outcome
input norm is simply a number assumed to
of the previous period. The planners adjusted
describe an efficient process of transformation of
this by reference to anticipated growth rates,
inputs into outputs. For example, suppose that
current economic policy, shortages and technical
the norm for the utilization of coal in the
progress. For nearly all products, the planned
production of one ton of steel is .r tons. Then the
output for next year was the anticipated output
efficient production of z tons of steel is assumed
for this year plus a few per cent added on. The
to require zx tons of coal.
advantages of incrementalism as a method of
The method of norms was widely used in
economic calculation were its simplicity, realism
Soviet planning, and considerable effort was
and compatibility with the functioning of a
devoted to updating them. Very detailed norm
hierarchical bureaucracy. Its disadvantages were
fixing took place for expenditures of fuel and
that it provided no method for making
energy. Much attention was devoted to the devel-
technically efficient or consistent decisions, nor
opment of norms for the expenditure of metal,
did it ensure that the
 Per capita
Norm consumption in
(kgs/head/ 1976 as % of
year) norm
120 128
Bread and bread products
Potatoes 97 123
Vegetables and melons 37 59
Vegetable oil and 7 85
margarine
82 68
Meat and meat products
Fish and fish products 18 101
Milk and milk products 434 78
Eggs 17 72

Sources: Weitzman ( ), Agababyan and Yakovleva

( , P-142)

housing, hospitals, schools and parks. Calcula-

tions of the desirable number of rooms, hospital
beds and school places per person are a familiar
tool of planning in welfare states.
There are two main problems with the norm
method of consumption planning. The first is
that of substitution between products. Although
consumers may well have a medically necessary
need for x grams of protein per day, they can
obtain these proteins from a wide variety of
foods. Second, consumers may choose to spend
their money ‘irrationally’, for example, to buy
spirits instead of children’s shoes.

Material Balances

A material balance is a balance sheet for a

particular commodity showing, on the one hand,
the economy’s resources and potential output,
and, on the other, the economy’s need for a
particular product. Material (and labour)
balances were the main methods used in
calculating production and distribution plans for
goods, supply plans and labour plans. Soviet
planners took great pride in the balance method
and considered it one of the greatest
achievements of planning theory and practice.
Material balances were drawn up for different
periods (for example, for annual or five
 Variants
I
Economic Calculation II III Countries
in Socialist IV V 3253
Production of steel in 109 115 121 128 136
1970 (millions
yearof periods), by different organizations (for and because of the assumption of constant coeffi-
tonnes)
example, Gosplan, Gossnab - the body responsi- cients, an input-output table can be utilized for
Variants
ble for allocating supplies of inputs
V - and the variant calculations.
I n in IV
ministries) and at different levels (for
11.4 example,
Engineering and metal 7.1 8.2
9.3 10.4
working
national and republican). The material balances
1
X=(I-A)~ Y
were
Light industry
also drawn
6.3 6.6
up with
6.8
different
7.0 7.2
degrees of
aggregation.7.1Highly
Food industry 7.3 aggregated
7.4 7.5 balances
7.6 were
On the assumption that A is given, X can be
drawn up for the Five Year Plans, and highly
calculated for varying values of Y. Variant calcu-
disaggregated balances by the chief administra-
lations of the structure of production were not
tions of Gossnab for annual supply planning.
undertaken with material balances because of
The aim of the material balance method was to
their great labour intensity. Variant calculations
ensure the consistency of the plans.
played a useful role in medium-term planning
Normally, at the start of the planning work,
because they enabled the planners to experiment
the anticipated availability of a commodity was
with a wide range of possibilities. The first major
not sufficient to meet anticipated requirements.
use of variant calculations of the structure of
To balance the two, the planners sought
production in Soviet national economic planning
possibilities of economizing on scarce products
was in connection with the 1966-70 Five Year
and substituting for scarce materials; they
Plan. Gosplan’s economic research institute
investigated the possibilities of increasing
analysed the results of various possible shares of
production or importing raw materials or
investment in the national income for 1966-70. It
equipment, or in the last resort they determined
became clear that stepping up the share of
the priority needs to be iulfilled by the scarce
investment in the national income would increase
commodity. Even with great efforts, achieving a
the rate of growth of the national income, but
balance was difficult. The complexity of an
that this would have very little effect on the rate
economy in which a great variety of goods are
of growth of consumption (because almost all of
produced by different processes, all of which are
the increased output would be producer goods).
subject to continuous technological change, was
The results of the calculations are set out in
often too great for anything more than a balance
Tables and .
that balanced only on paper. Hence it was
normal, during the ‘planned’ period, for the plan
to be altered, often repeatedly, as imbalances
came to light. Particularly important problems Economic Calculation in Socialist Countries,
Table 2 Output of steel on various assumptions
with the use of material balances were the highly
aggregated nature of the balances and their
interrelated nature.

Input-Output

A wide variety of input-output tables were regu-

Economic Calculation in Socialist Countries,
larly constructed in socialist countries. Ex post Table 3 Average annual growth rates of selected industries,
national tables in value terms, planning national 1966-1970(%)
tables in value and physical terms, regional
tables, and capital stock matrices were widely
constructed and used. An interesting and impor-
tant use concerned variant calculations of the
structure of production in medium-term
planning.
Because an input-output table can be Source: Ellman ( , p. 71)
represented by a simple mathematical model,
3254
The five variants are for the share of
investment in the national income, I being the
lowest and V the highest. A sharp increase in the
share of investment in the national income in the
Five Year Plan 1966-70 would have led to a
sharp fall in the share of consumption in the
national income, and only a small increase in the
rate of growth of consumption (within a Five
Year Plan period). What is very sensitive to the
share of investment in the national income is the
output of the producer goods industries, as
Tables and show.
These results are along the lines of what one
would expect on the basis of Fel’dman’s model,
but the input-output technique improves on
Fel’dman’s model since it enables the effect of
different strategies to be seen at the industry
level rather than merely in terms of macroeco-
nomic aggregates.
Another example of the use of input-output
for economic calculations concerns the statistical
data about the relations between industries
contained in the national ex post tables in value
terms. In his controversial 1968 book Mezhotraslevye svyazi seVskogo
khozyaistva, M. Lemeshev, then deputy head of the sector
for forecasting the development of agriculture of
the USSR Gosplan’s Economic Research
Institute, used the Soviet input-output table for
1959 as the basis for a powerful plea for more
industrial inputs to be made available to
agriculture.
He began by observing that from the 1959
input-output table it was clear that of the
current material inputs into agriculture in that
year only 23.4 per cent came from industry,
while 54.7 per cent came from agriculture itself
(feed, seed and so on). He argued that this was
most unsatisfactory. In the section on the
relationship between agriculture and
engineering Lemeshev argued that the supply to
agriculture of agricultural machinery was
inadequate, in the section on the relationship
between agriculture and the chemical industry
he argued that the supply of fertilizers was
inadequate, and in the section on agriculture and
electricity he argued that the supply of electricity
to the villages for both productive and
unproductive needs was inadequate. In addition,
in the section on the relationship between
agriculture and the processing industry he
Economic Calculation in Socialist Countries
not helping agriculture as it should do; for exam-
ple, it was sometimes impossible to accept vege-
tables (although the consumption of these in the
towns was below the norms) because of inade-
quate processing and distribution facilities. Fur-
thermore, he argued that the supply of
concentrated feed was inadequate and the pro-
cessing of milk wasteful. In view of the inade-
quate development of the food processing
industry, he argued for the development of pro-
cessing enterprises by the farms themselves.
The chapter on the productive relations
between agriculture and the building industry
was an extensive critique of the practice of pro-
ductive, and of housing and communal, building
in the villages. Lemeshev argued that the state
should take on responsibility for building on the
collective farms. The chapter on the relationship
between agriculture and transport was critical of
the shortage of river freight boats. The chapter
on investment argued that investment in
agriculture was inadequate, and that in the
period 1959-65 there was an unwarranted
increase in the proportion of investment in the
collective farms which they had to finance
themselves. He also argued that a greater
proportion of agricultural investment should be
financed by bank loans, and that as a criterion of
investment efficiency the recoupment period was
satisfactory. The concluding chapter was
concerned with improving the productive
relations between agriculture and the rest of the
economy. The author argued for improving
central planning by the use of input-output, for
replacing procurement plans by free contracts
between farms and the procurement organs (if a
shortage of a particular product threatened then
its price could be raised), and for the elimination
of the supply system (that is, the rationing of
producer goods) which hindered farms from
receiving the goods they wanted and sometimes
supplied them with goods that they did not want.
Lemeshev also argued for higher pay in agricul-
ture and for the reorganization of the labour
process within state and collective farms on the
basis of small groups which were paid by results.
This book was a good example of the use of
input-output to provide statistical data which
could be used, alongside other information, to
Economic Calculation in Socialist Countries
provide a description of important economic and 1981. In a very abbreviated and summary
relations and to support a case for important form, the 1981 version was as follows.
institutional and policy changes.
Project Evaluation
In the USSR of the 1930s, it was officially con-
sidered that there was no problem of project
evaluation to which economists could contribute.
The sectoral allocation of investment was a
matter for the central political leadership to
decide. It was they who decided in which sectors
and at which locations production should be
expanded. These decisions were based on the
experience of the more advanced countries, the
traditions of the Russian state (for example,
stress on railway building) and of the Bolshevik
movement (for example, stress on electrification
and on the metal-using industries) and on the
needs of defence. As far as decisions within
sectors were concerned, here the main idea was
to fulfil the plan by using the world’s most
advanced technology.
The practical study of methods for choosing
between variants within sectors was begun by engineers
in the electricity and railway industries. The problem
analysed was that of comparing the cost of alternative
ways of meeting particular plan targets. A classic example
of the type of problem considered was the choice
between producing electricity by a hydro station and by
a thermal station.
During Stalin’s lifetime, the elaboration by orthodox
economists and the adoption by the planners of
economic criteria for project evaluation were impossible
because they were outside Stalin’s conception of the
proper role of economists (apologetics). When
economists did make a contribution in this area, as was
done by Novozhilov, it was ignored. After Stalin’s death,
however, it became possible for Soviet economists to
contribute to the elaboration of methods of economic
calculation for use in the decisionmaking process. An
early and important example was in the field of project
evaluation. An official method for project evaluation was
adopted in 1960, and revised versions in 1964, 1966,
1969
3255
In evaluating investment projects, a wide
variety of factors have to be taken into account,
for example, the effect of the investment on
labour productivity, capital productivity,
consumption of current material inputs (such as
metals and fuel), costs of production,
environmental effects, technical progress, the
location of economic activity and so on. Two
indices which give useful synthetic information
about economic efficiency (but are not
necessarily decisive in choosing between
investment projects) are the coefficient of
absolute economic effectiveness and the
coefficient of relative economic effectiveness.
At the national level, the coefficient of
absolute effectiveness is defined as the
incremental output-capital ratio.
where E is the coefficient of absolute effectiveness
p
for a particular project, AFis the increase in
national income generated by the project, and I is
the investment cost. The value of E calculated in
p
this way for a particular investment has to be
compared with E , the normative coefficient of
a
absolute effectiveness, which is fixed for each
Five Year Plan and varies between sectors. In
the 11th Five Year Plan (1981-85) it was 0.16 in
industry, 0.07 in agriculture, 0.05 in transport
and communications, 0.22 in construction and
0.25 in trade.
If E > E
p a
then the project is considered efficient.
For calculating the criterion of absolute effec-
tiveness at the level of individual industries, net
output is used in the numerator instead of
national income. At the level of individual
enterprises and associations, in particular when
a firm’s own money or bank loans are the source
of finance, profit is used instead of national
income.
The coefficient of relative effectiveness is
used in the comparison of alternative ways of
producing particular products. In the two
products case
3256
E _ C 1 c,
Ki K I
where E is the coefficient of relative effectiveness,
C i is the current cost of the z'th variant, and /\, is
the capital cost of the ith variant.
IfE > En, where En is the officially
established normative coefficient of relative
economic efficiency, then the more capital
intensive variant is economically justified. In the
11th Five Year Plan, En was in general 0.12, but
exceptions were officially permitted in the range
0.08/0.10-0.20/0.25.
In the more than two variants case, they
should be compared according to the formula
Ci +E„Ki — > minimum
that is, choose that variant which minimizes the
sum of current and capital costs.
At one time a rationalist misinterpretation of
socialist planning was widespread. According to
this view, a planned economy was one in which
rational decisions were made after a
dispassionate analysis by omniscient and all-
powerful planners of all the alternative
possibilities. In such a system, the adoption of
rational criteria for project evaluation would
have been of enormous importance. Socialist
planning, however, was just one part of the
social relations between individuals and groups
in the course of which decisions were taken, all
of which were imperfect and many of which
produced results quite at variance with the
intentions of the top economic and political
leadership.
A good example of the factors actually
influencing investment decisions under state
socialism was the commencement ofthe construc-
tion of the Baoshan steel plant near Shanghai.
The site was apparently chosen because ofthe
political influence of a high-ranking Shanghai
party official. The location decision ignored the
fact that, because of the swampy nature of the
site, necessitating large expenditures on the
foundations, this was in fact the most expensive
of the sites considered. Veiy expensive, dogged
with cost overruns, involving major pollution
problems, the whole project was kept alive for
some time by a powerful
Economic Calculation in Socialist Countries
steel lobby. In due course, as a result of a
national policy reversal in Beijing, the second
phase was deferred and those involved publicly
criticized. To judge from its initial costs of
production, it produced gold rather than steel.
In general, the choice of projects owed more
to inter-organization bargaining in an
environment characterized by investment hunger
than it did to the detached choice of a cost-
minimizing variant. The development of new and
better criteria for project evaluation turned out
to be no guarantee that project evaluation would
improve since the criteria were often not in fact
used to evaluate projects. Their main function
was to provide an acceptable common language
in which various bureaucratic agencies
conducted their struggles. Agencies adopted
projects on normal bureaucratic grounds and
then tried to get them adopted by higher
agencies, or defended them against attack, by
presenting efficiency calculations using the
official methodology but relying on carefully
selected data.
Linear Programming and Extensions
Linear programming was discovered by the
Soviet mathematician Kantorovich in the late
1930s. Its relevance for Soviet planning was
widely discussed in the USSR in the 1960s and
extensive efforts were made actually to use it in
Soviet planning in the 1970s. Three examples of
its use follow.
Production Scheduling in the Steel Industry
Linear programming was discovered by
Kantorovich in the course of solving the
problem, presented to him by the Laboratory of
the all- Union Plywood Trust, of allocating
productive tasks between machines in such a way
as to maximize output given the assortment plan.
From a mathematical point of view, the problem
of optimal production scheduling for mbe mills
and rolling mills in the steel industry, which was
tackled by Kantorovich in the 1960s, is very
similar to the Plywood Trust problem, the
difference being its huge dimensions.
Economic Calculation in Socialist Countries
The problem arose in the following way. As
part of the planning of supply, Soyuzglavmetal
(the department of Gossnab concerned with the
metal industries), after the quotas had been spec-
ified, had to work out production schedules and
attachment plans in such a way that all the
orders were satisfied and none of the producers
received an impossible plan. In the 1960s an
extensive research programme was initiated by
the department of mathematical economics
(which was headed by Academician
Kantorovich) of the Institute of Mathematics of
the Siberian branch of the Academy of Sciences,
to apply optimizing methods to this problem.
The chief difficulties were the huge dimensions
of the problem and the lack of the necessary
data. About 1,000,000 orders, involving 60,000
users, more than 500 producers and tens of
thousands of products, were issued each year for
rolled metal. Formulated as a linear
programming problem it had more than a
million unknowns and 30,000 constraints.
Collecting the necessary data took about six
years. Optimal production scheduling was first
applied to the tube mills producing tubes for gas
pipelines (these were a scarce commodity in the
USSR). In 1970 this made possible an output of
tubes 108,000 tons greater than it would
otherwise have been, and a substantial reduction
in transport costs was also achieved.
The introduction of optimal production
scheduling into the work of Soyuzglavmetal was
only part of the work initiated in the late 1960s
on creating a management information and
control system in the steel industry. This was
intended to be an integrated computer system
which would embrace the determination of
requirements, production scheduling, stock
control, the distribution of output and
accounting. Such systems were widely
introduced in Western steel firms in the late
1960s. Work on the introduction of management
information and control systems in the Soviet
economy was widespread in the 1970s, but by the
1980s there was widespread scepticism in the
USSR about their usefulness. This largely
resulted from the failure to fulfil the earlier
exaggerated hopes about the returns to be
3257
Industry Investment Plans
In the state socialist countries investment plans
were worked out for the country as a whole, and
also for industries, ministries, departments, asso-
ciations, enterprises, republics, economic regions
and cities. An important level of investment
planning was the industry. Industry investment
planning is concerned with such problems as the
choice of products, of plants to be expanded,
location of new plants, technology to be used,
and sources of raw materials.
The main method used in the 1970s and 1980s
in the Council for Mutual Economic Assistance
(CMEA, known in the West as Comecon) coun-
tries for processing the data relating to possible
investment plans into actual investment plans
was mathematical programming. After extensive
experience in this field, in 1977 a Standard Meth-
odology for doing such calculations was adopted
by the Presidium of the USSR Academy of
Sciences.
The Soviet Standard Methodology presented
models for three standard problems. They were:
a static multi-product production problem with
discrete variables, a multi-product dynamic pro-
duction problem with discrete variables, and a
multi-product static problem of the production-
transport type with discrete variables.
The former can be set out as follows:
Let i = 1, ... , n be the finished goods or
resources, j = 1,... , m be the production units, r =
1, ... , Rj be the production technique in a unit, a j
r
t
be the output of good 1 = 1 , . . . , n or input of
resource i = n' + 1 , . . . , n, using technique r of
production in unit /; Cj are the costs of production
using technique r in unit /; D, is the given level of
output of good i,1=1,■■■,n;Pi is the total use of resource i,i=
n + 1 , . . . , n allocated to the industry; Z - is the
r
unknown intensity of use of technique r at unit j.
The problem is to find values of the variables
'/!■ that minimize the objective function
m Rj
E E<U d)
i=i i=i
3258
that is, minimize costs of production subject to
HI Rj
i=
j1r1
that is, each output must be produced in at least the
required quantities
HI Rj
YY a'uZj-Pi’ ' = «' + 1................................. n (3)
j=i r=l
that is, the total use of resources cannot exceed the
level allocated to the branch
RJ
YZi - 1’J = 1 ’ - ’ m
r1
Zj = 0 or 1, j = 1
that is, either a single technique of production
for unit j is included in the plan or unit j is not
included in the plan.
In order to illustrate the method, an
example will be given which is taken from the
Hungarian experience of the 1950s in working
out an investment plan for the cotton weaving
industry for the 1961-65 Five Year Plan. The
method of working out the plan can be presented
schematically by looking at the decision
problems, the constraints, the objective function
and the results.
The decision problems to be resolved were:
(a) How should the output of fabrics be
increased, by modernizing the existing
weaving mills or by building new ones?
(b) For part of the existing machinery, there
were three possibilities. It could be operated
in its existing form, modernized by way of
alterations or supplementary investments,
or else scrapped. Which should be chosen?
(c) For the other part of the existing machinery,
it could be either retained or scrapped.
What should be done?
(d) If new machines are purchased, a choice has
to be made between many types. Which
types
Economic Calculation in Socialist Countries
should be chosen, and how many of a particular
type should be purchased?
1...........n' (2) The constraints consisted of the output plan for
cloth, the investment fund, the hard currency quota,
the building quota and the material balances for
various kinds of yam. The objective function was to
meet the given plan at minimum cost.
The results provided answers to all the decision
problems. An important feature of the results was
the conclusion that it was cheaper to increase
production by modernizing and expanding existing
mills than by building new ones.
It would clearly be unsatisfactory to optimize
the investment plan of each industry taken in
(4)
isolation. If the calculations show that it is possible
to reduce the inputs into a particular industry below
1 (5) those originally envisaged, then it is desirable to
reduce planned outputs in other industries, or
increase the planned output of the industry in
question, or adopt some combination of these
strategies. Accordingly, the experiments in working
out optimal industry investment plans, begun in
Hungary in the 1950s, led to the construction of
multi-level plans linking the optimal plans of the
separate industries to each other and to the mac-
roeconomic plan variables. Multilevel planning of
this type was first developed in Hungary, but
subsequently spread to the other CM FA coun-
tries. Extensive work on the multi-level optimiza-
tion of investment planning was undertaken in
the USSR in connection with the 1976-90 long-
term plan. (The 1976-90 plan, like all previous
Soviet attempts to compile a long-term plan, was
soon overtaken by events. The plan itself seems
never to have been finished and was replaced by
ten- year guidelines for 1981-90.)
The Determination of Costs in the Resource
Sector
In view of the wide dispersion of production
costs in the resource sector, the use of average
costs (and of prices based on average costs) in
allocation decisions is likely to lead to serious
waste. An important outcome of the work of
Kantorovich and his school for practical policy
was (after a
Economic Calculation in Socialist Countries
long lag) official acceptance of this proposition
and of linear programming as a way of
calculating the relevant marginal costs. For
example, in 1979 in the USSR the State
Committee for Science and Technology and the
State Committee for Prices jointly approved an
official method for the economic evaluation of
raw material deposits. This was a prescribed
method for the economic evaluation of
exploration and development of raw material
deposits. What was new in principle about this
document was that it permitted the output
derived from the deposits to be evaluated either
in actual (or forecast) wholesale prices or in
marginal costs. For the fuel-energy sector, a lot
of work was done to calculate actual (and
forecast) marginal costs for each fuel at different
locations throughout the country and for
different periods. These figures were regularly
calculated on optimizing models (they were the
dual variables to the output maximizing primal)
and were widely used in planning practice for
many years.
Comparison with the West
An important method of economic calculation in
socialist countries was comparison with the
West. If a particular product or method of
production had already been introduced (or
phased out) in the West, this was generally
considered a good argument to introduce it (or
phase it out) in the socialist countries, subject to
national priorities and economic feasibility.
Obtaining advanced technology from abroad (by
purchase, Lend-Lease, reparations, espionage,
direct investment) was an integral part of
socialist planning, the importance of the different
elements varying over time. Comparisons with
the West were particularly important in an
economic system which lagged behind the
leading countries, lacked institutions which
automatically introduced innovations into
production (that is, profit-seeking business
firms), and found it difficult (because of the
ignorance of the planners, stable cost-plus prices
and the self- interest of rival bureaucratic
agencies) to notice, appraise realistically when
3259
Economic Calculation and Economic
Results
It is important not to exaggerate the influence of
methods of economic calculation on the perfor-
mance of an economy. The performance of an
economy is largely determined by external
factors (such as the world market), economic
policy (for example, the decision to import
foreign capital or to declare a moratorium),
economic institutions (like collective farms) and
the behaviour of the actors within the system
(for example, underestimation of investment
costs by initiators of investment projects). It is
entirely possible for an improvement in the
methods of economic calculation to coincide with
a worsening of economic performance (as
happened in the USSR in the Brezhnev period).
Realization of these facts led in the 1970s to a
shift from the traditional normative approach
(which concentrates on the methods of economic
calculation and which regards their
improvement as the main key to improved eco-
nomic performance and the main role of the
economist) in the study of planned economies, to
the systems and behavioural approaches.
Economic Calculation and Economic
Intuition
In view of bounded rationality, and the huge vol-
ume, and distorted nature, of the information
available to the central leadership, really existing
decision-making relied heavily on rales of thumb
and the ‘feel’ for reality of the top decision-
makers (sometimes known as ‘planning by feel’).
This could quickly lead to an equilibrium, but an
inefficient one.
See Also
3260 Economic Consequences of Weather, The

Bibliography Shlykov. V. 2004. The economics of defense in Russia and

the legacy of structural militarization. In The Russian
Agababyan. E.. and Ye Yakovleva, eds. 1979. Problemy militant ed. S. Miller and D. Trenin. Cambridge, .VIA:
raspredeleniya i rost narodnogo blagosostoyaniya. MIT Press.
Moscow: Nauka. Stalin. J. 1952. Concerning the errors of comrade L.D.
Birman, I. 1978. From the achieved level. Soviet Studies 30 Yaroshenko. In Economic problems of socialism in the
(2): 153-172. USSR. Moscow: Foreign Languages Publishing I louse.
Birman, I. 1996. Otraslevoe optimal'noe. Ch. 9 of Ya Tretyakova, A., and I. Birman. 1976. Input output analysis in
ekonomist. Novosibirsk: EKOR. the USSR. Soviet Studies 28 (2): 157 186.
Boltho. A. 1971. Foreign trade criteria in socialist econ- Weitzman. P. 1974. Soviet long term consumption planning:
omies. Cambridge: Cambridge University Press. Distribution according to rational need. Soviet Studies 26
Ellman. M. 1973. Planning problems in the USSR: The (3): 305 321.
contribution of mathematical economies to their solutionWorld Bank. 1992. China: Reform and the role of the plan in
1960-1971. Cambridge: Cambridge University Press. the 1990s. Washington, DC: World Bank.
Ellman. M. 1983. Changing views on central planning: 1958-
1983. ACES Bulletin [now Comparative Economic
Studies] 25(1): 11-34.
Gacs. J.. and M. Lacko. 1973. A study of planning behaviour
on the national-economic level. Economies of Planning
13: 91-119.
Giffen. J. 1981. The allocation of investment in the Soviet Economic Consequences of
Union. Soviet Studies 33 (4): 593 609. Weather, The
Granick. D. 1990. Planning as coordination. Ch. 3 of Chinese
state enterprises. Chicago: University of Chicago Press. Jordan Rappaport
Gregory, P. and M. Harrison. 2005. Allocation under dic-
tatorship: Research in Stalin's archives. Journal of
Economic Literature 43:721-761.
Komai, J. 1967. Mathematical planning of structural deci-
sions. Amsterdam: North-1 lolland.
Abst r act
Komai, J. 1980. Economics of shortage, 2 vols. Amsterdam:
North-1 lolland. Households in the United States and a
Komai, J. 1992. Planning and direct bureaucratic control ('h. number of other wealthy nations have been
7 of The socialist system. Oxford: Oxford University migrating to places with nice weather. This
Press. likely reflects an increase in the relative
Kueh, Y. 1985. Economic planning and local mobilization in
post-Mao (China. London: Contemporary China valuation of the weather’s direct contribution
Institute. to household utility. Several different
Kushnirsky, F. 1982. Soviet economic planning. 1965 1980. amenity' explanations are discussed that can
Boulder: Westview, ch. 4. account for the increased valuation and
Lazarev, V., and P. Gregory. 2003. Commissars and cars: A
case study in the political economy of dictatorship. ongoing move.
Journal oft Comparative Economies 31:1 19.
Keywor ds
Lemeshev, M. 1968. Mediotraslevve svvazi sel’skogo
khozyaistva. Moscow: Ekonomika. Compensating differentials; Consumption
Levine, II. 1959. The centralized planning of supply in Soviet amenities; Local growth; Migration;
industry. In Comparisons of the United States and Soviet
Weather
Economies. Washington, DC: Joint Economic
Committee, US Congress.
Malinovskii, B. 1995. Istoriva vychislitel’noi tekhniki v JE L Cl assi f i cat i ons

litsakh. Kyiv: KJT/A.S.K. R 11 ; R 12 ; R13; R23

Matekon. 1978. Standard methodology for calculations to
optimize the development and location of production in Introduction
the long inn. 15(1): 75-96.
Qian, Y., G. Roland, and C. Xu. 2000. Coordinating activities
under alternative organizational forms. In Planning, A cloudy day or a little sunshine har e as great an
shortage, and transformation, ed. E. .Vlaskin and A. influence on many constitutions as the most recent
Simonovits. Cambridge, .VIA: MIT Press. blessings or misfortunes.
Joseph Addison (1672 1719). English essayist, poet
and politician
Economic Consequences of Weather, The 3261

Don’t knock the weather. If it didn’t change once in heterogeneity, it is easy to see that farmland in
a while, nine out of ten people couldn’t start a places with weather most conducive to growing
conversation.
Kin Hubbard (1868-1930) American cartoonist, will be valued especially highly. Farmers,
humorist and journalist assumed to be mobile across locations, will bid
up the price ofproductive farmland until the
It is hard to find a research subject more
weather’s expected contribution to profits
important titan the weather. From ice ages to
becomes fully capitalized into land values. The
epic floods to endless droughts to malarial heat
higher productivity of farms in ideal-weather
to the present warming of the earth, human
locations simultaneously makes it possible to pay
welfare has always depended closely on it. Less
workers there relatively high wages while still
awe-inspiring but also important is that weather
attaining the profits that could be made
is a direct source of significant consumption.
elsewhere. Note that the farm workers in such
Nice weather underpins the enjoyment of most
high productivity locations are not necessarily
outdoor activities from picnics to sports games to
any better off than mobile farm workers
beach days, to an infinite set of other
elsewhere. General equilibrium considerations
possibilities. The discussion that follows will
imply that their higher wages will be offset by
focus primarily on this latter, consumption
higher prices for non-traded goods such as
dimension of weather. While such a focus may
housing.
seem shallow in the face of the significant
As with agriculture, the weather serves as a
challenges weather poses to humanity, those
productive input into numerous industrial pro-
challenges do not negate the fact that normal
cesses. Gunpowder, macaroni, tobacco, gum and
weather variations - the sorts that have been
chocolate are among the many products whose
experienced year after year by current and
production requires constant, low humidity.
recent generations - continue to be a large source
Inside weather conditions are thus an extremely
of consumption benefits.
important productive input. Of course, in pre-
The discussion below will argue that rising
sentday developed countries, inside and outside
incomes in the United States and other developed
weather are typically disconnected. But prior to
nations have increased households’ willingness to
air conditioning and central heating, inside
pay to live in a place with nice weather. As a
weather depended closely on outside weather.
result there has been a shift in population
Thus Oi ( ) argues that the spread of work
towards such places. Before we consider this
place air conditioning underpinned the rise of
consumption dimension of weather, however, a
manufacturing in the south of the United States.
brief discussion of the weather’s day-to-day
Nice weather also turns out to be empirically
contribution to production is warranted.
correlated with very-short-term stock market
returns (Saunders ; Hirshleifer and Shumway ).
Weather as a Production Amenity Specifically, daily measures of sunshine in cities
that host major stock exchanges are positively
Agriculture is the industry that most obviously
correlated with daily returns on those exchanges.
depends on weather as a productive input. This
In this case, weather’s contribution is potentially
dependence is multidimensional in the sense that
productive only for day traders with very low
temperature, humidity, cloud cover and rainfall
transaction costs. The hypothesized mechanism
- each over the entire growing season - all
is that nice weather uplifts traders’ mood and
matter. A large enough deviation by just one of
optimism. Such a mechanism has much more the
these can be sufficient to seriously impair yields.
flavour of a consumption amenity than a
To be sure, advancing agricultural science has
productive one.
allowed crops to thrive in a wider range of
weather conditions. But even loosened, the
constraints imposed by weather remain
Weather as a Consumption Amenity
significant.
Of course, different agricultural goods thrive Just as weather’s contribution to production
3262
housing, so too does its direct contribution to
household utility. But as a consumption amenity,
weather puts downward pressure on wages
rather than upward pressure.
The expected correlations from weather’s
role as a production amenity and as a
consumption amenity derive from the
compensating differential framework (Rosen ;
Roback ). An economy is assumed to be made up
of a number of geographically distinct labour
markets where households live and work and
firms produce. The labour market locations may
differ from one another with respect to
numerous exogenous production and
consumption amenities such as proximity to
navigable water, access to natural resources, low
risk of natural disasters, and - in a
multidimensional sense - the weather. Produc-
tion and consumption amenities may also be
endogenous, for instance if increasing returns to
scale lower input costs or expand the variety of
consumer goods. The assumed high mobility of
firms implies that they must be at least as profit-
able in their present location as they would be
anywhere else. The assumed high mobility of
households implies that they must derive at least
as much utility in their present location as they
would anywhere else.
The key to the compensating differential
framework is that prices - in particular for land,
labour and housing services - adjust to equate
profits and utility across the numerous locations.
In locations with high production amenities,
firms are willing to pay higher prices for inputs,
including for labour. These higher input prices
are required to lower what would otherwise be
higher profits than could be achieved from
locating elsewhere. Similarly, in locations with
high consumption amenities, households are
willing to accept lower wages and pay a higher
price for housing services. Such households thus
trade off lower tangible consumption of market
goods for higher intangible consumption of
amenities.
The empirical implementation of this model
typically focuses exclusively on households and
consumption amenities rather than firms and
production amenities. The reason is the difficulty
of observing the full range of firm input prices.
Notable exceptions include Gabriel and
Economic Consequences of Weather, The
( ) and Chen and Rosenthal ( ), which
treat housing service prices as a proxy for non-
labour input prices.
For households, the most common empirical
methodology is to separately regress micro data
of household income and a proxy for housing
service price on respective vectors of attributes
meant to control for differences in human capital
and differences in the quantity and quality of
housing services. The residuals from these
regressions can then be regressed on location-
specific attributes, including weather. Summing
the extra annual housing service cost implied by
a coefficient on a locational attribute in the
housing regression with the lost income implied
by the coefficient on the same locational
attribute in the income regression gives the
marginal consumption that a household forgoes
to obtain a small increase in that local attribute.
Estimated compensating differentials for
weather attributes from implementing this meth-
odology tend to be extremely large. For example,
the valuation per representative household for
one extra sunny day over the course of a year is
somewhere from US$21 (at 2005 prices) to $36.
The midpoint of this estimated range implies an
aggregate valuation of $57 million per year for a
metropolitan area with a population of 2 million.
Over 30 years using a three per cent discount
rate, the implied net present value is $560
million. Whether households really require such
a huge transfer to accept just a single extra
cloudy day per year seems questionable. Other
estimated weather valuations include one less
rainy day over the course of a year, $36 per
household; one less inch of precipitation, $-63 to
$37 per household; and one inch less snow per
year, $33 per household (Blomquist et al. ;
Gyourko and Tracy ; Stover and Leven ).
Heterogeneity of household preferences sug-
gests that these estimates may understate the
consumption benefits from weather. With
heterogeneity, it is no longer necessary that all
households be indifferent about where to live.
The distribution of wages and house prices
across locations that clears the labour, traded
goods, and housing markets will be driven in
large part by ‘marginal’ households, who tend to
value
Economic Consequences of Weather, The
consumption amenities by less than average.
‘Inframarginal’ households, in contrast, tend to
value at least some consumption amenities
highly. In order to live in a location where such
amenities are abundant, inframarginal
households are willing to accept a lower wage
and pay a higher housing-service price than is
actually required. Hence they enjoy a surplus
that is missed by the compensating valuations
above.
An even bigger empirical challenge to valuing
weather and other consumption amenities is the
difficulty of controlling for individual-specific
and house-specific characteristics. A low wage
may represent compensation for amenities, but it
also may represent low human capital. A high
expenditure on housing may compensate for high
amenities, but it also may reflect a high quality
and quantity of housing services being
purchased. The characteristics typically used as
controls when estimating the wage compensation
include age, experience, education, sex, industry
and occupation. For estimating the house price
compensation, typical controls include rooms,
bedrooms, units in structure, and appliances.
These sets of attributes miss substantial sources
of individual and housing-unit variation.
Probably most important for present purposes is
the difficulty of distinguishing between high
amenities and low human capital. The sorting of
human capital across metro areas suggests that
unobserved human capital characteristics may
be correlated with the weather. The
consequences of not sufficiently controlling for
individual and housing service characteristics
are evident in quality-of-life rankings of metro
areas based on compensating differentials, which
tend to contrast sharply with subjective rankings
(Rappaport )•
A complementary ‘quantity’ approach to the
compensating differential literature’s ‘price’
approach explicitly models population, capital
inputs, land and housing supply. As is intuitive,
high levels of consumption amenities attract
households to a location, resulting in higher pop-
ulation and population density. (Henceforth, I
shall make no distinction between the level of
population and its density.) The higher
population in turn supports the higher housing
3263
lower wages of the compensating equilibrium
(Haurin ; Rappaport ).
The seemingly obvious empirical implication
of the quantity approach is to regress a cross-
section of local population on exogenous local
attributes such as the weather to infer whether
such attributes are an amenity (with respect to
either production or consumption). However, the
extremely high persistence of local population
implies that the correlation of population with an
attribute might reflect an amenity contribution
in the distant past that no longer exists. Instead,
a crosssection of population growth rates can be
regressed on the exogenous attributes. The
resulting coefficients can be interpreted as
reflecting the accumulation of past changes of
the attributes’ amenity contributions (Mueser
and Graves ; Rappaport ). In other
words, a positive partial correlation between
population growth and a particular attribute
suggests that the attribute’s amenity
contribution increased - becoming either more
positive or less negative - in the intermediate
past. The high persistence of population growth
in the United States suggests that the
‘intermediate past’ probably reaches back at
least several decades (Greenwood et al. ;
Rappaport ; Glaeser and Gyourko ).
Empirically implementing the quantity
approach establishes that population growth in
the United States has been highly correlated with
nice weather. Growth has been fastest where
winters and summers are mild and the number
of rainy days is moderate. The quantitatively
strongest relationship, robust to numerous
controls, is a positive quadratic correlation of
growth with winter temperature. For the period
1970 to 2000, increasing January temperature
from one standard deviation below its sample
mean to one standard deviation above its sample
mean (from 29 °F to 54 °F) is associated with
faster growth of 1.3 per cent per year for US
counties (Rappaport ). Miami’s temperature in
January implies expected annual growth that is
3.4 per cent faster than that of US counties with
mean January temperature. For comparison, the
mean population growth rate of counties over
this period was 0.9 per cent per year.
3264
Population growth is negatively correlated
with summer temperature and humidity
(controlling for winter temperature, and robust
to the inclusion of numerous other attributes).
An increase in July heat index from one
standard deviation below its sample mean to one
standard deviation above its sample mean (from
87 °F to 109 °F) is associated with slower growth
of 0.5 per cent per year. An increase in relative
humidity from one standard deviation below its
sample mean to one standard deviation above its
sample mean (from 56 per cent to 75 per cent) is
associated with slower growth of 0.9 per cent per
year. Miami’s temperature and humidity in July
imply expected annual growth that is 0.7 per
cent slower that that of counties with mean heat
and humidity.
Finally, population growth is characterized
by a negative quadratic partial relationship with
the number of rainy days. Increasing the number
of rainy days by one standard deviation (25 days)
above the mean (94 days) leaves expected popu-
lation growth essentially unchanged. But increas-
ing rainy days by a second and then a third
standard deviation slows growth by 0.3 percent-
age points and then an additional 0.6 percentage
points. For Seattle, with an average of 182 rainy
days per year, annual expected population
growth is 1.3 percentage points lower than that
of a location with mean annual precipitation.
The weather accounts for a very large share
of the variation in local population growth rates.
The four weather variables just discussed,
entered linearly and quadratically, along with
annual precipitation entered similarly, can
account for 27 percent of the variation in US
county population growth from 1970 to 2000.
This is only slightly less than is accounted for by
dummies for each US state. For metro areas,
winter weather alone accounts for 44 percent of
the variation in growth from 1950 to 2000.
Results similar to those above hold for a num-
ber of nations, for a number of geograp hies
within them, and for a variety of time periods.
Similar partial correlations of growth with
weather characterize US metro area growth
from 1950 to 1980 (Mueser and Graves ). In
Europe, nice weather has been a major driver of
population flows from 1980 to 2000 within
countries
Economic Consequences of Weather, The
although not across them (Cheshire and Magrini
). And net migration among Japanese pre-
fectures from 1955 to 1990 was negatively corre-
lated with a measure of extreme temperature
(Barro and Sala-i-Martin ).
The partial conelations strongly suggest that
the amenity value of nice weather increased
beginning at some point in the intermediate past,
via either consumption or production. If the for-
mer, such places became inherently more desir-
able as the marginal utility from nice weather
rose relative to the marginal utility of private
consumption. If the latter, nice-weather places
became more desirable because firms there could
pay relatively higher wages.
The quantity framework allows for numerous
explanations, many complementary, of the
empirical migration to nice weather places. The
common element of these explanations is that
they posit a change in the valuation of some
aspect of weather’s amenity contribution, or else
a change in the valuation of an amenity
conelated with weather. One such explanation is
that the approximate sixfold rise in per capita
income over the course of the 20th century
lowered the marginal utility from the
consumption of private goods and services and
so increased the quantity of these that
households were willing to forgo in order to live
in a place with nice weather. Consistent with this
consumption amenity explanation, Costa and
Kahn ( ), using the compensating differential
framework, estimate that a representative house-
hold’s valuation of enjoying the weather of San
Francisco rather than that of Chicago increased
more than fivefold between 1970 and 1990.
This rising income explanation for the move
to nice weather might intuitively, but incorrectly,
be understood to depend on weather’s being a
luxury good. In fact, it depends only on there
being sufficient complementarity between
weather and private consumption in the
household utility function. Even with a
homothetic utility function over private
consumption and weather, an increase in income
requires a sufficient increase in the valuation of
nice weather to dissuade people from moving.
More specifically, if the elasticity of substitution
between private consumption and weather is
exactly 1 (Cobb Douglas), wages and
Economic Consequences of Weather, The
house service prices can adjust to maintain a
spatial equilibrium without any population
movement (Rappaport ). Essentially a rise in the
compensating price of nice weather can exactly
cancel an income-driven increase in demand for
nice weather. But if instead the elasticity of sub-
stitution between weather and private consump-
tion is less than 1, the incomedriven increase in
demand is stronger and the larger required
offsetting price increase can be supported only if
more people move to nice-weather places,
thereby driving up housing prices and driving
down wages to their general equilibrium values.
Conversely, an elasticity of substitution less than
1 will cause the increase in demand for nice
weather from increasing incomes to be
somewhat weaker. In this case, the required
increase in the compensating price is too low to
be sustained without some movement away from
places with nice weather. Intuitively, a broad,
tfp-based increase in wages across all locations
can increase the utility cost from not being
where wage rates are highest.
A first alternative amenity explanation, based
on production, is that the shift to nice weather
reflected the movement out of the agriculture
and manufacturing sectors. As the share of the
labour force employed in agriculture fell from 36
per cent in 1900 to 12 per cent in 1970 to 2 per
cent in 2000, the productive amenity contri-
bution of weather to the marginal product of
labour averaged over all workers probably
decreased greatly. Hence the valuation of
weather attributes directly increasing utility relative
to the valuation of weather attributes conducive
to growing would have increased. More recently,
as the manufacturing share of employment fell
from 25 per cent in 1970 to 14 per cent in 2000,
the opportunity cost of moving within the United
States from places with perceived less nice
weather has probably fallen. On reason is the
concentration of heavy manufacturing in the US
Midwest, in part due to the proximity of raw
materials and notwithstanding winters that are
colder and summers that are hotter than many
US households desire.
While the declines of agriculture and
manufacturing surely contributed to the move to
nice weather, they are unlikely to be the main
3265
cause. The partial correlation of population
growth with nice weather is mostly unaffected by
the inclusion of extensive controls for agriculture
and other industrial structure. Moreover, the
largest part of the move out of agriculture was
over by 1970, which is the start date for the
growth correlations reported above. Conversely,
the move to nice weather began in the 1920s,
when manufacturing employment was still grow-
ing vigorously.
A second alternative amenity story, based on
consumption, is that the move to nice weather
reflected the increased mobility and prosperity
of the elderly. Rather than the population as a
whole, it was primarily the elderly who increased
then- valuation of nice weather as it became part
of their locational choice set. The increase in
choice set followed from numerous trends,
including the passage of Social Security
(pensions for the elderly), increased longevity,
and falling transportation and communications
costs. Certainly, some warm-weather states such
as Florida and Arizona have attracted a
disproportionate number of elderly residents
from elsewhere. But the strength of the
correlation of growth with nice weather is nearly
the same for working-age individuals as it is for
seniors. Moreover, the move to nice weather
began long before the large increases in senior
longevity and prosperity.
A third, related, amenity explanation is that
for a broad swathe of the US population,
mobility costs fell over the course of the 20th
century. High moving costs allow for the
possibility of rents for those residing in nice-
weather places, with the negative compensating
differential settling lower (in absolute value)
than it would be with free mobility. To the extent
that mobility increased - for example, due to
falling transportation and communication costs -
nice-weather places would have grown
disproportionately fast until they reached their
free-mobility equilibrium. While this explanation
has intuitive appeal, the extent to which mobility
increased is unclear. The state-to-state gross
migration rate was approximately flat from 1947
to 1975, then fell slightly through 2000.
A fourth alternative consumption amenity
explanation for the move to nice weather is that
3266
it was caused by air conditioning. Air condition-
ing ameliorated the disamenity of hot and humid
summer weather, which in turn is correlated
with warm winter weather. Hence households no
longer needed to be compensated as much to live
in hot and humid places, which in turn should
have caused a shift in population towards such
places. Doubtless there is some truth to this
hypothesis, as many of the US metropolitan
areas that grew most rapidly from 1950 to 2000
have summer weather that would seem
insufferable without air conditioning (for
example, the daily high heat index in July for
Austin, Texas averages 118 ° F) . However, the
move to nice weather began decades before the
widespread diffusion of air conditioning.
Moreover, the negative partial correlation of
population growth with summer heat and
summer humidity is exactly the opposite of what
air conditioning is expected to cause. Also
tempering the air conditioning explanation is the
extremely rapid growth of coastal southern
California, where summer weather is relatively
mild.
An alternative, nonamenity explanation
argues that the correlation of population growth
with nice weather is largely a coincidence.
Glaeser and Tobio ( ) conclude that the
post-war movement to places with nice weather
arose from faster productivity growth in nice-
weather places accompanied by a high elasticity
of housing supply there. The latter was due to
some combination of plentiful land and minimal
government restrictions on building. The conclu-
sion that weather was not an important driver of
the population move to nice weather follows
primarily from wage and house price
compensating-differential regressions using data
from the 1950 through 2000 decennial censuses.
These regressions suggest that wages rose
quicker but house prices rose slower in places
with nice weather than elsewhere. Both of these
comparative growth rates suggest that house-
holds’ relative valuation of nice weather was
decreasing over this period.
Certainly, the convergence of productivity in
the US South to the national level was an
important aspect of the rapid growth of many
nice-weather places (Barro and Sala-i-Martin ,
; Caselli and Coleman ). But
Economic Consequences of Weather, The
in the absence of any increase in amenity
valuation, the relatively high density and con-
gestion that have come to characterize many
nice-weather cities would require productivity
there to surpass its level elsewhere, not just
converge to it.
Similarly, a relatively elastic housing supply
is certainly a necessary condition for the rapid
growth that was sustained over 50 years by a
number of nice-weather metro areas. In the
quantity model described above, the house
supply elasticity governs the magnitude of the
growth response to a change in amenities. But
the impetus for the growth is solely the amenity
change. Elastic housing supply, on its own, is not
sufficient. Many sparsely populated and
declining metro areas throughout the US
Midwest and deep South also have plentiful land,
light regulation, and in many cases an excess
supply of existing buildings.
An additional consideration is the generic
unreliability of the compensating differential
methodology. The estimated rising wages by
Glaeser and Tobio ( ) in nice-weather places
may partly reflect an upgrading of unobserved
human capital. The increase in the average skills
of workers in such metro areas may have been
faster than elsewhere. For example, workers who
moved to nice-weather places may have had
higher skills on average than the skills of
workers who already lived there. And slower-
than- expected house price growth might reflect
that the (negative) compensation for nice
weather is being paid, in part, by longer
commutes, increased traffic, and other sorts of
metro area congestion.
Conclusions
The conclusion that households are shifting
towards places with nice weather, at least for the
United States, is not very surprising. Indeed, the
US business magazine Forbes parodied some of the
research on the population shift to nice weather
with the headline, ‘Duh!’ (Kellner ). Much more
important is why households are doing so. The
explanations above together suggest that rising
incomes caused individuals to
Economic Consequences of Weather, The
sufficiently increase their valuation of weather as
a consumption amenity so as to require a shift in
population towards nice weather places. For the
increase in valuation to be sufficiently large,
weather must have been a complement to private
consumption rather than a substitute. The shift
towards nice weather was likely reinforced by
the change in industrial composition away from
agriculture and manufacturing, the increase in
productivity throughout the southern United
States, the spread of air conditioning, and the
increasing mobility and financial security of
seniors. Lastly, a high elasticity of housing
supply in many nice-weather places implied that
the population influxes required to support the
increased valuation were quite large.
An important implication of the income result
is that valuations of other local consumption
amenities are likely to have increased as well.
While local governments may be unable to affect
their local weather, they may want to consider
increasing the supply of other consumption
amenities in its place.
A last question is whether the increasing
valuation of nice weather and the shift in
population towards it are likely to continue.
Unambiguously, a continuing increase in income
will cause a continuing increase in the valuation
of nice weather. For the actual movement to nice
weather to continue, the increase in valuation
must be sufficiently large that it cannot be
supported by the existing distribution of
population across locations. With sufficient
complementarity between weather and private
consumption, theory suggests that the move can
continue forever, though at a diminishing pace
(Rappaport ). The increasingly swollen
populations of many nice-weather places put
downward pressure on their abilities to
elastically supply housing and address other
sorts of congestion. As housing supply becomes
less elastic and other sources of congestion rise, a
smaller increase in population can support a
given required increase in compensation for
local amenities. Consistent with a diminishing
shift, decade-by-decade regressions indeed show
that the move towards nice weather peaked in
the 1970s, and then slowed in each of the 1980s
and 1990s.
3267
See Also
► Compensating Differentials
► lousing :'J ... ply
► stems of Cities
► Urban Environment and Quality of Life
Acknowledgments The views herein are those of the author
and do not necessarily reflect the position of the Federal
Reserve Bank of Kansas City or the Federal Reserve System.
Thank you to Yi Li for excellent research assistance.
Bibliography
Barro, R.J., and X. Sala-I-Martin. 1991. Convergence across
states and regions. Brooking Papers on Economic
Activity 1: 107-182.
Barro, R.J., and X. Sala-I-Martin. 1992. Convergence.
Journal of Political Economy 100: 223-251.
Barro, R., and X. Sala-I-Martin. 1995. Economic growth.
New York: McGraw Hill.
Blomquist, G.C., M.C. Berger, and J.R Hoehn. 1988. New
estimates of quality of life in urban areas. American
Economic Review 78: 89-107.
Caselli, F., and W.J. Coleman II. 2001. The U.S. structural
transformation and regional convergence. Journal of
Political Economy 109: 584-616.
Chen, Y., and S.S. Rosenthal. 2008. Local amenities and life-
cycle migration: Do people move for jobs or fun?
Journal of Urban Economics 64: 519-537.
Cheshire, R, and S. Magrini. 2006. Population growth in
European cities: Weather matters - But only nationally.
Regional Studies 40: 23-37.
Costa, D.L., and M.E. Kahn. 2003. The rising price of
nonmarket goods. American Economic Review 93: 227-
232.
Gabriel, S.A., and S.S. Rosenthal. 2004. Quality of the
business environment versus quality of life: Do firms and
households like the same cities? Review of Economics
and Statistics 86: 438-444.
Glaeser, E., and J. Gyourko. 2005. Urban decline and durable
housing. Journal of Political Economy 113: 345-375.
Glaeser, E.L., and K. Tobio. 2008. The rise of the sunbelt.
Southern Economic Journal 74: 610-643.
Greenwood, M.J., G.L. Hunt, D.S. Rickman, and G.I. Treyz.
1991. Migration, regional equilibrium, and the estimation
of compensating differentials. American Economic
Review 81: 1382-1390.
Gyourko, J., and J. Tracy. 1991. The structure of local public
finance and the quality of life. Journal of Political
Economy 99: 774-806.
Haurin, D.R. 1980. The regional distribution of population,
migration, and climate. Quarterly Journal of Economics
95: 293-308.
3268
Hirshleifer, D., and T. Shumway. 2003. Good day sunshine: Stock
returns and the weather. Journal of Finance 58: 1009-1032.
Kellner, T. 2004. Duh! Forbes 173: 50.
Mueser, P.R., andP.E. Graves. 1995. Examining the role of
economic opportunity and amenities in explaining population
redistribution. Journal of Urban Economics 37: 176 200.
Oi, W.Y. 1997. The welfare implications of invention. In The
economics of new goods, ed. T.F. Bresnahan and R.J.
Gordon. Chicago: NBER/University of Chicago Press.
Rappaport, J. 2004. Why are population flows so persistent?
Journal of Urban Economics 56: 554-580.
Rappaport, J. 2007. Moving to nice weather. Regional Science
and Urban Economics 37: 375-398.
Rappaport, J. 2008. Consumption amenities and city population
density. Regional Science and Urban Economics 38: 533-
552.
Rappaport, J. 2009. The increasing importance of quality of life.
Journal of Economic Geography 9: 779-804.
Roback, J. 1982. Wages, rents, and the quality of life. Journal of
Political Economy 90: 1257-1278.
Rosen, S. 1979. Wage-based indexes of urban quality of life. In
Current issues in urban economics, ed. P. Miezkowski and
M. Straszheim. Baltimore: Johns Hopkins University Press.
Saunders Jr., E.M. 1993. Stock prices and Wall Street weather.
American Economic Review 83: 1337-1345.
Stover, M.E., and C.L. Leven. 1992. Methodological issues in the
determination of the quality of life in urban areas. Urban
Studies 2‘>: 737 754.
Economic Demography
Allen C. Kelley and Robert M. Schmidt
Abstract
Economic demography is an area of study that
examines the determinants and consequences of
demographic change, including fertility,
mortality, marriage, divorce, location
(urbanisation, migration, density), age, gender,
ethnicity, population size and population growth.
This article reviews and critically evaluates
important macroeconomic dimensions of the
‘population debates’ between the ‘optimists’ and
the ‘pessimists’ since 1950. It concludes with an
examination of demography in the popular
‘convergence’ growth models of the 1990s.
Economic Demography
Keywords
Adult equivalency; Ageing; Agricultural growth
and population change; Capital accumulation;
Convergence; Demographic drag; Demographic
gift; Demographic transition; Diffusion of
technology; Diminishing returns; Dismal
science; Economic demography; Economic
development; Economic growth; Education;
Endogenous growth; Fertility; Free rider
problem; Human capital; Innovation; Kuznets,
S.; Labour productivity; Learning - by-doing;
Life expectancy; Life-cycle modelling;
Malthus., T. R.; Mortality; Population density;
Population growth; Population size; Renewable
resources; Research and development; Rule of
law; Saving; Simon, J . L.; Subsistence;
Technical change
JEL Classifications
J10
Economic demography is an area of sUidy that
examines the determinants and consequences of
demographic change, including fertility, mortality,
marriage, divorce, location (urbanisation, migration,
density), age, gender, ethnicity', population size, and
population growth. An applied area of research,
economic demography draws upon the theoretical
and applied fields of economies. For example, the
determinants of fertility or migration primarily draw
upon microeconomic theory and labour economics,
while the consequences of population growth or
ageing primarily draw upon macroeconomic theory
and development economics.
The field has had a long tradition of controversy,
beginning with the publication in 1798 of An Essay
on the Principle of Population by the Reverend
Thomas Malthus. The basic Malthusian model is
founded on two propositions: (a) population, when
unchecked, increases at a geometric rate (for
example, 1, 2, 4, 8. . .) and (Z>) food, in contrast,
expands at an arithmetic rate (for example, 1, 2, 3,
4 .. . ). The result is a population trapped at a meagre
standard of living. Short of ‘preventive checks’
(birth control), population is constrained to live at
subsistence by ‘positive
Economic Demography
checks’ (deaths, war, famines and pestilence). In
later writings Malthus admitted the possibility of
‘moral restraint’ that could deter births, primarily
through the postponement of marriage. However, he
held little hope for a notable attenuation of the
‘natural passions’ of the working class.
While much of the controversy relating to Mal-
thusianism has focused on the determinants of
population growth, a second premise of his model
relates to its economic underpinnings: the
determinants of agricultural growth. Here Malthus
appealed to the historical law of diminishing returns
in agriculture. While this proposition engendered
relatively little dispute at the time, history has since
documented widespread and sometimes notable
improvements in agricultural technology. Indeed,
food production has represented an engine of
growth in many of the areas that Malthus
investigated. In some areas today, governments
worry about ‘excess’ food production that depresses
prices and farmers’ living standards. Unfortunately,
the pessimistic food-production predictions, when
confronted by rapid population growth, caused
economics to be dubbed the ‘dismal science’.
The enormous popularity of the Malthusian
ideas was the result of several factors: the model’s
simplicity and its explanation of poverty (the poor
failed to exercise moral restraint, ending up with
large families); the appeal of the message that
subsidising the poor is of questionable efficacy; and
the plausibility of the Malthusian argument given
the unexpected ‘population explosion’ revealed by
the 1801 census. These and other elements of the
‘Malthusian debate’ provide a useful taxonomy for
organising the present article.
Specifically, we highlight the macroeconomic
dimensions of the economic consequences of pop-
ulation growth since 1950. As with the early Mal-
thusian debates, an assessment of the
macroeconomic impacts of demographic change on
economic production has resulted in an outpouring
of research, which has spawned further debate.
There are periods when vigorous Malthusian-like
alarmism has carried the day; there are periods of
counter-challenges; and, since the mid-1980s, there
has been a productive ‘revisionist’ movement. In
short, the simplistic
3269
Malthusian notion of diminishing returns in pro-
duction has given way to more informed modelling
of economic-demographic interactions. An
assessment of the historical evolution of this liter-
ature will constitute the bulk of this review and
appropriately delimits the scope of our essay since a
wide range of important microeconomic themes are
taken up in other articles in this dictionary (see ► -
ertility in Developing Countries,” ►
T ion Making,” ► “Marriage and Divorce,” and -
and multiple articles dealing
with the topics of gender, ageing and mortality).
We begin by examining population impacts in
one-sector growth models. This leads nicely into a
more detailed assessment of factor accumulation,
and in particular, the impacts of demography on
saving, investment and technological change. This
is in turn followed by an analytical description of
the evolution of economic-demographic thinking
since 1950. Such a perspective exposes many of the
key analytical and empirical linkages of interest.
The article concludes with an examination of
‘convergence modelling’, a useful paradigm that
exposes the roles of changing demographic
structures that take place over the demographic
transition.
Theory: Modelling Economic-Demographic
Change
One-Sector Growth Models
The aggregate production function constitutes the
primary organising device for delineating the
impacts of demographic change on economic
growth. Within this model, labour productivity
depends on the availability of complementary fac-
tors of production (land, natural resources, human
and physical capital) and technology. If we assume,
for convenience, that labour is a constant fraction of
population, then population size directly affects
aggregate output.
In a production function with constant returns to
scale, an increase in population growth will lower
the average availability of other factors of
production - a ‘resource-shallowing’ effect, and,
through diminishing returns, reduce the growth of
worker productivity. Such an adverse
3270
demographic impact can be magnified (or
attenuated) if population growth diminishes (raises)
the growth rate of complementary factors.
In a standard growth model with factor inputs of
labour and capital, and a saving rate and pace of
technological change that are exogenous with
respect to population growth, demography affects
the long-run level but not the long-run growth rate
of output per capita. This is because the capital-
shallowing effect of increased population will
eventually reduce the capital per worker ratio to a
level sufficient to be maintained by a fixed rate of
saving. In this case, long-run growth is determined
by the pace of technological change. The
determinants of the ‘fixed’ saving rate and pace of
technology growth, both considered in more detail
below, are central to the analysis.
If one relaxes some of the assumptions of this
model, the impact of population growth on per
capita output growth can be ambiguous. Negative
impacts can arise through diminishing returns,
diseconomies of scale, and perhaps savings, while
positive impacts can arise through induced
technological change, economies of scale, and
possibly savings. Most economists believe that
adverse capital-shallowing impacts will dominate
positive feedback effects, although the magnitude of
the demographic impacts may not be all that large.
Saving
Possibly the most investigated linkage of population
growth to economic growth has been the impact of
demographic change on saving. Two perspectives
dominate.
Adult equivalency. Rapid (slow) rates of popu-
lation growth result in a disproportionate number of
children (elderly adults) who consume, but
contribute relatively little to, household income. In
recognising that these ‘dependents’ consume less
than a working-age adult, the notion of an ‘adult
equivalent’ consumer was bom. The financing of an
additional child’s ‘adult-equivalent’ consumption
has been hypothesised to be out of saving. Such a
view, however, has been challenged by
consideration of several offsetting alternatives.
Specifically, children may (a) substitute for other
forms of consumption, (b) contribute
Economic Demography
directly to household market and non-market
income, (c) encourage parents to work more (or
less), (d) stimulate the amassing (or reduction) of
estates, and (<?) encourage (or discourage) the
accumulation of certain types of assets (for
example, education or farm implements). The net
impact of changing dependency rates on saving is
therefore theoretically ambiguous. This is
particularly the case if one views human capital as
an investment financed in part by households and
governments. At any rate, empirical evidence
showing negative impacts of youth dependency on
saving are found in several studies.
The life-cycle. A second population-saving
linkage is based on a life-cycle formulation incor-
porated into a lifetime household utility function.
Specifically, households attempt to even out their
lifetime consumption by setting aside earnings
during working years to finance consumption by
their children as well as for their own retirement.
This formulation can yield positive or negative
impacts on aggregate saving depending on the
relative sizes of the dissaving youth and elderly
cohorts. While empirical evidence from life-cycle
modelling is mixed, those studies do tend to show
linkages between age structure and saving. How-
ever, the direction and magnitude of that impact
depends upon time and place. (See, for example,
Mason ; Higgins ; and Lee et al. .)
Population-Sensitive Government Spending
Government spending on population-sensitive
activities such as schooling (youth) and health
(elderly) has been alleged both to reduce saving and
to crowd out spending on relatively growth-
oriented investments. These two hypotheses con-
stitute the core of Ansley J. Coale and Edgar M.
Hoover’s ( ) path-breaking study of
India. While these premises are appealing, they
require qualification. Governments have many
options to accommodate population pressures.
Indeed, limited empirical evidence (for example,
Schultz ) has shown that education financing can be
met all or in part by (a) trade-offs within the public
sector, (b ) reductions in per pupil expenditures, and
(c) efficiency gains. While the second approach can
be expected to reduce the quality of
Economic Demography 3271

education (and therefore future productivity), the Chenery and Moises Syrquin ( ) draws upon
importance of population pressures on government the experience of 101 countries across the income
spending or educational quality is uncertain. spectrum over the period 1950-70. They find that
the structure of development reveals strong and
Technological Change: Density, Size pervasive scale effects (measured by population
and Endogenous Growth size) that vary by stage of development. Basically,
While development economists have for decades small countries develop a modem productive
harkened the pace of technological change as a industrial structure more slowly and later, while
(the?) major source of economic growth, most large countries have higher levels of accumulation
standard growth theory models take the rate of and (presumably) higher rates of technological
technological change as exogenous. With techno- change. Although these roles for demography may
logical change independent of demographic change, have been important historically, the impacts
population growth per se will have no impact on the plausibly have waned somewhat: (a) economies in
pace of economic growth in long- run equilibrium. infrastructure are judged to be substantially
By contrast, if technological change is all or in part exhausted in cities of moderate size; (b) speciali-
embodied in new investment, then a vintage sation through international trade provides a means
specification is appropriate whereby new capital is of gamering some or many of the benefits of size;
relatively more productive than old. In this set-up, and (c) scale effects are most prevalent in industries
population growth can be economic-growth with relatively high capital-labour ratios and such
enhancing by expanding the rate at which industries are inappropriate to the factor proportions
technology is incorporated into production. In yet of developing countries.
another specification, population growth can It is in agriculture where the positive benefits of
directly affect the rate of technological change population size have been most discussed. Higher
and/or its form (factor bias). Kenneth population densities can lower per unit costs and
J. Arrow ( ) has hypothesised that learning increase the efficiency of transport, irrigation,
by doing is quickened in an environment of rapid extension services, markets and communications
employment growth. (Glover and Simon ). Possibly the most cited work
A fourth linkage between technology and is that by Ester Boserup ( , ), who
demography is found in ‘endogenous growth’ observes that increasingly productive agricultural
models that relate the pace of technology directly to technologies are made economically attractive in
population size. In particular, the benefits of R&D response to higher land densities. While this is
are assumed to be available to all firms without probably tme, the issue becomes one of identifying
cost; that is, an R&D industry generates a non-rival the quantitative magnitude of such effects over
stock of knowledge. As a result, if we hold constant varying population sizes and in differing
the share of resources used for research, an increase institutional settings. One must be cautious in
in population size advances technological change attributing causation. For example, while high
without limit. This somewhat controversial population densities may have accounted for a
prediction has been qualified by models that portion of expanded agricultural output in recent
incorporate various firm- or industry-specific decades, in several important Asian countries these
constraints on R&D production. Such models densities were sufficiently high decades ago to
typically reduce, but do not eliminate, the positive justify the investments associated with the new
impacts of population size which, as in the technologies. Boserup in more recent writing has
embodiment models above, are manifested largely been less sanguine about the benefits of population
during the ‘transition’ to long- run equilibrium. size because densities appropriate to modem
Evidence on the roles of demographic- technologies in Asia are three to four times the
technology linkages and growth has been frag- average for Africa and Latin America.
mentary and sparse. A pioneering study by Hollis In short, a wide-ranging review of the literature
does not provide a strong consensus on the
3272 Economic Demography

quantitative linkages between the size and growth led to a mildly negative overall assessment that was
of population, on the one hand, and the pace of both cautious and qualified.
technological change and economic growth, on the The most notable feature of this report was its
other hand. methodology. More than any major study on pop-
ulation to that time, the UN Report embraced a
The Bottom Line methodology that would ultimately represent ele-
An evaluation of population growth on economic ments of modern-day ‘revisionism’. Specifically,
growth through the filter of formal economic- the report (a) downgraded the importance of pop-
growth modelling yields limited results: population ulation growth’s impact on economic growth by
growth affects the level but not the growth of per placing it on a par with several other determinants
capita income in long-run equilibrium. Moreover, of equal or greater impact; (b) assessed the conse-
the key determinants of long-run growth are saving quences of population over a long period of time;
and technology. Only if these factors depend on and (c) emphasised the importance of feedbacks
demographic change does population matter. This within and between the economic and political
somewhat constraining limitation of growth theory systems.
has caused researchers to branch out and explore a
host of economic-demographic interactions using Coale and Hoover, 1958
less formal paradigms. This blossoming literature The next major contribution to the population-
has been extensive, lively and sometimes impacts literature was provided by Ansley J. Coale
contentious. and Edgar M. Hoover in their 1958 book
Population Growth and Economic Development in
Low-Income Countries. Based on simulations of a
Evolution of Population-Impacts Thinking: mathematical model calibrated with Indian data,
1950-90 they concluded that India’s development would be
enhanced by lower population growth. This was due
Four major studies, two by the United Nations ( to the hypothesised adverse impacts of population
, ) and two by the National Academy on household saving. It was also proffered that
of Sciences ( , ), reveal well the evolu ‘unproductive’ investments in human capital (such
tion of thinking on population matters over the as health and education) would partially displace
period 1950-90. Three individual scholars, Coale investments in ‘relatively productive’ forms (such
and Hoover and Simon, also played prominent and as machines and factories). Economic growth would
important roles. (This section draws on Kelley •) diminish in response.
Empirically, the above hypotheses have not been
United Nations, convincingly established. While several studies
The 1953 United Nations report, Determinants and have exposed negative dependency-rate impacts on
Consequences of Population Trends, easily saving, there are others that show little or no impact.
represents the most important contribution to pop- Overall, the findings are mixed, with a tilt toward
ulation thinking since the writings of Malthus. supporting the Coale and Hoover formulation. (See
Unlike Malthus, however, the UN study was bal- section “ ’’ above for a
anced and exhaustive both in detail and in coverage. discussion of the trade-offs that households can
Some 21 linkages between population and the make to maintain saving in response to expanding
economy were taken up. For example, the impacts family size. )
of population on the economy can be: Similarly, there are alternative ways for gov-
(a) positive due to economies of scale and orga- ernments to organise and finance schooling in
nisation; (b) negative due to diminishing returns; or response to population pressines. Unfortunately,
(c) neutral due to technology and social progress. studies of this are limited, although one by T. Paul
An evaluation of these and other linkages Schultz ( ) finds no support for the Coale and
Hoover ( ) formulation.
Economic Demography
National Academy of Sciences, 1971
Arguably the most pessimistic assessment of the
consequences of population growth was a study
compiled by the National Academy of Sciences
(NAS). The panel’s final submission. Rapid Pop-
ulation Growth: Consequences and Policy Impli-
cations, issued in 1971, appeared in two volumes:
volume 1, Summary and Recommendations, and
volume 2, Research Papers. Unfortunately, the
Summary volume appeared to be more political than
academic in goal and orientation, and was not
faithful to many of the underlying research reports
assembled by the panel. Indeed, the Summary
volume highlighted some 25 alleged negative
consequences of population growth, whereas it
downplayed or eliminated impacts that could be
considered as ‘neutral’ or ‘favourable’. As a result,
the Summary represents an upper bound on the
negative consequences of population growth. (A
detailed documentation exposing the somewhat
controversial way in which the Summary was
compiled is provided by Kelley .)
What can be learned from the NAS study? First,
given its apparent bias and the lack of a systematic
vetting of volume 1 by members of the panel, it is
difficult to use that volume, either in full or in part.
However, the individual papers are available and
they, in total, offer a more balanced treatment.
Second, by its own acknowledgment, the study
focused on the short run when negative impacts of
population change are most likely to prevail. (‘We
have limited ourselves to relatively short term
issues’; 1971, p. vi.) By contrast, ‘direct’ (short-run)
impacts of demographic change are almost always
attenuated (and sometimes offset) by ‘indirect
feedbacks’ that occur over longer periods of time.
Thus the decision by the NAS panel to focus only
on the short-run direct impacts resulted in an overly
negative assessment of the consequences of popula-
tion growth.
Third, economists were underrepresented on
both the panel and in providing background reports.
This is relevant since economists have substantial
faith in the capacity of markets, individuals and
institutions to adjust in the face of population
pressures. Such adjustments, of course, take time
and they are not without cost.
3273
Finally, this NAS Report provides a striking
example of the difficulty of maintaining objectivity
when social science research enters the public
policy domain.
United Nations, 1973
In 1973 the United Nations weighed in with an
update of its previous seminal work (United
Nations ). In contrast to the broadly eclectic stance
in the earlier report, the new one ended with a mild
to moderate negative overall assessment of rapid
population growth. The authors were concerned
with the ability of agriculture to feed expanding
populations (a la Malthus) and the difficulty of
offsetting capital shallowing (a la Coale and
Hoover). Still, the 1973 Report, whose conclusions
are highly qualified, is not alarmist, nor is it all that
pessimistic. The reason for this moderate stance
was the exceptionally influential empirical finding
of Simon Kuznets ( , pp. 19-20, 63) that
notable negative corre
lations between population growth and per capita
output growth were largely absent in the data.
Given the strong priors of some contributors to the
UN study, a failure to find a negative association in
the aggregate data by a scholar with impeccable
credentials had a profound impact. Indeed, this
singular finding arguably kept the population
debate alive for yet another round of assessments in
the 1980s.
Revisionism, 1980s and Beyond
The 1980s representeda decade whenmanyofthe
underlying assumptions and conclusions of earlier
studies of population-development interactions
were subjected to critical scrutiny. The result was a
revisionist rendering that was both surprising and
controversial. Specifically, the revisionists
downgraded the prominence of population growth
as either a major source of, or a constraint on,
economic prosperity in the Third World. The basis
of this somewhat startling conclusion was the
revisionists’ methodology that (a) assessed the
consequences of demographic change over longer
periods of time and (b) expanded the analysis to
take into account indirect feedbacks within eco-
nomic and political systems. In general, empirical
assessments of population growth will be smaller
3274
(less negative or less positive) when using the
revisionist’s methodology than when focusing on
the short ran and ignoring feedbacks. On net, most
revisionists conclude that many, if not most, Third
World countries would benefit from slower
population growth.
Julian L. Simon, 1981
No one was more important in stimulating the new
round of debates in the 1980s than Julian L. Simon,
author of The Ultimate Resource ( ). This book
attracted enormous attention,
substantially because of two factors. First, it con-
cluded that population growth would likely provide
a positive impact on economic development of
many developed, and some less developed,
countries. Second, the book was accessible, well
written, and organised in a ‘debating’, confronta-
tional style. This included goading and prodding,
the setting up and knocking down of straw men,
and an examination of albeit popular, but somewhat
extreme, anti-natalist positions. Simon’s powerful
book helped spawn a group of survey articles in the
1980s.
What accounts for Simon’s positive assess-
ments? Simon was an early advocate of evaluating
the full effects of population over the intermediate
to long ran. He argued that the negative ‘direct’
impacts in the short ran will probably be moderated,
or sometimes overturned, when households,
businesses, and/or governments react to changing
prices which signal problems of resource scarcity.
Two important examples of responses to population
pressures can be cited: those relating to tech-
nological change and those relating to natural
resource scarcity, both highlighted by Simon.
Technological change. Simon hypothesised and
attempted to document that the pace of tech-
nological change, and its bias, can be stimulated by
population pressures. Technological change, in
turn, plays a central role in economic growth theory
and has been shown in sources-of-growth studies to
be a (the?) key to economic growth. Additionally,
with respect to population size impacts in general,
Simon observes that major social overhead projects
(for example, roads, communications and irrigation)
have benefited from expanded populations and
scale. (For more
Economic Demography
detail, see section “
” above.)
Resource depletion. Consider next the impacts
of population growth on natural resource depletion.
Theoretically an exhaustion of non-renewable
resources (for example, coal and minerals) would
appear to be inevitable in the long ran. However,
such a period may be in the indeterminably distant
future. By contrast, Simon argued that the most
relevant measure of resource scarcity is its price. He
prepared many graphs of US non-renewable
resource prices (deflated by price indexes in order
to focus on ‘real’ resource trends).
Surprisingly, virtually every resource has expe-
rienced a declining real price over lengthy periods
of time. This means, a la Simon, that resources are
becoming more abundant over time. It seems that
the more resources are used, the more abundant
they become! How can this happen? Simple. A
rising resource price, due in part to population
pressures, triggers several reactions that reduce or
even eliminate the apparent resource scarcity.
Specifically, in the short run, rising prices encour-
age an economising of the resource at every level of
production and consumption. In the longer run,
rising prices stimulate exploration, new methods of
extraction and process, and the search for
substitutes.
Nevertheless, Simon recognised that market
failures, institutional failures, and political factors
can all result in less-than-complete adjustments
when population and economic development press
against resource availabilities. This is particularly
the case with renewable resources (such as rain
forests, fisheries, the environment, and so forth)
where market or institutional failures are pervasive.
Without mechanisms to assign and maintain
property rights, internalise externalities, and address
free rider problems of public and quasi-public
goods, government regulation may be required to
safeguard renewable resources over time.
National Academy of Sciences, 1986
Some 15 years after the 1971 National Academy
Report that highlighted 25 negative consequences of
population growth, a new National Academy
Economic Demography 3275

Report was released. In contrast to the previous children on the capacity to save, and in turn to
study, the new report was balanced, eclectic and undertake productive investments; and (3) the
non-alarmist. A careful examination of its bottom inability to invest in schooling and health facilities.
line is instructive. Finally, the 1986 Report upgrades the concern
about population impacts on renewable natural
On balance, we reach the qualitative conclusion that
slower population growth would be beneficial to
resources (such as fishing areas and rain forests)
economic development of most developing coun- where property rights are difficult to assign and
tries. ( , p. 90; emphasis added) maintain. Overuse can result. It is recognised that
the problems of overuse are not solely due to
This qualified assessment reveals key features population growth per se, but rather institutional
found in most population assessments in the 1980s. failure. Cutting population growth by one half, or
Specifically: (a) there are both positive and negative even to zero, would not solve the problem. Rather it
impacts of demographic change (thus ‘on balance’); would slow the process and postpone the date of
(b) the magnitude of the net impacts cannot be resource exhaustion. Government policies are
determined given current evidence (thus needed to account for negative externalities and
‘qualitative’); (c) only the direction of the impact market failure. Slowing population growth provides
from high to low growth rates can be ascertained time for institutional response.
(thus ‘slower’ rather than ‘slow’); and (d) the net
impact varies from country to country, hr most
cases it will be negative; in some positive; and in New Paradigms for Modelling
others of little impact (thus, ‘most developing Demography's Role in Economic Growth:
countries’). 1990 and Beyond
What accounts for the dramatic turnaround in
the two National Academy assessments? Several As noted previously, Kuznets’s empirical finding of
factors can be advanced. First, the 1986 report an absence of notable negative correlations between
extends the short-run time horizon of the 1971 population growth and per capita output growth
report to examine individual and institutional influenced the population debate throughout the
responses to the initial impacts of population 1970s and 1980s. Simple correlations stimulated
change: conservation in response to scarcity, sub- research during the 1990s as well. This time,
stitution of abundant for scarce factors of produc- however, statistically significant negative
tion, innovation and adoption of technologies to correlations during the 1980s drove the discussion.
exploit profitable opportunities, and the like. These Interestingly, economic-demographic modelling
responses are considered to be pervasive and they continued in the ‘revisionist’ vein, incorporating
are judged to be important. According to the report positive and negative as well as short- and long-run
writers: ‘the key [is the] mediating role that human influences into an economic growth model. The
behaviour and human institutions play in the modelling challenge remains one of accommodating
relation between population growth and economic correlations that can be negative, positive or
processes’ ( , p. 4). insignificant depending upon time and place.
Second, the 1986 study was assembled almost
entirely by economists whose understanding of and Convergence Growth Models: A Framework
faith in markets to induce responses that modify for Assessing Demography's Impact
initial direct impacts of population change is far Renewed interest in modelling the impacts of
greater than that of other social and biological demographic change on economic growth coincided
scientists. with the emergence in the economic growth
Third, research accumulating over the 15 years literature of the ‘technology gap’ or ‘convergence’
between the two reports revealed a need to down- model. This model, formulated initially
grade: (1) the concern about non-renewable
resource exhaustion; (2) the adverse impact of
3276
by Barro and Sala-i-Martin ( ), has been used
widely to explore many hypothesised influences on
economic growth, including openness to trade, form
of government, and the mle of law. Since this type
of modelling highlights the dynamics of the
adjustment process, it is particularly relevant to
examining the impacts of major shifts in the
population’s age distribution associated with birth
and death rates that change systematically over the
demographic transition. As a result, economic
demographers have employed convergence
paradigms to explore demographic-economic
interactions.
Briefly stated, convergence models focus on the
pace at which countries move from their current
level of l a bo u r productivity to their long-run or
steady-state level of labour productivity. The model
assumes that all countries converge at the same rate
from their current to their long-run levels (which
can vary across countries and over time). The
greater the productivity gap, the greater are the gaps
of physical capital, human capital and technical
efficiency from their long-run levels. Large gaps
allow for ‘catching up’ through (physical and
human) capital accumulation, and technology
creation and diffusion across countries and over
time. Indeed, many empirical studies indicate that
growth rates do slow down as a country approaches
its long- run productivity level, especially those
studies that provide for country- and period-specific
conditions that influence the long-run level of
labour productivity.
Since long-run labour productivity is
unobservable, empirical implementations of the
model substitute a vector of ‘conditioning’ variables
thought to influence long-mn labour productivity.
The actual specification of these conditioning
variables varies notably. Consider two of their many
representations. The first, by Barro ( ), highlights
inflation, government
consumption ratios, the mle of law, the form of the
political system, terms of trade, human capital, the
total fertility rate, and life expectancy at birth (a
proxy for health). The second formulation, by
Bloom and Williamson ( ), highlights two
categories of growth-rate determinants: economic
structure variables (natural resources, schooling,
Economic Demography
access to ports, location in the tropics, whether
landlocked, and extent of coastline); and economic
and political policies (openness to trade, quality of
institutions, and government savings share of GDP).
Clearly there are many defensible perspectives on
variable choice, and much is yet to be learned about
the appropriate configuration of conditioning
variables that influence long-run productivity levels.
Alternative Demographic Renderings
Within a Convergence Framework
The 1990s witnessed attempts by various
researchers to model demography in a manner that
accommodates both the insignificant correlations of
the 1960s and 1970s as well as the significant
negative correlations of the 1980s and 1990s. Three
different approaches are described here. All three
employ a convergence-type growth model and all
employ a broad set of countries spanning the
income spectrum.
Modelling through aggregate measures of fer-
tility and mortality. B a r r o ( ) i n c l u d e s
two
demographic aggregate measures among his list of
conditioning variables, the total fertility rate (TFR)
and life expectancy. Barro’s formulation thus has
demography impacting the long-run equilibrium
level of per capita income. The TFR captures, for
example, the adverse capital- shallowing impact of
more rapid population growth as well as the
resource opportunity costs of bringing up children.
Furthermore, while Barro treats life expectancy as a
human capital proxy for health, demographers
consider it to be a demographic variable. Both are
statistically significant, with a higher TFR
inhibiting, and longer life expectancy enhancing
economic growth.
Modelling through population growth compo-
nents. K e l l e y a n d S c h m i d t ( )
decompose
population growth by examining two components
(births and deaths) and by modelling their con-
temporaneous and lagged impacts. This approach
allows for disparate impacts of fertility and mor-
tality as well as negative short-run effects (costs of
high birth and death rates) and positive long-run
effects (favourable impacts of past births on current
labour force growth and declining mortality).
Consistent with Kuznets’s earlier work, they
Economic Demography
found an absence of a net demographic impact on
economic growth in the 1960s and 1970 - the
separate impacts of births and deaths are notable but
offsetting. Consistent with empirical work of the
early 1990s, they found negative impacts
throughout the 1980s. These negative correlations
were in part the result of (a) rising short-run costs of
high birth rates, (b) declining benefits of mortality
reduction, and (c) insufficient labour force entry
from past births to offset these increased costs.
Modelling through differential age-structure
growth. In a series of papers beginning in the late
1990s, several Harvard economists argued for a
demographic rendering that incorporates not only
population growth but also labour growth (see, for
example, Bloom and Williamson ; and Bloom et
al. ). They note that, while theorists conceptualise
the economic growth process in labour productivity
terms, empirical growth models are generally
specified in per capita terms. This makes no
difference when population and labour grow at the
same rate, but does when they grow at different
rates.
The authors argue that the post-war period was
exactly such a time since during that period demo-
graphic transitions took place in different countries
at different times and at different paces. At various
stages of the demographic transition, the population
and working ages (used within this framework as a
proxy for labour) can grow at very different rates. In
a predictable pattern, the population initially grows
faster, then slower, and then faster than the
working-aged population during the transition from
a high-fertility, high- mortality to a low-fertility,
low-mortality demographic steady-state
equilibrium. (For an historical evolution of
economic, sociological, and biological factors
during the demographic transition, see R.A.
Easterlin .)
Without allowing for differential growth rates of
the population and working ages, demographic
coefficient estimates (mainly population growth)
will be biased. In that case the population-growth
coefficient captures net demographic impacts that
can be positive, negative, or neutral, depending
upon time and place. Bloom and Williamson ( )
demonstrate this point for a broad cross
3277
section of countries over the period 1965-90 in a
convergence model that also includes fife expec-
tancy as a human capital variable. Consistent with
some studies, their simple demographic rendering
results in a positive but insignificant coefficient for
the population growth rate. When supplemented by
the working-age growth rate, however, that
coefficient turns negative and the coefficient for the
working-age growth rate is positive, both
statistically significant.
Effectively, the Harvard economists append an
accounting structure to translate labour productivity
impacts into per capita terms. The resulting
demographic specification is elegant in its sim-
plicity, incorporating only two demographic vari-
ables that have unambiguous predicted coefficient
values of- 1 (for population rate of growth, Ngr) and
+1 (for working-age population rate of growth,
WAgr) when used to expose demography’s impact
on income growth per capita relative to income
growth per working- age population. In that context,
demography exerts its primary impact on the pace at
which the long-run equilibrium is reached (Bloom
and Williamson , p. 419) rather than on the long-
run equilibrium level of productivity.
This is an intriguing specification. The inter-
pretation is clear: if labour force growth exceeds
population growth, then the rate of per capita
income growth is boosted by demography. The
Harvard economists label this phenomenon the
‘demographic gift’ that may be reaped for several
decades after the onset of fertility decline as new
labour force entrants from earlier large birth cohorts
outpace fertility. The ‘gift’ was large throughout the
1965-90 period for Japan and other Asian Tigers
because of the early and rapid pace of their
demographic transition. Of course, the converse of
the ‘gift’ began to be felt in the 1990s as new labour
force entry from smaller birth cohorts was outpaced
by labour force exit of the ageing population. The
model predicts productivity outpacing per capita
income growth over several decades into the future
in these Asian (and other) countries.
Note that the qualitative predictions are based
on theoretically determined coefficients on WAgr
and Ngr of +1 and — 1, respectively. To the extent
3278
that estimated coefficients deviate from +1 and —
1, WAgr and Ngr play an additional role in the
determination of the long-run productivity level.
The Harvard studies provide some guidance in this
area. In their earlier study. Bloom and Williamson (
) estimate coefficients that dif
fer significantly from +1 and —1. However, in a
later study that further elucidates the accounting,
Bloom et al. ( ) find no significant difference
from those values. If that is the case, then the model
at once makes an important contribution and is
somewhat narrower than many in the literature
which admit both short-run and long-run impacts of
demographic change as a part of the theoretical
structure. Yet modelling demography in growth
equations tends to be both imprecise and ad hoc. In
contrast, the Bloom and Williamson model is
relatively clear in interpretation, and it targets the
shorter-run impacts that are of primary interest to
policymakers.
The Bottom Line
Bloom and Williamson ( ) estimated that as
much as one-third of the average per capita income
growth rate in East Asian countries over the period
1965-90 is explained by population dynamics.
Kelley and Schmidt ( ) evaluated
eight distinct demographic renderings within a
convergence model using a consistent set of con-
ditioning variables - those described above for
Barro’s variant. Among others, these renderings
included Barro’s TFR; a ‘naive’ variant predating
the 1990s work that simply includes Ngr; a ‘com-
ponents’ model (contemporaneous and lagged birth
rates and the death rate: Kelley and Schmidt ); two
variants of the Harvard transitions framework; and
demographic extensions to several variants.
Kelley and Schmidt ( ) find that on aver
age, across all eight demographic formulations and
over their frill 86-country sample (covering the full
income spectrum), approximately 21 per cent of the
combined impacts on change in the per capita
income growth rate is accounted for by changes in
the demographic variables in the various models.
What is striking about this result is that the 21 per
cent is fairly stable across all eight demographic
renderings, from one that is quite
Economic Demography
simplistic (Ngr only) to those that incorporate
short-, intermediate- and long-term population
effects. On the one hand, this should not be terribly
surprising because of the interconnectedness of all
of the demographic measures. On the other hand,
while population matters, it is still important to
determine why.
Although there is an emerging consensus that
the magnitude of the impacts of population growth
have been sizeable (for example, 21 per cent
globally and as much as 33 per cent in East Asia ),
the reasons why this is the case are still both
contestable and not well understood. Are the
demographic determinants primarily longer-run
impacts, or are they mainly shorter-run transitional
dynamics that are diminishing? Will the so-called
‘demographic gift’ of these dynamics in the past
reveal themselves as a ‘demographic drag’ in the
future, deriving from reduced fertility, slow
population growth and ageing? Or will a new
mechanism reveal itself? For example, (a) will
future modelling better expose the components of
labour force change (for example, utilisation rates,
age- and/or gender-specific participation rates); and
(b) will fertility and mortality be endogenously
specified to better reveal the dynamics of the
demographic transition about which the field of
economic demography has much to say? Whatever
the outcome, the stage is set for another round of
research, pinning down the results of the past with
the goal of understanding the future.
See Also
► amily I ecision Making
► "ertility in Developing Countries
► Marria end Divorce
► tetirer
Bibliography
Arrow, K.J. 1962. The economic implications of learning
by doing. Review of Economic Studies 29: 155-173.
Barro, R.J. 1997. Determinants of economic growth: A cross-
country empirical study. Cambridge, MA: MIT Press.
Economic Development and the Environment
Barro, R.J., and X. Sala-i-Martin. 1991. Convergence.
Journal of Political Economy 100: 223-251.
Birdsall, N., A.C. Kelley, and S. Sinding (eds.). 2001.
Demography matters: Population change, economic
growth and poverty in the developing world. Oxford:
Oxford University Press.
Bloom, D.E., and J.G. Williamson. 1998. Demographic
transitions and economic miracles in emerging Asia.
World Bank Economic Review 12: 419-455.
Bloom, D.E., D. Canning, and P. Malaney. 2000. Demo-
graphic change and economic growth in Asia. Population
and Development Review 26: 257-1990.
Bosemp, E. 1965. Conditions of agricultural growth.
Chicago: Aldine.
Bosemp, E. 1981. Population and technological change.
Chicago: University of Chicago Press.
Chenery, H., and M. Syrquin. 1975. Patterns of development:
1950-1970. Oxford: Oxford University Press.
Coale, A.J., and E.M. Hoover. 1958. Population growth and
economic development in low-income countries.
Princeton: Princeton University Press.
Easterlin, R.A. 1978. The economics and sociology of
fertility: A synthesis. In Historical studies of changing
fertility, ed. C. Tilly. Princeton: Princeton University
Press.
Glover, D.R., and J.L. Simon. 1975. The effect of population
density on infrastructure: The case of road building.
Economic Development and Cultural Change 23: 453-
468.
Higgins, M. 1998. Demography, national savings, and
international capital flows. International Economic
Review 39: 343-369.
Kelley, A.C. 1988. Economic consequences of population
change in the Third World. Journal of Economic Liter-
ature 26: 1685-1728.
Kelley, A.C. 2001. The population debate in historical
perspective: Revisionism revisited. In Population mat-
ters: Demographic change, economic growth, and pov-
erty in the developing world, ed. N. Birdsall, A. Kelley,
and S. Sinding. Oxford: Oxford University Press.
Kelley, A.C., andRM. Schmidt. 1995. Aggregate population
and economic growth correlations: The role of the
components of demographic change. Demography 32:
543-555.
Kelley, A.C., and R.M. Schmidt. 2001. Economic and
demographic change: A synthesis of models, findings and
perspectives. In Population Matters : Demographic
Change, Economic Growth, and Poverty in the
Developing World, ed. N. Birdsall, A.C. Kelley, and S.
Sinding. Oxford: Oxford University Press.
Kuznets, S. 1960. Population change and aggregate output. In
Demographic and economic change in developed
countries, National Bureau of Economic Research.
Princeton: Princeton University Press.
Lee, R.D., A. Mason, and T. Miller. 2001. Saving, wealth and
population. In Population matters: Demographic change,
economic growth, and poverty in the developing world,
ed. N. Birdsall, A.C. Kelley, and S. Sinding. Oxford:
Oxford University Press.
3279
Malthus, T.R. 1798. An essay on the principle of popula-
tion, , 1970. Harmondsworth: Penguin.
Mason, A. 1987. National saving rates and population
growth: A new model and new evidence. In Population
growth and economic development: Issues and evidence,
ed. D.G. Johnson and R.D. Lee. Madison: University of
Wisconsin Press.
National Academy of Sciences. 1971. Rapid population
growth: Consequences and policy implications, vol. 1:
Summary and recommendations; vol. 2: Research
papers. Baltimore: Johns Hopkins University Press.
National Academy of Sciences. 1986. Population growth and
economic development: Policy questions. Washington,
DC: National Research Council.
Schultz, T.P. 1987. Schooling expenditures and enrollments
1960-1980: The effects on income, prices and population
growth. In Population growth and economic development
issues and evidence, ed. D. Gale Johnson and R.D. Lee.
Madison: University of Wisconsin Press.
Simon, J.L. 1981. The ultimate resource. Princeton: Princeton
University Press.
Simon, J.L. 1996. The ultimate resource 2. Princeton:
Princeton University Press.
Srinivasan, T.N. 1988. Modeling growth and economic
development Journal of Policy Modeling 10: 7-28.
United Nations. 1953. The determinants and consequences of
population trends. New York: United Nations.
United Nations. 1973. The determinants and consequences of
population trends. New York: United Nations.
Economic Development and
the Environment
Ian Coxhead
Abstract
Economic development in low-income econo-
mies is initially highly resource-intensive.
Resource depletion and pollution damage is
often estimated to reduce ‘real’ GDP growth by
between one and two per cent per year. Growth
and structural change alter the environment-
development nexus in nonlinear fashion. Policy
reforms, global market integration, and
institutional development all alter the propensity
for growth to generate environmental damage.
The emergence of new trade patterns among
developing countries has created
3280
new challenges in the measurement and analysis
of development-environment interactions.
Larger developing economies are now emerging
as major sources of emissions that contribute to
global climate change.
Keywords
Biodiversity; Comparative advantage; Conser-
vation; Economic development and the envi-
ronment; Environmental economics;
Environmental Kuznets curve; Greenhouse gas
emissions; Growth and international trade;
Heckscher-Ohlin trade theory; Import
substitution; Income effects; Natural resources;
Non-use amenities; North-South economic
relations; Pollution; Pollution haven hypothesis;
Poverty alleviation; Ricardian trade theory;
Social norms; Structural change; Sustainable
development
JEL Classifications
010
Economic development depends on sustained per
capita income growth and entails dramatic changes
in production structure. In low-income economies,
growth typically stimulates markets and promotes
the evolution of institutions that constrain
behaviour according to social norms. The
expansion of trade in relation to GDP is another
common accompaniment to growth. Each of these
has effects on ‘the environment’, which in a
developing-country setting refers not only to
phenomena such as water and air quality but also,
importantly, to natural resource stocks such as
forests, fisheries and soils.
Conversely, changes in environmental quality,
including resource stock drawdowns, may affect
economic development in a dynamic interaction.
This feedback is hard to quantify; however, the
World Bank's World Development Indicators series
now includes ‘adjusted’ national accounts data
reporting GDP and savings net of the implied value
of resource depletion and environmental damage
(Bolt et al. ). These indicate that environmental
damage can reduce GDP growth by as much as one
to two per cent per year. On a
Economic Development and the Environment
broader scale, growth of large low-income econ-
omies like China and India is beginning to have
ramifications not only for their own environmental
conditions, but also for the global environment
through transboundary pollution spillovers and
greenhouse gas (GHG) emissions.
The welfare of the poor in low-income countries
is intimately linked to their access to environmental
assets, and especially to the natural resource base.
Despite this, the central concerns of environmental
and resource economics - the economic costs of
pollution and natural resource depletion - have only
recently begun to be linked to models of economic
development. Publication of the so-called
Brundtland Report (WCED ) was a watershed
event; since then, ‘no account of economic
development would be regarded as adequate if the
environmental-resource base were absent from it’
(Dasgupta and Maler , p. 2734).
Growth in low-income economies is inevitably
associated with higher resource demands and
increased pollution intensity per unit of income
generated. Other things equal, more economic
activity generates more environmental damage
monotonically through a scale effect. The rela-
tionship may be nonlinear, however. As income
grows, environmental damage per unit of additional
income may initially rise, then decline. This
conjecture, known as the environmental Kuznets
curve (EKC), posits that scale effects dominate all
other influences on the growth-environment
relationship at low income levels, but that, as
incomes rise, changes in the composition of
production, technological improvements, and
income-elastic preferences for conservation and a
cleaner environment become more influential
(Grossman and Krueger ). Institutional and legal
constraints on pollution and resource depletion,
initially so weak as to create a form of open access
for polluters and resource depleters, may also
evolve or be applied with greater vigour as incomes
increase, whether due to income effects or to
increased recognition of limits to growth imposed
by pollution and resource scarcity (Stokey ).
Despite the heuristic value of EKC, however,
empirical tests in low-income economies are
plagued by data and measurement
Economic Development and the Environment
problems. Most notably, there is no robust evidence
of an EKC for resource-depleting activities such as
deforestation.
Changes in production structure and factor
demands are also inherent to development. The
most prominent manifestation of structural change
in low-income countries is the relative decline of
agricultural and resource sectors as contributors to
GDP and employment. This has clear
environmental implications when the majority of
the population is initially dependent on the natural
resource base. In capital-scarce economies, forest
and land conversion for agriculture and the
exploitation of fisheries and other resource stocks
are standard strategies for increasing labour
productivity and generating surpluses. Accordingly,
early stages of development are characterized by
rapid resource depletion - most visibly in the form
of tropical deforestation. Such processes are abetted
by conditions of open access (Barbier )•
Whether the depletion rate eventually slows - a
prerequisite for sustainable development - depends
largely on the extent to which surpluses are used to
build capacity in secondary and tertiary industries
making more intensive use of reproducible
resources such as labour, technology and human
capital. In this way, the central story of structural
change in low-income economies is intimately
linked to the evolution of demands on the
environmental and natural resource base. Sustained
growth leads to a relative reduction in dependence
on natural resources, and thus makes it easier for
society to agree to promote conservation,
biodiversity retention and non-use amenities.
Conversely, macroeconomic failures, often in
combination with rapid population growth, high
transactions costs and market failures, can lead low-
income economies into unsustainable cycles of
poverty, resource over-exploitation, and insti-
tutional failure.
Trade is another influential source of structural
change. Early development policies stressing import
substitution and de-emphasizing trade have, in most
countries, been supplanted by greater outward
orientation. Trade-to-GDP ratios have risen and
domestic prices have tended to converge on world
market prices, thus altering
3281
domestic production and investment incentives.
With the exception of resource-poor East Asian
countries like Korea and Taiwan, the pursuit of
comparative advantage in low-income countries
initially means expanded exports of tropical agri-
culture, forestry and fisheries and of resource- based
semi-manufactures such as sawnwood. Both the
growth of global demand and the pro-trade effects
of policy reforms encourage accelerated resource
drawdowns; unless property rights and externalities
are adequately dealt with, these are likely to occur at
socially excessive rates (Coxhead and Jayasuriya ).
A related idea known as the pollution haven
hypothesis posits that weak environmental laws and
unresolved externalities may lead developing
countries to specialize in pollution-intensive
industrial activities (Copeland and Taylor ).
Whereas early policy advice to developing
countries typically stressed the desirability of
exploiting resource wealth to create jobs and earn
foreign exchange, contemporary concerns about
exhaustibility and the integrity of ecological sys-
tems have led to more cautious counsel and an
emphasis on sustainable development. Such advice,
however, is often difficult to implement as policy in
the face of pressures to promote growth and
alleviate poverty in the current generation.
New issues in the development-environment
relationship continue to emerge as economies grow
and become more globalized. Traditionally, trade-
environment analyses used Ricardian or Heckscher-
Ohlin models of North-South interactions in which
welfare growth in resource-abundant South is
contingent on trade witli industrialized North and on
domestic externalities or market failures (for
example, Chichilnisky ). However, South-South
trade - or, in the case of China's emergence as a
major market for resource exports from Asia,
Africa, and Latin America, ‘East-South’ trade - is
now growing much faster than trade of the North-
South type. South-South trade is a form of
internationally fragmented production in which
primary products or semimanufactures are exported
from one low-income country to another to be used
in production of final goods. The latter low-income
economy thus moves to ‘clean’ growth based on
labour-intensive
3282
manufactures, while growth in the former becomes
more resource-intensive. Countries in the South
may have comparative advantage in either clean or
dirty goods - or both. Conventional models and
measures for evaluating environmental costs of
growth must be adapted to such new modalities.
Other new trends reflect the growing global
influence of large developing economies. In poor
countries, about 50 per cent of carbon dioxide
emissions (the primary sources of GHGs) comes
from land conversion. But total emissions increase
rapidly with energy demands driven by growth,
urbanization and industrialization. According to the
International Energy Agency, China accounted for
13 per cent of global energy-related CO2 emissions
in 2006, and is expected to overtake the USA as the
largest CO2 source by 2009; India is now following
a similar path (IEA ). Under the 1997 Kyoto
Protocol, these economies are not required to limit
GHG emissions. But, even if they do take major
steps to limit pollution intensity, scale effects of
then- growth will ensure that global pollution
externalities will continue to expand for the
foreseeable future. In turn, concerns over the global
environmental consequences of growth in low-
income countries will find increasingly forceful
expression in international negotiations not only on
the environment but also on trade and other forms
of international integration.
See Also
► Climate Change, Economics of
► Environmental Economics
► Environmental Kuznets Curve
► Poverty Alleviation Programmes
Bibliography
Barbier, E.B. 2005. Natural resources and economic devel-
opment. Cambridge: Cambridge University Press.
Bolt, K., M. Matete, and M. Clemens. 2002. Manual for
calculating adjusted net savings. Environment Depart-
ment, World Bank: Mimeo.
Chichilnisky, G. 1994. North-South trade and the global
environment. American Economic Review 84: 851-874.
Economic Epidemiology
Copeland, B.R., and M.S. Taylor. 1994. North-South trade
and the environment. Quarterly Journal of Economics
109: 755-787.
Coxhead, I., and S.K. Jayasuriya. 2003. The open economy
and the environment: Development, trade and resources
in Asia. Cheltenham/Northampton: Edward Elgar.
Dasgupta, P., and K.-G. Maler. 1995. Poverty, institutions
and the natural resource base. In Handbook of develop-
ment economics, ed. J. Behrman and T.N. Srinivasan,
Vol. 3A. Amsterdam: North-Holland.
Grossman, G.M., and A.B.. Krueger. 1993. The environ-
mental impacts of a North American Free Trade
Agreement. In The US-Mexico free trade agreement, ed.
P. Garber. Cambridge: MIT Press.
IEA (International Energy Agency). 2006. World energy
outlook 2006. Paris: IEA.
Stokey, N. 1998. Are there limits to growth? International
Economic Review 39: 1—31.
WCED (World Commission on Environment and Devel-
opment). 1987. Our common future. Oxford/New York:
Oxford University Press.
Economic Epidemiology
Tomas J. Philipson
Abstract
The economic analysis of epidemiological
issues has different implications for disease and
its optimal control from those of traditional
analysis of such issues. It views undesirable
disease occurrence as the result of self-
interested behaviour in the presence of con-
straints. Unlike with methods used in public
health, the effects and desirability of disease-
reducing public interventions are then evaluated
in terms of how they improve the private
behaviour essential to controlling disease. Eco-
nomic epidemiology has been applied to a wide
range of topics, including infectious diseases
such as AIDS, and also to non-infectious
behaviour such as smoking, obesity, and crime.
Keywords
Economic epidemiology; Health subsidy;
Morbidity; Mortality; Programme evaluation;
Public health; Rational epidemics
Economic Epidemiology
JEL Classifications
II
The fast-growing literature on the economic
analysis of epidemiological issues (see Philipson ,
for a review) delivers very different implications
about disease occurrence and its optimal control
from those of traditional analysis of the same issues
in the field of public health. At the risk of vastly
oversimplifying the positive component of the
public health approach, the traditional analysis
comprises empirical methods and analysis aimed at
identifying and quantifying the effects of ‘risk
factors’ on health outcomes. These factors are
typically defined as covariates that negatively affect
the measured health outcomes - for example, the
effects of smoking on lung cancer or the effects of
obesity on heart disease. Thereafter, the normative
component of the public health approach is
concerned with attempts to reduce the measured
risk factors, whether through private or public
intervention, and to thereby improve health
outcomes.
This approach drastically differs from that of
economic epidemiology, which attempts to explain
undesirable disease occurrence as the result of self-
interested behavioiu in the presence of constraints.
The effects and desirability of disease- reducing
public interventions are then evaluated in terms of
how they improve the private behaviour essential to
controlling disease in the first place. In some sense,
the public health approach aims to improve health,
whereas the economic approach aims to improve
economic efficiency, even if that does not
necessarily improve health. Just as closing
highways would improve health but impair eco-
nomic efficiency, the two approaches often clash in
desired interventions. The public health approach,
therefore, more often favours public intervention,
and sometimes simply assumes that the existence of
a health problem is sufficient cause for intervention,
potentially because it lacks a theory about how
private incentives affect the observed level of
disease across time and populations.
3283
Economic Epidemiology and Infectious
Disease
Infectious diseases cause roughly one-third of all
deaths worldwide and represent the primary cause
of mortality in the world. Historically, the share of
worldwide mortality due to infectious diseases has
been even greater, although data tend to be less
reliable for earlier periods. Morbidity and mortality
from infectious diseases such as tuberculosis,
malaria and acute respiratory infection have always
been at the forefront of public policy in developing
countries, where infectious diseases accounted for
nearly one-half of mortality in the 1990s.
Worldwide concern about infectious disease has
received renewed interest in public policy
discussions given the disastrous impacts of
HIV/AIDS and the potential threat of bird flu. Like
most communicable diseases, especially those that
are potentially fatal, HIV has incited an extensive
governmental response, consisting of regulatory
measures, subsidies for research, education,
treatment, testing and counselling. Here we review
the main contributions of economic epidemiology in
predicting both the short- and the long-run
behaviour of infectious disease, as well as the
effects and desirability of public health
interventions that attempt to reduce such disease.
Philipson and Posner ( ) provide the first
systematic analysis of rational infectious disease
epidemics in the context of AIDS. Kremer ( )
analyses the effects of a reduction in the number of
one’s sexual partners on the growth of disease. The
predictions of such models rely crucially on the
prevalence elasticity of private demand for
prevention against disease, that is, the degree to
which prevention increases in response to disease
occurrence. Prevalence-elastic behaviour has dif-
ferent implications for the susceptibility to infection
than standard epidemiological models of disease
occurrence as discussed in Philipson ( ).
Evidence of the degree of prevalence-
elastic demand is discussed in Ahituv et al. ( ) and
Auld ( , ). Oster
( ) attempts to explain the lack of
3284
prevalence-elastic demand in Africa by the com-
peting risks that lower the demand for prevention in
that part of the world. Lakdawalla et al. (
provide evidence that demand is sensitive to overall
risk, both in terms of prevalence and the cost of
infection as when reduced by new medical
technologies.
This type of prevalence-elastic behaviour has
two major implications. First, growth of infectious
disease is self-limiting because it induces preventive
behaviour. Second, since the decline of a disease
discourages prevention, initially successful public
health efforts actually make it progressively harder
to eradicate infectious diseases. Geoffard and
Philipson ( ) discuss a very
general result concerning the inability of private
markets to eradicate disease when demand is
prevalence-elastic because a disappearing disease
implies less prevention. Barrett ( ) and Barrett
and Hoel ( ) also analyses the implications of
economic efficiency for optimal eradication. See
also Gersovitz and Hammer ( , ,
Regarding the value of public health interven-
tions, Mechoulan ( ) analyses the prevalence
and efficiency implications of HIV testing.
Geoffard and Philipson ( ) argue that eradica
tion is never Pareto optimal when only the current
generation is considered. However, the missing
market is dynamic: future generations cannot pay
vaccine producers for the benefit they derive from
the producers’ product. Brito et al. (
analyse the nonstandard efficiency implications of
mandatory vaccinations.
Moreover, the prevalence elasticity of demand
lowers the price elasticity of demand, which implies
that Pigouvian-style subsidies to stimulate
prevention may have only limited success. This
occurs because demand rises among those who are
subsidized and falls among those who are not - in
the extreme case, total demand is inelastic to
subsidies. In addition, prevalence competes with
public interventions in inducing protective activity,
which makes the timing of the public intervention a
crucial factor in determining its economic
efficiency. If the subsidy is not prompt enough, the
growth in prevalence will have already induced
protection.
Economic Epidemiology
A growing literature examines the optimal con-
trol of infectious diseases in the presence of anti-
) biotic resistance (see, for example, Laxminarayan
and Brown ; Laxminarayan and Weitzman ;
Laxminarayan ; and Horowitz and Moehring ). The
standard, positive external effect of treating more
individuals with an infectious disease is partly or
fully offset by the negative external effect induced
by increased antibiotic resistance. The R&D
problem induced by external consumption effects
such as antibiotic resistance is discussed in
Philipson et al. ( ).
Economic epidemiology has also considered the
welfare losses induced by disease, the welfare
effects of R&D in developing new methods of
prevention and treatment (Philipson ), and how
these contrast with cost-of-illness studies of disease
burden.
Spread of Economic Epidemiology
). to Other Fields
Several other topics have grown out of this more
systematic analysis of infectious disease by econ-
omists. One strand is the analysis of public health-
related issues such as obesity (Philipson and
Posner ; Lakdawalla et al. ). The addictive aspect of
obesity is analysed by Cawley ( ). Empirical
studies explaining the observed
) growth in obesity, whether it includes a rise in
caloric intake or fall in caloric expenditure, include
Cutler et al. ( ). Chou et al. ( )
and Rashad and Grossman ( ) analyse the
co-variation between the growth of obesity and
smoking and fast-food establishments in the United
States. The important and rich set of issues raised
by growth in obesity promises a usefiil role for
economic analysis.
Another area in which economic analysis of
epidemiological issues has emerged is the economic
analysis of clinical trials (see, for example,
Philipson and DeSimone ; Philipson and Hedges ;
Malani ). This literature deals with the non-
traditional aspects of programme evaluation that are
unique to clinical trials - for example, the blinding
of subjects. Economic
Economic Epidemiology
analysis of clinical trials differs from bio-statistical
analysis in that subjects are assumed to act in their
best interest rather than be passively observed.
The stark difference between economic expla-
nations of disease occurrence on the one hand and
the evaluation of public interventions aimed at
limiting disease on the other implies that economics
may have a very useful role to play in under-
standing these issues.
See Also
► Tealf:.
Bibliography
Ahituv, A., J. Hotz, and T. Philipson. 1996. Is AIDS self-
limiting? Evidence on the prevalence elasticity of the
demand for condoms. Journal of Human Resources 31:
869-898.
Auld, M.C. 2003. Choices, beliefs, and infectious disease
dynamics. Journal of Health Economics 22: 361-377.
Auld, M.C. 2006. Estimating behavioral response to the
AIDS epidemic. Contributions to Economic Analysis and
Policy, 5(1), Article 1.
Barrett, S. 2003. Global disease eradication. Journal of the
European Economic Association 1: 591-600.
Barrett, S., and M. Hoel. 2004. Optimal disease eradication.
Working Paper No. 604, FEEM.
Brito, D., E. Sheshinski, and M. Intriligator. 1991. Exter-
nalities and compulsory vaccinations. Journal of Public
Economics 45: 69-90.
Cawley, J. 1999. Obesity and addiction. Ph.D. dissertation,
Department of Economics, University of Chicago.
Chou, S.Y., M. Grossman, and H. Sailer. 2004. An economic
analysis of adult obesity, results from the behavioral risk
factor surveillance system. Journal of health Economics
23: 565-587.
Cutler, D.M., E.L. Glaeser, and J.M. Shapiro. 2003. Why
have Americans become more obese? Journal of Eco-
nomic Perspectives 17(3): 93-118.
Geoffard, P.Y., and T. Philipson. 1996. Rational epidemics
and their public control. International Economic Review
37: 603-624.
Gersovitz, M., and J.S. Hammer. 2003. Infectious diseases,
public policy, and the marriage of economics and epi-
demiology. The World Bank Research Observer 18: 129-
157.
Gersovitz, M., and J.S. Hammer. 2004. The economical
control of infectious diseases. Economic Journal 114: 1-
27.
3285
Gersovitz, M., and J.S. Hammer. 2005. Tax/subsidy policies
toward vector-borne infectious diseases. Journal of
Public Economics 89: 647-674.
Grossman, M., and I. Rashad. 2004. The economics of
obesity. Public Interest 156: 104-112.
Horowitz, B.J., and B.H. Moehring. 2004. How property
rights and patents affect antibiotic resistance. Health
Economics 13: 575-583.
Kremer, M. 1996. Integrating behavioral choice into epi-
demiological models of the AIDS epidemic. Quarterly
Journal of Economics 111: 549-573.
Lakdawalla, D., T. Philipson, and J. Bhattacharya. 2005.
Welfare enhancing technological change and the growth
of obesity. American Economic Review 95: 253-258.
Lakdawalla, D., N. Sood, and D. Goldman. 2006. HIV
breakthroughs and risky sexual behavior. Quarterly
Journal of Economics 121: 1063-1102.
Laxminarayan, R. 2002. How broad should the scope of
antibiotics patents be? American Journal of Agricultural
Economics 84: 1287-1292.
Laxminarayan, R., and G.M. Brown. 2001. Economics of
antibiotics resistance: A theory of optimal use. Journal of
Environmental Economics and Management 42: 183-206.
Laxminarayan, R., and M.L. Weitzman. 2002. On the
implications of endogenous resistance to medications.
Journal of Health Economics 21: 709-718.
Malani, A. 2006. Identifying placebo effects with data from
clinical trials. Journal of Political Economy 114: 236-256.
Mechoulan, S. 2004. HIV testing, a Trojan horse? Topics in
Economic Analysis and Policy, 4(1), article 18.
Oster, E. 2006. HIV and sexual behavior change: Why not
Africa? Working paper, Graduate School of Business,
University of Chicago.
Philipson, T. 1995. The welfare loss of disease and the theory
of taxation. Journal of Health Economics 14: 387-396.
Philipson, T. 2000. Economic epidemiology and infectious
disease. In Handbook of Health Economics, ed. T. Culyer
and J. Newhouse. Amsterdam: North- Holland.
Philipson, T., and J. DeSimone. 1997. Experiments and
subject sampling. Biometrika 84: 618-632.
Philipson, T., and L. Hedges. 1998. Subject evaluation in
social experiments. Econometrica 66: 381-409.
Philipson, T., S. Mechoulan, and A.B.. Jena. 2006. IP and
external consumption effects: Generalizations from health
care markets. Working Paper No. 11930. Cambridge,
MA: NBER.
Philipson, T., and R.A. Posner. 1993. Private choices and
public health, an economic interpretation of the AIDS
epidemic. Cambridge, MA: Harvard University Press.
Philipson, T., and R.A. Posner. 2003. The long run growth of
obesity as a function of technological change. Per-
spectives in Biologv and Medicine 46(3): 87-108.
Rashad, I., and M. Grossman. 2004. The economics of
obesity. Public Interest 156: 104-112.
3286
Economic Freedom
Alan Peacock
Economic freedom describes a particular condition
in which the individual finds himself as a result of
certain characteristics in his economic environment.
Taking a simple formulation of decision-making in
which it is assumed that the individual maximizes
his satisfaction both as a consumer of private and
government goods and services and as a supplier of
factor services, his position may be depicted as
follows:
Ma xU‘ = Ui(xi,qk,ai)
(1)
Subject to
pck ■ d + T\ = Y‘ = p (a‘) = pi ■ d ( 2 )
where .r, is a vector of ‘private goods’, qk is a vector
of goods supplied by government, a' is a vector of
factor inputs, pk is a vector of product prices for
private goods, pi is a vector of factor prices, Tk is
net tax liability of individual 1 (tax obligation less
transfers), Y is personal income before tax of
individual' and subscript k denotes an exogenously
determined variable.
Assuming the budget constraint ( ) is exactly
satisfied, the individual maximizes his satisfaction
solving for the vector of private-goods consumption
in terms of their prices, disposable income and
predetermined levels of public goods available for
consumption, where goods prices and factor prices,
quantities of factor inputs and tax liabilities are
either known or predicted by the individual.
Economic freedom requires that the various
terms in the budget constraint reflect the absence of
‘preference or restraint’ (Adam Smith) on the
individual. Therefore pk is a vector of product prices
which result from the operation of competitive
market forces with the individual being free to
choose between alternatives. Similarly,/?)? must
Economic Freedom
be characterized by competition in the factor market
with the individual being ‘free to bring both his
industry and capital into competition with those of
any other man or order or men’ (Adam Smith).
There is less certainty concerning the constraints
placed on Tk and qk. Some writers would argue that
economic freedom requires a pre-established limit
on the values of Tk and qk either expressed or
implied in a country’s constitution (Nozick’s
‘minimal state’; see Nozick ). Others would argue
that within a system of democratic government it
should be possible to devise voting systems through
which individuals express their preferences for
values of Tk and qk which simulate if they do not
replicate the competitive market in the private
sector (see Buchanan ). All agree, however, that
economic freedom is not compatible with large
values of T and q in relation to values of .r', mainly
because a large public sector increases the
monopoly power of public servants both as
suppliers of public goods and factor services to
produce them and encourages the growth of private
monopolies as a defence against public monopsony
buying.
Economic Freedom and Libertarian
Philosophy
There are features of this attempt at a ‘technical’
definition which may be called in question and
which must be considered later, but it will be
recognizable to those economists who have elevated
economic freedom to an important goal in its own
right and have claimed that it is the most important
means for ensuring that the economy develops at
the right ‘tempo’. Discussion of the usefulness of
the concept of economic freedom, therefore, centres
in these two libertarian propositions.
The first proposition is contained in a striking
passage in Book III of his Essay on Liberty. J.S.
Mill wrote:
He who lets the world, or his own portion of it, choose
his plan of life for him, has no need of any other faculty
than the ape-like one of imitation. He who chooses to
plan for himself, employs all his faculties. He must use
observation to see, reasoning
Economic Freedom
and judgment to foresee, activity to gain materials
for decision, discrimination to decide and, when he
has decided, firmness and self-control to hold to his
deliberate decision... It is possible that he might be
guided on some good path, and kept out of harm’s
way, without any of these things. But what will be
his comparative worth as a human being? It really is
of importance, not only what men do, but also what
manner of men they are that do it. (Mill )
The passage captures the essence of the liber-
tarian view of the good society, clearly implying
that it requires that individuals should accept the
necessity for choosing and for recognizing their
responsibility for making choices. It must simul-
taneously require that, to develop the capacity for
choosing, individuals must have the widest possible
freedom of choice in the acquisition and disposal of
resources. Two further conclusions follow.
The only restriction on economic freedom
experienced by the individual should be when such
freedom harms others.
The individual is not accountable to society for
his actions and this, together with the different and
changing preferences of individuals, makes liber-
tarians distance themselves from attempts to
establish a ‘social welfare function’ (cf. Rowley and
Peacock ).
The second proposition maintains that economic
freedom brings the added bonus of promoting the
economic welfare of both the individual and of
society. Economic freedom encourages the
individual to ‘better his condition’ (Smith ) by
exploiting opportunities for specialization and gains
from trade which will be fully realized through the
spontaneous emergence of markets. Not only is
economic freedom regarded as the only material
condition compatible with human dignity but it is
also a necessary condition for the economic growth
of the economy and for its adjustment to the
changing preference structures of its members in
response to market forces. The market is a
‘discovery process’ (Hayek ) in which participants
adjust to change giving rise to the notion of the
‘invisible hand’ which coordinates human
economic actions automatically without recourse to
government intervention. Pace Hahn ( ) much discussion to the possibilities of‘privatizing’
and others, libertarians do not
attach importance to a general equilibrium
3287
solution, attained by the operation of competitive
market forces (cf. Barry ). Indeed, though some
exceptions will be noted below, it is claimed by
supporters of the doctrine of economic freedom that
disturbance of the natural process of exchange by
government intervention assumes knowledge of the
intricacies of the economy which is vouchsafed to
no one, but there is no guarantee that officials, who
maximize their private interests like everyone else,
would be willing to maximize some social optimum
even if they knew how to do so.
It was clearly recognized, by Hume and Smith
for example, that for markets to work efficiently
there must be a well-defined system of property
rights and that costs of contracting between indi-
viduals in order to benefit from gains-from-trade
would need to be minimized. The promotion of
market efficiency was therefore bound to require
some government intervention. No specialization or
gains-from-trade would take place in a society in
which there was no machinery for settling disputes
and for preserving law and order. Acceptance of
coercive intervention, however, requires that the
‘rule of law’ prevails. The law must be prospective
and never retrospective in its operation, the law
must be known and, as far as possible, certain, and
the law must apply with equal force to all
individuals without exception or discrimination.
The state could also have a role in reducing the
costs of contracting both by the removal of barriers
to trade and to factor mobility and by the positive
encouragement to the reduction in the costs of
transport. In this latter respect Adam Smith
supported reduction in the ‘expense of carriage’ by
state financing of road building and supervision of
financial methods to promote road maintenance and
improvement.
At no stage therefore in the development of the
doctrine of economic freedom, as understood by
economists, was it regarded as synonymous with
‘laissez-faire’. At the same time, the role of the state
in respect of the promotion of economic freedom
was and has remained strictly limited in libertarian
thinking. Indeed, some modem libertarians devote
even such traditional functions of the state as the
maintenance of law and order.
3288
Some Problems Raised by the Concept
of Economic Freedom
The most obvious question posed to libertarians by
those who are sceptical of their position is that the
system of economic freedom is silent on the
question of the distribution of property rights. In
terms of our simple model, what principle should
determine the values of Y.......... Y......... Y1 which,
when aggregated, would describe some initial
distribution of income as measured, say, by the
shape of the Lorenz Curve? What reason have we
for supposing that the 'optimal' distribution of
income would emerge from the process of eco-
nomic exchange between individuals?
The answer to this question does not find lib-
ertarians speaking with one voice. The problem is
not one of principle, for the ultimate test to them is
how far any government intervention represents a
restriction of freedom. The problem is one of
interpretation. It would be difficult today to find
libertarians who would object to government
intervention designed to assure protection to those
who are severely deprived. Thus Hayek has argued
that so long as ‘a uniform minimum income is
provided outside the market to all those who, for
any reason, are unable to earn in the market an
adequate maintenance, this need not lead to a
restriction of freedom, or conflict with the Rule of
Law’. This still leaves room for much disagreement
among libertarians as to the precise level of the
minimum and how to decide on who is entitled to
receive it. Some supporters of the libertarian
position, including the present author, would go
much further and argue, along with J.S. Mill, that
concentrations of wealth sustained over lengthy
time periods can endanger economic freedom, not
to speak of political freedom, by the association of
such concentrations with the concentration of power
of wealthy individuals over the less fortunate.
If the concept of economic freedom cannot
embrace some precise guidance about the extent to
which economic exchanges should be interfered
with, it certainly places limits on the form of that
interference. Thus libertarians, to the extent that
they accept the need for a state-guaranteed
minimum standard of living, prefer the use of
Economic Freedom
money transfers to individuals rather than the
provision of social services below or at zero cost,
that is to say the economic condition of individuals
in receipt of state support should be reflected in
reduction in Vk (whose value may have to be
negative) rather than an increase in qk. Thus it is
argued that individuals then retain responsibility for
the purchase of goods and services designed to
promote their own welfare and that the power of the
state over the individual by bureaucratic
dictatorship of preferences and by the lack of
incentives in the public sector to economize in
resource use is circumscribed.
A more severe test for the practicality of liber-
tarian measures, designed to permit some redistri-
bution without increasing the power of the state,
arises in the case of any attack on the concentration
of wealth. Clearly, a system of inheritance taxation
which results in the transfer of capital from the
private to the public sector would not conform to
libertarian thinking, not only because this would
discourage private saving but also because it would
build up the power of the state. A system of taxation
would have to be devised which not only did not
discourage accumulation of private capital but also
simultaneously encouraged legators to disseminate
capital in favour of those with little capital. It is a
long time since libertarians have plucked up the
courage to try to develop such a system, given that
eminent public finance specialists have failed in
their attempts to fulfil these requirements.
The second major question arises from the
persistent objection of Marxists and other Socialist
writers that the system of economic freedom, as
depicted by the libertarians, fails to solve the
problem of ‘worker alienation’. It may be that the
system of economic freedom can allow employees
alone or in combination with others to influence the
price of factor inputs (pf) and the work/leisure
combination (a1), variables which play a crucial part
in individual welfare. The fact remains that the
system of property rights, which libertarians
support, includes the individual ownership of
capital and the use of capitalistic methods of
production which imply an authority relationship
between employer and worker. The hierarchical
order at the place ofwork
Economic Governance
seems at complete variance with the independence
of economic action attributed to the individual by
the supporters of economic freedom.
Reactions to this argument by libertarians are
sometimes reminiscent of the Scots preacher who,
on recognizing a theological difficulty in his ser-
mon, recommended his congregation to look the
difficulty squarely in the face and pass it by.
However, even Socialist writers, notably the
prominent Marxist Ota Sik ( ), have recog
nized that the alternative to market capitalism -
collectivist production - does not solve the problem
for it is not synonymous with democratization at the
shop-floor level. In other words, the basis of
alienation is technological and not institutional.
Some libertarians, notably Mill, have made
common cause with Socialists by arguing that
alienation must not be taken to be an inevitable
consequence of productive activity. Mill sought one
solution in the encouragement of firms owned and
managed by the labour force, but still subject to
competition. Utopian Socialists have claimed that
the only solution is to reject altogether the
technology which imposes hierarchical relations in
the first place. Both ‘solutions’ are still the subject
of living debate in both the professional and
political arena.
See Also
► Economic Harmony
► t" interest
► Jtilitar.ia.nic.:
Bibliography
Barry, N.P. 1985. In defense of the invisible hand. The Cato
Journal 5(1): 133-148.
Buchanan, J.M. 1975. The limits of liberty: Between Anarchy
and Leviathan. Chicago/London: University of Chicago
Press.
Hahn, F. 1982. Reflections on the invisible hand. Lloyds Bank
Review 144: 1-21.
Hayek, F.A. 1979. Law, legislation and liberty, vol. 3. London:
Routledge & Kegan Paul.
Mill, J.S. 1859. Essay on liberty. Oxford: Oxford University
Press, 1942.
Mill, J.S. 1871. Principles of political economy, Book IV, eh.
VII. Toronto: University of Toronto Press, 1965.
3289
Nozick, R. 1974. Anarchy, state and utopia. Oxford: Basil
Blackwell.
Peacock, A. 1979. The economic analysis of government and
related themes. Oxford: Martin Robertson, chs. 5 and 6.
Rowley, C., and A. Peacock. 1975. Welfare economics: A
liberal re-interpretation. London: Martin Robertson.
Sik, O. 1974. The shortcomings of the Soviet economy as
seen in Communist ideologies. Government and Oppo-
sition 9(3): 263-276.
Smith, A. 1776. An inquiry into the nature and causes of the
wealth of nations, Book IV, ch. 9 and Book V. Ed. R.H.
Campbell and A.J. Skinner. Oxford: Clarendon Press,
1976.
Economic Governance
Avinash K. Dixit
Abstract
Economic governance consists of the processes
that support economic activity and economic
transactions by protecting property rights,
enforcing contracts, and taking collective action
to provide appropriate physical and
organizational infrastructure. These processes
are carried out within institutions, formal and
informal. The field of economic governance
studies and compares the performance of dif-
ferent institutions under different conditions,
the evolution of these institutions, and the
transitions from one set of institutions to
another.
Keywords
Accountability; Arbitration; Authoritarianism and
economic development; Better Business Bureaus
(USA); Coase, R.; Colonization; Commitment;
Common law vs civil law; Common-pool resoiuce
management; Corporate governance; Corruption;
Corruption control; Credit rating agencies;
Democracy and economic development; Democracy:
parliamentary vs presidential; Discriminating align-
ment hypothesis; Economic governance; Electoral
rales; Ethnic trading networks; Evolution of
institutions; Exploitation; Formal vs
3290
informal (private) contract enforcement; For-
profit private institutions; Governance; Growth
and governance; Incentive compatibility;
Inequality; Information verifiability; Institutions
and transition; International contracts; Mafia;
Market institutions; Monitoring mechanisms;
Opportunism; Political stability; Prisoner’s
Dilemma; Property rights enforcement;
Regulatory burden; Relation-based vs rule-
based governance; Reputation mechanisms;
Rule of law; Slavery; Social networks; Social
norms; Subgame perfection; Title registration;
Transaction cost economics; Trust; Voice;
World Bank; World Trade Organization;
Yakuza
JEL Classifications
D02; HI 1; K14; P26; P51
Formal and informal institutions arise and evolve to
underpin economic activity and exchange by
protecting property rights, enforcing contracts, and
collectively providing physical and organizational
infrastructure. The field of economic governance
studies and compares these institutions: state
politico-legal institutions, private ordering within
the law (credible contracting, arbitration), for-profit
governance (credit-rating agencies, organized
crime), and social networks and norms. Private
institutions can outperform the state’s legal system
in obtaining and interpreting relevant information,
and imposing social sanctions on the violators of
norms. But private institutions are often limited in
size; as economic activity expands, a transition
towards more formal institutions is usually
observed.
Concepts and Taxonomies
The term ‘governance’ has exploded from obscurity
to ubiquity in economics since the 1970s. A search
of the EconLit database shows clear evidence of
this explosion. In the relevant categories (title,
keywords and abstracts), there are just five
occurrences of the word from 1970 to 1979. The
number jumps to 112 for the 1980s and 3,825
Economic Governance
for the 1990s. Since 2000 to the time of this writing
(December 2005), there are already 7,948.
The Oxford English Dictionary gives several
definitions of the word ‘governance’: (a) the action
or manner of governing; controlling, directing, or
regulating influence; control, sway, mastery; the
state of being governed; good order;
(b) the office, function, or power of governing;
authority or permission to govern; that which
governs; (c) the manner in which something is
governed or regulated; method of management,
system of regulations; a rale of practice, a disci-
pline; and (d) the conduct of life or business; mode
of living; behaviour, demeanour; discreet or
virtuous behaviour; wise self-command. These
diverse meanings allow the word to be used (and
sometimes misused) for almost any context of
economic decision-making or policy.
Two areas of application merit special mention.
One is corporate governance. This analyses the
internal management of a corporation - organiza-
tional structure and the design of incentives for
managers and workers - and the rales and pro-
cedures by which the corporation deals with its
shareholders and other stakeholders.
The second is economic governance;
Williamson ( ) expresses its theme as the
‘study of good order and workable arrangements’.
This includes the institutions and organizations that
underpin economic transactions by protecting
property rights, enforcing contracts, and organizing
collective action to provide the infrastructure of
rales, regulations, and information that are needed
to lend feasibility or workability to the interactions
among different economic actors, individual and
corporate. Different economies at different times
have used different institutions to perform these
functions, with different degrees of success. The
field of economic governance studies and compares
these different institutions. It includes theoretical
models and empirical and case studies of the
performance of different institutions under different
circumstances, of how they relate to each other, of
how they evolve over time, and of whether and how
transitions from one to another occur as the nature
and scope of economic activity and its institutional
requirements change.
Economic Governance
Corporate governance and economic governance
are connected because the boundary of a
corporation is itself endogenous, determined by the
same considerations of information and com-
mitment costs that raise problems of internal orga-
nization as well as those of property and contract
(Coase ). Specifically, the nature of transaction
costs may make it more efficient to handle some
problems of governance by merging the two parties,
for example by vertical integration (Williamson ,
). But it is analytically
convenient to separate the two. This article concerns
economic governance. To avoid constant repetition,
I will simply call it ‘governance’ here unless some
explicit reference to corporate governance is
relevant.
Governance was neglected by economists for a
long time, perhaps because they expected the
government to provide it efficiently. However,
experience with less developed and reforming
economies, and observations from economic
history, have led economists to study non-
governmental institutions of governance.
Governance is not a field per se; it is an orga-
nizing or encompassing concept that bears on issues
in many fields, including institutions and
organizational behaviour, economic development
and growth, industrial organization, law and eco-
nomics, political economy, comparative economic
systems, and various subfields of these.
We can organize the subject by classifying insti-
tutions along different dimensions. As is usual with
such taxonomies, these are conceptual categories to
help organize our thinking and analysis. In reality,
there are significant differences within each cate-
gory and overlaps across categories.
The first dimension concerns the purpose of the
institution. The categories are: (a) protection of
property rights against theft by other individuals
and usurpation by the state itself or its agents, (b)
enforcement of voluntary contracts among indi-
viduals, and (c) provision of the physical and
regulatory infrastructure to facilitate economic
activity and the functioning of the first two cate-
gories of institutions. We might also consider a
fourth category, namely, the deep institutions that
are essential to avoid serious cleavages or alienation
that threaten the cohesion of the society
3291
itself. But this has not been studied in this context
so far.
The second dimension concerns the nature of
the institution. The categories are: (a) the formal
state institutions that enact and enforce the laws,
including the legislature, police, judiciary and
regulatory, agencies, (b) institutions of private
ordering that function under the umbrella of state
law, for example various forums for arbitration,
(c) private for-profit institutions that provide
information and enforcement, and (cl) self-
enforcement within social or ethnic groups and
network. My discussion is organized in sections
along this dimension.
A third dimension distinguishes institutions that
arise and evolve organically from those that are
designed purposively; self-enforcing groups are
often organic while the first three categories in the
second dimension usually require some measure of
design. This matters for the evolution of institutions
of governance (see Greif especially ch. 6;
Williamson p. 1).
Formal Institutions of the State
There is broad agreement that the qualify of insti-
tutions of governance significantly affects economic
outcomes. The importance of protecting property
rights, both from other individuals and from
predation by the state itself, is generally recognized
and documented (for example, De Soto ). But
serious disputes about the precise measures of
qualify of institutions, and about many details of the
causal mechanisms by which they affect economic
outcomes, remain.
At the broadest level, the distinction is between
democracy and authoritarianism, each of which
comes in many different varieties. Democracy has
many normative virtues, but its worth in governance
is less clear. Barro ( , p. 61) finds an
inverse U-shaped relationship between economic
growth and a continuous measure of democracy -
‘more democracy raises growth when political
freedoms are weak, but depresses growth when a
moderate amount of freedom is already established’
- but the fit is relatively poor. Persson ( ),
using cross-sectional as well as
3292
panel data, finds that the crude distinction between
democratic and non-democratic forms of
government is not enough. The precise form of
democracy matters for policy design and economic
outcomes: ‘parliamentary, proportional, and
permanent democracies seem to foster the adoption
of more growth-promoting structural policies,
whereas ... presidential, majoritarian, and temporary
democracy do not’ (Persson , p. 22). However,
Keefer ( , p. 10),
after surveying a wide-ranging literature on elec-
toral rules and legislative organizations, concludes
that they affect policies but are not a crucial
determinant of success: ‘electoral rales ... almost
surely do not explain why some countries grow and
others do not’, and ‘the mere fact that developing
countries are more likely to have presidential forms
of government is unlikely to be a key factor to
explain slow development. ’
Democracy can be important for governance
because its reliance on rales and procedures pro-
vides citizens with protection against predation by
the state or its agents. Indeed, the elite, which might
otherwise prefer to rale unconstrained, may find it
in its own interest to make a credible commitment
not to steal from the population by creating and
fostering democracy (Acemoglu ; Acemoglu and
Robinson ). Greif et al. ( ) discuss how groups of
traders
(guilds) in late medieval Europe took collective
action to counter rulers’ incentives to violate their
members’ property rights.
Even in a democracy, agents of the state may
pursue their private interests using corruption,
complex regulations to extract rent, and favouritism.
In fact, an emerging literature argues that economic
growth, at least in its early stages, is better
promoted under suitably authoritarian regimes.
Glaeser et al. ( ) argue that less
developed countries that achieve economic success
do so by pursuing good policies, often under
dictatorships, and only then do they democratize.
While these conclusions are controversial, these
authors’ criticisms of the measures of institutions
used in the research that argues for the primacy of
institutions in general, and of democracy in par-
ticular, are telling. Giavazzi and Tabellini (
find a positive feedback between economic and
Economic Governance
political reform, but they also find that the sequence
of reforms matters, and countries that implement
economic liberalization first and then democratize
do much better in most dimensions than those that
follow the opposite route. In practice, of course, it is
difficult to ensure ex ante that an authoritarian ruler
will implement good governance.
Many different measures of institutional quality
exist. World Bank researchers Kaufman et al. ( ,
which contains citations to their ear
lier work) have constructed six: (a) Voice and
Accountability - measuring political, civil and
human rights; (b) Political Instability and Violence -
measuring the likelihood of violent threats to, or
changes in, government, including terrorism; (c)
Government Effectiveness - measuring the
competence of the bureaucracy and the quality of
public service delivery; (d) Regulatory Burden -
measuring the incidence of market-unfriendly
policies; (e) Rule of Law - measuring the quality of
contract enforcement, the police, and the courts, as
well as the likelihood of crime and violence; and (/)
Control of Corruption - measuring the exercise of
public power for private gain, including both petty
and grand corruption and state capture. Of these,
(e), (f) and also (b) concern the most basic
institutions for protection of property rights and
enforcement of contracts, (a) relates to governance
because voice and accountability can reduce the
severity of the agency problem between the citizens
and the agencies of the state, and (c) and (d) pertain
to what I called provision of the infrastructure of
governance. Conceptually they are a mixed bag; the
quality of some of them can itself depend on the
quality of other more basic ones, and some are
closer to being measures of effects than of causes.
Their method of construction relies on subjective
perceptions, and is subject to error. But when used
with caution, they have proved significant as
explanatory variables in empirical studies of eco-
nomic growth, and for observing changes in gov-
ernance quality over time in specific countries.
Corruption and regulatory burdens are major themes
of the World Bank’s research on governance in
many countries (see World Bank Institute, website).
)
Economic Governance
Empirical estimations of the level or growth of
GDP on various measures of institutional quality
confront many conceptual and econometric prob-
lems. Researchers have tackled the issue of reverse
causation by using various instruments, such as the
nationality of colonizers (Hall and Jones ),
mortality among colonizers
(Acemoglu et al. ), and whether a colony had rich
mineral resources or climatic and soil conditions
conducive to plantation agriculture and a large or
dense native population, or was sparsely populated
and poor in the 1500 s (Engerman and Sokoloff ;
Acemoglu et al. ). The general idea is that in the
former circumstances the European colonizers
established institutions of slavery and inequality to
facilitate the exploitation of labour on a large scale,
whereas in the latter conditions, where the
colonizers had to exert their own effort, their
institutions provided the correct incentives and
became conducive to longer-term economic suc-
cess. The debate on the factual and econometric
validity and the economic interpretation of these
findings is fierce and continuing; Hoff (
surveys and discusses this literature in detail.
La Porta et al. ( , ) contrast different
legal traditions for protecting the rights of small
shareholders. If such protection is poor, that will
inhibit the flows of capital to its most efficient uses.
They find that systems based on common law are
better in this regard than those based on civil law.
But Rajan and Zingales ( ) and
Lamoreaux and Rosenthal ( ) argue that in
practice there was little difference between the
systems during critical periods of industrialization.
These debates are sure to continue, and this
section will get out of date very quickly.
At the international level, formal governance
works through bodies like the World Trade Orga-
nization. Their members are sovereign countries;
therefore their procedures must be subject to self-
enforcement in repeated interactions, whether
through bilateral or multilateral sanctions. These
institutions are therefore basically similar to the
social networks discussed below. See Maggi ( ) and
Bagwell and Staiger ( ) for
detailed analyses.
3293
Private Institutions
The policing functions for property right protection
supplied by the state are often supplemented by
private security systems that serve specific clients
and purposes - firms employ or hire security
personnel, gated communities and neighbourhoods
have private (hired or volunteer) patrols. These
generally merely supplement the functions of the
police for their specific context and work
cooperatively with the police, but the two may clash
if the private security system goes beyond its
permissible functions.
Private institutions of contract enforcement
similarly coexist with formal law, and become
essential when the latter is weak or nonexistent.
Explicit or implicit private contractual arrangements
are also important for assignment of property rights
as a part of Coasean contracting for efficient
outcomes. Therefore, analyses of private institutions
often focus on the governance of contracts.
The basic problem of contract enforcement is
) control of opportunism. If one or both parties have
to make transaction-specific investments, the other
can attempt to secure a greater part of the benefit by
reneging or demanding renegotiation. The prospect
of this can jeopardize the potentially mutually
beneficial deal in the first place. Williamson ( ,
) pioneered the analysis
of this issue under the title of transaction cost
economics.
Information constitutes a major source of
advantage for private ordering over formal law.
Enforcement of a contract in a court requires
offering proof of misconduct by the other party in
the dispute; the relevant information must be
verifiable to outsiders. Therefore, formal contracts
can stipulate actions by the parties conditional only
on verifiable information. Other or more detailed
information may be observable to the parties
themselves, or can be inferred by specialist insiders
to the industry, but cannot be verified to non-
specialist judges or juries of the state’s legal system,
or can be verified only at excessive cost.
The informational advantage of private ordering
may be offset by a disadvantage in enforcement.
Informal arrangements must be made to
3294
overcome each participant’s temptation to behave
opportunistically at the others’ expense. Different
methods of this kind underlie the various institu-
tions of informal governance, and achieve different
degrees of success. Some are able to exert coercion
for immediate punishment of misbehaviour. Others
create long-run costs, typically in the form of
exclusion from future participation or worse future
opportunities, to offset the short- run advantages of
opportunism. This is the standard theory of self-
enforcing cooperation in repeated Prisoner’s
Dilemmas. The following sections discuss some of
these alternatives.
Private Ordering with Formal Law
in the Background
Perhaps the most remarkable thing about formal
legal institutions and mechanisms for the enforce-
ment of commercial contracts is how rarely they are
actually used. Business transactions often do have
underlying formal contracts, but when disputes arise
recourse to the law is often the last resort. Other
private alternatives are tried first; these include
bilateral negotiation, arbitration by industry experts,
and so on. Filing a suit in a formal court of law
often signals the end of a business relationship.
Most actual practice in business contracting is
therefore better characterized as ‘private ordering
under the shadow of the law’ (Macaulay ;
Williamson
pp. 95-100, 121-2).
If one of the parties to an ongoing informal
relationship behaves opportunistically, the most
common alternative is to fall back on a formal
contract based on verifiable contingencies alone.
Suppose an outcome based on a tacit understanding
of what each party should do in any one exchange
(including good-faith negotiation to adapt to
changing circumstances) yields both of them higher
payoffs than does a formal contract. Consider the
implicit arrangement where, if one party deviates
from the agreed course of action to its own
advantage and to the detriment of the other, their
future exchanges will be governed by the formal
contract. This yields a subgame-perfect (credible)
equilibrium of the repeated game if
Economic Governance
each party’s one-time gain from opportunism does
not exceed the capitalized value of the future
difference of payoffs between the tacit and the
formal contracts. Williamson ( , p. 2)
expresses this well: ‘continuity can be put in jeop-
ardy by defecting from the spirit of cooperation and
reverting to the letter.’
When such relationship-based implicit
contracting prevails, partial improvement in the
formal system can worsen the outcome, due to a
problem of the second-best. The partial improve-
ment raises the payoffs the two parties could get
from the fallback formal contract. This in turn
reduces the future cost of a current deviation from
the implicit contract or spirit of cooperation. It
tightens the incentive-compatibility constraints, and
therefore worsens what can be achieved by
relational contracting (Baker et al. ; Dixit , ch. 2).
Arbitration comes in two prominent forms. One
is industry-specific, based on expert knowledge of
insiders. More information is verifiable in such
settings; therefore richer contracts specifying
actions for more detailed contingencies become
feasible. In many industries there is a large
common-knowledge basis of custom and practice,
which may even make it unnecessary to write down
a contingent contract in great detail. Arbitration can
also provide an opportunity for the parties to
communicate and renegotiate adaptations to new
circumstances. Formal legal systems often
recognize these advantages of expert arbitration,
and courts stand ready to enforce the decisions of
arbitrators if the losing party tries to evade the
sanctions. However, industry arbitrators often have
severe sanctions at their own disposal; they can
essentially drive the miscreant out of business, and
even ostracize him or her from the social group of
that business community. Examples of arbitration
institutions include Bernstein’s ( )
classic study of the diamond industry. For further
discussion and modelling, see Dixit ( , ch. 2)
and Williamson ( , p. 14).
The other prominent forums of arbitration deal with
international contracts (Dezalay and Garth ;
Mattli ). There are several of these, specializing in
different legal traditions. They lack direct power to
enforce their decisions, but
Economic Governance
are backed by treaties that ensure enforcement by
national courts. These forums do not have industry-
specific knowledge, their processes can be slow and
costly, and their decisions can be somewhat
arbitrary. But parties in transnational transactions
may prefer them to either country’s courts,
suspecting that these will be biased in favour of
their own nationals.
For-Profit Private Institutions
If the state is unwilling to protect certain kinds of
property or enforce certain kinds of contracts (for
example in illegal activities), or is unable to do so
(for example in weak and failing states), or is itself
predatory, then private institutions can emerge to
perform these functions for a profit. Organized
crime often fills the niches uncovered by the state.
Gambetta ( ), Bandiera ( ) and
others argue that the Mafia emerged in just such a
situation to fill the vacuum of protection in late
19th-century Sicily. Landowners began to hire
guards of former feudal lords, and even the toughest
among bandits, to protect their property. Gambetta
describes how the Mafia’s role expanded to
providing contract enforcement in illegal or grey
markets. Similarly, the Japanese Yakuza was
instrumental in organizing markets at the end of the
Second World War in August and September 1945
when the Japanese state had collapsed (Dower , pp.
140-8), and mafias grew in Russia after the collapse
of the Soviet regime (Varese ).
Gambetta ( , p. 19) argues that this ‘busi
ness of protection’ is the core business of the Mafia.
It may engage in other activities using in-house
protection, but that is just downstream vertical
integration - the opposite of upstream integration
where an ordinary business firm has its in-house
security department. A transaction- cost analysis of
the internal organization of mafias, and of their
vertical integration decisions, may provide an
interesting link between economic governance and
corporate governance. Another dimension in which
the protection business can expand is extortion;
although private protectors may be welcome when
state protection
3295
has collapsed, ‘protectors, once enlisted, invariably
overstay their welcome’ (Gambetta p. 198).
The Mafia can provide contract enforcement
because, even though two traders may not have
sufficiently frequent dealings with each other to
achieve good outcomes in an ongoing bilateral
relationship, each trader can be a regular customer
of the enforcer. This converts multiple one-shot
Prisoner’s Dilemma games among the whole group
of traders into several bilateral repeated games of
each trader with the enforcer. The intermediary can
provide information (keeping track of previous
contract violations and informing a customer of the
history of a potential trading partner) and/or actual
punishment if a customer’s trading partner violates
their contract. The information role of the Mafioso
is similar to that of credit rating agencies and Better
Business Bureaus in the United States. Dixit ( , ch.
4)
constructs a model of such for-profit governance,
and establishes the conditions for an equilibrium
with for-profit private enforcement. These are lower
bounds on the shares of the surplus that the
customer and the Mafioso must have, so as to
overcome the trader’s temptation to cheat and the
Mafioso’ temptation to double-cross the customer.
Milgrom et al. ( ) have a related and
complementary model of private judges at medieval
European trade fairs. They specify the game of each
trade, and investigation in the event of cheating, in
greater detail, but do not examine the issue of the
judges’ honesty.
Group Enforcement Through Social
Networks and Norms
Any institution of contract enforcement must solve
three key problems: (a) detection of opportunistic
deviations from the contractually stipulated
behaviour, (b) preservation and dissemination of
information about the histories of the participants’
behaviour, and (c) inflicting appropriate
punishments to reduce future payoffs of any
deviators. The first is often constrained by the
available technology of monitoring, although
institutions and regulations such as reporting
3296
requirements and auditing can improve the tech-
nology. The second and third problems are best
resolved in bilateral ongoing relationships: each
party has a natural incentive to detect and remember
the other’s cheating, and can punish the other by
breaking off the relationship. However, governance
is often needed in groups each of whose members
interacts frequently with someone else in the group,
but not necessarily bilaterally with the same person
every time. Now remembering and transmitting
information about your current partner’s behaviour
to others, and refusing a potentially beneficial deal
because the counterparty has cheated someone else
in the past, are privately costly activities and
therefore require their own governance
mechanisms.
Formal state institutions of governance can solve
these problems by fiat; the legal system compels the
whole group of traders to commit to good behaviour
by subjecting themselves to detection and
punishment if they cheat. A third- party supplier of
information or enforcement serves similar
functions. In the case of a Mafia enforcer, anyone
who trades with a customer of the Mafioso subjects
himself to the grim punishment if he cheats. In the
case of a Better Business Bureau, a firm that joins
the organization thereby gives hostage to its own
good behaviour: if it misbehaves it will get a poor
rating or blacklisting. Transactions vary in their
characteristics; therefore we should expect the
effectiveness of such reputation mechanisms to vary
also, and should not expect universal success from
any one.
An institution of social networks and norms can
solve the problems of information and punishment
in a decentralized manner. Each participant can
transmit information about his or her current trading
partner’s behaviour to others in the group to whom
he or she is linked. And each can play his or her
assigned part in punishment, typically by refusing to
trade, if he or she gets matched with a potential
partner who is known to have misbehaved in past
dealings with others in the group. Incentives to
transmit information or refuse potentially good
trades can be established by a norm that regards
refusal to do so as itself a punishable offence, as in
Abreu’s ( ) people who have sufficiently good
penal codes for repeated games; see Calvert
Economic Governance
( , ). Extrinsic incentives may even be
unnecessary if people have sufficiently strong nat-
ural instincts to punish social cheaters, as found by
Fehr and Gachter ( ).
Numerous empirical and case studies of gov-
ernance based on social relations have been
conducted; space constraints allow mention of only
a few. Greif’s ( ) historical analysis of
Maghribi traders’ system of communication and
collective punishment is well known. So is
Ostrom’s ( ) synthesis of the evidence on
common-pool resource management; she empha-
sizes the importance of local knowledge and com-
munication, of appropriately designed (generally
graduated) punishments, and of incentives for
individuals to perform their assigned roles and
actions in the system. Fafchamps ( ) studies
and compares many different market institutions in
Africa; his work highlights the importance of
designing systems appropriate to the conditions of
each country or group. Ensminger ( )
describes a similarly rich complex of arrangements
for trade and employment relationships among the
Orma tribe of Kenya, and examines how formal
institutions of property right enforcement including
title registration can interact dysfunctionally with
traditional arrangements based on family and tribal
connections. Johnson et al. () present and analyse
findings from
survey research in former socialist economies. Of
particular interest are the links between evolving
formal and informal governance. Even without a
backup of courts, trust in bilateral relationships can
build quickly in response to good experiences. New
or transient customers are more likely to be offered
credit if courts work better, but the effectiveness of
courts becomes largely irrelevant for the
functioning of established relationships. Casella and
Rauch ( ) study the role of ethnic networks in inter
national trade.
Li ( ) points out a key difference between
the costs of operating such a system and those of
formal governance. A relation-based system of
networks and norms has low fixed costs, but high
and rising marginal costs. Trading on a small scale
naturally starts among the most closely connected
Economic Governance
communication and common understanding to
sustain honesty. No fixed costs need be incurred to
establish any formal rules or mechanisms of
enforcement. But as trade expands, potential part-
ners added at the margin are almost by definition
less well-connected, making it harder to commu-
nicate information with them and to ensure their
participation in any punishments. By contrast,
formal or rule-based governance has high fixed
costs of setting up the legal system and the infor-
mation mechanism, but once these are incurred,
marginal costs of dealing with strangers are low.
Therefore, relation-based governance is better for
small groups and rule-based governance better for
large groups. Greif’s ( ) comparison
between the relation-based system of Maghribi
traders and the formal institutions of Genoese
traders supports this theory. Dixit ( , ch. 3)
constructs a formal model that compares relation-
based and rule-based systems. This characterizes
the maximum size of a self- enforcing group, and
finds that, when the group exceeds this critical size,
the maximum scope of sustainable honesty shrinks
absolutely. The intuition is as follows. At the
critical size, each trader is indifferent between
honesty and cheating when dealing with the most
distant person. When more traders are added, this
weakens the communication between the previously
marginal person and other almost equally distant
ones, tipping the balance toward cheating.
Kranton ( ) models individuals who can
either choose bilateral long-lived self-enforcing
trading relationships or search for one-time trading
partners in an anonymous market with external
enforcement. The market thus provides the outside
opportunity in the repeated game of bilateral trade.
If more people trade in the anonymous market, it
becomes thicker and offers better prospects for
successful search. Then parties in bilateral
relationships have better outside opportunities,
which makes it harder to sustain tacit cooperation
there, further increasing the relative attraction of the
market. Therefore the system can have multiple
equilibria - no one uses the market because no one
else uses it, or everyone uses the market because
everyone else does - and can get locked into a
Pareto-inferior equilibrium.
3297
Evolution and Transformation
of Governance Institutions
A persistent theme in this survey has been that
different governance institutions are optimal for
different societies, for different kinds of economic
activity, and at different times. Changes in under-
lying technologies of production, exchange and
communication change the relative merits of dif-
ferent methods of governance. As the volume and
scope of trade expand, formal institutions generally
become superior to informal ones, but informal ones
serve useful roles under the shadow of formal ones
even in the most advanced economies and sectors.
All this raises the question of whether we should
expect institutions to adapt and evolve optimally.
Williamson’s famous ‘discriminating alignment
hypothesis’ says that transactions, with their
different attributes, align with institutions, with their
different costs and competencies; see his recent
exposition ( , p. 6). This gives
ground for optimism for synergistic evolution of the
need for governance and the institutions that supply
it. Others are less sanguine. North ( )
and others argue that institutional change is subject
to long delays due to resistance by organized
interests favouring the status quo, problems of
coordinating collective action to bring about a
discrete change in equilibrium, and so on. Dixit (
, pp. 79-85) discusses some of these prob
lems for transition from relation-based to rule-
based contract enforcement. Eggertson ( )
gives a dramatic example of how institutions
restricting fishing and requiring costly mutual
insurance persisted in Iceland for centuries after
they had become obstacles to good economic
performance.
I believe that a balanced approach is needed,
recognizing the tendency towards synergistic
alignment but also the obstacles to its realization.
The net outcome will depend on many specifics of
each context. Understanding and predicting the
process requires a combination of approaches: case-
based and analytical, inductive and deductive. Greif
( ) discusses, develops and applies
such methodologies using historical studies of trade
in medieval Europe.
3298
See Also
► .operation
► Zc:: orations
► Growth and Institutions
► Up ?ro
► n, Economic Analys e ..
► Law, Public Enforcement of
► Market Institutions
► Economics 58: 19-47.
► Cc ml. Norms
► Spontaneous Order
► Transition and Institutions
Acknowledgments I thank Tore Ellingsen, Diego Gam- betta,
Karla Hoff, Eva Meyersson-Milgrom, Dani Rodrik, Oliver
Williamson, and the editors for comments on previous drafts,
and the National Science Foundation for research support.
Bibliography
Abreu, D. 1986. Extremal equilibria of oligopolistic super-
games. Journal of Economic Theory 39: 191-225.
Acemoglu, D. 2003. Why not a political Coase Theorem?
Social conflict, commitment and politics. Journal of
Comparative Economics 31: 620-652.
Acemoglu, D., S. Johnson, and J. Robinson. 2001. The
colonial origins of comparative development: An
empirical investigation. American Economic Review 91:
1369-1401. "
Acemoglu, D., S. Johnson, and J. Robinson. 2002. Reversal of
fortune: Geography and institutions in the making of the
modem world income distribution. Quarterly Journal of
Economics 117: 1231-1294.
Acemoglu, D., and J. Robinson. 2005. Economic origins of
dictatorship and democracy. New York: Cambridge
University Press.
Bagwell, K., and R. Staiger. 2003. The economics of the
world trading system. Cambridge, MA: MIT Press.
Baker, G., R. Gibbons, and K. Murphy. 1994. Subjective
performance measures in optimal incentive contracts.
Quarterly Journal of Economics 109: 1125-1156.
Bandiera, O. 2003. Land refomr, the market for protection and
the origins of the Sicilian Mafia: Theory and evidence.
Journal of Law, Economics, and Organization 19:218-
244.
Barro, R. 1999. Determinants of economic growth: A cross-
country empirical study Cambridge, MA: MIT Press.
Bernstein, L. 1992. Opting out of the legal system: Extra-
legal contractual relations in the diamond industry. The
Journal of Legal Studies 21: 115-157.
Economic Governance
Calvert, R. 1995a. The rational choice theory of social
institutions: Cooperation, communication, and coordi-
nation. In Modern political economy: Old topics, new
directions, ed. J. Banks and E. Hanushek. Cambridge:
Cambridge University Press.
Calvert, R. 1995b. Rational actors, equilibrium, and social
institutions. In Explaining social institutions, ed. J.
Knight and I. Sened. Ann Arbor: University of Michigan
Press.
Casella, A., and J. Rauch. 2002. Anonymous market and
group ties in international trade. Journal of International
Coase, R. 1937. The nature of the firm. Economica 4: 386-
406.
De Soto, H. 2000. Mystery of capital: Why capitalism
triumphs in the west and fails everywhere else. New York:
Basic Books.
Dezalay, Y., and B. Garth. 1996. Dealing in virtue: Inter-
national commercial arbitration and the construction of a
transnational order. Chicago/London: University of
Chicago Press.
Dixit, A. 2004. Lawlessness and economics: Alternative
modes of governance. Princeton: Princeton University
Press.
Dower, J. 1999. Embracing defeat: Japan in the wake of
world war II. New York: W.W. Norton.
Eggertson, T. 2005. Imperfect institutions: Possibilities and
limits of reform. Ann Arbor: University of Michigan
Press.
Engerman, S., and K. Sokolofif. 2002. Factor endowments,
inequality, and paths of development among New World
economies. Economia 3: 41-109.
Ensminger, J. 1992. Making a market: The institutional
transformation of an African society. New York:
Cambridge University Press.
Fafchamps, M. 2004. Market institutions in Sub-Saharan
Africa: Theory and evidence. Cambridge, MA: MIT
Press.
Fehr, E., and S. Gachter. 2000. Cooperation and punishment
in public goods experiments. American Economic Review
90: 980-994.
Gambetta, D. 1993. The Sicilian Mafia: Die business of
private protection. Cambridge, MA: Harvard University
Press.
Giavazzi, F., and G. Tabellini. 2005. Economic and political
liberalizations. Journal of Monetary Economics 57: 1297-
1330.
Glaeser, E., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer.
2004. Do institutions cause growth? Journal of Economic
Growth 9: 271-303.
Greif, A. 1993. Contract enforceability and economic insti-
tutions in early trade: The Maghribi traders’ coalition.
American Economic Review 83: 525-548.
Greif, A. 1994. Cultural beliefs and the organization of
society: A historical and theoretical reflection on col-
lectivist and individualist societies. Journal of Political
Economy 102: 912-950.
Greif, A. 2006. Institutions and the path to the modern
economy: Lessons from medieval trade. New York:
Cambridge University Press.
Economic Growth
Greif, A., P. Milgrom, and B. Weingast. 1994. Coordination,
commitment, and enforcement: The case of the merchant
guild. Journal of Political Economy 102: 745-776.
Hall, R., and C. Jones. 1999. Why do some countries produce
so much more output than others? Quarterly Journal of
Economics 114: 83-116.
Hoff, K. 2003. Paths of institutional development: A view
from economic history. World Bank Research Observer
18:205-226.
Johnson, S., J. McMillan, andC. Woodruff. 2002. Courts and
relational contracts. Journal of Law, Economics, and
Organization 18: 221-277.
Kaufman, D., A. Kraay, and M. Mastruzzi. 2005. Governance
matters IV: Updated governance indicators 1996-2004.
Washington, DC: World Bank research paper.
Online. Available at
. Accessed 20 Apr 2006.
Keefer, P. 2004. What does political economy tell us about
economic development - And vice versa? Annual Review
of Political Science 7: 247-272.
Kranton, R. 1996. Reciprocal exchange: A self-sustaining
system. American Economic Review 86: 830-851.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. Vishny.
1998. Law and finance. Journal of Political Economy
106: 1113-1155.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. Vishny.
1999. The quality of government. Journal of Law,
Economics, and Organization 15: 222-279.
Lamoreaux, N., and J.-L. Rosenthal. 2005. Legal regime and
contractual flexibility: A comparison of business’s
organizational choices in France and the United States
during the era of industrialization. American Law and
Economics Review 7: 28-61.
Li, J.S. 2003. Relation-based versus rule-based governance:
An explanation of the East Asian miracle and Asian
crisis. Review of International Economics 11: 651-673.
Macaulay, S. 1963. Non-contractual relationships in business:
A preliminary study. American Sociological Review 28:
55-70.
Maggi, G. 1999. The role of multilateral institutions in
international trade cooperation. American Economic
Review 89: 190-214.
Mattli, W. 2001. Private justice in a global economy: From
litigation to arbitration. International Organization 55:
919-947.
Milgrom, P., D. North, and B. Weingast. 1990. The role of
institutions in the revival of trade: The law merchant,
private judges, and the Champagne fairs. Economics and
Politics 2: 1-23.
North, D. 1990. Institutions, institutional change, and eco-
nomic performance. Cambridge: Cambridge University
Press.
Ostrom, E. 1990. Governing the commons: The evolution of
institutions for collective action. Cambridge/New York:
Cambridge University Press.
Persson, T. 2005. Forms of democracy, policy, and economic
development, Working Paper No. 11171. Cambridge,
MA: NBER.
3299
Rajan, R., and L. Zingales. 2003. The great reversals: The
politics of financial development in the twentieth century.
Journal of Financial Economics 69: 5-50. Varese, F. 2001.
The Russian Mafia: Private protection in a new market
economy. Oxford: Oxford University Press. Williamson, O.
1975. Markets and hierarchies: Analysis and antitrust
implications. New York: Free Press. Williamson, O. 1995.
The mechanisms of governance.
New York: Oxford University Press.
Williamson, O. 2005. The economics of governance.
American Economic Review 95: 1-18.
World Bank Institute. Governance and anti-corruption.
Online. Available at
. Accessed 20 Apr 2006.
Economic Growth
Peter Howitt and David N. Weil
Abstract
Economic growth is the increase in a country’s
standard of living over time. Growth economists
study how living standards differ across
countries as well as across time. This article
discusses some of the broad facts of economic
growth and some of the main approaches to its
study.
Keywords
Adjustment costs; Capital accumulation; Cobb-
Douglas functions; Competition policy; Creative
destruction; Cultural determinants of growth;
Demographic dividend; Development
accounting; Dutch disease; Economic growth;
Endogenous growth; Export promotion; Fertil-
ity; General purpose technology; Geographical
determinants of growth; Growth accounting;
Health; Human capital; Import substitution;
Increasing returns; Inequality (global); Innova-
tion; Institutional determinants of growth;
Learning by doing; Malthus, T. R.; Mortality;
Natural capital; Neoclassical growth theory;
Patents; Physical capital; Policy determinants of
growth; Population growth; Production
functions; Productivity growth; Research and
development; Resource curse; Savings and
3300 Economic Growth

investment; Schooling; Technical change; the rest of the world, with the exception of Japan, in
Technical complementarities; Technology; what is often referred to as the ‘Great Divergence’.
Technology spillovers; Trade and growth As shown in Fig. below, the proportional gap in per
capita GDP between the richest group of countries
and the poorest group (as classified by Maddison)
JEL Classifications grew from three in 1820 to 19 in 1998. Pritchett (
D4; DIO ) tells a similar story, estimating
that the proportional gap between the richest and
Economic growth is typically measured as the
poorest countries grew more than fivefold from
change in per capita gross domestic product (GDP).
1870 to 1990.
Sustained long-term economic growth at a positive
This widening of the cross-country income
rate is a fairly recent phenomenon in human history,
distribution seems to have slowed during the second
most of it having occurred in the last 200 years.
half of the 20th century, at least among a large
According to Maddison’s ( )
group of nations. Indeed, Fig. , which is drawn on a
estimates, per capita GDP in the world economy
proportional scale, shows that with the acceleration
was no higher in the year 1000 than in the year 1,
of growth in Asia there has been a narrowing of the
and only 53 per cent higher in 1820 than in 1000,
spread between the richest and the second poorest
implying an average annual growth rate of only
group since 1950. Evans ( ) shows a narrowing of
one- nineteenth of one per cent over the latter 820-
the top end of the
year period. Some time around 1820, the world
distribution (that is, among Organisation for Eco-
growth rate started to rise, averaging just over one-
nomic Cooperation and Development, OECD,
half of one per cent per year from 1820 to 1870, and
countries) over the period. However, not all coun-
peaking during what Maddison calls the ‘golden
tries have taken part in this convergence process, as
age’, the period from 1950 to 1973, when it
the gap between the leading countries as a whole
averaged 2.93 per cent per year. By 2000, world per
and the very poorest countries has continued to
capita GDP had risen to more than 8.5 times its
widen. In Fig. the gap between the Western
1820 value.
Offshoots and Africa grew by a factor of 1.75
Growth has been uneven not only across time
between 1950 and 1998. Likewise, the proportional
but also across countries. Since 1820, living stan-
income gap between Mayer-Foulkes’s ( )
dards in Western Europe and its offshoots in North
richest and poorest convergence groups
America and the Antipodes have raced ahead of
grew by a factor of 2.6 between 1960 and 1995.

Economic Growth,
Per capita GDP, 1650-
Fig. 1
2000, 1990 international
dollars (Source: Maddison (
))
Economic Growth
Jones ( ) argues that continuing divergence
of the poorest countries from the rest of the world
does not imply rising income inequality among the
world’s population, mainly because China and
India, which contain about 40 per cent of that
population, are rising rapidly from near the bottom
of the distribution. Indeed, Sala-i- Martin ( ) that makes the marginal product asymptote to zero
shows, using data on within-country
income distributions, that the cross-individual dis-
tribution of world income narrowed considerably
between 1970 and 2000, even as the cross-country
distribution continued to widen somewhat. But
between-country inequality is still extremely
important; in 1992 it explained 60 per cent of
overall world inequality (Bourguignon and Mor-
rison ). Another reason that growth economists are
typically more concerned with the crosscountry
than the cross-individual distribution is that many of
the determinants of economic growth vary across
countries but not across individuals within
countries.
The Production Function Approach
The main task of growth theory is to explain this
variation of living standards across time and coun-
tries. One way to organize one’s thinking about the
sources of growth is in terms of an aggregate pro-
duction function, which indicates how a country’s
output per worker y depends on the (per worker)
stocks of physical, human, and natural capital,
represented by the vector k, according to
y=f{k,A),
(1)
where A is a productivity parameter. Economic
growth, as measured by the growth rate of y,
depends therefore on the rate of capital accumu-
lation and the rate of productivity growth. Similarly,
countries can differ in their levels of GDP per capita
either because of differences in capital or because of
differences in productivity. Much recent work on
the economics of growth has focused on trying to
identity the relative contributions of these two
fundamental factors to differences in growth rates
or income levels among countries.
3301
Modem growth theory started with the neo-
classical model of Solow ( ) and Swan
( ), who showed that in the long run growth
cannot be sustained by capital accumulation alone.
In their formulation, the diminishing marginal
product of capital (augmented by an Inada condition
as capital grows) will always terminate any
temporary burst of growth in excess of the growth
rate of labour-augmenting productivity. But this
perspective has been challenged by more recent
endogenous growth theory. In the AK theory of
Frankel ( ) and Romer ( ),
growth in productivity is functionally dependent on
growth in capital, through learning by doing and
technology spillovers, so that an increase in
investment rates in physical capital can also sustain
a permanent increase in productivity growth and
hence in the rate of economic growth. In the
innovation- based theory that followed AK theory,
the Solow model has been combined with a
Schumpeterian theory of productivity growth, in
which capital accumulation is one of the factors that
can lead to a permanently higher rate of productivity
growth (Howitt and Aghion ).
Capital
Having introduced the production function in a
general sense, we now examine the accumulation of
different types of capital in more detail, and then
turn to an assessment of the relative importance of
factor accumulation and productivity in explaining
income differences among countries and growth
over time.
Physical Capital
Physical capital is made up of tools, machines,
buildings, and infrastructure such as roads and
ports. Its key characteristics are, first, that it is
produced (via investment), and second that it is in
turn used in producing output. Physical capital
differs importantly from technology (which, as is
discussed below, is also both produced and pro-
ductive) in that physical capital is rival in its use:
only a limited number of workers can use a single
piece of physical capital at a time.
3302
Differences in physical capital between rich are
poor countries are very large. In the year 2000, for
example, physical capital per worker was 148,091
dollars in the United States, 42,991 dollars in
Mexico, and 6270 dollars in India. These large
differences in physical capital are clearly contrib-
utors to income differences among countries in a
proximate sense. That is, if the United States had
India’s level of capital it would be a poorer country.
The magnitude of this proximate effect can be
calculated by using the production function. For
example, using a value for capital’s share of
national income of 1/3 (which is consistent with the
findings of Gollin ( ) for a cross-section of
countries), the ratio of capital per worker in the
United States to that in India would by itself explain
a ratio of income per capita in the two countries of
7.9 ( = (148,091/6270)1/3).
Differences in physical capital among countries
can result from several factors. First, countries may
differ in their levels of investment in physical
capital relative to output. In an economy closed to
external capital flows, the investment rate will equal
the national saving rate. Saving rates can differ
among countries because of differences in the
security of property rights, due to the availability of
a financial system to bring together savers and
investors, because of government policies like
budget deficits or pay-as-you- go old age pensions,
differences in cultural attitudes towards present
versus future consumption, or simply because
deferring consumption to the future is a luxury that
very poor people cannot afford.
A second factor that drives differences in
investment rates among countries is the relative
price of capital. The price of investment goods in
relation to consumption goods is two to three times
as high in poor countries as in rich countries. If one
measures both output and investment at
international prices, investment as a fraction of
GDP is strongly correlated with GDP per capita
(correlation of 0.50), and poor countries have on
average between one half and one quarter of the
investment rate of rich countries. When investment
rates are expressed in domestic prices, the
correlation between investment rates and GDP per
capita falls to 0.05 (Hsieh and Klenow ).
Economic Growth
But levels of capital can also differ among
countries for reasons that have nothing to do with
the rate of accumulation. Differences in pro-
ductivity (the A term in Eq. ) will produce different
levels of capital even in countries with the same
rates of physical capital investment. Similarly,
differences in the accumulation of other factors of
production will produce differences in the level of
physical capital per worker.
Human Capital
Human capital refers to qualities such as education
and health that allow a worker to produce more
output and which themselves are the result of past
investment Like physical capital, human capital can
earn an economic return for its owner. However, the
two types of capital differ in several important
respects. Most significantly, human capital is
‘installed’ in a person. This makes it very difficult
for one person to own human capital that is used by
someone else. Human capital investment is a sig-
nificant expense. In the United States in the year
2000, spending by governments and families on
education amounted to 6.2 per cent of GDP;
forgone wages by students were of a similar
magnitude.
Information on the productivity of human cap-
ital can be derived from comparing wages of
workers with different levels of education. So called
‘Mincer regressions’ of log wage on years of
education, controlling by various means for bias due
to the endogeneity of schooling, yield estimated
returns to schooling of about ten per cent per year.
In the year 2000, the average schooling of workers
in advanced countries was 9.8 years and among
workers in developing countries 5.1 years. Applying
a rate of return of ten per cent implies that the
average worker in the advanced countries supplied
56 per cent more labour input because of this
education difference. If laboin’s share in a Cobb-
Douglas production function is two-thirds, this
would imply that education differences would
explain a factor of 1.35 difference in income
between the advanced and developing countries,
which is very small relative to the observed gap in
income. Allowing for differences in school quality
increases somewhat the income differences
explained by human capital in the form of
schooling.
Economic Growth
A second form of human capital is health. The
importance of health as an input into production can
be estimated by looking at microeconomic data on
how health affects individual wages. Health
differences between rich and poor countries are
large, and in wealthy countries worker health has
improved significantly over the last 200 years
(Fogel ). Weil ( ), using the
adult survival rate as a proxy for worker health,
estimates that eliminating gaps in worker health
among countries would reduce the log variance of
GDP per worker by 9.9 per cent.
Natural Capital
Natural capital is the value of a country’s agricul-
tural and pasture lands, forests and subsoil
resources. Like physical and human capital, natural
capital is an input into production of goods and
services. Unlike other forms of capital, however, it
is not itself produced.
Natural capital per worker and GDP per worker
are positively correlated, but the link is much
weaker than for the other measures of capital
discussed above. The poor performance of many
resource-rich countries has led many observers to
identify a ‘resource curse’ by which the availability
of natural capital undermines other forms of capital
accumulation or reduces productivity. Among the
suggested channels by which this happens are that
resource booms lead countries to raise consumption
to unsustainable levels, thus depressing saving and
investment (Rodriguez and Sachs ); that
exploitation of natural resources
suppresses the development of a local
manufacturing sector, which holds back growth
because manufacturing is inherently more techno-
logically dynamic than other parts of the economy
(this is the so called Dutch disease); and that
economic inefficiencies are associated with political
competition or even civil war to appropriate the
rents generated by natural resources.
Population and Economic Growth
Population affects the accumulation of all three
forms of capital discussed above, and through them
the level of output per worker. Rapid
3303
population growth dilutes the quantities of physical
and human capital per worker, raising the rates of
investment and school expenditure required to
maintain output per worker. The interaction of nat-
ural capital with population growth is at the centre
ofthe model of Malthus ( ). For a fixed stock of
natural capital, higher population lowers output per
capita. Combined with a positive feedback from the
level of income to population growth, this resource
constraint produces a stable steady state level of
output per capita and, with technology fixed, a
stable level of population as well. This Malthusian
feedback is the explanation for the long period of
nearly constant living standards that preceded the
Industrial Revolution (Galor and Weil ). Because of
resource-saving technological progress, as well as
expansion of international trade, which allows
countries to evade resource constraints, the
interaction of population and natural capital is much
less important today than in the past, with the
exception of very poor countries that are reliant on
subsistence agriculture.
In addition to its effect on the level of factors of
production per worker, population also matters for
economic growth because demographic change
produces important changes in the age structure of
the population. A reduction in fertility, for example,
will produce a long period of reduced dependency,
in which the ratio of children and the elderly, on the
one hand, to working age adults, on the other, is
temporarily below its sustainable steady state level.
This is the so-called ‘demographic dividend’ (see
population ageing).
In addition to these effects of population on the
level of income per capital, there is also causality
that mns from the economic to the
demographic. Over the course of economic devel-
opment, countries generally move through a
demographic transition in which mortality rates fall
first, followed by fertility rates. While the decline in
mortality is easily explained as a consequence of
higher income and technological progress, the
decline in fertility is not fully understood. Among
the factors thought to contribute to the decline in
fertility are falling mortality, a shift along a quality-
quantity trade-off due to rising returns to human
capital, the rise of women’s relative wages, the
reduced importance
3304 Economic Growth

of children as a means of old age support, and explanation for income differences. Similarly, dif-
improvements in the availability of contraception. ferences in productivity growth are the most
important explanation for differences in income
growth rates among countries. And as a theoretical
Growth Accounting and Development matter, the Solow model shows that as long as there
Accounting are decreasing returns to capital per worker,
productivity growth can be the only source of long-
The discussion above makes clear that stocks of term growth. The question is: what explains these
different forms of capital are positively correlated changes over time and differences in the level of
with GDP per capita. Similarly, as countries grow, productivity? Over the long term it is natural to
levels of capital per worker grow as well. It is associate productivity growth with technological
natural to ask whether these variations in capital are change. However, especially as an explanation for
sufficiently large to explain the matching variations differences in productivity at a given point in time,
in growth. The techniques of growth accounting a second possibility is that productivity differences
(Solow ) and development accounting (Klenow and reflect differences not in technology, in the sense of
Rodriguez-Clare ; Hall and Jones ) attempt to give inventions, blueprints, and so on, but rather
quantitative answers to this question. Using a differences in how economies are organized and use
parameterized production function and measures of available technology and inputs. We label this
the quantities of human and physical capital, one second contributor to productivity as ‘efficiency’.
can back out relative levels of productivity among
countries and rates of productivity growth within a Technology
country. Technology consists of the knowledge of how to
Caselli ( ) presents a review of develop transform basic inputs into final utility. This
ment accounting along with his own thorough knowledge can be thought of as another form of
estimates. His finding is that if human and physical capital, an intangible intellectual capital. What
capital per worker were equalized across countries, distinguishes technology from human or physical
the variance of log GDP per worker would fall by capital is its non-rival character. For example, the
only 39 per cent. In other words, the majority of knowledge that a particular kind of com will be
variation in income is due to differences in immune to caterpillars, or the knowledge of how to
productivity, not factor accumulation. Differences produce a 3 GHz CPU for a portable computer, can
in productivity growth, rather than differences in the be used any number of times by any number of
growth of physical and human capital, are also the people without diminishing anyone’s ability to use
dominant determinants of differences in income it again. By contrast, if you drive a lorry for an
growth rates among countries (Weil , eh. 7; Klenow hour, or if you employ the skills of a doctor for an
and Rodriguez- Clare ); differences in productivity hour, then that lorry or those skills are not available
levels among countries are striking. For example, to anyone else during that hour.
comparing the countries at the 90th and 10th Different growth theories have different
percentiles of the income distribution (which differ approaches to modelling the accumulation of
in income by a factor of 21), the former would pro- technology - that is, technical progress. According
duce seven times as much output as the latter with to neoclassical theory, for example, the relationship
equal quantities of human and physical capital. between technology and the economy is a one-way
street, with all of the causation running from
technology to the economy. It portrays technical
Productivity, Technology and Efficiency progress as emanating from a scientific progress
that operates outside the realm of economics, and
Development accounting shows that productivity thus takes the rate of technical progress as being
differences among countries are the dominant given exogenously.
Economic Growth
This neoclassical view has never been accepted
universally. Specialists in economic history and the
economics of technology have generally believed
that technical progress comes in the form of new
products, new techniques and new markets, which
do not spring directly from the scientific laboratory;
instead they come from discoveries made by private
business enterprises, operating in competitive
markets, and motivated by the search for profits.
For example, the transistor, which underlies so
much recent technological progress, was discovered
by scientists working for the AT&T telephone
company on the practical problem of how to
improve the performance of switch boxes that were
using vacuum tubes. Rosenberg ( ) describes
many other examples
of scientific and technological breakthroughs that
originated in profit-oriented economic activity.
What kept this view of endogenous technology
from entering the mainstream of economics until
recently was the difficulty of incorporating
increasing returns to scale into dynamic general
equilibrium theory. Increasing returns arise once
one considers technology as a kind of capital that
can be accumulated, because of its non-rival nature;
that is, the cost of developing a technology for
producing a particular product is a fixed set-up cost,
which does not have to be repeated when more of
the product is produced. Once the technology has
been developed then there should be at least
constant returns to scale in the factors that use that
technology, on the grounds that if you can do
something once then you can do it twice. But this
means that there are increasing returns in the broad
set of factors that includes the technology itself.
Increasing returns creates a problem because it
generally implies that a competitive equilibrium
will not exist, at least not without externalities.
These technical difficulties were overcome by
the new ‘endogenous growth theory’ introduced by
Romer ( ) and Lucas ( ), which incor
porated techniques that had been developed for
dealing with increasing returns in the theories of
industrial organization and international trade. The
first generation of endogenous growth theory to
enter the mainstream was the ‘AAK theory’,
according to which technological progress takes
3305
place as a result of externalities in learning to
produce capital goods more efficiently. The second
generation was the innovation-based theory of
Romer ( ) and Aghion and Howitt ( ),
which emphasizes the distinction between tech-
nological knowledge and other forms of capital, and
analyses technological innovation as a separate
activity from saving and schooling.
Historically, technical progress has engendered
much social conflict, because it involves what
Schumpeter ( ) called ‘creative destruction’;
that is, new technologies render old technologies
obsolete. As a result, technical progress is a game
with losers as well as winners. From the handloom
weavers of early 19th century Britain to the former
giants of mainframe computing in the late 20th
century, many people’s skills, capital equipment
and technological knowledge have been devalued
and their livelihoods imperilled by the same
innovations that have created fortunes for others.
The destructive side of technical progress shows
up most clearly during periods when a new ‘general
purpose technology’ (GPT) is being introduced. A
GPT is a basic enabling technology that is used in
many sectors of the economy, such as the steam
engine, the electric dynamo, the laser or the
computer. As Lipsey et al. ( ) have emphasized, a
GPT typically
arrives only partially formed, creates technological
complementarities and opens a window on new
technological possibilities. Thus it is typically
associated with a wave of new innovations.
Moreover, the period in which the new GPT is
diffusing through the economy is typically a period
of rapid obsolescence, costly learning and
wrenching adjustment. Greenwood and Yorukoglu (
) argue that the productivity
slowdown of the 1970s is attributable to the arrival
of the computer, and Howitt ( ) argues
that the rapid obsolescence generated by a new GPT
can cause per capita income to fall for many years
before eventually paying off in a much higher
standard of living.
New technologies are often opposed by those
who would lose from their introduction. Some of
this opposition takes place within the economic
sphere, where workers threaten action against
3306
firms that adopt labour-saving technologies and
firms try to pre-empt innovations by rivals. But
much of it also takes place within the political
sphere, where governments protect favoured firms
from more technically advanced foreign
competitors, and where people sometimes vote for
politicians promising to preserve traditional ways of
life by blocking the adoption of new technologies.
The leading industrial nations of the world spend
large amounts on R&D for generating innovations.
In the United States, for example, R&D
expenditures constituted between 2.2 and 2.9 per
cent of GDP every year from 1957 to 2004. But not
much cutting-edge R&D takes place outside a small
group of countries. In 1996, for example, 73 per
cent of the world’s R&D expenditure, as measured
by UNESCO, was accounted for by just five
countries (in decreasing order of R&D expenditure
they are the United States, Japan, Germany, France
and United Kingdom). In the majority of countries
that undertake very little measured R&D,
technology advances not so much by making
frontier innovations as by implementing tech-
nologies that have already been developed
elsewhere. But the process of implementation is not
costless, because technologies tend to be context-
dependent and technological knowledge tends to be
tacit. So implementation requires an up-front
investment to adapt the technology to a new
environment (see, for example, Evenson and
Westphal ). This investment plays the same role
analytically in the implementing country as R&D
does in the original innovating country.
Implementation is important in accounting for
the patterns of cross-country convergence and
divergence noted above. This is because a country
in which firms are induced to spend on implemen-
tation have what Gerschenkron ( ) called an
‘advantage of backwardness’. That is, the further
they fall behind the world’s technology frontier the
faster they will grow with any given level of
implementation expenditures, because the bigger is
the improvement in productivity when they
implement any given foreign technology. In the
long run, as Howitt ( ) has shown, this force
can cause all countries that engage in R&D or
implementation to grow at the same rate, while
Economic Growth
countries in which firms are not induced to make
such investments will stagnate. But technology
transfer through implementation expenditures is no
guarantee of convergence, because the technologies
that are being developed in the rich R&D-
performing countries are not necessarily appropriate
for conditions in poor implementing countries
(Basil and Weil ; Acemoglu and Zilibotti ) and
because financial constraints may prevent poor
countries from spending at a level needed to keep
pace with the frontier (Aghion et al. ).
Efficiency
The efficiency with which a technology is used is
not likely to play a major role in accounting for
long-run growth rates, because there is a finite limit
to how high you can raise living standards simply
by using the same technologies more efficiently.
But there is good reason to believe that differences
in efficiency account for much of the cross-country
variation in the level of productivity.
Inefficiencies take several different forms.
Economic resources are sometimes allocated to
unproductive uses, or even unused, as when union
featherbedding agreements kick in. Resources can
be misallocated as the result of taxes, subsidies and
imperfect competition, all of which create
discrepancies between marginal rates of
substitution. Technologies can be blocked by those
who would lose from their implementation and have
more market power or political influence than those
who would win.
The distinction between differences in technol-
ogy and differences in efficiency is often unclear.
Suppose firms in country A are using the same
machinery and the same number of workers per
machine as in country B, but output per worker is
higher in A than B. This may appear to be an
obvious case of inefficiency, since the technology
embodied in the machines used by workers in the
two countries is the same. But maybe it is just that
people in country B lack the knowledge of how best
to use the machines, in which case it may actually
be a case of differences in technology. As an
example, General Motors has had little success in
their attempts to emulate the manufacturing
Economic Growth
methods that Toyota has deployed successfully for
many years even in then US operations.
Moreover, identical technologies will have dif-
ferent effects in different countries, because of
differences in language, raw materials, consumer
preferences, workers expectations and the like. Eino
Disney, for example, was plagued initially with
laboin disputes when it first opened its park in the
outskirts of Paris in 1987. It took the American
managers several years to realize that the problem
was not recalcitrant workers but rather that French
workers consider it an intolerable indignity to be
forced to wear items such as mouse ears when
serving the public. A minor adjustment in
amusement park technology was needed to make it
as productive in France as it had been in the United
States.
Deeper Determinants of Growth
Even if we knew how much of the cross-country
variation in growth rates or income levels to attri-
bute to different kinds of capital or to technology or
efficiency, we would still be faced with the deeper
question of why these differences in capital and
productivity arise. A large number of candidate
explanations have been offered in the literature.
These candidates can be classified into four broad
categories: geography, institutions, policy and
culture.
Geographical differences are perhaps the most
obvious. As Sachs ( ) has emphasized, coun
tries that are landlocked, that suffer from a haz-
ardous disease environment and that have difficult
obstacles in the way of internal transport, will
almost certainly produce at a lower level than
countries without these problems, even if they use
the same technology and the same array of capital.
In addition, the lower productivity of these
countries will serve to reduce the rate of return to
accumulating capital and to generating new
technologies.
Institutions matter because of the way they
affect private contracts and also because of the way
they affect the extent to which the returns to
different kinds of investments can be appropriated
by the government. The origin of a country’s legal
3307
system has been shown by La Porta et al. ( ) to
have an important effect on private contracts. In
particular, these authors show that countries with
British legal origins tend of offer greater protection
of investor and creditor rights, which in turn is
likely to affect both capital accumulation and
investment in technology by making outside finance
more easily available.
Because long-term productivity growth requires
technical progress, it depends on political,
institutional and regulatory factors that affect the
way the conflict between the winners and losers of
technical progress will be resolved, and hence affect
the incentives to create and adopt new technologies.
For example, the way intellectual property is
protected will affect the incentive to innovate,
because on the one hand no one will want to spend
resources creating new technologies that his or her
rivals can easily copy, while on the other hand a
firm that is protected from competition by patent
laws that make it difficult for rivals to innovate in
the same product lines will be under less pressure to
innovate. Likewise, a populist political regime may
erect barriers to labour- saving innovation, resulting
in slower technical progress.
Economic policies matter not only because of
the way they affect the return to investing in capital
and technology but also because of the
inefficiencies that can be created by taxes and
subsidies. But how these policies affect economic
growth can vary from one country to another. In
particular, Aghion and Howitt ( ) have argued
that growth-promoting policies in technologically
advanced countries are not necessarily growth-
promoting in poorer countries, because innovation
and implementation are affected differently by the
same variables. For example, tighter competition
policy in a relatively backward country might retard
technology development by local firms that will be
discouraged by the threat of foreign entry, whereas
in more advanced countries firms will be spurred on
to make even greater R&D investments when
threatened by competition.
As this example suggests, international trade is
one ofthe policy domains most likely to matter for
growth and income differences, because of the
3308
huge productivity advantage that is squandered by
policies that run counter to comparative advantage,
because protected firms tend to become tech-
nologically backward firms, and because for many
countries international trade is the only way for
firms to gain a market laige enough to cover the
expense of developing leading-edge technologies.
So it is probably no accident that export promotion
has been a prominent feature of all the East Asian
countries that began escaping from the lower end of
the world income distribution towards the end of the
20th century, whereas import substitution was a
prominent feature of several Latin American
countries that fell from the upper end of the
distribution early in the 20th century.
Culture is a difficult factor to measure. In prin-
ciple, however, it is capable of explaining a great
deal of cross-country variation in growth, because a
society in which people are socialized to trust each
other, to work hard, to value technical expertise and
to respect law and order is certainly going to be
thriftier and more productive than a society in
which these traits do not apply. Recent work has
begun to quantify the role of culture using measures
of social capital, social capability, ethno- linguistic
fractionalization, religious belief, the spread of
Anglo-Saxon culture and many other variables.
See Also
► Economic Growth, Empirical Regularities in
► Economic Growth in the Very Long Run
► Endogenous Growth Theory
► Growth /: :
► Growth and Institutions
► Growth Take-Offs
► >;vel/.
► Population Ageing
Bibliography
Acemoglu, D., and F. Zilibotti. 2001. Productivity
differences. Quarterly Journal of Economics 116: 563-
606.
Economic Growth
Aghion, P., and P. Howitt. 1992. A model of growth through
creative destruction. Econometrica 60: 323-351.
Aghion, R, and P. Howitt. 2006. Appropriate growth policy:
An integrating framework. Journal of the European
Economic Association 4: 269-314.
Aghion, P., P. Howitt, and D. Mayer-Foulkes. 2005. The
effect of financial development on convergence: Theory
and evidence. Quarterly Journal of Economics 120: 173-
222.
Basu, S., and D.N. Weil. 1998. Appropriate technology and
growth. Quarterly Journal of Economics 113: 1025-1054.
Bourguignon, F., and C. Morrison. 2002. Inequality among
world citizens: 1820-1992. American Economic Review
92: 727-744.
Caselli, F. 2005. Accounting for cross-country income
differences. In Handbook of economic growth, ed. P.
Aghion and S.N. Durlauf, vol. 1. Amsterdam: North-
Holland.
Evans, P. 1996. Using cross-country variances to evaluate
growth theories. Journal of Economic Dynamics and
Control 20:1027-1049.
Evenson, R.E., and L.E. Westphal. 1995. Technological
change and technology strategy. In Handbook of devel-
opment economics, ed. T.N. Srinivasan and J. Behrman,
vol. 3A. Amsterdam: Elsevier.
Fogel, R. 1997. New findings on secular trends in nutrition
and mortality: Some implications for population theory.
In Handbook of population and family economics, ed.
M.R. Rosenzweig and O. Stark, vol. 1A. Amsterdam:
North-Holland.
Frankel, M. 1962. The production function in allocation
and growth: A synthesis. American Economic Review 52:
995-1022.
Galor, O., and D.N. Weil. 2000. Population, technology, and
growth: From Malthusian stagnation to the demographic
transition and beyond. American Economic Review 90:
806-828.
Gerschenkron, A. 1952. Economic backwardness in historical
perspective. In The Progress of Underdeveloped Areas, ed.
B.F. Hoselitz. Chicago: University of Chicago Press.
Gollin, D. 2002. Getting income shares right. Journal of Political
Economy 110: 458^474.
Greenwood, J., and Yorukoglu, M. 1997. 1974. Carnegie-
Rochester Conference Series on Public Policy 46: 49-95.
Hall, R., and C. Jones. 1999. Why do some countries produce so
much more output per worker than others? Quarterly Journal
of Economics 114: 83-116.
Howitt, P. 1998. Measurement, obsolescence, and general purpose
technologies. In General purpose technologies and
economic growth, ed. E. Helpman. Cambridge, MA: MIT
Press.
Howitt, P. 2000. Endogenous growth and cross-country income
differences. American Economic Review 90: 829-846.
Howitt, P, and P. Aghion. 1998. Capital accumulation and
innovation as complementary factors in long-run growth.
Journal of Economic Growth 3: 111-130.
Economic Growth in the Very Long Run 3309

Hsieh, C.-T., and P. Klenow. 2007. Relative prices and Weil, D.N. 2007. Accounting for the effect of health
relative prosperity. American Economic Review 97. on economic growth. Quarterlv Journal of Economics 122: 1265-1306.
Jones, C.I. 1997. On the evolution of the world income
distribution. Journal of Economic Perspectives 11(3): 19-36.
Klenow, P., and A. Rodriguez-Clare. 1997. The
neoclassical revival in growth economics: Has it
gone too far? In NBER macro annual, ed. B. Bemanke and J.
Rotemberg. Cambridge, MA: MIT Press. Economic Growth in the Very
La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and Long Run
R.W. Vishny. 1998. Law and finance. Journal of Political Economy
106: 1113-1155.
Oded Galor
Lipsey, R.G., K.I. Carlaw, and C.T. Bekar. 2005. Economic
transformations: General purpose technologies and long term economic growth. New York:
Oxford University Press.
Lucas, R.E.Jr. 1988. On the mechanics of economic
development. Journal of Monetary Economics 22: 3-42.
Maddison, A. 2001. The world economy: A millennial perspective, Development Abstract
centre studies. Paris: OECD. The evolution of economies during the major
Malthus, T.R. 1798. An essay on the principle of popula- portion of human history was marked by Mal-
tion, as it affects the future improvement of society with
thusian stagnation. The transition from an epoch
remarks on the speculations of Mr Godwin, M.
Condorcet, and other writers. London: printed for J. of stagnation to a state of sustained economic
Johnson in St Paul’s Churchyard. growth has shaped the contemporary world
Mayer-Foulkes, D. 2002. Global divergence. Documento de economy and has led to the great divergence in
Trabajo del CIDE, SDTE 250, Division de Economia.
income per capita across the globe in the past
Mexico: CIDE.
Pritchett, L. 1997. Divergence, big-time. Journal of Economic two centuries. This article examines the process
Perspectives 11(3): 3-17. of development over the course of human
Rodriguez, F., and J.D. Sachs. 1999. Why do resource- history in light of recent advances in unified
abundant economies grow more slowly? Journal of Economic Growth 4:
growth theory.
277-303.
Rorner, P.M. 1986. Increasing returns and long-run
growth. Journal of Political Economy 94: 1002-1037. Keywords
Rorner, P.M. 1990. Endogenous technological change.
Capital accumulation; China; Colonialism;
Journal of Political Economy 98: S71—102.

Rosenberg, N. 1981. How exogenous is science? In Inside the
black box: Technology
ed. N. Rosenberg. New York: Cambridge
economics,
University Press.
Sachs, J.D. 2003. Institutions don't rule: Direct effects of
geography on per capita income (Working Paper No.
9490). Cambridge, MA: NBER.
Sala-i-Martin, X. 2006. The world distribution of
income: Falling poverty and ... convergence,
period. Quarterly Journal of Economics 121: 351-397.
Schumpeter, J.A. 1942. Capitalism, socialism and democracy. New York:
Harper.
Solow, R.M. 1956. A contribution to the theory of eco-
nomic growth. Quarterly Journal of Economics 70: 65-94. ""
Solow, R.M. 1957. Technical change and the aggregate
production function. Review of Economics and Statistics 39: 312-320.
Swan, T.W. 1956. Economic growth and capital accumu-
lation. Economic Record 32: 334-361.
Weil, D.N. 2005. Economic growth. Boston: Addison- Weslcy.
Convergence clubs; Demographic transition;
and Economic growth in the very long run; Educa-
tion; Endogenous growth; Exogenous growth;
Fertility; Globalization; Great divergence;
Growth take-offs; Human capital; Income
growth; Industrial Revolution; Industrialization;
Inequality; Land; Land productivity; Literacy;
Malthusian stagnation; Marriage; Micro
foundations; Population growth; Post-
Malthusian regime; Sustained growth regime;
Technological progress; Unified growth theory
JEL classifications
Nl; 04; Oil; 014; 033; 040; Jll; J13
The evolution of economies during the major por-
tion of human history was marked by Malthusian
stagnation. Technological progress and population
growth were minuscule by modem standards, and
3310
the average growth rates of income per capita in
various regions of the world were even slower due
to the offsetting effect of population growth on the
expansion of resources per capita.
In the past two centuries the pace of techno-
logical progress increased significantly in associ-
ation with the process of industrialization. Various
regions of the world departed from the Malthusian
trap and experienced a considerable rise in the
growth rates of income per capita and population.
Unlike episodes of technological progress in the
pre-industrial Revolution era that failed to generate
sustained economic growth, the increasing role of
human capital in the production process in the
second phase of industrialization ultimately
prompted a demographic transition, liberating the
gains in productivity from the counterbalancing
effects of population growth. The decline in the
growth rate of population and the enhancement of
human capital formation and technological progress
paved the way for the emergence of the modem
state of sustained economic growth. Variations in
the timing of the transitions from a Malthusian
epoch to a state of sustained economic growth
across countries lead to a considerable rise in the
ratio of GDP per capita between the richest and the
poorest regions of the world from 3:1 in 1820 to
18:1 in 2000 (see Fig. ).
The transition from stagnation to growth and the
associated phenomenon of the great divergence
have been the subject of intensive research in the
growth literature in recent years (Galor and
Economic Growth in the
Very Long Run, Fig. 1 The
evolution of regional
income per capita, 1-2000
(Source: Maddison ( ))
0
Economic Growth in the Very Long Run
Weil , ; Galor and Moav ; Lucas
; Hansen and Prescott ; Jones ; Hazan and
Berdugo ; Doepke ; Lagerlof
, ; Galor and Mountford , ).
The inconsistency of exogenous and endogenous
growth models with some of the most fundamental
features of the process of development has led to a
search for a unified theory that would unveil the
underlying microfoundations of the growth process
in its entirety, and would capture in a single
framework the epoch of Malthusian stagnation that
characterized most of human history, the
contemporary era of modem economic growth, and
the driving forces that triggered the recent transition
between these regimes.
The advance of unified growth theory was
fuelled by the conviction that the understanding of
the contemporary growth process would be fragile
and incomplete unless growth theory were based on
proper microfoundations that reflect the various
qualitative aspects of the growth process and their
central driving forces. Moreover, it has become
apparent that a comprehensive understanding of the
hurdles faced by less developed economies in
reaching a state of sustained economic growth
would remain obscure unless the factors that
prompted the transition of the currently developed
economies into a state of sustained economic
growth could be identified and modified to account
for the differences in the growth structure of less
developed economies in an interdependent world.
Economic Growth in the Very Long Run
Unified growth theory explores the fundamental
factors that generated the remarkable escape from
the Malthusian epoch and their significance in
understanding the contemporary growth process of
developed and less developed economies.
Moreover, it sheds light on the perplexing phe-
nomenon of the great divergence in income per
capita across regions of the world in the past two
centuries. It suggests that the transition from stag-
nation to growth is an inevitable outcome of the
process of development. The inherent Malthusian
interaction between the level of technology and the
size and the composition of the population
accelerated the pace of technological progress and
ultimately raised the importance of human capital in
the production process. The rise in the demand for
human capital in the second phase of
industrialization and its impact on the formation of
human capital as well as on the onset of the
demographic transition brought about significant
technological advances along with a reduction in
fertility rates and population growth, enabling
economies to convert a larger share of the fruits of
factor accumulation and technological progress into
growth of income per capita, and paving the way for
the emergence of sustained economic growth.
Differences in the timing of the take-off from
stagnation to growth across countries (for example,
England’s earlier industrialization in comparison
with China) contributed significantly to the great
divergence and to the emergence of convergence
clubs. These variations reflect initial differences in
geographical factors and historical accidents and
their manifestation in diversity in institutional,
demographic, and cultural factors, trade patterns,
colonial status, and public policy. In particular, once
a technologically driven demand for human capital
emerged in the second phase of industrialization,
the prevalence of human capital-promoting
institutions determined the extensiveness of human
capital formation, the timing of the demographic
transition, and the pace of the transition from
stagnation to growth. Thus, unified growth theory
provides the natural framework of analysis in which
variations in the economic performance across
countries and regions could be examined based on
the effect of
3311
variations in educational, institutional, geograph-
ical, and cultural factors on the pace of the transi-
tion from stagnation to growth.
The Process of Development
The process of economic development has been
characterized by of three fundamental regimes: the
Malthusian epoch, the post-Malthusian regime, and
the sustained growth regime.
The Malthusian Epoch
During the Malthusian epoch that characterized
most of human history, humans were subjected to a
persistent struggle for existence. Resources gen-
erated by technological progress and land expansion
were channeled primarily towards an increase in the
size of the population, with a minor long-run effect
on income per capita. Improvements in the
technological environment or in the availability of
land generated temporary gains in income per
capita, leading eventually to a larger but not richer
population. Technologically superior countries
ultimately had denser populations but their standard
of living did not reflect the degree of their
technological advancement.
During the Malthusian epoch the average
growth rate of output per capita was negligible and
the standard of living did not differ greatly across
countries. The average level of income per capita in
the world during the first millennium fluctuated
around $450 per year (in 1990 international dollars)
and the average growth rate of output per capita was
nearly zero (Maddison ). This state of Malthusian
stagnation persisted until the end of the 18th
century. In the years 1000-1820, the average level
of income per capita in the world economy was
below $670 per year, and the average growth rate of
the world income per capita was minuscule,
creeping at a rate of about 0.05 per cent per year.
Nevertheless, income per capita fluctuated
significantly within regions, deviating from their
sluggish long-run trend over decades and sometimes
centuries.
Population growth over this era followed the
Malthusian pattern as well. The gradual increase in
income per capita during the Malthusian epoch was
3312
associated with a monotonic increase in the average
rate of growth of world population. The slow pace
of resource expansion in the first millennium was
reflected in a modest increase in the population of
the world from 231 million people in 1 CE to 268
million in 1000 CE: a minuscule average growth
rate of 0.02 per cent per year. The more rapid (but
still very slow) expansion of resources in the period
1000-1500 permitted the world population to
increase by 63 per cent, from 268 million in 1000 to
438 million in 1500; a slow 0.1 per cent average
growth rate per year. Resource expansion over the
period 1500-1820 had a more significant impact on
the world population, which grew 13 8 per cent
from 438 million in 1500 to 1041 million in 1820:
an average pace of 0.27 per cent per year.
Variations in population density across countries
during the Malthusian epoch reflected primarily
cross-country differences in technology and land
productivity. Due to the positive adjustment of the
population to an increase in income per capita,
differences in technology or in land productivity
across countries resulted in variations in population
density rather than in the standard of living. For
instance, China’s technological advancement in the
period 1500-1820 permitted its share
ofworldpopulation to increase from 23.5 per cent
to 36.6 per cent, while its income per capita in the
beginning and the end of this time interval
remained approximately $600 per year.
The Post-Malthusian Regime
During the post-Malthusian regime, the pace of
technological progress markedly increased in
association with the process of industrialization,
triggering a take-off from the Malthusian trap. The
growth rate of income per capita increased signif-
icantly but the positive Malthusian effect of income
per capita on population growth was still
maintained, generating a sizeable increase in pop-
ulation growth that offset some of the potential
gains in income per capita.
The take-off of developed regions from the
Malthusian regime occurred at the beginning of the
19th century and was associated with the Industrial
Revolution, whereas the take-off of less developed
regions occurred towards the beginning of the 20th
century and was delayed
Economic Growth in the Very Long Run
in some countries well into the 20th century. During
the post-Malthusian regime the average growth rate
of output per capita increased significantly and the
standard of living began to differ considerably
across countries. The average growth rate of output
per capita in the world soared from 0.05 per cent per
year during the period 1500-1820 to 0.53 per cent
per year in the years 1820-70, and 1.3 per cent per
year during the period 1870-1913. The timing of the
take-off and its magnitude differed across regions.
The take-off from the Malthusian epoch and the
transition to the post-Malthusian regime occurred in
western Europe, the Western offshoots (that is, the
United States, Canada, Australia and New Zealand),
and eastern Europe at the beginning of the 19th
century, whereas in Latin America, Asia and Africa
it occurred towards the beginning of the 20th
century.
The rapid increase in income per capita in the
post-Malthusian regime was channeled partly
towards an increase in the size of the population.
During this period, the Malthusian mechanism
linking higher income to higher population growth
continued to function. However, the effect of higher
population on the dilution of resources per capita
was counteracted by accelerated technological
progress and capital accumulation, allowing income
per capita to rise despite the offsetting effects of
population growth.
The western European take-off along with that
of the Western offshoots brought about a sharp
increase in population growth in these regions and
consequently a modest rise in population growth in
the world as a whole. The subsequent take-off of
less developed regions, and the associated increase
in their rates of population growth, brought about a
significant rise in population growth in the world.
The rate of population growth in the world
increased from an average rate of 0.27 per cent per
year in the period 1500-1820 to 0.4 per cent per
year in the years 1820-70, and to 0.8 per cent per
year in the time inteival 1870-1913. Despite the
decline in population growth in western Europe and
the Western offshoots towards the end of the 19th
century and the beginning of the 20th century, the
delayed take-off of less developed regions, and the
Economic Growth in the Very Long Run
significant increase in their income per capita prior
to their demographic transitions, generated a further
increase in the rate of population growth in the
world to 0.93 per cent per year in the period 1913-
50, and 1.92 per cent per year in the period 1950-
73. Ultimately, the onset of the demographic
transition in less developed economies during the
second half of the 20th century reduced population
growth rates to 1.66 per cent per year in the 1973-
98 period (Maddison ).
It appears that the significant rise in income per
capita in the post-Malthusian regime increased the
desired number of surviving offspring and thus,
despite the decline in mortality rates, fertility
increased significantly so as to enable households to
reach this higher desired level of surviving
offspring. Fertility was controlled during this
period, despite the absence of modem contraceptive
methods, partly via adjustment in marriage rates.
Increased fertility was achieved by earlier female
age of marriage, and a decline in fertility by a delay
in the marriage age.
The take-off in the developed regions was
accompanied by a rapid process of industrialization.
Per-capita level of industrialization increased
significantly in the United Kingdom, rising 50 per
cent over the 1750-1800 period, quadrupling in the
years 1800-60, and nearly doubling in the time
period 1860-1913. Similarly, per capita level of
industrialization accelerated in the United States,
doubling in the 1750-1800 as well as 1800-60
periods, and increasing sixfold in the years 1860-
1913. A similar pattern was experienced in
Germany, France, Sweden, Switzerland, Belgium
and Canada. The take-off of less developed econ-
omies in the 20th century was associated with
increased industrialization as well. However, during
the 19th century these economies experienced a
decline in per capita industrialization, reflecting the
adverse effect of the sizeable increase in population
on the level of industrial production per capita as
well as the forces of globalization and colonialism,
which induced less developed economies to
specialize in the production of raw materials (Galor
and Mountford , ).
The acceleration in technological progress dur-
ing die post-Malthusian regime and the associated
increase in income per capita stimulated the
3313
accumulation of human capital in the form of liter-
acy rates, schooling, and health. The increase in the
investment in human capital was induced by the rise
in income per capita, as well as by qualitative
changes in the economic environment that increased
the demand for human capital and induced house-
holds to invest in the education of their offspring.
In the first phase of the Industrial Revolution,
human capital had a limited role in the production
process. Education was motivated by a variety of
reasons, such as religion, enlightenment, social
control, moral conformity, socio-political stability,
social and national cohesion, and military
efficiency. The extensiveness of public education
was therefore not necessarily correlated with
industrial development, and it differed across
countries due to political, cultural, social, historical
and institutional factors. In the second phase of the
Industrial Revolution, however, the demand for
education increased, reflecting the increasing skill
requirements in the process of industrialization. The
economic interests of capitalists were a significant
driving force behind the implementation of
educational reforms (Galor and Moav ). The process
of industrialization has been characterized by a
gradual increase in the relative importance of
human capital in less developed economies as well
and educational attainment increased significantly
across all less developed regions in the post-
Malthusian regime.
The Sustained Growth Regime
The acceleration in the rate of technological pro-
gress in the second phase of industrialization, and
its interaction with human capital formation, trig-
gered a demographic transition, paving the way to a
transition to an era of sustained economic growth.
In the post demographic-transition period, the rise
in aggregate income due to technological progress
and factors accumulation was no longer
counterbalanced by population growth, permitting
sustained growth in income per capita in regions
that experienced sustained technological progress
and accumulation of physical and human capital.
The transition of the developed regions of
western Europe and the Western offshoots to the
state of sustained economic growth occurred
3314
towards the end of the 19th century, and their
income per capita in the 20th century has advanced
at a stable rate of about two per cent per year. The
transition of some less developed countries in Asia
and Latin America occurred towards the end of the
20th century. Africa, in contrast, is still struggling
to make this transition.
The transition to a state of sustained economic
growth was characterized by a gradual increase in
the importance of the accumulation of human
capital relative to physical capital as well as with a
sharp decline in fertility rates. In the first phase of
the Industrial Revolution (1760-1830), capital
accumulation as a fraction of GDP significantly
increased whereas literacy rates remained largely
unchanged. Skills and literacy requirements were
minimal, the state devoted virtually no resources to
raise the level of literacy of the masses, and workers
developed skills primarily through on-the-job
training (Green ; Mokyr ). Consequently, literacy
rates did not increase during the period 1750-1830
(Sanderson ).
In the second phase of the Industrial Revolution,
however, the pace of capital accumulation subsided,
skills became necessary for production and the
education of the labour force markedly increased.
The investment ratio in the UK, which increased
from six per cent in 1760 to 11.7 per cent in 1831,
remained at around 11 per cent on average in the
years 1856-1913 (Crafts ). In contrast, the average
years of schooling of males in the labour force that
did not change significantly until the 1830s tripled
by the beginning of the 20th century. The drastic
rise in the level of income per capita in England as
of 1865 was associated with an increase in school
enrolment of ten-year-old children from 40 per cent
in 1870 to 100 per cent in 1900. Moreover, total
fertility rate in England sharply declined over this
period from about five in 1875, to nearly two in
1925.
The demographic transition swept the world in
the course of the 20th century. The unprecedented
increase in population growth during the post-
Malthusian regime was reversed and the demo-
graphic transition brought about a significant
reduction in fertility rates and population growth in
various regions of the world, enabling
Economic Growth in the Very Long Run
economies to convert a larger share of the fruits of
factor accumulation and technological progress into
growth of income per capita. The demographic
transition enhanced the growth process via three
channels: (a) reductions in the dilution of the stocks
of capital and natural resources, (b ) enhancements
in human capital formation, and (c) changes in the
age distribution of the population, temporarily
increasing the size of the labour force relative to the
population as a whole.
The timing of the demographic transition dif-
fered significantly across regions. The reduction in
population growth occurred in Western Europe, the
Western offshoots, and eastern Europe towards the
end of the 19th century and in the beginning of the
20th century, whereas Latin America and Asia
experienced a decline in the rate of population
growth only in the last decades of the 20th century.
Africa’s population growth, in contrast, has been
rising steadily.
The process of industrialization was character-
ized by a gradual increase in the relative importance
of human capital in the production process. The
acceleration in the rate of technological progress
gradually increased the demand for human capital,
inducing individuals to invest in education, and
stimulating further technological advancement.
Moreover, in developed as well as less developed
regions, the onset of the process of human capital
accumulation preceded the onset of the
demographic transition, suggesting that the rise in
the demand for human capital in the process of
industrialization and the subsequent accumulation
of human capital played a significant role in the
demographic transition and the shift to a state of
sustained economic growth.
Notably, the reversal of the Malthusian relation
between income and population growth during the
demographic transition corresponded to an increase
in the level of resources invested in each child. For
example, literacy rate among men in England was
stable at around 65 per cent in the first phase of the
Industrial Revolution and increased significantly
during the second phase, reaching nearly 100 per
cent at the end of the 19th century. In addition, the
proportion of children aged 5 to 14 in primary
schools increased from 11 per cent in 1855 to 74 per
cent in 1900.
Economic Growth in the Very Long Run
A similar pattern is observed in other European
societies (Flora, Kraus and Pfenning ).
The process of industrialization was character-
ized by a gradual increase in the relative importance
of human capital in less developed economies as
well. Educational attainment increased significantly
across all less developed regions. Moreover, in line
with the pattern that emerged among developed
economies in the 19th century, the increase in
educational attainment preceded or occurred
simultaneously with the decline in total fertility
rates.
The Great Divergence
The differential timing of the take-off from stag-
nation to growth across countries and the
corresponding variations in the timing of the
demographic transition led to a great divergence in
income per capita as well as population growth.
Inequality in the world economy was negligible till
the 19th century. The ratio of GDP per capita
between the richest region and the poorest region in
the world was only 1.1:1 in 1000, 2:1 in 1500 and
3:1 in 1820. In the past two centuries, however, the
ratio of GDP per capita between the richest group
(Western offshoots) and the poorest region (Africa)
has widened considerably from a modest 3:1 ratio in
1820, to 5:1 ratio in 1870, 9:1 ratio in 1913, 15:1
ratio in 1950, and 18:1 ratio in 2001.
An equally momentous transformation occurred
in the distribution of world population across
regions. The earlier take-off of western European
countries increased the amount of resources that
could be devoted for the increase in family size,
permitting a 16 per cent increase in the share of
their population in the world from 12.8 per cent in
1820 to 14.8 per cent in 1870. However, the early
onset in the western European demographic
transition and the long delay in the demographic
transition of less developed regions, well into the
second half of the 20th century, led to a decline in
the share of western European population in the
world, from 14.8 per cent in 1870 to 6.6 per cent in
1998. In contrast, the prolongation of the post-
Malthusian period among less developed regions, in
association with the delay in their demographic
transition well into the second half
3315
of 20th century, channelled their increased
resources towards a significant increase in their
population. Africa’s share of world population
increased from seven per cent in 1913 to 12.9 per
cent in 1998, Asia’s share of world population
increased from 51.7 per cent in 1913 to 57.4 per
cent in 1998, and Latin American countries
increased their share in world population from two
per cent in 1820 to 8.6 per cent in 1998.
Unified Growth Theory
Galor and Weil ( ) advanced a unified growth
theory that captures the three regimes that have
characterized the process of development as well as
the fundamental driving forces that generated the
transition from an epoch of Malthusian stagnation
to a state of sustained economic growth. The theory
replicates the observed time paths of population,
income per capita, and human capital, generating:
(a) the Malthusian oscillations in population and
output per capita during the Malthusian epoch, (b)
an endogenous take-off from Malthusian stagnation
that is associated with an acceleration in
technological progress and is accompanied initially
by a rapid increase in population growth, and (c) a
rise in the demand for human capital, followed by a
demographic transition and sustained economic
growth. These qualitative patterns are confirmed in
the calibration of the theory by Lagerlof ( ).
The theory proposes that in early stages of
development economies were in the proximity of a
stable Malthusian equilibrium. Technology
advanced rather slowly, and generated proportional
increases in output and population. The inherent
positive interaction between population and
technology in this epoch, however, gradually
increased the pace of technological progress, and
due to the delayed adjustment of population, output
per capita advanced at a minuscule rate. The slow
pace of technological progress in the Malthusian
epoch provided a limited scope for human capital in
the production process and parents, therefore, had
no incentive to reallocate resources towards human
capital formation of their offspring.
3316
The Malthusian interaction between technology
and population accelerated the pace of tech-
nological progress and permitted a take-off to the
post-Malthusian regime. The expansion of
resources was partially counterbalanced by the
enlargement of population, and the economy was
characterized by rapid growth rates of income per
capita and population. The acceleration in techno-
logical progress eventually increased the demand
for human capital, generating two opposing effects
on population growth. On the one hand, it eased
households’ budget constraints, allowing the
allocation of more resources for raising children.
On the other hand, it induced a reallocation of
resources towards child quality. In the post-
Malthusian regime, due to the modest demand for
human capital, the first effect dominated, and the
rise in real income permitted households to increase
the number as well the quality of their children.
As investment in human capital took place, the
Malthusian steady-state equilibrium vanished and
the economy started to be attracted by the gravi-
tational forces of the modem growth regime. The
interaction between investment in human capital
and technological progress generated a virtuous
circle: human capital generated faster technological
progress, which in turn further raised the demand
for human capital, inducing further investment in
child quality, and eventually triggering the onset of
the demographic transition and the emergence of a
state of sustained economic growth.
The theory suggests that the transition from
stagnation to growth is an inevitable outcome of the
process of development. The inherent Malthusian
interaction between the level of technology and the
size of the population accelerated the pace of
technological progress, and ultimately raised the
importance of human capital in the production
process. The rise in the demand for human capital
in the second phase of the Industrial Revolution and
its impact on the formation of human capital as well
as on the onset of the demographic transition
brought about significant technological
advancements along with a reduction in fertility
rates and population growth, enabling economies to
convert a larger share of the fruits of factor
Economic Growth in the Very Long Run
accumulation and technological progress into
growth of income per capita, and paving the way for
the emergence of sustained economic growth.
Quantitative analysis of unified growth theories
(Doepke ); Lagerlof ) indeed suggest that the rise in
the demand for human capital was a significant
force behind the demographic transition and the
emergence of a state of sustained economic growth.
Variations in the timing of the transition from
stagnation to growth and thus in economic perfor-
mance across countries reflect initial differences in
geographical factors and historical accidents and
their manifestation in diversity in institutional,
demographic, and cultural factors, trade patterns,
colonial status, and public policy. In particular, once
a technologically driven demand for human capital
emerged in the second phase of industrialization,
the prevalence of human capital- promoting
institutions determined the extensiveness of human
capital formation, the timing of the demographic
transition, and the pace of the transition from
stagnation to growth.
The theory proposes that the growth process is
characterized by stages of development and it
evolves nonlinearly. Technological leaders expe-
rienced a monotonic increase in the growth rates of
their income per capita. Their growth was rather
slow in early stages of development, increased
rapidly during the take-off from the Malthusian
epoch, and continued to rise, often stabilizing at
higher levels. In contrast, technological followers
that made the transition to sustained economic
growth experienced a non-monotonic increase in the
growth rates of their income per capita. Their
growth rate was rather slow in early stages of
development, but increased rapidly in the early
stages of the take-off from the Malthusian epoch,
boosted by the adoption of technologies from the
existing technological frontier. However, once these
economies reached the technological frontier, their
growth rates dropped to the level of the
technological leaders. Hence, consistently with
contemporary evidence about the existence of
multiple growth regimes (Durlauf and Quah ), the
differential timing of the take-off from stagnation to
growth across economies generated convergence
clubs characterized
Economic Growth Non-linearities 3317

by a group of poor countries in the vicinity of the Lagerlof, N. 2003. From Malthus to modem growth: the three
Malthusian equilibrium, a group of rich countries in regimes revisited. International Economic Review 44:
755-777.
the vicinity of the sustained growth equilibrium, and Lagerlof, N. 2006. The Galor-Weil model revisited: a
a third group in the transition from one club to quantitative exploration. Review of Economic Dynamics
another. 9: 116-142.
Lucas, R.E. 2002. The industrial revolution: Past and future.
Cambridge, MA: Harvard University Press.
Maddison, A. 2001. The world economy: A millennia per-
See Also spective. Paris: OECD.
Mokyr, J. 1993. The new economic history and the industrial
revolution. In The British industrial revolution: An
economic perspective, ed. J. Mokyr. Boulder, CO:
Westview Press.
Sanderson, M. 1995. Education, Economic Change and
Society in England 1780-1870. Cambridge: Cambridge
Bibliography University Press.

Crafts, N.F.R. 1985. British economic growth during the

industrial revolution. Oxford: Oxford University Press.
Doepke, M. 2004. Accounting for fertility decline during the
transition to growth. Journal of Economic Growth 9: 347-
383. Economic Growth Non-linearities
Durlauf, S.N., and D. Quah. 1999. The new empirics of
economic growth. In Handbook of macroeconomics, ed. Chih Ming Tan
J.B. Taylor and M. Woodford. Amsterdam: North-
Holland.
Flora, P., F. Kraus, and W. Pfenning. 1983. State, Economy
and Society in Western Europe 1815-1975. Chicago: St.
James Press. Abstract
Galor, O. 2005. From stagnation to growth: unified growth
theory .In Handbook ofeconomic growth, ed. P. Aghion Nonlinearities in growth have important
and S.N. Duraluf. Amsterdam: North-Holland. implications for cross-country income
Galor, O., and O. Moav. 2002. Natural selection and the inequality. In particular, they imply that
origin of economic growth. Quarterly Journal of Eco- countries may spend long periods of time in a
nomics 117: 1133-1192.
Galor, O., and O. Moav. 2006. Das human kapital: A theory low-growth poverty trap. However, finding
of the demise of the class structure. Review of Economic evidence of such nonlinearities in the data and
Studies 73: 85-117. accounting for their emergence pose unique
Galor, O., and A. Mountford . 2003. Trading population for challenges to researchers.
productivity. Working paper, Brown University.
Galor, O., and A. Mountford. 2006. Trade and the great Keywords
divergence: The family connection. American Economic Balanced growth; Conditional convergence;
Review 96: 299-303. Convergence; Convergence clubs; Demographic
Galor, O., and D.N. Weil. 1999. From Malthusian stagnation
transition; Diffusion of technology; Economic
to modem growth. American Economic Review 89: 150-
154. development; Economic growth nonlinearities;
Galor, O., and D.N. Weil. 2000. Population, technology and Increasing returns; Industrialization; Market
growth: from the Malthusian regime to the demographic size; Multiple-growth regimes; Neoclassical
transition and beyond. American Economic Review 110:
growth theory; Neoclassical production
806-828.
Green, A. 1990. Education and state formation. New York: function; Nonconvexity; Poverty traps; Spillover
St. Martin’s Press. effects; Stages of theory of growth; Strategic
Hansen, G., and E. Prescott. 2002. Malthus to Solow. complementarities; Structural change; Take-off;
American Economic Review 92: 1205-1217.
Technological progress
Hazan, M., and B. Berdugo. 2002. Child labor, fertility and
economic growth. Economic Journal 112: 810-828.
Jones, C.I. 2001. Was an industrial revolution inevitable?
Economic growth over the very long run. Advances in JEL Classifications
Macroeconomics 1: 1-43. 04
3318
Nonlinear Growth Models
Nonlinear growth models are characterized by a
country’s subsequent performance being critically
dependent upon its initial conditions. In particular,
these models tend to imply that countries which
have unfavourable initial conditions may either
experience substantial periods of time in low-
growth/low-income poverty traps or be altogether
caught in one. In some cases, it has been explicitly
suggested that active (exogenous) policy
interventions may be necessary in order to kick-start
a country into a more favourable equilibrium.
Nonlinear growth models can be broadly classified
into two classes: structural change (or ‘stages of
development’) models, and models that emphasize
endogenous technological development and cross-
country interactions in terms of technological
diffusion.
Structural change models focus on the (internal)
transformations of an economy as it transits through
critical phases or ‘stages’ (see Lewis ; Rostow )
leading to industrialization. The aim of this work is
to clarify the conditions for such transitions to
occur. Early work in the economic development
literature (see Rosenstein-Rodan ; Nurkse ;
Scitovsky ; Fleming ; formalized by Murphy et al. )
emphasized the importance of increasing returns
and the size of the market in industrialization. The
key idea behind this view is that countries could be
locked in a no-industrialization trap because of the
small size of the market for each sector of the
economy. No single sector can achieve growth on
its own. However, the growth of one sector results
in the enlargement of markets for other sectors. The
enlargement of markets then encourages investment
and growth in the corresponding sectors. These
spillover effects and strategic complementarities
imply that a ‘big push’ - that is, coordinated
investments (or ‘balanced growth’) across sectors -
may be sufficient to push the economy out of the
trap and into a ‘take-off towards industrialization.
Other models are explicitly informed by the analysis
of historical data (see Maddison ), and emphasize
the importance of explaining simultaneously both
historical patterns
Economic Growth Non-linearities
of other state variables associated with growth and
growth itself. An important recent work that models
the demographic transition in growth take-offs is
Galor and Weil ( ). Because
these models require that certain conditions be met
before countries are able to achieve take-off, those
who do not meet these requirements could find
themselves trapped in a phase of economic
stagnation for extended periods of time.
The second class of models focuses on the role
of technological progress in growth. In particular,
the emphasis of these models is on the diffusion of
technology from countries which are technological
leaders to less developed countries. Lucas ( ) is a
seminal work in this area (see also
Basu and Weil ; Parente and Prescott ; Howitt and
Mayer-Foulkes ). Particular attention has been paid
to exploring the channels through which less
advanced countries imitate or adopt technologies in
leader countries. If there are no barriers to
technological diffusion across countries, then these
models typically predict that rich and poor countries
would gradually converge in per capita income.
However, if such barriers exist, then countries may
differ in their ability to adopt technologies leading
to the creation of ‘clusters’ of countries defined by a
set of common barriers to technological adoption.
Countries within each of these clusters or
‘convergence clubs’ converge to common levels of
mean per capita income. Nevertheless, the per
capita incomes across convergence clubs need never
converge and the polarization of per capita incomes
across countries may be permanent.
Growth Empirics
In both classes of models, therefore, the primary
concern is that countries may become separated -
perhaps permanently - into multiple growth regimes
corresponding to different levels of long-mn per
capita income. The fact that nonlinear growth
models imply that global inequality may be
persistent has sparked major advances in the area of
cross-country growth empirics. Driven by such
concerns, the central preoccupation of growth
empirics has been to evaluate the
Economic Growth Non-linearities
conditions under which poor countries catch up
with rich ones or fail to do so. Initial work along
these lines focused on the concept of ‘conditional
convergence’. Conditional convergence is said to
occur if permanent per capita income differences
between countries can be accounted for solely by
structural differences (and not initial conditions).
Researchers initially argued that because condi-
tional convergence was predicted by the canonical
neoclassical growth model (see Ramsey ; Solow ;
Swan ; Cass ; Koopmans ) whereas nonlinear
growth models potentially predict dependence on
initial conditions, tests for conditional convergence
could be used to discriminate between these classes
of theories.
Following Mankiw et al. ( ) and Barro and
Sala-i-Martin ( ), the canonical way such tests
were conducted was to first construct a linearized
version of the neoclassical growth model about the
(unique) steady state with average growth rates
across a time period as the dependent variable, and
measures of physical and human capital, population
growth rates, and initial per capita income as
covariates. Researchers then applied the linearized
neoclassical model to cross-country data with the
aim of testing to see whether the data supported a
negative coefficient on initial per capita income. A
finding of a negative coefficient on initial per capita
income was taken to imply that, conditional on
countries having similar structural characteristics
(as defined by the set of covariates), poorer
countries would close the income gap with the rich
- that is, conditional convergence.
An important outcome of the, oftentimes heated
(see Sala-i-Martin ), convergence debates of the
1990s was precisely to weaken the idea that such
tests of convergence could be interpreted as model
selection tests. In a highly influential work, Bernard
and Durlauf ( ) data to suggest that countries grouped according to
strongly disputed the interpretation of such ‘con-
ditional convergence’ tests by pointing out that
these tests were not able to discriminate against a
class of nonlinear growth theories that have dra-
matically different ergodic implications from the
neoclassical model. The class of models they were
referring to was developed by Azariadis and Drazen
( ). Azariadis and Drazen extended
the spillover models of Lucas ( ) and Romer
3319
( ) and showed that, if (local) nonconvexities
in the production function were sufficiently strong,
then countries that are similar in all aspects except
for initial conditions may nevertheless be organized
into multiple growth regimes, each of which
corresponds to a different steady state for long-run
per capita income.
Bernard and Durlauf showed that the multiple-
regimes Azariadis-Drazen model was theoretically
consistent with a finding of conditional convergence
in the data. Therefore, even in the narrowly
restricted sense of countries being structurally
similar, the finding of a negative coefficient to
initial income in the data was no guarantee that
countries would converge to a common steady state.
Galor ( ) lent further support to the
relevance of the Azariadis-Drazen model by arguing
that standard ways of augmenting the traditional
Solow model increased the likelihood that the true
data-generating process followed a multiple-
regimes rather than a single steady-state model.
Clearly, evidence of multiple regimes and
nonlinearities in growth raises questions about
misspecification in empirical studies that assume
that all countries follow the same growth process,
and casts doubt on inferences and policy recom-
mendations that are drawn from these studies.
The work by Bernard and Durlauf has spurred a
large quantity of research searching for the exis-
tence of multiple-growth regimes. One direction of
this new research has been to argue that the finding
of parameter heterogeneity in the neoclassical
model may be suggestive of the existence of
multiple growth regimes. In a seminal work,
Durlauf and Johnson ( ), employing a classi
fication and regression tree methodology,
implemented a version of Azariadis and Drazen’s
model and showed that there was evidence in the
initial per capita income and literacy rates cor-
respond to four different growth regimes. Their
work has inspired a long list of confirmatory works
using a wide variety of econometric approaches (for
example, Bloom et al. ; Canova ; Durlauf et al. ;
Kourtellos ; Liu and Stengos ; and Tan ).
While there now is a strong consensus in the
literature that there exists substantial heterogeneity
3320 Economic Growth Non-linearities

across countries, it should be emphasized that this overcome the convergence effect of diminishing
finding is only suggestive of multiple-growth regimes marginal returns to produce divergence and bimodality
and is not conclusive evidence of it. These in cross-country incomes nevertheless. Without an
heterogeneities could arise because of small deviations in explicit theory to explain the observed income
the specification of the production function (see divergence, there is also the question of whether the
Masanjala and Papageorgiou ) which need not bimodality in the cross-country income distribution is a
correspond to multiple- growth regimes. Further, even transitional or permanent feature of growth (see Galor ;
within the context of Azariadis-Drazen model, if non- Lucas ).
convexities in the production function are not strong
enough, the finding of parameter heterogeneity would
not imply the existence of multiple regimes (see Durlauf Conclusion
and Johnson , Figure 2).
An alternative approach to investigating the Nonlinearities in growth have been highly influen-
existence of multiple regimes or convergence clubs tial in shaping the thinking of both growth
has focused on the evolution of the world theorists and empiricists in recent years. The work
distribution of per capita income. The aim of this on multiple-growth regimes and the world income
research has been to look for evidence of emerging distribution suggests that there may exist growth
multimodality (typically, bimodality) in the world factors strong enough to overcome the decreasing
income distribution. A secondary aim has been to marginal productivity of the neoclassical produc-
evaluate the degree of churning within the tion function, thereby producing increasing
multimodal distribution. If the world income dis- inequality across countries. Nevertheless, while an
tribution is characterized by emerging multi- increasingly large body of work finds evidence that
modality with little evidence of countries moving freely is suggestive of growth nonlinearities, many
within the distribution (that is, churning), then this questions remain open and are the subject of current
finding would suggest, in a manner analogous research. What are the factors that are responsible
with the finding of multiple- growth regimes, that for generating multiple growth regimes or
global income inequality is real, intensifying and convergence clubs? Are the effects of these factors
persistent in nature. In fact, these are the precise transient or permanent? If the former, what are the
findings by Quah ( ). By applicable timescales? This area of research con-
estimating transition probabilities for the cross- tinues to be promising and fruitful.
country per capita income distribution, Quah finds
emerging ‘twin peaks’ in the world income
distribution as well as substantial persistence within See Also
the distribution. Quah’s seminal work has been
confirmed by subsequent work (for example,
Bianchi ; Fiaschi and Lavezzi ; and Paap and van
Dijk ) even though there had been questions about
the robustness of his initial methodology (see
Kremer et al. ).
While the findings of the ‘twin peaks’ literature
have been suggestive of growth nonlinearities and
multiple equilibria, it is not definitive. It is quite
possible, for instance, that the aggregate production
functions across countries actually exhibit
decreasing marginal productivity of capital, so that Bibliography
there is only one steady state. However, other
growth factors are sufficiently strong to Azariadis, C., and A. Drazen. 1990. Threshold externalities in
economic development. Quarterly Journal of Economics
105: 501-526.
Economic Growth, Empirical Regularities in
Barro, R.J., and X. Sala-i-Martin. 1992. Convergence.
Journal of Political Economy 100: 223-251.
Basu, S., and D.N. Weil. 1998. Appropriate technology and
growth. Quarterly Journal of Economics 113: 1025-1054.
Bernard, A., and S. Durlauf. 1996. Interpreting tests of the
convergence hypothesis. Journal of Econometrics 71:
161-173.
Bianchi, M. 1997. Testing for convergence: Evidence from
nonparametric multimodality tests. Journal of Applied
Econometrics 12: 393^109.
Bloom, D., D. Canning, and J. Sevilla. 2003. Geography and
poverty traps. Journal of Economic Growth 8: 355-378.
Canova, F. 2004. Testing for convergence clubs in income
per capita: A predictive density approach. International
Economic Review 45: 49-77.
Cass, D. 1965. Optimum growth in an aggregative model of
capital accumulation. Review of Economic Studies 32:
233-240.
Durlauf, S., and P. Johnson. 1995. Multiple regimes and
cross-country growth behavior. Journal of Applied
Econometrics 10: 363-384.
Durlauf, S., A. Kourtellos, and A. Minkin. 2001. The local
Solow growth model. European Economic Review 45:
928-940.
Fiaschi, D., and M. Lavezzi. 2003. Distribution dynamics and
nonlinear growth. Journal of Economic Growth 8: 355-
378.
Fleming, J.M. 1955. External economies and the doctrine of
balanced growth. Economic Journal 65: 241-256.
Galor, O. 1997. Convergence? Inferences from theoretical
models. Economic Journal 106: 1056-1069.
Galor, O., and D.N. Weil. 2000. Population, technology, and
growth: From the Malthusian regime to the demographic
transtition and beyond. American Economic Review 90:
806-828.
Howitt, P., and D. Mayer-Foulkes. 2005. R&D, implemen-
tation, and stagnation: A Schumpeterian theory of con-
vergence clubs. Journal of Money, Credit and Banking
37: 147-177.
Koopmans, T.C. 1965. On the concept of optimal growth. In
The econometric approach to development planning.
Amsterdam: North Holland.
Kremer, M., A. Onatski, and J. Stock. 2001. Searching for
prosperity. Carnegie-Rochester Conference Series on
Public Policy 55: 275-303.
Kourtellos, A. 2005. Modeling parameter heterogeneity in
cross-country growth regression models. Working paper,
Department of University of Cyprus.
Lewis, A. 1956. The theory of economic growth. London:
Allen & Unwin.
Liu, Z., and T. Stengos. 1999. Non-linearities in crosscountry
growth regressions: A semiparametric approach. Journal
of Applied Econometrics 14: 527-538.
Lucas, R. 1988. On the mechanics of economic development.
Journal of Monetary Economics 22: 3-42.
Lucas, R. 2000. Some macroeconomics for the 21st cen- tury.
Journal of Economic Perspectives 11: 159-168.
3321
Maddison, A. 2004. The world economy: Historical statis-
tics. Paris: Development Studies Centre, OECD.
Mankiw, N.G., D. Romer, and D. Weil. 1992. A contribution
to the empirics of economic growth. Quarterly Journal
of Economics 107: 407^137.
Masanjala, W., and C. Papageorgiou. 2004. The Solow model
with CES technology: Nonlinearities and parameter
heterogeneity. Journal of Applied Econometrics 19: 171-
202.
Murphy, K., A. Shleifer, and R. Vishny. 1989. Industriali-
zation and the big push. Journal of Political Economv
97: 1003-1026.
Nurkse, R. 1953. Problems of capital f'onnation in underde-
veloped countries. New York: Oxford University Press.
Paap, R., and H. van Dijk. 1998. Distribution and mobility of
wealth of nations. European Economic Review 42: 1269-
1293.
Parente, S.L., and E.C. Prescott. 1994. Barriers to technology
adoption and development. Journal of Political Economy
102: 298-321.
Quah, D.T. 1993. Empirical cross-section dynamics for eco-
nomic growth. European Economic Review 37: 426^434.
Ramsey, F. 1928. A mathematical theory of saving. Eco-
nomic Journal 38: 543-559.
Romer, PM. 1986. Increasing returns and long-run growth.
Journal of Political Economy 94: 1002-1037.
Rosenstein-Rodan, P. 1943. Problems of industrialization of
Eastern and South-Eastern Europe. Economic Journal
53: 202-211.
Rostow, W.W. 1960. The stages of economic growth. Oxford:
Oxford University Press.
Sala-i-Martin, X. 1996. The classical approach to conver-
gence analysis. Economic Journal 106: 1019-1036.
Scitovsky, T. 1954. Two concepts of external economies.
Journal of Political Economy 62: 143-151.
Solow, R. 1956. A contribution to the theory of economic
growth. Quarterly Journal of Economics 70(1): 65-94.
Swan, T.W. 1956. Economic growth and capital accumu-
lation. The Economic Record 32: 334-361.
Tan, C.M. 2005. No one true path to development:
Uncovering the interplay between geography, institu-
tions, and ethnic fractionalization in economic devel-
opment. Mimeo: Tufts University.
Economic Growth, Empirical Regularities
in
Steven N. Durlauf and Paul A. Johnson
Abstract
The evolution of economic growth theory
throughout the post-war period has been deeply
influenced by the effort to explain
3322
broad patterns in cross-country behaviour. We
discuss some of the salient empirical regularities
associated with neoclassical and new growth
economics and consider the shift in focus that
has occurred. We first describe the stylized facts
of Kaldor that played an important role in the
assessment of neoclassical growth models. Next,
we consider how a switch in focus to a different
class of regularities is associated with the new
growth economics that began in the 1980s and
dominates contemporary research.
Keywords
Economic growth; Elasticity of substitution;
Endogenous growth; Factor shares in national
income; Kaldor, N; Laboiu productivity;
Labour’s share of income; Model uncertainty;
Neoclassical growth theory
JEL Classification
047
The evolution of economic growth theory through-
out the post-war period has been deeply influenced
by the effort to explain broad patterns in cross-
country behavioiu. In this entry, we discuss some of
the salient empirical regularities associated with
neoclassical and new growth economics and con-
sider the shift in focus that has occurred. We first
describe the role of empirical regularities in neo-
classical growth theory as it emerged in the 1950s.
Next, we consider how a switch in focus to a
different class of regularities is associated with the
new growth economics that developed in the 1980s
and continues to dominate contemporary research.
Finally, we assess this shift. Durlauf et al. ( )
contains details of the data and
methods used to substantiate the claims made here.
Empirical Regularities and Neoclassical
Growth
Neoclassical growth theory is commonly associated
with Kaldor’s ( ) well-known ‘stylized
facts’ of long-run economic behavioiu, which
Economic Growth, Empirical Regularities in
primarily focused on the invariance of long run
behaviour for advanced economies. Four of his six
facts -(F) the constancy of the growth rate of output
per worker over long time horizons, (2) the
constancy of the growth rate of capital which is
lower than the growth rate of the laboiu supply, (3)
the absence of any systematic trends in the capital-
output ratio and (4) the constancy of the rate of
profit (and, by implication with the other facts,
factor shares in national income) - emphasize
common behavioiu across countries. Only the fifth
and sixth facts - the presence of substantial
differences in output per worker across countries,
and the positive relationship between the rate of
profit and the investment - output ratio - focus on
heterogeneity. Kaldor (1957) cites the prediction of
constant factor shares as an important test of
alternative growth models. An important empirical
study at the time was Klein and Kosobud (1961)
who investigated constancy by testing for a trend in
labour’s share, finding none using US data from
1900 to 1953.
While these facts are generally cited as a
motivation of neoclassical growth models, their
actual relationship to the theory is in fact more
complicated. In Solow ( ), for example, the
objective is the explanation of long-run economic
growth and the constancy of factor shares is only
mentioned in passing as an implication of the Cobb-
Douglas technology. Indeed Solow ( ) criticizes the
literature studying the con
stancy of factor shares for lacking a precise notion
of constancy given that exact constancy cannot
reasonably be expected. Bronfenbrenner ( )
argues that, for a wide range of values of
the elasticity of substitution between capital and
labour, and for reasonable variation in the capital-
labour ratio, the theoretical variation in factor shares
is consistent with that observed. He concludes that
the constancy or otherwise of factor shares is not
useful in the assessment of (distribution) theories.
Put differently, the first three of Kaldor’s stylized
facts seem most important to understanding the
motivation of the neoclassical program; Solow (
, p. 4) (in a
discussion originally published in 1970) remarks
that growth theory is largely
Economic Growth, Empirical Regularities in
devoted to analyzing the properties of steady states
and to finding out whether an economy not initially
in a steady state will evolve into one ...
How do Kaldor’s stylized facts appear from the
vantage point of modem empirical growth research?
Barro and Sala-i-Martin ( , Divergence in living standards over the 1960-
pp. 12—16) assess the concordance of Kaldor’s
stylized facts with the data and conclude that, with
the exception of the constant rate of profit, each of
the first five holds ‘reasonably well’ for developed
economies. They cite evidence suggesting some
tendency for the real rate of return to decline in
some economies. The evidence they present, and
that which we discuss below, shows that, at least as
far as it concerns the rate of growth of labour
productivity, the sixth of Kaldor’s facts also fits
well with the data.
Kaldor’s stylized facts are therefore of contem-
porary use in understanding long- ran output
behaviour. That said, the facts are no longer central
to the research efforts in growth economics as other
regularities (or the lack thereof) have become the
primary focus of research. We therefore turn to
those regularities that have become the focus of
contemporary work.
Empirical Regularities and the New
Growth Economics
The renaissance in growth theory associated with
the rise of endogenous growth models was
influenced by interest in the determinants of het-
erogeneity in growth experiences. While not usually
called stylized facts, there is a set of general
propositions about heterogeneity that have been
very important in influencing research. The most
prominent global features evident in the data are the
divergence in living standards over the past three
centuries and the large disparities in living
standards at the end of the twentieth century. By
modem standards, all countries were poor in 1700
but since then sustained growth, first in the United
Kingdom and parts of Western Europe, and more
recently in the United States and parts of the Asia-
Pacific region, has resulted in large crosscountry
differences in living standards. In 2000 average
GDP per worker in some countries was
3323
about one-fiftieth that in the United States while
more than 40% of the world’s population lived in
countries with average levels of GDP per worker of
no more than ten per cent of that in the United
States.
2000 period is also evident in the large group of
countries covered by the Penn World Tables (PWT)
(Heston et al. ). While a substantial group of
countries has exhibited prolonged growth over this
period, there remains a large mass of countries at
the bottom of the distribution. One result was a
hollowing out of the middle of the distribution - a
phenomenon labelled ‘twin peaks’ by Quah ( ;
).
Moreover, there is strong persistence within the
cross-country income distribution with a Spearman
rank correlation of 0.84 between GDP per worker in
1960 and that in 2000. This degree of correlation is
not peculiar to the PWT data. Easterly et al. ( )
report a rank correlation of 0.82
between GDP per capita in 1988 and that in 1870
for the 28 countries in Maddison ( ). This
sense of a lack of mobility is reinforced by Bianchi
( ), who found that very few of the
possible crossings from one end of the distribution
to the other actually occurred between 1970 and
1989.
The persistence in levels of GDP per worker
contrasts sharply with the wide cross-country var-
iation in the growth rates of GDP per worker
especially for those countries with relatively low
levels of GDP per worker in 1960. The data show
scant support for the proposition that the countries
of the world are converging to a common level of
income per person or for the belief that poor
countries have always grown slowly. Both growth
‘miracles’ - countries exhibited consistently strong
growth over the 1960-2000 period - and growth
‘disasters’ - countries that did poorly, often having
negative average growth rates - are present in the
data. East and South East Asian countries are well
represented among the former group while the later
is dominated by countries in sub-Saharan Africa.
Taiwan, for example, grew at an average annual rate
of over six per cent during this 40-year period and
increased GDP per worker by a factor of 11 in the
process. Hong Kong,
3324
Korea and Singapore were not far behind in either
respect. By contrast, Mauritania, Senegal, Chad,
Mozambique, Madagascar, Zambia, Mali, Niger,
Nigeria, the Central African Republic, Angola and
the Democratic Republic of the Congo all had
negative average growth over this period.
For most countries, the average growth rate
from 1980 to 2000 was lower than that from 1960 to
1980. The notable exceptions to this observation are
China and India. Moreover, past growth does not
seem to be a good predictor of future growth as, for
example, the correlation between growth in 1960-
1980 and that in 1980-2000 is just 0.40. Easterly et
al. ( ) sug
gest that the lack of persistence in growth rates
indicates the importance of good luck in economic
development. Nevertheless, the cross-decade cor-
relations in growth rates have tended to increase
during the 1960-1980 period, indicating a sorting of
countries into distinct groups of winners and losers.
There seems to be little relationship between the
1960 level of GDP per worker and subsequent
average growth rates. The cross-country dispersion
of growth rates tends to fall as initial income rises
largely due to the rarity of poor performance among
the countries with relatively high levels of GDP per
worker in 1960. There is, however, a close
relationship between geographical group
membership and economic growth between 1960
and 2000. As alluded to above, the countries of sub-
Saharan Africa performed poorly over this period,
with three-quarters of them growing at an average
annual rate of less than just 1.3%. The countries in
South and Central America did somewhat better
with three-quarters of them having grown at an
average of less than 1.5%. Among the East and
South East Asian countries, three- quarters grew at
an average rate of over 3.8%, and a similar fraction
of the South Asian countries grew at over 1.9%.
Many of the poor countries of the world were
unable to break out of stagnation between 1960 and
2000. A country growing at two per cent per year
for 40 years would enjoy a 120% increase in
income per person over that period. Yet, between
1960 and 2000, about a quarter of countries never
exceeded their 1960 income level by more than
Economic Growth, Empirical Regularities in
60%, and about ten per cent of countries never
exceeded their 1960 level by more than 30%. One
reason for this stagnation is the disposition of some
economies to large, abrupt output collapses. About
half of countries experienced a 3-year output
collapse of 15% or more between 1960 and 2000.
Over the same period, the largest 3-year output
collapse in the United States was 5.4%, and in the
United Kingdom 3.6%, both in 1979-1982.
In sum, there are large cross-countries dispar-
ities in GDP per worker and hence in living stan-
dards. These disparities have grown wider since
1960 and the middle of cross-country income
distribution has thinned since 1960. There is sub-
stantial immobility in a country’s position in the
distribution. Growth rates are much less persistent
and have tended to fall since 1980. In general, the
countries of sub- Saharan Africa performed poorly
over the 1960-2000 period. The countries in South
and Central America did somewhat better while the
South Asian countries did better still. The East and
South East Asian countries did best of all.
The Changing Empirical Focus of Growth
Economics
The two sets of empirical regularities we have
described, while appearing to differ greatly in terms
of their implications for understanding the
determinants of the growth process, may in fact be
reconciled. A key difference between neoclassical
and modem growth economics is its domain of
explanation: whereas neoclassical theory
attempted to understand the long-run behaviour of
advanced industrialized economies, the new growth
economics attempts to understand worldwide
growth patterns. As a result, the differences
between the advanced industrialized economies and
the rest of the world take on primary importance.
Lucas ( , pp. 2-3) describes his moti
vation as
to see whether modem growth theory could also be
adapted for use as a theory of economic development
Adaptation of some kind was evidently necessary:
The balanced path of growth theory, with
Economic Growth, Empirical Regularities in
constant income growth, and the assumed absence of
population pressures, obviously did not fit all of
economic history or even all the behavior that can be
seen in today’s world. The theory is, and was
designed to be, a model of the recent past of a subset
of countries.
Thus, as the domain of inquiry in growth economics
has evolved, the stylized facts of interest have
shifted to identifying features of international
divergence rather than international convergence.
Further, the effort to identify patterns that char-
acterize the differences in crosscountry growth
experiences has led to empirical research that
focuses on the identification of particular factors in
generating the divergence. Theoretical work in
growth economics moved away from the traditional
emphasis on factor accumulation and towards the
analysis of a wide range of social, historical,
geographic, and political factors as soinces of cross-
country heterogeneity. For example, a major strand
of contemporary research focuses on the ways that
institutional quality affects growth and
development; see Acemoglu et al. ( ) for a detailed
survey. The richness of
the modem growth literature has led to the wide-
spread use of regression methods to allow for the
simultaneous consideration of multiple growth
determinants, with a focus on identifying which
determinants in fact matter.
The move towards regression methods as the
basis for empirical growth research has altered the
nature of the sorts of regularities that link data and
theory. It is still the case that theoretical analyses
are often motivated by the identification of a
bivariate relationship between some factor of
interest and growth rates. However, relationships of
this type do not represent basic growth regularities
in the way that Kaldor’s stylized facts did. The
reason for this transition is that the different growth
factors that have expanded the domain of growth
economics are typically mutually consistent (Brock
and Durlauf ) and so the empirical significance of
one factor can only be assessed when others are
considered as well. Put differently, the finding of a
bivariate relationship, or lack thereof, can always be
rationalized as reflecting a failure to control for
other factors.
3325
As a result, the empirical regularities that mat-
ter for contemporary research, such as the coeffi-
cient relating a measure of institutional quality to
growth, are derivative from statistical analyses of
the entire growth process. But statistical models of
growth are subject to many forms of model uncer-
tainty, ranging from uncertainty about the appro-
priate theories to employ to uncertainty about the
empirical measurement of the qualitative factors
identified by a theory to uncertainty about the
details of the statistical specification of a model;
see Durlauf and Quah ( ) and Durlauf
et al. ( ) for a delineation of these issues.
Model uncertainty has meant that there is rela-
tively little consensus on the empirically salient
determinants of growth and so little consensus on
which regularities should be of primary interest.
Thus current growth economics has been
handicapped as different papers identify different
salient empirical regularities, with inadequate
attention to the robustness of such claims. The
development of sturdy inferences about the
growth process thus represents a very active area
of current work.
See Also
► iconom
► .
rivlv .
► Tro r
► >. el
Bibliography
Acemoglu, D., S. Johnson, and J. Robinson. 2005. Institu-
tions as the fundamental cause of long-run growth. In
Handbook of economic growth, ed. P. Aghion and
S. Durlauf. Amsterdam: North-Holland.
Barro, R., and X. Sala-i-Martin. 2004. Economic growth,
2nd ed. Cambridge, MA/London: MIT Press.
Bianchi, M. 1997. Testing for convergence: Evidence from
nonparametric multimodality tests. Journal of Applied
Econometrics 12: 393-409.
Brock, W., and S. Durlauf. 2001. Growth empirics and
reality. World Bank Economic Review 15: 229-72.
Bronfenbrenner, M. 1960. A note on relative shares and the
elasticity of substitution. Journal of Political Economy
68: 284-7.
Durlauf, S., and D. Quah. 1999. The new empirics of
economic growth. In Handbook of macroeconomics,
3326
ed. J. Taylor and M. Woodford. Amsterdam: North-
Holland.
Durlauf, S., P. Johnson, and J. Temple. 2005. Growth
econometrics. In Handbook of economic growth, ed. P.
Aghion and S. Durlauf. Amsterdam: North-Holland.
Easterly, W., M. Kremer, L. Pritchett, and L. Summers. 1993.
Good policy or good luck? Country growth performance
and temporary shocks. Journal of Monetary Economics
32: 459-83.
Heston, A., R. Summers, and B. Aten. 2002. Penn world
table version 6.1. Philadelphia: Center for International
Comparisons at the University of Pennsylvania (CICUP).
Kaldor, N. 1957. A model of economic growth. Economic
Journal 67: 591-624.
Kaldor, N. 1961. Capital accumulation and economic growth.
In The theory of capital, Proceedings of a Conference
held by the International Economic Association, ed. F.
Lutz and D. Hague. London: Macmillan.
Klein, L., and R. Kosobud. 1961. Some econometrics of
growth: Great ratios of economics. Quarterly Journal of
Economics 125: 173-98.
Lucas, R. 2002. Lectures on economic growth. Cambridge,
MA: Harvard University Press.
Maddison, A. 1989. The world economy in the 20th Century.
Paris: OECD.
Quah, D. 1996. Twin peaks: Growth and convergence in
models of distribution dynamics. Economic Journal 106:
1045-55.
Quah, D. 1997. Empirics for growth and distribution:
Stratification, polarization, and convergence clubs.
Journal of Economic Growth 2(1): 27-59.
Solow, R. 1956. A contribution to the theory of economic
growth. Quarterly Journal of Economics 70: 65-94.
Solow, R. 1958. A skeptical note on the constancy of relative
shares. American Economic Review 48: 618-31.
Solow, R. 2000. Growth theory: An exposition. Oxford:
Oxford University Press.
Economic Harmony
Israel M. Kirzner
This term has been introduced frequently into economic
discussion, and especially into discussions concerning
the history of economic thought. Yet there seems to be a
good deal of ambiguity as to what it is to mean.
Moreover, there has developed considerable
disagreement concerning the
Economic Harmony
centrality of the ‘harmony’ idea to the develop-
ment of economic thought, and similar disagree-
ment concerning the extent to which the classical
economists, in particular, are to be seen as
harmony-theorists. We will return a little later to
distinguish various different senses that have
been attached to the term ‘harmony’ in
economics. For each of these different senses,
however, acceptance of the harmony thesis has
been held to imply a favourable stance towards a
policy of laissez-faire. It is thus not surprising that
18th- century precursors of the notion of harmony
have been discovered in Cantillon and in Quesnay
(Schumpeter , p. 234). And we are not sur prised to
find some writers emphasizing the harmony ideas
they see in the classical economists, especially in
Adam Smith (Halevy —4, p. 89; Heimann , p. 65),
while others vehemently question the unqualified
identification of these writers with harmony theories
(Robbins , pp. 22-9; Samuels , pp. 6-8; Sowell , pp.
16f). It was in the middle of the 19th century that
the best-known writings appeared concerning
economic harmony. The term appeared in the title
of two books by the American economist Henry C.
Carey (Carey , ).
These works were followed by a general treatise
stressing the same theme (Carey -60). The term also
appeared in the title of a book by the French
economic writer Frederic Bastiat ( ).
For a (muted) defence of Bastiat against widespread
19th-century charges that his work in this respect
was a crude plagiarism of Carey, see Teilhac ( , pp.
100-113), who points to the
inspiration that both Carey and Bastiat received
from J.B. Say. Subsequent references to harmony
theories in economics generally tended to be crit-
ical, as economists began to argue (from the latter
decades of the 19th century into the 20th century )
for greater state intervention in market economies
on perceived grounds of economic efficiency or
economic justice. During most of the 20th century
economists, even when they have defended the
efficiency and justice of markets, have generally not
couched their arguments explicitly in terms of
harmony theory. Even Ludwig von Mises who, as
we shall see, was an important exception to this last
generalization, relegated the notion of
Economic Harmony
harmony to a distinctly subsidiary role in his
system. Recent re-awakened attention to 18th-
century theories of spontaneous order, especially as
rediscovered and expanded in the work of Hayek,
has not had the effect of reintroducing the term
‘economic harmony’ to current usage. We turn now
to take notice of the several different (although
certainly interrelated) senses in which this term has
been used during the history of economics.
Harmony as Flowing from Divine
Providence
A harmony ‘theory’ is not, in this sense, one that
flows out of economic science; rather it represents
an attitude of (usually religious) optimism and faith,
which itself suggests and guides the course of
scientific investigation.
Just as Kepler was inspired by the doctrine of
harmony in the spheres to discover the laws which
govern the orbits of the planets, so the early
economists were inspired by the doctrine that there
is a harmony of interests in a society to formulate
economic laws (Streeten
p. 208).
It was from this sense of the term that Lord
Robbins vigorously dissociated the classical school.
It was this optimistic doctrine that came to be
referred to contemptuously by the German term
‘Harmonielehre’. Archbishop Whately, who in 1832
set up a chair of political economy at Trinity
College, Dublin, was an influential harmony
theorist in this sense. He saw the purpose of the
chair as that of combatting the irreligious
implications, as he saw them, of Ricardian eco-
nomics. The early Dublin professors ‘were under
pressure to present an optimistic or harmonious
picture of how the market economy operates’ and
the resulting critical attitude towards Ricardian
theory reflected ‘these extrascientific concerns’
(Moss , p. 153). A variant of this approach to the
harmony doctrine was the Enlightenment view, in
which Deistic philosophy perceived a natural order
as responsible for ‘predetermined harmony’
(Mises , p. 239; Heimann , P- 49).
3327
Harmony Theory as the Doctrine
of Maximum Satisfaction
When major neoclassical economists such as Mar-
shall ( , p. 470) and Wicksell ( , p. 73)
referred to harmony theorists, they evidently had in
mind those who believed that economic theory
demonstrates that free competitive markets generate
maximum total satisfaction for society as a whole.
‘Harmony theory’ thus referred to a very specific
conclusion of economic science, a conclusion
central to welfare economics, but a conclusion
whose validity both Marshall and Wicksell were
concerned to refute. Of special concern, in this
context, was the issue of whether the new marginal
utility doctrines had been successfully deployed by
Jevons, or by Walras, to arrive at ‘harmony’
conclusions similar to those that had been reached,
on other grounds, by Bastiat.
Parallel to this sense of harmony was that which
attributed ethical virtues to the distributive results
of competitive markets. Thus J.B. Clark’s
demonstration of the justice of marginal-
productivity incomes is seen as ‘harmony doctrine’
(Myrdal , p. 148).
Harmony Doctrine as the Denial
of Class Conflict
One sense in which harmony doctrines have been
understood throughout the history of economics is
that in which it is sought to demonstrate the mutual
compatibility of the interests of the various
individuals and groups in society. In particular,
such doctrines tend to dismiss the notion of inherent
class conflict under capitalism. A 20th-century
economist who has himself emphasized this idea of
harmony of interests in the market society, put the
genesis of this idea as follows:
When the classical economists [asserted ‘the
theorem of the harmony of the rightly understood
interests of all members of the market society’ they
were stressing] two points: First, that everybody is
interested in the preservation of the social division
of labour, the system that multiplies the
productivity of human efforts. Second, that in the
3328
market society consumers’ demand ultimately
directs all production activities (Mises p. 674).
Mises, indeed, saw these ideas as important
results of economic science, having wide applica-
tion. ‘There is no conflict between the interests of
the buyers and those of the sellers, between the
interest of the producers and those of the con-
sumers’ (Mises , p. 357). Only in the special case of
resource monopoly ownership may it happen that
the ‘emergence of monopoly prices... creates a
discrepancy between the interests of the monopolist
and those of the consumers’ (Mises
, p. 680).
Harmony and the Spontaneous Order
Tradition
Since the early 1940s F.A. Hayek has succeeded in
drawing the attention of economists and others to a
line of social analysis since the 18th century, an
approach often termed the ‘spontaneous order
tradition’. The emphasis, in this tradition, is on the
evolution of institutions and social outcomes ‘which
are indeed the results of human action, but not the
execution of any human design’ (Ferguson , p. 187,
cited in Hayek , p. 96). There is no doubt that the
term ‘economic harmony’ has often been applied as
an expression of belief in the possibility and social
benignity of undesigned social outcomes. To some
extent, of course, this sense of the term overlaps
those listed above, but the emphasis here is not in
the denial of conflict, not on any particular welfare
theorem, certainly not on any religiously based
optimism, but on the counter-intuitive possibility of
orderly results emerging without deliberate design
from the spontaneous interplay of independently
acting individuals. ‘Order’ in this context has come
to mean ‘mutually reinforcing expectations’. The
following reference to this notion of harmony
expresses this usage of the term:
The great general mle governing human action at
the beginning, namely that it must conform to fair
expectations, is still the scientific rale. All the forms
of conduct complying with this rule are consistent
with each other and become the
Economic Harmony
recognized customs. The body of custom there-
fore tends to become a harmonious system (Carter
, p. 331, cited in Hayek , p. 169).
The above survey has been confined to notions
of economic harmony believed to be achieved
spontaneously, ‘naturally’, without design. For the
sake of completeness it should perhaps be noted
that the term ‘harmony’ has occasionally been used
to describe the objective of deliberate social policy.
Thus a well-known debate was initiated by E.
Halevy in his claim that Bentham and the
philosophical Radicals subscribed to two partly
contradictory principles: the ‘economic’ principle of
‘natural identity’ (i.e. harmony) of interests, and the
‘juristic’ principle of the ‘artificial identification of
interests’ (Halevy -4, pp. 15, 17, 489). Lord
Robbins, in disputing Halevy concerning any
contradiction in the Benthamite position, refers to
the juristic principle as contending it to be ‘the
function of the legislator to bring about an artificial
harmonization of interest’ (Robbins , pp. 190f).
While occasional references may be found to
harmony sought to be artificially accomplished, the
term has, in general, been associated almost
invariably with harmony achieved undeliberately in
a decentralized system.
See Also
References
Bastiat, F. 1850. Les harmonies economiques. Paris:
Guillaumin.
Carey, H.C. 1836. The harmony of nature. Philadelphia:
Carey, Lea & Blanchard.
Carey, H.C. 1852. The harmony of interests, agricultural,
manufacturing, and commercial, 2nd ed. New York:
Myron Finch.
Carey, FI.C. 1858-60. Principles of social science. Phila-
delphia: J.B. Lippincott.
Carter, J.C. 1907. Law, its origin, growth and function.
New York/London: G.P. Putnam’s Sons.
Ferguson, A. 1767. An essay on the history of civil society.
London.
Flalevy, E. 1901-4. The growth of philosophic radicalism.
Translated from the French by M. Morris, 1928. Boston:
Beacon, 1955.
Economic History 3329

Hayek, F.A. 1967. Studies in philosophy, politics and eco- applying theory to history. But its emphasis on
nomics. Chicago: University of Chicago Press. data analysis retained a bridge to older tradi-
Hayek, F.A. 1973. Law, legislation and liberty, Rules and
Order, vol. I. Chicago: University of Chicago Press. tions. As economists have rediscovered an
Heimann, E. 1945. History of economic doctrines. An interest in long-term economic growth, often
introduction to economic theory. New York: Oxford applying traditional institutional approaches,
University Press.
there is continuing evidence of rapprochement.
Marshall, A. 1920. Principles of economics, 8th ed. London:
Macmillan, 1936.
Moss, L.S. 1976. Mountifort Longfield: Ireland s first pro- Keywords
fessor of political economy. Ottawa: Green Hill. American Economics Association; Anthropo-
Myrdal, G. 1932. The political element in the development of
economic theoty. Translated from the Gemran by P.
metric history; Cliometrics; Economic devel-
Streeten. Cambridge, MA: Harvard University Press, opment; Economic growth; Economic history;
1954. Economic History Association in the United
Robbins, L. 1952. The theory of economic policy in English States; Historical School (German); Great
classical political economy. London: Macmillan, 1965.
Samuels, W.J. 1966. The classical theory of economic policy.
Depression; Industrial Revolution; Kuznets, S.;
Cleveland/New York: World. Marshall, G.; Marx, K.; Mathematical models;
Schumpeter, J.A. 1954. History of economic analysis. New Mercantilism; New economic history; North, D.;
York: Oxford University Press. Physiocrats; Royal Economic Society; Smith,
Sowell, T. 1974. Classical economics reconsidered.
A.; Total factor productivity; Weber, M
Princeton: Princeton University Press.
Streeten, P. 1954. Recent controversies. Appendix to Myrdal
(1932).
Teilhac, E. 1936. Pioneers of American economic thought in
the nineteenth centuiy. Translated from the French by
E.A.J. Johnson (1936), reprinted, New York: Russell and JEL Classifications
Russell, 1967 NO
von Mises, L. 1949. Human action: A treatise on economics,
3rd ed. Chicago: Regnery, 1966. Economic history is a sub-discipline within eco-
Wicksell, K. 1901. Lectures on political economy, Vol. I. nomics and, to a lesser degree, within history,
Translated from the Swedish by E. Classen. London:
whose main focus is the study of economic growth
Routledge and Kegan Paul, 1934.
and development over time. It is to be distinguished
from the history of economic thought, a branch of
intellectual history .
Studies in economic growth, whether historical
or contemporary, develop and analyse quantitative
Economic History
measures of increases in output and output per
Alexander J. Field capita, emphasizing in particular changes in saving
rates and rates of technological innovation and their
consequences. Economic development is a larger
and more encompassing rubric, also including
consideration of the role of cultural changes and
Abstract
changes in formal institutions.
Economic history focuses on the historical study Economic history has its origins in two main
of growth and development. Originating in the traditions. The first is the German historical school,
German historical school and studies of the a group of scholars in the 19th and early 20th
Industrial Revolution in England, it became centimes, including Gustav Schmoller and Max
professionally differentiated from economics Weber, who ranged widely over human history with
proper with the establishment of associations in special emphasis on the consequences of
Britain (1926) and the United States (1941). As institutional variation for economic as well as
economics continued on its increasingly political performance. The second tradition stems
mathematical and ahistorical path in the 1960s, from the efforts of a group of writers who
the ‘new economic history’ advocated
3330
viewed the complex of innovations in steam power,
iron manufacture, and textiles in late 18th-century
Britain as an epochal event - an industrial revolution
- equivalent in its significance for human welfare to
the Neolithic revolution which gave birth to
agriculture around ten millennia earlier. The study
of the causes, dimensions and consequences of the
emergence of sustained increases in per capita
incomes - what Simon Kuznets ( ) called
modem economic
growth - along with a focus on the consequences of
institutional variation, continues to define much of
what economic historians do.
Although historians have practised their craft at
least since the time of Herodotus (the fifth century
BC), economics emerged as a separate social
science with the work of Adam Smith or, perhaps,
as some have argued, that of the Mercantilists and
the Physiocrats. Classical economists, with the
notable exception of Ricardo, were almost all also
historical economists. The reader of Smith, Mill,
Marx, or even Marshall ploughs through thick
volumes in which propositions in economic theory
are embedded in often lengthy descriptions of
historical events or the course of economic history.
Throughout most of the 19th century, the divide
between economists and economic historians was
weak.
With the professionalization of economics that
picked up speed in the 20th century (the American
Economics Association was founded in 1885; the
Royal Economic Society in 1890), economic history
began to emerge as a distinct and to some degree
separate sub-discipline. The Economic History
Society was founded in Britain in 1926; the
Economic History Association in the United States
in 1941. The trend towards a separate identity
accelerated in the third quarter of the 20th century,
with the increasing emphasis within economics on
formal mathematical modelling and the weakening
within the general profession of ties to historical
traditions. In the economic history societies, in con-
trast, those trained as historians as well as
economists remained active; in Britain, distinc-
tiveness was accentuated by the establishment of
separate university departments of economic
history.
Economic History
As the intellectual paths taken by economics and
economic history seemed increasingly to diverge, a
countervailing intellectual movement known
variously as ‘cliometrics’ or the ‘new economic
history’ emerged. Its pioneers knew then- history,
but emphasized by argument and example that, if
economic history was to remain influential within
economics, it had to make more use of formal
models as well as place increased emphasis on
quantitative (rather than just qualitative) data and
more advanced statistical techniques (econometrics)
to analyse them.
The use of mathematical models was anathema
within historical traditions, but by the 1960s widely
accepted in economics. Thus, the new economic
history represented something of a gauntlet thrown
down to those trained in history or allied with its
traditions. The push for quantitative data analysis,
in contrast, was more crosscutting in the challenges
it implied. Many traditional economic historians
had in fact examined such data, although the
statistical techniques they used were often quite
rudimentary. Within some economic circles, on the
other hand, an emphasis on data was becoming
suspect. Here, some scholars were comfortable with
the evolution of economic theory as a branch of
applied mathematics, constrained and judged by the
rules of logic and consistency, but governed in its
realism, if at all, by intuition rather than systematic
empirical inquiry.
The effort to force formal theory upon tradi-
tional economic history often lacked acknowl-
edgement that the relation between economic
history and formal theory might usefully be a two-
way street. The emphasis on data analysis, in
contrast, offered a bridge between economics and
economic history. It helped reaffirm within
economics the importance of empirical inquiry, and
encouraged those historically trained to become
more sophisticated in their statistical analyses.
Nevertheless, the stress on quantitative data
could not help but draw attention away from eco-
nomic history’s traditional concern with legal and
institutional variation, where the source documents
were almost uniformly qualitative. How would this
theme, one of the defining features of
Economic History
economic history since its inception, survive the
new economic history? The initial ‘solution’ was to
try to make institutions endogenous. Blending a
mix of influences from technologically determin-
istic Marxism to the emerging law and economics
and public choice literatures, a number of scholars
suggested that institutions could be understood as
epiphenomenal: reflective of more fundamental
givens. The high point of such efforts was probably
the short book by North and Thomas ( ).
These efforts, however, gradually disintegrated
under the force of the ad hoc twists required to
make the framework consistent with known his-
torical evidence (Field ), and even proponents such
as North eventually backed away from this agenda.
Formal rules often vary where technologies and
endowments are similar, and are often similar when
more fundamental givens differ, and such variation
has consequences for economic performance. Had
the endogenization initiative been successful, it
would have eliminated from economic history one
of the most important perspectives it offers to
general economics.
The old economic historians had taken it as
obvious that, at critical historical junctures, changes
in formal institutions such as laws or constitutions
had powerful influences on the course of a country
or region’s economic development, and that these
changes were not always predictable ex ante. The
breakdown of the former Soviet Empire, and the
opportunities afforded to Western scholars actually
to influence the design of formal institutions, gave a
powerful impetus to returning to thinking about
such designs as consequential, and increasingly this
perspective came to be reflected in research by
scholars who did not necessarily think of
themselves as economic historians.
If the main subject of economic history con-
tinues to be the history of economic growth and
development, the influence of variations in formal
and informal institutions in both the private and
public arenas will remain an important theme.
These institutions and a broader economic culture
help structure the environment in which individuals
pursue their interests. But the success of an
economy in raising output and output per person
3331
also depends on available technologies, on the size,
composition, and characteristics of the labour force,
on natural resources, and on the accumulation of
physical capital. The study of the evolution of these
inputs suggests some of the other themes around
which economic historians organize their work. In
particular, there is a rich tradition, particularly in the
United States, examining issues in and applying
methods from modem labour economics within an
historical context.
The basic agenda of economic history has not
changed since the first edition of The New Pal-
grave. Interest in the causes and dimensions of the
Industrial Revolution, for example, remains strong,
particularly in Britain. But the field has evolved in
new directions, with several discem- able trends.
First, scholars have concerned themselves with a
broadening range of topics under the umbrella of
growth and development. In the 1960s and 1970s,
especially in the United States, railroads and slavery
dominated much of the discussion. In recent
decades, it is not possible to point to one or two
issues around which research and discussion has
coalesced to the same degree. Instead, there have
been a number of new initiatives; one example
would be the growing exploitation of
anthropometric data to make inferences about
variation in standards of living.
Associated with this has been a broadening of
the scope of the discipline, both in terms of the
countries in which economic history research is
conducted and in the geographical range of topics,
which extends, somewhat more so than in earlier
decades, beyond Western Europe and North
America to Asia, Latin America, Australia, and
Africa. One illustration of this has been a range of
cross-national studies, exploring such issues as
economic convergence.
A third trend has been a growing willingness to
think of the 20th century as an historical epoch in its
own right. When the new economic history began,
the Second World War had barely ended and the
Great Depression was recent history. The main
focus of research was the 18th and 19th centuries.
Treating the 20th century as an historical period
promises to reduce the gap between economics and
economic history. The Great
3332 Economic Impact of the Olympic Games

Depression, of course, continues to attract attention, cultural variation. Doctoral training with a spe-
but interest in the 20th century is beginning to cialization in economic history is well suited to
expand beyond this. The data and events of recent imparting such knowledge and the skills for
decades can now more easily be seen in an acquiring it, capabilities that will remain essential
historical context. The result can be a smoother in developing improved theory and policy in the
continuum between topics understood as economic area.
history and the analysis of contemporary data.
Placing more recent developments within a
longer-mn perspective has already begun to pay See Also
important dividends. Many trends that economists
and economic historians expected at mid-century
would characterize the 20th century as a whole
moderated, became erratic, or in some cases
reversed themselves in the last quarter of the cen-
tury (Field ). In 1950, for example, it looked as if
the United States (and other countries) would
continue to experience decreases in wealth and
income inequality, robust and perhaps rising shares
of union membership in the labour force, a growing Bibliography
role for government, and a continuing high
Caimcross, A. 1989. In praise of economic history. The
contribution of total factor productivity (TFP) Economic History Review 42: 173-185.
growth to growth in output per hour. In fact, Field, A. 1981. The problem with neoclassical institutional
inequality has generally increased, union economics. Explorations in Economic Histoty 18: 174-
198.
membership has fallen, and TFP growth basically
Field, A. 1987. The future of economic histoty. Boston:
disappeared in much of the developed world Kluwer-Nijhoff.
between 1973 and the 1990s. The size and role of Field, A. 2001. Not what it used to be: The Cambridge
government, which many predicted would continue economic history of the United States, vols. II and HI.
Journal of Economic History 61: 806-818.
to expand, has in fact displayed a more complex
Kuznets, S. 1966. Modern economic growth: rate, structure,
dynamic. and spread. New Haven: Yale University Press.
A fourth and related trend has been a McCloskey, D. 1987. Econometric history. Basingstoke:
reinvigoration within mainstream economics of Macmillan Education.
North, D., and R. Thomas. 1973. The rise of the Western
interest in what has always been a primary subject
world. Cambridge: Cambridge University Press.
of economic history: economic growth. Much of Solow, R. 1985. Economic history and economics. American
economic theory in the 1950s and 1960s modelled Economic Review 75: 328-331.
production and allocation within a static economy.
The revived interest in the study of growth, com-
bined with the growing willingness of economists to
adopt traditional institutional approaches, reflects
the persisting influence of the original concerns and Economic Impact of the Olympic Games
approaches of economic history within the larger
profession. Andrew Zimbalist
Whatever the labels people apply to themselves
and others, if we want better understanding of the
processes of growth and development, we will
continue to need scholars familiar with how to work
with data and interpret the influences on economic Abstract
outcomes of institutional, political, and The Olympic Games are among the largest and
most visible sporting events in the world. Every
two years, the world’s best athletes
Economic Impact of the Olympic Games
from some 200 countries come together to
compete in lavish new venues in front of thou-
sands of spectators. Hundreds of millions of
sports fans worldwide watch the Games on
television. Although Pierre de Coubertin, who
founded the modem Olympics in the late 19th
century, may have had altruistic, idealistic
notions of pure amateur competition, unsullied
by financial motivations, the Olympic Games
have become a big business. The participants
are effectively professional athletes; the orga-
nizers are highly compensated, professional
bureaucrats; hosting the Games involves huge
construction and renovation projects that take
nearly a decade to complete, and these expen-
ditures are usually justified by claims of
extraordinary economic benefits that will accrue
to the host city or region as a direct result of
hosting the Games. This article examines the
financing of the Olympic Games, explores how
the awarding of the Games has become a high-
stakes contest, and analyzes the costs of running
the Games and their economic impact on the
host city and nation.
Keywords
Economics of sport; International Olympic
Committee (IOC); Major events; Multiplier;
Olympic bid; Olympic Games; Sport finance;
Sport infrastructure
JEL Classifications
L31; L83; L88; R1
Financing the Olympics
The modem Olympic Games began in 1896, but it
was not until 1976 that a watershed event shook up
the financing model for the Games and set the
Olympics on its current economic course. In that
year the city of Montreal hosted the Summer
Games, which were originally predicted to cost
$124 million. In fact, Montreal incurred a debt of
$2.8 billion (approximately $10 billion in 2010
dollars) that was only finally paid off in 2005
3333
(Burton 2003). Annual debt service created a large
budgetary hole for the city for three decades.
By the end of the Montreal Games, the 1980
Olympics had already been set for Moscow, but no
city wanted to bid for the right to host the 1984
Games. After some scrambling, Los Angeles agreed
to host the Games, but only on the condition that it
took on no financial ob I igati on. With no
alternative, the International Olympic Committee
(IOC) accepted this condition and Los Angeles was
awarded the 1984 Summer Games on 1 July 1978.
That year also marked the first significant
relaxation of Olympic amateur rales under then IOC
president Lord Killanin. Rule 26 of the Olympic
Charter was modified so that athletes were allowed
openly to earn money from endorsements, if the
money went to their national sports federation or
their country’s National Olympic Committee
(NOC). The receiving organization was then
permitted to pay the athlete’s expenses, including
‘pocket money’. ‘Broken-time’ payments for time
away from the athlete’s regular job were also
authorized if the athlete had a regular job. But the
rale continued to declare that professional athletes
were ineligible.
During the 1980-2001 reign of IOC President
Juan Antonio Samaranch complete professionali-
zation and commercialization of the Olympics were
realized. In 1982, the amateur rales were revised to
permit payments into a trust fund that provided
expenses during the athlete’s active career - and
substantial sums thereafter. Eventually, decisions
about accepting professionals were left to the
International Federation (IF) of each sport. The new
professional era was heralded during the 1992
Games in Barcelona, when the USA sent its ‘Dream
Team’ of National Basketball Association (NBA)
stars which went on to win the gold medal.
Nominally, for the 2004 Games in Athens, boxing
was the only sport that did not accept professionals,
but even this distinction is dubious, because the
National Olympic Committees (NOCs) of many
countries gave their boxing medalists cash prizes.
These changes also led to increased commer-
cialization and increased TV and sponsorship
money, which in turn led to corruption and
3334
scandal within the IOC. Before 1980, the 112 IOC
representatives had to pay their own way to cities
bidding for the Games. Within a year they were
receiving first-class tickets and all expenses, as well
as lavish entertainment. Outrageous tales of the
excesses enjoyed by IOC representatives abounded.
Revelations of bribery and cormption around the
Salt Lake City bid for the 2002 Games plunged the
IOC into scandal: six members were expelled, four
resigned and 10 were warned. Since then, the IOC
has reformed itself by reducing the number of
voters and by officially declaring an outright ban on
gifts.
Meanwhile, the modest financial success of the
1984 Games in Los Angeles led to a new era of
international competition among cities to host the
Games. The relative success of Los Angeles, how-
ever, was sui generis. Los Angeles had very little
construction expense and the chair of the Los
Angeles Organizing Committee for the Olympic
Games (LAOCOG), Peter Ueberroth, was able to
raise substantial sums by selling sponsorships to
corporations. LAOCOG generated a modest surplus
(just over $300 million) and reset the Olympic
financial model for less public and more private
financing.
Nonetheless, other host cities have found it
impossible to procure the same proportion of pri-
vate support and have relied upon large public
expenditures. Several billion dollars of public
monies were committed in Seoul (1988), Barcelona
(1992), Nagano (1998), Sydney (2000), Athens
(2004) and Beijing (2008). In some cases, the local
OCOG ran a modest surplus (20% of any surplus
must be shared with the IOC; Preuss , guarantee to cover cost overruns) and the federal
p. 194), but the local government laid out billions of
dollars to help finance the activities of the OCOG.
In the case of Athens, for instance, the public
investment exceeded $10 billion - some of this
public investment resulted in improved, more
modem infrastructure for the city, but some of it
resulted in white elephants. Many facilities built
especially for the Games go unused or underutilized
after the period of Olympic competition itself, while
requiring tens of millions of dollars annually to
maintain and occupying increasingly scarce real
estate. Public investment
Economic Impact of the Olympic Games
for the 2008 Beijing Olympics exceeded $40
billion.
Salt Lake City Olympiad chief and former
governor of Massachusetts Mitt Romney questioned
whether US cities should enter bids to host the
Olympic Games, stating that they were increasingly
driven by ‘giganticism’ with the addition of new
sports and more frills.
Present Day Financial Arrangements
The IOC presents the financing of the Olympic
Games in terms of the related organizations: the
local organizing committee (OCOG), the NOC, the
IFs, and itself. The OCOG budget is not the same as
the budgetary impact on the local city that hosts the
games. The local city and its regional and national
government may provide billions of dollars of
subsidies to the OCOG, and the OCOG may report
a surplus. This surplus has little meaning regarding
the budgetary impact on public bodies from hosting
the Games.
For instance, for the recent Games hosted in the
USA, the federal government provided $1.3 billion
in Salt Lake City in 2002; $609 million in Atlanta
in 1996; and $75 million in Los Angeles in 1984
(all reckoned in 1999 prices) (Ungar , p. 5). For the
2010 Winter Games in Vancouver, in addition to
the provincial government of British Columbia and
the federal government of Canada each budgeting
$9.1 million to help finance the bidding process, the
provincial government was scheduled to put up
$1.25 billion to finance the Games (and provide a
government was budgeted to contribute another
$330 million. The city of Vancouver was budgeted
to contribute $170.3 million (
accessed 22 August 2007). Not surprisingly,
financing did not work out as planned, in part due to
the worldwide recession of 2008-09. In fact, the
IOC provided a $423 million subsidy to the
Vancouver Organizing Committee (VANOC) and
‘sources said that the IOC agreed to the first- of-its-
kind bailout because without it, spending for the
games would have come to a screeching halt and
major cutbacks would have been
Source
1993-96 1997-2000
Broadcast Economic Impact of the Olympic
1,251 Games 1,845
TOP programme 279 579
Economic Impact of the Olympic
Dom. Sponsorship 534 Games, Table655
1 Olympic movement revenue
Ticketing 451 625
Licensing 115 66
Total 2,630 3,770
Sources'. IOC, 2006 Olympic Marketing Fact File,
File, p. 26; IOC, 2008, Media Marketing Guide, p. 4
n e c e s s a r y ’ ( Sports Business Daily ).
VANOC also received an $87 million public bailout
loan (Sports Business Journal ) and
Standard & Poor’s lowered the city of Vancouver’s
credit rating due to Olympic financing shortfalls,
raising the city’s borrowing costs (Sports Business
Daily ). The New York
Times reported that: ‘The immediate legacy for this
city of 580,000 is a nearly $1 billion debt from
bailing out the Olympic Village development.
Beyond that, people in Vancouver and British
Columbia have already seen cuts in services like
education, health care and arts financing from their
provincial government, which stuck with many
other Olympics-related costs’ (Austen )•
Moreover, it is common practice for the OCOG
budget to consist entirely (or almost entirely) of
operating, as opposed to capital, expenditures
(Preuss , p. 195). Nonetheless, to the extent that the
OCOG receives funding from the IOC or from
private sources, the lower will be the financing
burden that falls on the local, state and national
government that hosts the Games. What follows,
then, is a discussion of how the IOC distributes the
revenue that is collected from the staging of each
Olympic Games.
Table presents the total revenue that accrues to
the IOC or any of its constituent organizations
during each quadrennial Olympic cycle, consisting
of one Winter and one Summer Games, ft shows a
healthy revenue growth in each of the major
categories, with television revenues the largest
single source of revenue by a factor of 3. The TOP
marketing program consists of 11 companies that
hold exclusive category sponsorships as the official
Olympic company.
2001-04 2005-08
2,230 2,570 3335
663 866
796 (current US $ millions)
1,555
411 274
87 185
4,187 5,450
, p. 16; IOC, 2010 Olympic Marketing Fact
TOP (The Olympic Partner) Programme reve-
nues go 50% to the local OCOGs, 40% to the NOCs
and 10% to the IOC (International Olympic
Committee ). Broadcast revenue goes 49% to
the host OCOG and 51% to the IOC, which in turn
distributes the lion’s share of this revenue to the
NOCs and IFs. Prior to 2004, the host OCOGs
received 60% of broadcast revenue. Beginning in
2012, it has been determined that OCOGs will
receive a fixed amount, rather than a fixed per-
centage, as broadcast revenues continue to rise
(Preuss , p. 100). Overall, the IOC has retained 8%
of Olympic revenue; the remaining 92% has been
shared by the OCOGs, NOCs and IFs.
Table depicts the astronomical growth in
television broadcasting revenue for the Summer and
Winter Games since 1960. Not surprisingly, the
largest share of broadcast revenue comes from the
USA. For instance, for the 2004 Athens Games, the
IOC contract with NBC yielded $793.5 million, or
53.1% of the total. Following the US rights fee
were Hu rope ($394 million), Japan ($155 million),
Australia ($50.5 million), Canada ($37 million) and
South Korea ($15.5 million). All told, there were 80
rights holders televising the Athens Games to 220
countries and 2 billion potential viewers worldwide.
Ten thousand media personnel were on hand to
cover the Games (International Olympic Committee
,pp. 51-54).
The US share of total media rights has trended
downward over time, from 83.4% during 1986-89,
to 60% during 2001-04, to 52.6% dining 2009-12
(Sports Business Journal ). Because of the high US
share, the US Olympic Committee (USOC) has
received a
Summer Winter
Olympic Games Broadcast revenue Olympic Games Broadcast revenue
1960 Rome 1.20 1960 Squaw Valley 0.05
1964 Tokyo 1.58 1964 Innsbruck 0.94
1968 Mexico City 9.75 1968 Grenoble 2.61
1972 Munich 17.79 1972 Sapporo 8.48
1976 Montreal 34.86 1976 Innsbruck 11.63
1980 Moscow 87.98 1980 Lake Placid 20.73
1984 Los Angeles 286.91 1984 Sarajevo 102.68
1988 Seoul 402.60 1988 Calgary 324.90
1992 Barcelona 636.06 1992 Albertville 291.93
1996 Atlanta 898.27 1994 Lillehammer 352.91
2000 Sydney 1,331.55 1998 Nagano 513.49
2004 Athens 1,494.03 2002 Salt Lake 738.00
2008 Beijing 1,739.00 2006 Torino 831.00

Sources'. IOC, 2006 Marketing Fact File, p. 46; IOC, 2010 Market Fact File, p. 27; IOC, 2008, Media Marketing Guide, pp.
4, 6

disproportionate share of the total collected fees.
Out of 205 NOCs in 2009, USOC received 12.75%
of all media rights fees from the Olympics. (This
share had been 10% until 1996, when it was
raised. ) In 2009, after sharp disagreement, a
negotiation between USOC and the IOC led to a
new agreement that in 2020 the USOC share would
be lowered again. The new level, however, was not
agreed upon.
OCOGs do not cover all their expenses from the
above sources. For instance, the Nagano OCOG in
1998 had revenues of $990 million, of which
approximately $435 million came from the IOC.
Similarly, the Salt Lake OCOG had revenues of
$1,348 billion, of which approximately $570
million came from the IOC (International Olympic
Committee , pp. 82-83; from the Salt Lake Games
revenues, the IOC also provided $305 million to the
NOCs.).
Economic Results
Economic theory would suggest that any expected
local economic benefit would be bid away as cities
compete with each other to host the Games. More
precisely, with perfect information the city with the
highest expected gain could win the Games by
bidding $1 more than the expected gain to the
second highest city. Such an outcome could yield a
small benefit to the winning city, but this would
require perfect information and an open market
bidding process. In fact, the bidding process is not
done in dollar amounts, but comes rather in the
form of providing facilities and guaranteeing
financing and security. In the post- 9/11 world,
security costs are far from trivial. Total security
costs in Athens in 2004 came to $1.4 billion, with
40,000 security people; Beijing in 2008 was
projected to have over 80,000 security personnel
working the Games.
It is also widely acknowledged that the bidding
process is laden with political considerations.
Moreover, the bidding cities are more likely to be
motivated by gains to particular private interests
within the city (developers, construction companies,
hotels, investment bankers, architects, real estate
companies, etc.) than by a clear sense that the city
as a whole will benefit economically.
In contrast, the IOC views its principal role as
promoting sport, not economic development. It
requires buildings and infrastructure to be financed
with non-Olympic money (Preuss , p. 195).
Accordingly, even though a local OCOG may
break even or have a small surplus, the greatest
likelihood is that the city itself (and state and
national governments) experiences a fiscal deficit
Economic Impact of the Olympic Games
from the Games (of course, not all OCOGs manage
to break even: the Albertville OCOG lost $57
million: Burton 2003, p. 3). On the one hand, the
only tax revenue that would accrue to host gov-
ernmental bodies would be from incremental sales
and income resulting from hosting the Olympic
Games. (Other taxes, such as real estate taxes,
might come into play depending on the local tax
system and whether or not the Games affected real
estate values, positively or negatively. See, for
instance, Ahlfeldt and Maennig ( ), which
finds a positive impact of sports facilities on real
estate values within two miles of a new facility.)
The evidence on this score is not encouraging. On
the other hand, hosting governmental bodies,
together with any private support, must pay for
facility constmction, upgrade and infrastructural
improvements necessitated by the Games. It must
also pay for the opening and award ceremonies,
transportation of the athletes to the various venues,
entertainment, a telecommunications/ broadcasting
centre, and security, among other things. Naturally,
to the extent that some of this public spending is on
productive infrastructure, these fiscal deficits may
prove to be beneficial in the long term to the
economy.
The initial publicized budgets of the OCOGs
invariably understate both the ultimate cost to the
OCOG and, to a much greater degree, the total cost
of staging the Games. The former escalates for
several reasons:
1. Construction costs inflate significantly as land
values increase with growing scarcity during the
roughly ten-year cycle of Olympic host
planning, bidding, selection and preparation.
2. It is usually in the interest of the bidding team to
under-represent the true costs, as they seek
public endorsement.
3. As the would-be host city enters into competi-
tion with other bidders, there is a natural ten-
dency to match their competitors’ proposals and
to embellish their original plans.
The total cost escalates because it includes
infrastructure and facility costs, whereas the pub-
licized OCOG budget includes only operating costs.
The infrastructure and facility costs usually
3337
form the largest component of total expenses, and
often do so by a substantial margin.
Thus Athens initially projected that its Games
would cost $1.6 billion, but they ended up costing
closer to $16 billion (including facility and infra-
structure costs). Beijing’s projected budget was
$1.6 billion, but ended up in total costing over $40
billion (inclusive of facility and infrastructure
costs). The 2014 Winter Games in Sochi, Russia,
were initially budgeted at around $12 billion; the
projected price tag in late 2009 reached $33 billion.
Of this, $23 billion would come from public
sources (Sports Business Daily ).
London expected its 2012 Games to cost under
$4 billion, but they are now projected to cost over
$19 billion (Carlin ; Simon ; Sports Business
Daily ). As expenses have escalated for London,
some of the projects have been scaled back, such as
the abandonment of the planned roof over the
Olympic Stadium. The stadium was originally
projected to cost $406 million and will end up
costing over $850 million. Further, its constmction
will be financed by taxpayers and the government
has been unsuccessful in its effort to find a soccer
or a rugby team to be the facility’s anchor tenant
after the 2012 Games. This will saddle the British
taxpayers with the extra burden of millions of
dollars annually to keep the facility operating. It is
little wonder that the London Olympics Minister
Tessa Jowell stated: ‘Had we known what we know
now, would we have bid for the Olympics? Almost
certainly not'. (Sports Business Daily ( ),
citing a story in
Daily Telegraph (2008). The Olympic Village was
to be privately financed, but the plan fell through
and will instead cost the taxpayers nearly $1 billion.
The government hopes that the apartments will be
sold after the Games and the financing will be
recouped.)
In a world where total revenue from the Games
is in the neighborhood of $4-$5 billion for the
Summer Olympics and roughly half that for the
Winter Games, costs above these levels mean that
someone has to pay. (To be sure, the Winter Games
involve fewer participants, fewer venues and less
constmction; hence the cost of these Games is
lower than for the summer Games.) While private
companies often contribute a share
3338 Economic Impact of the Olympic Games

of the capital costs (beyond the purchase of spon- Games, it is unlikely in the extreme to come in the
sorships), host governmental bodies usually pick up form of improving the budgets of local gov-
a substantial part of the tab. Moreover, as we have ernments. This raises the question of whether there
seen, not all the money generated at the Games are broader, longer-term or less tangible economic
stays in the host city to pay for the Games; rather, gains that accrue from hosting the Olympic Games.
close to half the money goes to support the
activities of the IFs, the NOCs and the IOC itself.
Thus, while the Sydney OCOG in 2000 reported How Do the Olympic Games Affect
that it broke even, the Australian state auditor the Economy?
estimated that the Games’ true long-term cost was
In general, sporting events produce two types of
$2.2 billion. In part, this was because it is now
economic benefit: direct and indirect. Direct eco-
costing $30 million a year to operate the 90,000-
nomic benefits include net spending by tourists who
seat Olympic Stadium. The story was little different
travel from out of town to attend the event;
for the 2004 Games in Athens, where maintenance
spending on capital and infrastructure construction
costs on the Olympics facilities in 2005 will
related to the event; long-run benefits - for example
reportedly come in around $124 million and there
lower transportation costs attributable to an
appears to be little to no local interest being
improved road or rail network - generated by this
expressed in the two Olympic soccer stadiums.
infrastructure; and the effect of hosting a sporting
According to one report, 21 of the 22 stadiums built
event on local security markets, primarily stock
for the 2004 Summer Games in Athens were
markets. Indirect benefits include possible
unoccupied in 2010.
advertising effects that make the host city or coun-
Similarly, the 1992 Olympics in Barcelona
try more visible as a potential tourist destination or
generated a reported surplus of $3 million for the
business location in the future and increases in civic
local organizing committee, but it created a debt of
pride, local sense of community, and the perceived
$4 billion for the central Spanish government and
stature of the host city or country relative to other
of $2.1 billion for the city and provincial
cities or countries.
governments. (The total reported cost of the Bar-
Among the direct economic benefits generated
celona Games was $9.3 billion, of which private
by the Olympic Games, tourist spending is probably
sources covered $3.2 billion and public sources
the most prominent. From Table , an average of 5.1
covered $6.1 billion (Burton 2003, p. 39). For a
million tickets were sold for the past six Summer
related account of large public expenditures on
Olympic Games, including almost six million
infrastructure for the Salt Lake City Olympics, see
tickets to the 1984 Games in Los Angeles. The
Bartlett and Steele ( ), who reported that
Winter Games are considerably smaller, averaging
the US government spent $1.5 billion of taxpayers’
1.3 million tickets over the past five Winter
money on the purchase of land, road construction,
Olympics. Even though most spectators buy tickets
sewers, parking lots, housing, buses, fencing, a light
to multiple events, so that selling five million
rail system, airport improvements, and security
tickets does not mean that there are five million
equipment, inter alia. Some have argued that a part
spectators, and many of the tickets are sold to local
of these expenditures would have occurred even if
residents, especially for the Summer Games, which
Salt Lake City had not hosted the Olympics.) The
typically take place in large metropolitan areas, a
Nagano Organizing Committee (Winter Games
sporting event of this size and scope has the
1998) showed a $28 million surplus, but the various
potential to attract a significant number of visitors
units of Japanese government were left with an $11
from outside the host city. Also, since the Games
billion debt (Burton and O’Reilly ).
are often spread over more than two weeks, these
For all of the foregoing reasons, if there is to be
visitors may spend a significant amount of time in
an economic benefit from hosting the Olympic
the host area,
Games
Tickets sold (millions)
Economic Impact of the
1984 Los Angeles 5.7Olympic Games
1988 Calgary 1.6
1988 SeoulEconomic Impact of3.3 the Olympic Games, Table 3
Ticket
1992 Albertville 0.9
1992 Barcelona 3.0
1994 Lillehanrnrer 1.2
1996 Atlanta 8.3
1998 Nagano 1.3
2000 Sydney 6.7
2002 Salt Lake 1.5
2004 Athens 3.8
2008 Beijing 6.5
Sources: IOC, 2006 Marketing Fact File, p. 60; IOC, 2010 Marketing Fact File, p. 39
generating substantial spending in the lodging, and
food and beverage sectors.
The Olympic Games require large spending on
constructing and updating venues. In addition to
venue construction, hosting the Olympic Games
often requires expansive infrastructure to move the
participants, officials, and fans to and from the
venues. Host cities and regions have also spent
considerable sums on roads and airport con-
struction, as well as on the renovation and con-
struction of public transportation systems (Essex
and Chalkley ). In less developed cities, the
building of a modem telecommunications capacity
also represents a substantial investment.
After the construction period, Olympics-
generated infrastructure can provide the host met-
ropolitan area or region with a continuing stream of
economic benefits in the form of reduced pro-
duction costs and prices charged by local businesses
(Rephann and Isserman ). The indirect economic
benefits generated by the Olympic Games are
potentially more important than the direct benefits,
and also more difficult to quantify. One possible
indirect benefit is the advertising effect of the
Olympic Games. Many Olympic host metropolitan
areas and regions view the Olympics as a way to
raise their profile on the world stage. If hosting the
Olympic Games leads tourists who would not have
otherwise considered this to be a destination to visit
the host city or region, then this advertising effect
can generate economic benefits over a long period
of time.
% capacity Revenue to OCOG (Smillion)
83 156 3339
78 32
75 revenues (current36$)
75 32
80 79
87 26
82 425
89 74
88 551
95 183
72 228
96 185
The Evidence on Economic Impact
The evidence on the economic impact of the
Olympic Games falls into four categories: (1) ret-
rospective evidence based on econometric analysis;
(2) case studies of individual Olympic Games; (3)
evidence derived from computable general
equilibrium (CGE) models; and (4) ‘multiplier-
based’ estimates of future economic impact. Note
the important temporal element associated with
each of these types of evidence. The ‘multiplier-
based’ estimates are prospective; these studies are
basically forecasts of economic benefits that will
take place at some time in the future. Because this
type of evidence is a forecast, it should be judged
by the same criteria as any other economic forecast.
The other three types of evidence are retrospective.
They are based on an examination of what actually
happened in the past when a metropolitan area or
region hosted the Olympic Games. This
fundamental difference between ‘multiplier-based’
estimates and other types of evidence is critically
important for understanding the differences in
estimates of the economic impact of the Olympic
Games.
Econometric-based evidence on the economic
impact of sporting events uses historic data on the
performance of the local economy before, during
and after the event takes place. This approach uses
statistical methods to determine how much of the
past local economic activity could be attributed to
the sporting event, and
3340
how much would have taken place without the
sporting event occurring.
Case studies of the economic impact use a
similar approach to the econometric method. This
approach examines past indicators of economic
activity, but does not use sophisticated regression
techniques. Case studies often examine a broader
set of economic indicators than econometric studies,
and use unconditional statistical tests like tests of
differences in means, crosstabs, or chi square tests
of statistical independence.
CGE models are complex representations of the
entire economy, including sectors that are not
related to sporting events. These models explicitly
account for the interconnected nature of the econ-
omy. Because the Olympic Games are large-scale
events, involving significant numbers of partici-
pants, officials, staff and spectators, the effects of
the Games may spread beyond the immediate area
and affect a number of distinct sectors of the
economy in different ways. CGE models can
account for complex economic effects, such as the
effect of the additional borrowing needed to finance
venue and infrastructure construction on the
availability (or price) of hinds to finance other
construction projects in the economy. CGE models
can also explicitly account for long-run economic
effects.
The basic idea behind multipliers is straight-
forward, and emerges horn input-output models of
the economy. When a consumer purchases a $ 1
pack of gum at a local store, the economic effects of
that transaction extend well beyond the consumer
handing a dollar to the cashier, who places that
dollar in the till. Some of that $1 in spending finds
its way into the pocket of the cashier, in the form of
wages; some folds its way into the pocket of the
store owner; some into the pocket of the driver who
delivered the gum; and so on. If the clerk, store
owner and delivery person live in the local
community, then this money is further distributed in
the local economy as these individuals pay rent, buy
groceries, and so on. A multiplier is an analytical
device used to estimate the broad economic impact
of each dollar spent in the local economy in terms
of the total amount of additional revenues earned by
firms, the total amount of
Economic Impact of the Olympic Games
personal income, and the number of jobs generated
in the local economy.
Estimating the economic impact of a sporting
event using the multiplier approach is relatively
simple in theory. First, estimate the number of
people who attend the sporting event; second,
estimate the amount of spending by these attendees;
third, apply a multiplier to this spending to estimate
the broad, overall impact of this spending on the
economy. However, on closer examination, this
process requires a significant amount of
discretionary input on the part of the researcher, and
coming up with accurate estimates of several of
these components is not a straightforward process.
The Problem with Multipliers
Multiplier-based estimates of the economic impact
potentially have several problems; See, for one, the
discussion in Crompton ( ).
First, multiplier-based estimates stem from the
estimate of the number of attendees. New economic
impact can only be generated by the spending of
spectators, participants and officials from outside
the host area. Estimating the total number of
attendees is much easier than estimating the number
of attendees from outside the host area. From
Table , 8.3 million tickets were sold to events at the
1996 Atlanta Summer Olympic Games. Some of
these tickets were clearly sold to residents of
Atlanta. Buthowmany of these 8.3 million tickets
were purchased by Atlantans?
Further complicating the process are ‘time
switchers’ and ‘casuals’. “Time switchers” are
attendees who would have visited the host area at
some other time, for some other reason, but instead
choose to visit the host area during the sporting
event. ‘Casuals’ are attendees who visit the host
area at the same time as the sporting event for some
other reason and decide to attend the event out of
convenience. The spending by both types of
attendees needs to be removed from the economic
impact estimate, as it cannot be directly attributable
to the sporting event. Failure to remove this
spending leads to overestimates of the economic
impact generated by the event.
Economic Impact of the Olympic Games
Second, multiplier-based economic impact
estimates fail to account for crowding out. In many
cases, the host area for the Olympic Games is a
tourist destination in its own right; tourists would
visit this area even if the Olympic Games were held
elsewhere; London, for example, is a major tourist
destination. Crowding out takes place when outside
visitors attending the Olympic Games buy hotel
rooms, meals and other travel-related goods and
services that would have been purchased by other
visitors absent the Olympic Games. Crowding out
implies that each dollar of new economic impact
estimated by multiplier-based methods needs to be
offset by some corresponding lost economic impact
that was crowded out, or else the net economic
impact will be overstated.
It is extremely difficult to determine how much
crowding out actually takes place when an area
hosts the Olympic Games. However, one study
found that gate arrivals at the Atlanta airport during
the 1996 Summer Games were identical to gate
arrivals in the same months in 1995 and 1997,
implying that quite a few to mists to Atlanta were
crowded out by the 1996 Games (Porter ). In late
2004, Athens tourism officials were estimating
about a 10% drop in summer tourism in 2004 due to
the Olympics. The Utah Skier Survey found that
nearly 50% of non-residents would stay away from
Utah in 2002 due to the expectation of more crowds
and higher prices. The Beijing Tourism Bureau pro-
jected that the number of visitors to the city in
August 2008 dining the Games would not be greater
than in August 2007. An additional problem for
Beijing was that in order to abate the city’s intense
pollution during the summer months, the
government ordered many of the city’s factories
closed leading up to and during the Games, and it
imposed severely restrictive driving regulations.
Third, multiplier-based estimates overlook the
displacement phenomenon. Some local residents
may choose to leave town to avoid the congestion
during the Games. The displaced people spend
money outside the local area that they would have
spent locally absent the Games. For instance, a
survey in Barcelona indicated that fully
3341
one-sixth of the city’s residents planned to travel
outside the city during the 1996 Olympics.
Fourth, multiplier-based estimates depend crit-
ically on the selection of the multiplier. Economic
theory does not provide exact guidance on the size
of the multiplier to use in any particular application.
The size of the multiplier used is at the discretion of
the analyst. This creates an incentive for researchers
to systematically choose large multipliers in order
to generate large estimates of the economic impact
of sporting events.
Despite all these problems, the majority of
published estimates of the economic impact of the
Olympic Games come from multiplier-based
estimates. Multiplier-based estimates are widely
used because, relative to the other approaches
discussed above, this approach requires little data,
little technical expertise, and very little in the way
of computing power. Multiplier-based estimates are
relatively cheap to produce and easy to manipulate.
Considering the size and prominence of the
event, relatively little objective evidence on the
economic impact of the Olympic Games exists.
Much of the existing evidence has been developed
by the host cities or regions, which have a vested
interest in justifying the large expenditures on the
games that were documented above. These ‘pro-
motional’ studies, which have produced widely
disparate estimated impacts of between $40 million
and $16 billion for different Olympic Games
between 1984 and 2006, suffer from the flaws
discussed, and should be viewed sceptically. For a
more detailed discussion of these studies, see
Humphreys and Zimbalist ( ).
Estimates of the economic impact of the Olym-
pic Games derived from academic research
published in peer-reviewed journals tend to be more
reliable. Only a few such studies, however, exist
(Hotchkiss et al. 2003; Jasman and Maennig 2008;
Feddersen and Maennig 2008; Lybbert and Hilmany
2000; Teigland 1999; Ritchie and Smith 1991).
The results of these studies present a consistent
picture of the economic impact of hosting the
Olympic Games on regions. Some jobs will be
created as a result of hosting the Games. However,
there appears to be no detectable effect on income,
3342
suggesting that existing workers do not benefit
from the Games. Moreover, the overall economic
impact of hosting the Games depends on the overall
labour market response to the new jobs created by
the Games. When taking into account the overall
labour market situation, the net impact of the
Games on a region may not be positive. The
negative impact on regional income found by the
study that examined four North American regions is
consistent with a negative overall labour market
response to hosting the Games. Furthermore, the
long-run impacts on tourism in the host region may
be overstated, based on evidence from
Lillehammer. Clearly, the weak results from aca-
demic research on the economic impact of hosting
the Olympic Games call into question the reported
economic impact from the promotional studies.
Some economists also have looked beyond
income and employment measures for evidence that
hosting the Olympic Games has an economic
impact on the host economy. One area examined is
stock markets (Berman et al. ; Veraros et al. ). The
relationship between hosting the Olympic Games
and stock markets is straightforward. To the extent
that hosting the Olympic Games generates any
benefits, including tangible economic benefits
associated with increased tourism, or intangible
benefits like national pride, sporting benefits,
increased visibility etc., stock markets should be
efficient mechanisms for valuing these benefits far
into the future and discounting them back to the
present. Positive benefits, if present, may be
capitalized into stock prices at the time that the
Games are awarded.
Research on the effect of hosting the Olympic
Games on stock markets exploits the nature of the
process through which the Games are awarded.
Until the announcement of the winning city is
made, there is considerable uncertainty about who
will be awarded the games, and the contest is
winner-take-all. The announcement about the
winner of the Games takes place at a specific time
(seven years prior to the Games) and represents a
natural experiment in stock prices.
The existing evidence is mixed. The announce-
ment that Sydney would host the 2000 Summer
Olympic Games produced modest increases in
stock returns in a limited number of industries:
Economic Impact of the Olympic Games
building materials, developers and contracts, and
engineering. The announcement that Athens would
host the 2004 Summer Olympic Games produced a
short-term, significant increase in overall stock
returns on the Athens Stock Exchange, but had no
impact on the Milan Stock Exchange. Milan was
one of the cities in the running for the 2004
Summer Games. Stock returns in construction
related industries on the Athens Stock Exchange
increased more than other sectors following the
announcement, suggesting that much of the
economic benefit accrues to this sector.
This evidence is limited to only two Olympic
Games, and the increases in stock returns reported
in the studies are modest, short-term, and primarily
limited to the construction industry and related
sectors of the economy. The empirical model used
to analyze stock returns on the Athens Stock
Exchange explains only 6% of the observed var-
iation in returns. Overall, the evidence from this
literature suggests that stock markets do not fore-
cast large positive economic impacts flowing from
the Olympic Games. While the idea that hosting the
Olympic Games affects stock returns may appear
important to the general public, a careful reading of
this literature reveals that the underlying effects are
small, transitory and limited to a few sectors of the
economy. This evidence does not support net
economic impact to the host city or region on the
order of those publicized in promotional reports.
Further, to the extent that hosting the Games may
produce substantial fiscal deficits and growing
public debt, the long-term effect on securities
markets may well be negative.
Can the Olympic Games Be an Economic
Success?
This review reveals relatively little evidence that
hosting the Games produces significant economic
benefits for the host city or region. If the economic
gains are modest, or perhaps non-existent, what can
host cities and regions do to maximize the potential
gains from hosting the Olympic Games? A careful
examination of past experiences suggests two
important avenues for leveraging the
Economic Impact of the Olympic Games
Olympic Games: host cities or regions need to make
careful land use decisions and exploit the post-
Olympic Games use of new and renovated facilities
and infrastructure.
Land is an increasingly scarce resource both in
the large urban areas that typically host the Summer
Games and in the mountainous areas that host the
Winter Games. Hosting the Olympic Games
requires a significant amount of land for facilities,
the Olympic Village, housing for the media and
staff, accommodations for spectators, and parking.
Unsuccessful Games leave behind legacies of
seldom or never used structures taking up valuable
land.
Successful Games, like the 1984 Los Angeles
Summer Games, utilize existing facilities as much
as possible, consuming as little scarce urban land as
possible. The stadium used for the opening and
closing ceremonies in the 1996 Atlanta Games was
reconfigured to a baseball stadium immediately
following the conclusion of the Games. The bullet
train built for the Nagano Games greatly reduced
the travel time between that city and Tokyo.
Tying up scarce land for seldom-used Olympic
venues in both urban areas and alpine recreation
areas cannot be an optimal use of this valuable
resource. Olympic planners need to design facilities
that will be useful for a long time after the Games
are over, and are constructively integrated into the
host city or region.
Clearly, the impact of the Olympic Games will
vary according to the differing levels of develop-
ment in the host city and country. Properly planned,
hosting the Games can catalyze the construction of
a modem transportation, communications and sport
infrastructure. Such a potential benefit is bound to
be greater for less developed areas. But even in
such areas, hosting the Games will require a
significant outlay of public hinds to finance the
infrastructural improvements. These improvements
can also be made without hosting the Games. Thus
it is relevant to ask whether the planning for the
Olympics produces an optimal use of scarce public
monies. It is also relevant to consider that in many
circumstances the public policy process is so
gridlocked that needed infrastructural investments
may be delayed for years, if
3343
not decades, without the Olympic catalyst and that
the Games do provide at least some capital to
facilitate the completion of desirable projects.
Conversely, in more developed regions, where
land is even more scarce during the initial bid
planning (and destined to become scarcer still over
the ten-year period of Olympic planning, bidding,
selection and preparation) and labour and resource
markets are tight, hosting the Games can occasion a
gross misuse of land as well as provoke wage and
resource price pressure leading to higher inflation.
Finally, it is important to recognize that hosting
the Olympic Games may generate significant non-
pecuniary benefits to the host city or region. The
residents of the host city or region are likely to
derive significant pride and sense of community
from hosting the Games. Their homes are the focus
of the world’s attention for a brief but intense
period. The planning and work required to host the
Games takes considerable time and effort, and
much of the hard work is done by volunteers.
Pulling off such a huge endeavour is a source of
considerable local and national pride. These factors
are both important and valuable, even though
researchers find it difficult to place a dollar value
on them.
Some recent research has attempted to quantify
the value of the non-pecuniary benefits generated
by the Olympic Games (Atkinson et al. ).
Economists have used the Contingent Valuation
Method (CVM) to place a dollar value on such
diverse intangible benefits as cleaning up oil spills
in pristine wilderness areas and preserving green
space in urban areas. The basic approach in CVM is
to elicit people’s willingness to pay for some
intangible through hypothetical questions involving
referendum voting or changes in taxes. A recent
estimate of the total willingness to pay for the
intangible benefits generated in the United
Kingdom from hosting the 2012 Summer Games
was in excess of d2 billion.
In the end, the economic and non-economic
value to hosting the Olympic Games is a complex
matter, likely to vary from one situation to another.
Simple conclusions are impossible to draw.
Prospective hosts of future Games would do well to
steer clear of the inevitable Olympic
3344 Economic Integration

hype and to take a long, hard and sober look at the Rephann, T., and A. Isserman. 1994. New highways as
long-run development goals of their region. economic development tools: An evaluation using quasi-
experimental matching methods. Regional Science and
This article is adapted from Chapter 6 in Urban Economics 24(6): 728.
Zimbalist, A. 2010. Circling the Bases: Essays on Simon, B. 2006. Cost of Canadian 2010 Winter Olympics
the Challenges and Prospects of the Sports escalates. Financial Times, 6 February.
Sports Business Daily. 2008a. 14 November.
Industry. Temple University Press, Philadelphia.
Sports Business Daily. 2008b. 14 November.
Sports Business Daily. 2009a. 27 October.
Sports Business Daily. 2009b. 14 January.
Sports Business Daily. 2009c. 30 September.
See Also Sports Business Journal. 2009a. 7-13 September.
Sports Business Journal. 2009b. 10-18 October, p. 8.
► Zz Ungar, B.L. 2000. Olympic Games: Federal government
provides significant funding and support. Collingwood:
Diane Publishing Co..
Veraros, N., E. Kasimati, and D. Dawson. 2004. The 2004
Bibliography Olympic Games announcement and its effect on the
Athens and Milan stock exchanges. Applied Economics
Ahlfeldt, G., and W. Maennig. 2009. Impact of sports arenas Letters 11: 749-753.
on land values: Evidence from Berlin. The Annals of
Regional Science 44(2): 205-227.
Atkinson, G., S. Mourato, and S. Szymanski. 2008. Are we
willing to pay enough to ‘back the bid?’: Valuing the
intangible benefits of hosting the summer Olympic Economic Integration
Games. Urban Studies 45: 419-444.
Austen, I. 2010. A $1 billion hangover from an Olympic
Bela Balassa
party. New York Times, 25 February.
Bartlett, D.L., and J.B. Steele. 2001. Snow job. Sports
Illustrated 21(December): 79-98.
Berman, G., R. Brooks, and S. Davidson. 2000. The Sydney
Olympic Games announcement and Australian stock
market reaction. Applied Economics Letters 7: 781-784. In everyday parlance, integration is defined as
Burton, R. and O’Reilly, N. 2009. Consider intangibles when bringing together of parts into a whole. In the
weighing Olympic host city benefits. Sports Business economic literature, the term ‘economic integration’
Journal, 7-13 September, p. 33.
Carlin, B. 2007. Olympic budget trebles to d9.3bn. Daily
does not have such a clear-cut meaning. At one
Telegraph, 15 March. Daily Telegraph 2008. 13 extreme, the mere existence of trade relations
November. between independent national economies is con-
Crompton, J. 1995. Economic impact analysis of sports sidered as a form of economic integration; at the
facilities and events: Eleven sources of misapplication.
Journal of Sport Management 9: 15.
other, it is taken to mean the complete unification
Essex, S., and B. Chalkley. 2004. Mega-sporting events in of national economies.
urban and regional policy: A history of the Winter Economic integration is defined here as process
Olympics. Planning Perspectives 19(1): 205. and as a state of affairs. Considered as a process, it
Humphreys, B., and A. Zimbalist. 2008. The financing and
economic impact of the Olympic Games. In The business
encompasses measures designed to eliminate
of sports, ed. B.R. Humphreys and D.R. Howard. discrimination between economic units that belong
Westport: Praeger. to different national states; viewed as a state of
International Olympic Committee. 2007. 2006 Olympic affairs, it represents the absence of various forms of
Marketing Fact File.
discrimination between national economies.
. Accessed 28 Oct
2007. Economic integration may take several forms
Porter, P. 1999. Mega-sports events as municipal investments that represent various degrees of integration. In a
: A critique of impact analysis. In Sports economics: free trade area, tariffs (and quantitative import
Current research, ed. J. Fizel, E. Gustafson, and L.
restrictions) among participating countries are
Hadley. Westport: Prager.
Preuss, H. 2003. The economics of staging the Olympic eliminated, but each country retains its own tariffs
Games. Northampton: Edward Elgar Publishing. against non-members. Establishing a customs
Economic Integration
union involves, apart from the suppression of intra-
area trade barriers, equalizing tariffs on imports
from non-member countries.
A common market goes beyond a customs
union, inasmuch as it also entails the free movement
of factors of production. In turn, an economic union
combines the suppression of restrictions on
commodity and factor movements with some
degree of harmonization of national economic
policies, so as to reduce discrimination owing to
disparities in these policies. Finally, total economic
integration means the unification of economic
policies, culminating in the establishment of a
supra-national authority whose decisions are
binding for the member states.
History
The first important case of economic integration
was the German Zollverein in the 19th century,
which subsequently led to total economic integra-
tion through the unification of the German states
with the establishment of the Deutsches Reich. In
the 20th century, the creation of the Benelux cus-
toms (1948) and subsequently economic (1949)
union, comprising Belgium, Luxemburg, and the
Netherlands, represented the first step towards
European economic integration. It was followed by
the establishment of the European Coal and Steel
Community (1953) and the European Economic
Community or EEC (1958), both comprising
Belgium, France, Italy, Luxemburg, the
Netherlands, and West Germany.
Austria, Denmark, Norway, Portugal, Sweden,
Switzerland, and the United Kingdom founded the
Eiuopean Free Trade Association or EFTA in 1960,
with Finland participating first as an associate and
later as a full member. In turn, Denmark and the
United Kingdom left EFTA and, together with
Ireland, entered the European Economic
Community in 1968; Greece became a member
ofthe EEC in 1978, and Portugal and Spain joined
in 1986.
In Eastern Europe, the Council for Mutual
Economic Assistance or CMEA was established in
1948, with the participation of the Soviet Union,
Bulgaria, Czechoslovakia, Hungary,
3345
Poland, and Romania. Albania and East Germany
joined shortly thereafter; subsequently, Cuba and
Mongolia became full members while Albania
ceased to participate in CMEA activities.
There have been a number of attempts at eco-
nomic integration in developing countries. Some
were to involve the establishment of a free trade
area, such as the Latin American Free Trade Asso-
ciation (1960) comprising Argentina, Bolivia,
Brazil, Chile, Colombia, Ecuador, Mexico, Peru,
Uruguay, and Venezuela; others were designed to
become customs unions, such as the West African
Customs Union (1959), including the Ivory Coast,
Mali, Mauritania, Niger, Senegal, and Upper Volta.
In 1960, the Central American Common Market was
established, with Costa Rica, Guatemala, Honduras,
Nicaragua, and El Salvador as members; in turn, the
East African Common Market, comprising Kenya,
Tanzania, and Uganda and subsequently
transformed into the East African Economic
Community (1967), was designed to become an
economic union. None of these attempts has come
to fruition, however, as barriers to intra-area trade
have not been fully eliminated or have subsequently
been restored.
Trade Creation and Trade Diversion
Viner’s The Customs Union Issue ( ) was the
first important contribution to the theory of eco-
nomic integration. Viner investigated the impact of
a customs union on trade flows and distinguished
between the ‘trade-creating’ and the ‘trade-
diverting’ effects of a union. In the first case, there
is a shift from domestic to partner country sources
of supply of a particular commodity; in the second
case, the shift occurs from non-member country to
partner country sources of supply.
Trade creation increases economic welfare,
inasmuch as higher-cost domestic sources of supply
are replaced by lower-cost imports from partner
countries that were previously excluded by the
tariff. In turn, trade diversion has a welfare cost
since tariff discrimination against non-member
countries, attendant on the establishment of the
customs union, leads to the replacement of lower-
cost sources of supply in
3346
these countries by higher-cost partner country
sources.
The net welfare effects of the customs union will
depend on the amount of trade created and diverted
as well as on differences in unit costs. In a partial
equilibrium framework, under constant costs, there
will be a welfare gain (loss) if the amount of trade
created, multiplied by differences in unit costs
between the home and the partner countries,
exceeds (falls short of) the amount of trade diverted,
multiplied by differences in unit costs between the
partner and the non-member countries.
Meade ( ) further considered the effects of
a customs union on intercommodity substitution,
involving the replacement of domestic products by
partner country products (trade creation) and the
replacement of products of non-member countries
by partner country products (trade diversion). As in
the case of substitution among the sources of supply
of a particular commodity (production effects), trade
creation involves a welfare improvement, and trade
diversion the deterioration of welfare, in the event
of substitution among commodities (consumption
effects).
The separation of production and
consumption effects does not imply the absence
of interaction between the two. Substitution
among sources of supply will affect the pattern of
consumption through changes in the prices paid
by the consumer. Also, intercommodity
substitution will lead to modifications in the
pattern of production by changing the prices
received by producers.
At the same time, as Lipsey and Lancaster (
-57) first noted, production and
consumption
effects - and the theory of customs unions in
general - should be considered as special cases of
the theory of the second best. Assuming that the
usual conditions for a Pareto optimum are
fiilfilled, free trade will lead to efficient resource
allocation while pre-union, as well as the post-
union, situations are sub-optimal because tariffs
exist in both cases. In the abstract, then, one
cannot make a judgement as to whether
establishing a customs union will increase or
reduce welfare. Nevertheless, a consideration of
certain factors may provide a presumption as to
the possible direction of the welfare effects of a
union.
Economic Integration
Factors Influencing the Welfare Effects
of a Custom Union
Lipsey ( ) suggested that the welfare effects of
a customs union will depend on the relative
importance in home consumption of goods pro-
duced domestically and imported from non-member
countries prior to the establishment of the union.
Ceteris paribus, the larger the share of domestic
goods and the smaller the share of goods imported
from non-member countries, the greater is the
likelihood of an improvement in welfare following
the union’s establishment. Such will be the case
since substitution of partner country products for
domestic products entails trade creation and their
substitution for the products of non-member
countries involves trade diversion.
These propositions are consistent with
Tinbergen’s ( ) conclusion that increases in
the size of a customs union will augment the
probability of favourable welfare effects; in the
limiting case, the customs union includes the entire
world, which is equivalent to free trade. Applying
the argument that gains are obtained through the
enlargement of a union because of increased
possibilities for the reallocation of production, it
also follows that the gains are positively correlated
with increases in the market size of the participating
countries (e.g. small countries will gain more from
participation in a customs union than large
countries).
Viner further considered the implications that
differences in production structures among the
member countries have for the welfare effects of a
customs union. He suggested that the more
competitive (the less complementary) is the pro-
duction structure of the member countries, the
greater is the chance that a customs union will
increase welfare.
This proposition reflects the assumption that
countries with similar production structures tend to
replace domestic goods by competing imports from
partner countries following the establishment of a
customs union, while differences in the production
structure within the union lead to substitution of
partner country products for lower- cost products
originating in non-member
Economic Integration
countries (the latter conclusion does not hold if the
union includes the low-cost producer).
The welfare effects of a customs union will also
depend on transportation costs. Ceteris paribus, the
lower are transport costs among the member
countries, the greater will be the gains from their
economic integration. Thus, the participation of
neighbouring countries in a union, with greater
possibilities for trade creation across their borders,
will offer advantages over the participation of
faraway countries that tends to promote trade
diversion.
The height of tariffs will further affect the
potential gains and losses derived from a customs
union. High pre-union tariffs against the future
member countries will increase the possibility of
trade creation, and hence gains in welfare, follow-
ing the establishment of the union while low tariffs
against non-member countries will reduce the
chances for trade diversion. But, these conclusions
have little relevance under the application of the
most-favoured-nation clause that entails providing
equal tariff treatment to all countries before the
customs union is established.
Customs Unions vs. Unilateral Tariff
Reductions
In the Viner-Meade-Lipsey analysis, participation in
a trade-creating customs union was considered as a
means to reduce the distorting effects of the
country’s own tariffs. This argument was carried to
its logical conclusion in contributions by Cooper
and Massell ( ) and Johnson ( ) who
suggested that participation in a customs union is
inferior to the unilateral elimination of tariffs, which
leads to greater trade creation without giving rise to
trade diversion.
The same authors claimed that the reasons for
the establishment of customs unions he in the gains
participating countries may obtain in furthering
non-economic objectives, and considered
preference for industry as such an objective. They
further assumed that this objective can be pursued at
a lower cost in the framework of the larger market
of a customs union than in the country’s own
domestic market.
3347
As Johnson noted, the formation of a customs
union in the pursuit of the stated objective pre-
supposes that the member countries are at a com-
parative disadvantage in the production of industrial
goods vis-a-vis the rest of the world. Cooper and
Massed ( ) identified such coun
tries with developing countries, further suggesting
that the economic planners of these countries are
willing to accept some reduction in national income
in order to assure increases in industrial production.
The question remains as to why there is a
preference for industry. Johnson ( ) expressed
the view that such preference may reflect nationalist
aspirations and rivalry with other countries; the
power of industrial firms and workers to increase
their incomes; or the bedef that industrial activity
involves beneficial extemadties. The last point,
however, implies that there is no need to introduce
non-economic considerations to obtain the Cooper-
Mansell-Johnson result; the desirability of a
customs union may be established in economic
terms, provided that it permits obtaining
externalities that cannot be achieved otherwise.
A further question is if unilateral tariff reduc-
tions will be superior to a customs union in the
absence of a preference for industry or beneficial
externalities. The Wonnacotts ( ) showed that
this may not be the case if one admits the existence
of tariffs in partner and in non-member countries
prior to the formation of the customs union.
The elimination of tariffs by partner countries
will provide benefits to the home country as it can
now sell at a higher price in partner country mar-
kets. This gain will be larger the higher is the pre-
union tariff in the partner countries and will further
be affected by tariffs in the non-member countries.
This is because, in selling in partner country
markets free of duty, home country producers avoid
paying the tariff in non-member countries.
Finally, Cooper and Massed ( ) noted that
a subsidy-union, with each participating country
subsidizing its own industrial production, is supe-
rior to a customs union. This conclusion follows
since the consumption cost of the tariff can be
3348 Economic Integration

avoided if the prices of industrial products in the objectives and beneficial externalities. Such
union are maintained at the world market level improvements also favour a customs union over
through subsidies. However, production subsidi- unilateral tariff reductions, which would lead to the
zation may be done by each country individually, deterioration of the terms of trade of the country
with the attendant welfare benefits, without par- concerned.
ticipating in a union. Other things being equal, the larger the union
the greater will be its gain, and hence the loss to
non-member countries, through terms of trade
Multi-country Analysis of a Customs changes. This is because, ceteris paribus, the larger
Union the union the higher will be the elasticity of its
reciprocal demand for foreign products and the
Traditionally, the welfare effects of a customs union
lower the elasticity of reciprocal demand on the part
were considered from the point of view of a single
of non-member countries for the union’s products.
country. Yet, these effects may differ among
The extent ofterms-of-trade effects will further
member countries, depending on their production
depend on the height of tariffs before and after the
structure, location, the height of pre-union tariffs,
establishment of a customs union. As Vanek ( )
and other characteristics. In fact, one member
first showed, a customs union will not
country may obtain a gain and another a loss, when
involve a loss to non-member countries, while
any attempt to aggregate gains and losses
benefiting its own members, if the union’s external
encounters the well-known difficulties of interna-
tariff level is sufficiently lower than the pre-union
tional welfare comparisons.
tariffs of the member countries.
The distribution of welfare gains and losses in a
Vanek’s proposition was formulated in a three-
customs union will be further affected by changes in
country, two-commodity (3 x 2) model. It has
the terms of trade. The establishment of a union
subsequently been extended to a general case, under
may give rise to price changes in trade between the
which compensatory payments to non-member
member countries, even if the prices at which trade
countries were also introduced (Kemp and Wan ).
takes place with non-member countries remain
At the same time, these propositions indicate a
unchanged (the case of the ‘small’ union).
theoretical possibility rather than a likely outcome,
In the more general case, prices in trade with
since customs unions have shown little inclination
non-member countries will also vary. Now, while
to compensate non-member countries for losses
trade diversion involves a welfare loss to the
attendant upon the union’s establishment.
member countries of a customs union under
3x3 models represent an intermediate case
unchanged terms of trade, this loss may be offset by
between 3x2 and m x n models. They permit
a welfare gain due to improvements in the terms of
introducing a greater number of possible trade
trade attendant on trade diversion. Conversely,
patterns, differential tariffs, complementarity and
whereas under the assumption of unchanged terms
substitution in consumption, with a large number of
of trade the welfare of non-member countries is
marginal conditions in production and con-
unaffected by the establishment of a customs union,
sumption, as well as intermediate products (Lloyd ).
non-member countries will lose owing to the
The 3x3 model is thus richer in content than the 3x2
adverse impact of trade diversion on their terms of
model. Despite attempts made at introducing new
trade. This may be interpreted as the result of a shift
terminology (Collier ), however, adding a third
in the union members’ reciprocal demand curve for
commodity does not appear to have materially
products originating in non-member countries.
affected the basic propositions of customs union
Improvements in the terms of trade thus provide
theory. This conclusion may also find application to
reasons for the establishment of a customs union
m x n models.
even in the absence of non-economic
Economic Integration 3349

Free Trade Areas area. Other things being equal, then, their self-
interest would tend to encourage member countries
In a free trade area, maintaining different tariffs to reduce their own tariffs.
among member countries on the products of non-
members introduces the possibility of trade
deflection. Furthermore, production and investment Factor Movements
deflection may occur if one admits trade in
intermediate products. The deflection of investment may occur within a
There wil l be trade deflection if imports enter country or may involve international capital
the free trade area via the member country which movements. In the first case, it affects the allocation
applies the lowest tariff. Transportation costs apart, of the country’s own capital among industries; in
this is equivalent to adopting a tariff equal to the the second case, it influences the international
lowest tariff for each commodity in any of the allocation of capital.
member countries. Under the assumption of The last point leads to the case of common
imchanged terms of trade, the deflection of trade markets where, by definition, the full mobility of
will increase welfare in the member countries by factors is assured. Meade ( ) first analysed the
limiting the extent of trade diversion. Removing welfare effects of the movement of factors of
this assumption, trade deflection will affect the production in an integrated area. He concluded that
distribution of welfare between member and non- free factor movement will increase the gains
member countries by reducing the terms of trade obtained in a union by reducing the relative scar-
gain (loss) for the former (latter). cities of the factors of production. This conclusion
Production deflection will occur if the manu- reflected the assumption that the conditions for
facture of products containing imported inputs factor price equalization through trade are not
shifts to countries which have lower tariffs on these fulfilled.
inputs, because differences in tariffs outweigh If factors of production were not free to move
differences in production costs. The deflection of between member and non-member countries, there
production will have unfavourable effects on will be no welfare loss due to factor movements
welfare, since the pattern of productive activity will among member countries to correspond to trade
not follow lines of comparative advantage but rather diversion in commodity trade. In the event of such
differences in duties. factor movements, however, an analogous case to
The deflection of production may also affect the trade creation and trade diversion occurs if the
pattern of investment. Other things being equal, movement of factors were subject to taxes prior to
investors will establish factories in countries with the establishment of a union and these taxes have
lower tariffs on imported inputs. Again, adverse been removed among union members. And, in any
welfare effects will ensue because investments case, there will be indirect effects on welfare to the
respond to tariff differences rather than to extent that factor movements substitute for trade.
differences in production costs. These effects may involve welfare losses to non-
The deflection of trade, production, and invest- member countries as the newly- established
ment represent unintended effects of free trade productions substitute for imports from them.
areas. To avoid such an eventuality member coun-
tries of free trade areas have imposed country of
origin rules. These rules limit the freedom of intra- Economies of Scale
area trade to commodities that incorporate a certain
proportion of domestic products or undergo a Economic integration may lead to lower costs
particular process of transformation in one of the through increases in the volume of plant output. For
member countries. The application of origin rules various types of equipment, such as containers,
limits, but does not entirely eliminate, trade, pro- pipelines, and compressors, cost is a
duction and investment deflection in a free trade
3350 Economic Integration

function of the surface area whereas capacity is separate plants, each of which enjoys economies of
related to volume; per unit costs decline with scale, thereby resulting in cost reductions.
increases in output in the case of bulk transactions
as well as for nonproportional activities such as
design production planning, research, and the col- Competition and Technological Change
lection and channelling of information; inventory
Economic integration will also create the conditions
holdings do not need to increase proportionately
for more effective competition (Scitovsky ). By
with output; larger output warrants the application
increasing the number of firms each producer
of technological methods that call for the use of
considers as his competitors, the opening of national
specialized equipment or assembly-line production;
frontiers will contribute to the loosening of
and large-scale production may be necessary to
monopolistic and oligopolistic market structures in
ensure the optimum use of various kinds of
the individual countries. At the same time, there is
indivisible equipment.
no contradiction between gains from economies of
Corden ( ) showed that the traditional con
scale and increased competition, since a wider
cepts of trade creation and trade diversion will be
market can sustain a larger number of efficient units
relevant in the case of economies of scale on the
(Balassa ).
plant level but new concepts are added: the cost-
Greater competition may have beneficial effects
reduction effect and the trade-suppression effect.
through improvements in manufacturing efficiency
The former refers to reductions in average unit costs
as well as through technological change. While the
as domestic output expands following the
former has no place in traditional theory, which
establishment of the union; the latter refers to the
postulates the choice of the most efficient
replacement of cheaper imports from non-member
production methods among those available to the
countries by domestic production under economies
firm, it may assume considerable importance in
of scale. In Corden’s view, a net benefit is likely to
countries whose markets have been sheltered from
ensue as the cost-reduction effect tends to outweigh
foreign competition.
the trade-suppression effect.
The stick and the carrot of competition also
Plant size and unit costs are not necessarily
provides inducement for technological progress in
correlated in the case of multiproduct firms. In such
the member countries. In particular, increased
instances, costs may be lowered by reducing
competition may stimulate research activity aimed
product variety through specialization in an inte-
at developing new products and improving
grated area, which permits lengthening production
production methods. Finally, economic integration
runs for individual products.
may contribute to the transmission of technological
The advantages of longer production runs derive
knowledge by increasing the familiarity of
from improvements in manufacturing efficiency
producers with new products and technological
along the Teaming curve’ as cumulated output
processes originating in the partner countries.
increases; the lowering of expenses involved in
It has been suggested, however, that gains from
moving from one operation to another that involves
competition, and from economies of scale, may be
the resetting of machines, the shifting of labour, and
obtained through unilateral trade liberalization and
the reorganization of the work process; and the use
that the gains are predicated on the response of
of special-purpose machinery in the place of general
economic agents to the stimulus provided by
purpose machinery.
competition (Krauss ). While the validity of the
Apart from product or horizontal specialization,
second point depends on factors which are
there are possibilities for vertical specialization by
particular to each country, the first neglects the
subdividing the production process among
gains obtained through the increases in output
individual establishments in an integrated area. As
associated with sales in the markets of partner
the sales of the final product increase, parts, com-
countries.
ponents, and accessories may be manufactured in
Economic Integration
Policy Harmonization
Policy differences among the member countries
may influence trade flows and factor movements,
thereby modifying the welfare effects of economic
integration. Industrial policies, social policies, fiscal
policies, monetary policies, and exchange rate
policies are relevant in this context (Balassa ).
Industrial policies may involve granting credit
preferences and/or tax benefits across the board or
to particular activities. ‘Horizontal’ policies that are
applied across-the-board do not create distortions,
unless the conditions under which they are provided
favour one activity over another. By contrast,
‘vertical’ measures are granted to particular
activities and thereby introduce distortions, which
may counteract the effects of the elimination of
intra-area tariffs.
Intercountry differences in social policies will
not give rise to distortions, provided that social
benefits are financed from the contributions of
employers and employees. Nor are these
conclusions affected if factor mobility is introduced
into the analysis as long as the employees regard the
resulting social benefits as part of their
compensation.
The situation is different if social benefits are
financed from general tax revenue. This case is
equivalent to a wage subsidy that favours labour-
intensive activities. Correspondingly, differences in
tiie mode of financing social security among the
member countries will introduce distortions in
resource allocation. This conclusion is strengthened
if consideration is given to factor movements that
respond to international differences in labour costs.
The elimination of vertical measures of indus-
trial policy and the equalization ofthe conditions of
financing social security will reduce distortions in
resource allocation as well as differences in tax
burdens among the member countries. Differences
in the tax burden may remain, however, owing to
national preferences as to the provision of collective
goods. The effects of such differences on factor
movements will depend on the spending of the tax
proceeds. But, there may be ‘supply-side’ effects,
with a lower tax burden providing incentives for
work effort and risk taking.
A further question is if, for a given tax burden,
intercountry differences in reliance on indirect
3351
taxes and income taxes will distort competition.
Under the destination principle, indirect taxes are
rebated on exports and imposed on imports without
such adjustments occurring in regard to income
taxes. Nevertheless, distortions in the conditions of
competition will not ensue as flexibility in exchange
rates will offset differences in rates of indirect
taxes.
The application of the origin principle, with
indirect taxes levied on production irrespective of
the country of sale, in one country and that of the
destination principle in another will similarly be
offset through exchange rate flexibility. Such will
not be the case, however, if cascade-type taxation
applied in one country and value added taxation in
another, with the former raising the tax burden on
industries that go through several stages of
fabrication, each of which is subject to tax.
Eliminating this source of distortion would
necessitate the adoption of value added taxation in
all member countries of a union.
While exchange rate flexibility is necessary to
offset intercountry differences in systems of taxa-
tion, it has been proposed that fixed exchange rates
be established following the creation of a union. But
such an action is predicated on the coordination -
and eventual unification - of monetary and fiscal
policies, since otherwise pressures are created for
exchange rate changes. The fixity of exchange rate
should thus be considered as the final outcome of
policy coordination rather than an intermediate step
in economic integration (Balassa ).
See Also
► in storm : .. .
► List, Friedrich (1789-1846)
Bibliography
Balassa, B. 1961. The theory of economic integration.
Homewood: Richard D. Irwin.
Balassa, B. 1975. Monetary integration in European Common
Market. In European economic integration, ed. B. Balassa,
175-220. Amsterdam: North-Holland. Collier, R 1979. The
welfare effects of a customs union: An anatomy. Economic
Journal 83: 84-87.
3352 Economic Interpretation of History

Cooper, C.A., and B.F. Massed. 1965a. A new look at into three stages, each defined by a different type of
customs union theory. Economic Journal 75: 742-747.
economy. If Polanyi is right in suggesting that
Cooper, C.A., and B.F. Massed. 1965b. Towards a general
theory of customs unions for developing countries.reciprocity, redistribution and the market each
Journal of Political Economy 73: 461-476. defined a different kind of society, this is, in a way,
tantamount to saying that the economy is primary,
Corden, W.M. 1972. Economies of scale and customs union
theory. Journal of Political Economy 80: 465-475. and thus his work constitutes a species of the
Johnson, HLG. 1965. An economic theory of protectionism,
economic interpretation of history. Nevertheless,
tariff bargaining, and the formation of customs unions.
Journal of Political Economy 73: 256-283. despite the importance of Polanyi’s work and the
Kemp, M.C., and HLY. Wan Jr. 1976. An elementary prop-possibility of other rival economic interpretations,
osition concerning the formation of customs unions.
Marxism remains the most influential, the most
Journal of International Economics 6: 95-97.
important, and perhaps the best elaborated of all
Kraus s, M.B. 1972. Recent developments in customs union
theories, and we shall concentrate on it.
theory: An interpretative survey. Journal of Economic
Literature 10: 413-436. One often approaches a theory by seeing what it
Lipsey, R.G. 1960. The theory of customs unions: A general
denies and what it repudiates. This approach is quite
survey. Economic Journal 70: 496-513.
frequently adopted in the case of Marxism, where it
Lipsey, R.G., and K.J. Lancaster. 1956-7. The general theory
is both fitting and misleading. We shall begin by
of second best. Review of Economic Studies 24, 11-32.
adopting this approach, and turn to its dangers
Lloyd, RJ. 1982. The theory of customs unions. Journal of
International Economics 12: 41-63. subsequently.
Meade, J.E. 1953. Problems of economic union. London:
Marxism began as the reaction to the romantic
Aden & Unwin.
idealism of Hegel, in the ambience of whose
Meade, J.E. 1955. The theory of customs union. Amsterdam:
North-Holland. thought the young Karl Marx reached maturity. This
Scitovsky, T. 1958. Economic theory and Western European
no doubt is the best advertised fact about the origin
integration. London: Allen & Unwin.
of Marxism. The central point about Hegelianism
Tinbergen, J. 1957. Customs unions: Influence of their size on
was that it was acutely concerned with history and
their effect. Zeitschrift der gesamten Staatswis- senschaft
113: 404-414. social change, placing these at the centre of
Vanek, J. 1965. General equilibrium of international dis-
philosophical attention (instead of treating them as
crimination. The case of customs unions. Cambridge:
mere distractions from the contemplation of
Harvard University Press.
timeless objects, which had been a more frequent
Viner, J. 1950. The customs union issue. New York: Carnegie
Endowment for International Peace. philosophical attitude); and secondly, it taught that
Wonnacott, P., and R. Wonnacott. 1981. Is unilateral tariff
history was basically determined by intellectual,
reduction preferable to a customs union? The curious case
spiritual, conceptual or religious forces. As Marx
of the missing foreign tariffs. American Economic Review
71: 704—714. and Engels put it in The German Ideology, ‘The
Young Hegelians are in agreement with the Old
Hegelians in their belief in the rule of religion, of
concepts, of an abstract general principle in the
existing world’ (Marx and Engels -1846, p. 5).
Now the question is - why did Hegel and
followers believe this? If it is interpreted in a
Economic Interpretation of History
concrete sense, as a doctrine claiming that the ideas
Ernest Gellner of men determined their other activities, it does not
have a great deal of plausibility, especially when
put forward as an unrestricted generalization. If it is
formulated - as it was by Hegel - as the view that
some kind of abstract principle or entity dominates
Marxism does not possess a monopoly of the history, the question may well be asked: what
economic interpretation of history. Other theories of evidence do we have for the very existence of this
this kind can be formulated - for instance that which mysterious poltergeist
can be found in the very distinguished work of Karl
Polanyi, dividing the history of mankind
Economic Interpretation of History
allegedly manipulating historical events? Given the
fact that the doctrine is either implausible or
obscure, or indeed both, why were intelligent men
so strongly drawn to it?
The answer to this may be complex, but the
main elements in it can perhaps be formulated
simply and briefly. Hegelianism enters the scene
when the notion of what we now call culture enters
public debate. The point is this: men are not
machines. When they act, they do not simply
respond to some kind of push. When they do
something, they generally have an idea, a concept,
of the action which they are performing. The idea or
conception in turn is part of a whole system. A man
who goes through the ceremony of marriage has an
idea of what the institution means in the society of
which he is part, and his understanding of the
institution is an integral part of his action. A man
who commits an act of violence as part of a family
feud has an idea of what family and honour mean,
and is committed to those ideas. And each of these
ideas is not something which the individual had
excogitated for himself. He took it over from a
corpus of ideas which differ from community to
community, and which change over time, and which
are now known as culture.
Put in this way, the ‘conceptual’ determination
of human conduct no longer seems fanciful, but on
the contrary is liable to seem obvious and trite. In
various terminologies (‘hermeneutics’, ‘struc-
turalism’, and others) it is rather fashionable now-
adays. The idea that conduct is concept-saturated
and that concepts come not singly but as systems,
and are carried not by individuals but by on-going
historic communities, has great plausibility and
force. Admittedly, those who propose it, in Hegel’s
day and in ours, do not always define their position
with precision. They do not always make clear
whether they are merely saying that culture in this
sense is important (which is hardly disputable), or
claiming that it is the prime determinant of other
things and the ultimate source of change, which is a
much stronger and much more contentious claim.
Nonetheless, the idea that culture is important and
pervasive is very plausible and suggestive, and
Hegelianism can be credited with being one of the
philosophies which, in its
3353
own peculiar language, had introduced this idea. It
is important to add that Hegelianism often speaks of
‘Spirit’ in the singular; our suggestion is that this
can be interpreted as culture, as the spirit of the age.
This made it easy for Hegelianism to operate as a
kind of surrogate Christianity: those no longer able
to believe in a personal god could tell themselves
that this had been a parable on a kind of guiding
historical spirit. For those who wanted to use it in
that way, Hegelianism was the continuation of
religion by other means.
But Hegelianism is not exhausted by its sense of
culture, expressed in somewhat strange language. It
is also pervaded by another idea, fused with the first
one, and one which it shares with many thinkers of
its period: a sense of historical plan. The turn of the
18th and 19th centuries was a time when men
became imbued with the sense of cumulative
historical change, pointing in an upward direction -
in other words, the idea of Progress.
The basic fact about Marxism is that it retains
this second idea, the ‘plan’ of history, but aims at
inverting the first idea, the romantic idealism, the
attribution of agency to culture. As the two founders
of Marxism put it themselves in The German
Ideology (pp. 14-15),
In direct contrast to German philosophy which
descends from heaven to earth, here we ascend from
earth to heaven ... We set out from real active men,
and on the basis of their real life-process we
demonstrate the development of the ideological
reflexes and echoes of this life-process ... Morality,
regligion, metaphysics, all the rest of ideology and
their corresponding forms of consciousness, thus no
longer retain the semblance of independence. They
have no history, no development; but men,
developing their material production and their
material intercourse, alter, along with their real
existence, their thinking and the products of their
thinking. Life is not determined by consciousness,
but consciousness by life.
Later on in the same work, the two founders of
Marxism specify the recipe which, according to
them, was followed by those who produced the
idealistic mystification. First of all, ideas were
separated from empirical context and the interests
3354
of the rulers who put them forward. Secondly, a set
of logical connections was found linking successive
ruling ideas, and their logic is then meant to explain
the pattern of history. (This links the concept-
saturation of history to the notion of historic design.
Historic pattern is the reflection of the internal
logical connection of successive ideas.) Thirdly, to
diminish the mystical appearance of all this, the
free-floating, self-fransforming concept was once
again credited to a person or group of persons.
If this kind of theory is false, what then is true?
In the same work a little later, the authors tell us:
This sum of productive forces, forms of capital and
social forms of intercourse, which every individual
and generation finds in existence as something given,
is the real basis of ... the ... ‘essence of
man’____These conditions of fife, which different
generations find in existence, decide also whether or
not the periodically recurring revolutionary convul-
sion will be strong enough to overthrow the basis of
all existing forms. And if these material elements of
a complete revolution are not present... then, as far as
practical developments are concerned, it is absolutely
immaterial whether the ‘idea’ of this revolution has
been expressed a hundred times already...
(p. 30).
The passage seems unambiguous: what is
retained is the idea of a plan, and also the idea of
primarily internal, endogenous propulsion. What
has changed is the identification ofthe propulsion,
of the driving force ofthe transformation. Change
continues to be the law of all things, and it is
governed by a plan, it is not random; but the
mechanism which controls it is now identified in a
new manner.
From then on, the criticisms ofthe position can
really be divided into two major species: some
challenge the identification of the ruling mecha-
nism, and others the idea of historic plan. As the
most dramatic presentation of Marxist development,
Robert Tucker’s Philosophy and Myth in Karl Marx
( , p. 123) puts it:
Marx founded Marxism in an outburst of Hegelizing.
He considered himself to be engaged in ... [an]... act
of translation of the already discovered truth ... from
the language of idealism into
that of materialism______ Hegelianism itself was
latently or esoterically an economic interpretation of
history. It treated history as ‘a history of
Economic Interpretation of History
production’ ... in which spirit externalizes itself in
thought-objects. But this was simply a mystified
presentation of man externalizing himself in material
objects.
This highlights both the origin and the validity
or otherwise of the economic interpretation of
history. Some obvious but important points can be
made at this stage. The Hegel/Marx confrontation
owes much of its drama and appeal to the extreme
and unqualified manner in which the opposition is
presented. This unqualified, unrestricted
interpretation can certainly be found in the basic
texts of Marxism. Whether it is the ‘correct’
interpretation is an inherently undecidable question:
it simply depends on which texts one treats as final -
those which affirm the position without restriction
and without qualification, or those which contain
modifications, qualifications and restrictions.
The same dilemma no doubt arises on the
Hegelian side, where it is further accompanied by
the question as to whether the motive force, the
spirit of history, is to be seen as some kind of
abstract principle (in which case the idea seems
absurd to most of us), or whether this is merely to
be treated as a way of referring to what we now
term culture (in which case it is interesting and
contentious).
One must point out that these two positions, the
Hegelian and the Marxist, are contraries, but not
contradictories. They cannot both be true, but they
can perfectly well both be false. A world is easily
conceivable where neither of them is true: a world
in which social changes sometimes occur as a
consequence of changes in economic activities, and
sometimes as a consequence of strains and stresses
in the culture. Not only is such a world conceivable,
but it does really rather look as if that is the kind of
world we do actually live in. (Part of the appeal of
Marxism in its early days always hinged on
presenting Hegel-type idealism and Marxism as two
contradictories, and ‘demonstrating’ the validity of
Marxism as a simple corollary of the manifest
absurdity of strong versions of Hegelianism.) In this
connection, it is worth noticing that by far the most
influential (and not unsympathetic) sociological
critic of Marx is Max Weber, who upholds precisely
this kind of
Economic Interpretation of History
position. Strangely enough, despite explicit and
categorical denials on his own part, he is often
misrepresented as offering a return to some kind of
idealism (without perhaps the mystical idea of the
agency of abstract concepts which was present in
Hegel). For instance, Michio Morishima, in Why
has Japan ‘Succeeded’? ( , p. 1),
observes: ‘Whereas Karl Marx contended that
ideology and ethics were no more than reflections ...
Max Weber... made the case for the existence of
quite the reverse relationship. ’ Weber was sensitive
to both kinds of constraint; he merely insisted that
on occasion, a ‘cultural’ or ‘religious’ element
might make a crucial difference.
Connected with this, there is another important
theoretical difference to be found in Weber and
many contemporary sociologists. The idea of the
inherent historical plan, which had united Hegel and
Marx, is abandoned. If the crucial moving power of
history comes from one source only, though this
does not strictly speaking entail that there should be
a plan, an unfolding of design, it nevertheless does
make it at least very plausible. If that crucial
moving power had been consciousness, and its aim
the arrival at self-consciousness, then it was natural
to conclude that with the passage of time, there
would indeed be more and more of such
consciousness. So the historical plan could be seen
as the manifestation of the striving of the Absolute
Spirit or humanity, towards ever greater awareness.
Alternatively, if the motive force was the growth of
the forces of production, then, once again, it was
not unreasonable to suppose that history might be a
series of organizational adjustments to expanding
productive powers, culminating in a full adjustment
to the final great flowering of our productive capac-
ity. (Something like that is the essence of the
Marxist vision of history. )
If on the other hand the motive forces and the
triggers come from a number of sources, which
moreover are inherently diverse, there is no clear
reason why history should have a pattern in the
sense of coming ever closer to satisfying some
single criterion (consciousness, productivity, con-
gruence between productivity and social ethos, or
whatever). So in the Weberian and more modem
vision, the dramatic and unique developments of
3355
the modem industrial world are no longer seen as
the inevitable fulfilment and culmination of a
potential that had always been there, but rather as a
development which only occurred because a certain
set of factors happened to operate at a given time
simultaneously, and which would otherwise not
have occurred, and which was in no way bound to
occur. Contingency replaces fatality.
So much for the central problem connected with
the economic interpretation of history. The question
concerning the relative importance of conceptual
(cultural) and productive factors is the best known,
most conspicuous and best advertised issue in this
problem area. But in fact, it is very far from obvious
that it is really the most important issue, the most
critical testing ground for the economic theory of
history. There is another problem, less immediately
obvious, less well known, but probably of greater
importance, theoretically and practically. That is the
relative importance of productive and coercive
activities.
The normal associations which are likely to be
evoked by the phrase ‘historical materialism’ do
indeed imply the downgrading of purely conceptual,
intellectual and cultural elements as explanatory
factors in history. But it does not naturally suggest
the downgrading of force, violence, coercion. On
the contrary, for most people the idea of coercion by
threat or violence, or death and pain, seems just as
‘realistic’, just as ‘materialistic’ as the imperatives
imposed by material need for sustenance and
shelter. Normally one assumes that the difference
between coercion by violence or the threat of
violence, and coercion by fear of destitution, is
simply that the former is more immediate and works
more quickly. One might even argue that all
coercion is ultimately coercion by violence: a man
or a group in society which coerces other members
by controlling the food supply, for instance, can
only do it if they control and defend the store of
food or some other vital necessity by force, even if
that force is kept in reserve. Economic constraint, it
could be argued (as Marxists themselves argue in
other contexts), only operates because a certain set
of rales is enforced by the state, which may well
remain in the background. But economic constraint
is in this way parasitic on
3356
the ultimate presence of enforcement, based on the
monopoly of control of the tools of violence.
The logic of this argument may seem persuasive,
but it is contradicted by a very central tenet of the
Marxist variant of the economic theory of history.
Violence, according to the theory, is not
fundamental or primary, it does not initiate funda-
mental social change, nor is it a fundamental basis
of any social order. This is the central contention of
Marxism, and at this point, real Marxism diverges
from what might be called the vulgar image
possessed of it by non-specialists. Marxism stresses
economic factors, and downgrades not merely the
importance of conceptual, ‘superstruc- tural’ ones,
but equally, and very significantly, the role of
coercive factors.
A place where this is vigorously expressed is
Engels’s ‘Anti-Duhring’ ( ):
... historically, private property by no means makes
its appearance as the result of robbery or violence.
... Everywhere where private property developed,
this took place as the result of altered relations of
production and exchange, in the interests of
increased production and in furtherance of
intercourse - that is to say, as a result of economic
causes. Force plays no part in this at all. Indeed, it is
clear that the institution of private property must be
already in existence before the robber can appropri-
ate another person’s property... Nor can we use either
force or property founded on force to explain the
‘enslavement of man for menial labour’ in its most
modem form - wage labour.... The whole process is
explained by purely economic causes; robbery, force,
and the state of political interference of any kind are
unnecessary at any point whatever (Bums , pp. 267-
9).
Engels goes on to argue the same specifically in
connection with the institution of slavery:
Thus force, instead of controlling the economic
order, was on the contrary pressed into the service of
the economic order. Slavery was invented. It soon
became the predominant form of production among
all peoples who were developing beyond the
primitive community, but in the end was also one of
the chief causes of the decay of that system (ibid., p.
274).
Engels a little earlier in the same work was on
slightly more favourable ground when he discussed
the replacement of the nobility by the bourgeoisie as
the most powerful estate in the land. If physical force
were cmcial, how should the peaceful merchants and
producers have prevailed over the professional
warriors? As Engels puts it: ‘During the
Economic Interpretation of History
whole of this stmggle, political forces were on the side of
the nobility...’ (ibid., p. 270).
One can of course think of explanations for this
paradox: the nobility might have slaughtered each
other, or there might be an alliance between the
monarchy and the middle class (Engels himself
mentioned this possibility, but does not think it
constitutes a real explanation) and so forth. In any
case, valid or not, this particular victory of pro-
ducers over warriors would seem to constitute a
prima facie example of the non-dominance of force
in history. The difficulty for the theory arises when
the point is generalized to cover all social orders
and all major transitions, which is precisely what
Marxism does.
Engels tries to argue this point in connection
with a social formation which one might normally
consider to be the very paradigm of the domination
by force: ‘oriental despotism’. (In fact, it is for this
very reason that some later Marxists have
maintained that this social formation is incompat-
ible with Marxist theory, and hence may not exist.)
Engels does it, interestingly enough, by means of a
kind of functionalist theory of society and
government: the essential function, the essential
role and duty, of despotic governments in hydraulic
societies is to keep production going by looking
after the irrigation system. As he puts it:
However great the number of despotic governments
which rose and fell in India and Persia, each was
fully aware that its first duty was the general main-
tenance of irrigation throughout the valleys, without
which no agriculture was possible (Bums p. 273).
It is a curious argument. He cannot seriously
maintain that these oriental despots were always
motivated by a sense of duty towards the people
they governed. What he must mean is something
like this: unless they did their ‘duty’, the society in
question could not survive, and they themselves, as
its political parasites, would not survive either. So
the real foundation of ‘oriental despotism’ was not
the force of the despot, but the functional
imperatives of despotically imposed irrigation
systems. Economic need, as in the case of slavery,
makes use of violence for its own ends, but violence
itself initiates or maintains nothing. This
Economic Interpretation of History
interpretation is related to what Engels says a little
further on. Those who use force can either aid
economic development or accelerate it, or go
against it, which they do rarely (though he admits
that it occasionally occurs), and then they them-
selves usually go under: ‘Where... the internal
public force of the country stands in opposition to
economic development... the contest has always
ended with the downfall of the political power’
(Bums , p. 277).
We have seen that Engels’s materialism is curi-
ously functional, indeed teleological: the economic
potential of a society or of its productive base
somehow seeks out available force, and enlists it on
its own behalf. Coercion is and ought to be the slave
of production, he might well have said. This
teleological element is found again in what is
perhaps the most famous and most concise
formulation of Marxist theory, namely certain
passages in Marx’s preface to A Contribution to
‘The Critique of Political Economy’ ():
A social system never perishes before all the pro-
ductive forces have developed for which it is wide
enough; and new, higher productive relationships
never come into being before the material conditions
for their existence have been brought to maturity
within the womb of the old society itself. Therefore,
mankind always sets itself only such problems as it
can solve; for when we look closer we will always
find that the problem itself only arises when the
material conditions for its solution are already
present, or at least in the process of coming into
being. In broad outline, the Asiatic, the ancient, the
feudal, and the modem bourgeois mode of
production can be indicated as progressive epochs in
the economic system of society (Bums , p.372).
The claim that a new order does not come into
being before the conditions for it are available, is
virtually a tautology: nothing comes into being
unless the conditions for it exist. That is what
‘conditions’ mean. But the idea that a social system
never perishes before it has used up all its potential
is both strangely teleological and disputable. Why
should it not be replaced even before it plays itself
out to the full? Why should not some of its potential
be wasted?
It is obvious from this passage that the purpo-
sive, upward surge of successive modes of
3357
production cannot be hindered by force, nor even
aided by it. Engels, in ‘Anti-Diihring’, sneers at
rulers such as Friedrich Wilhelm IV, or the then
Tsar of Russia, who despite the power and size of
their armies are unable to defy the economic logic
of the situation. Engels also treats ironically Herr
Diihring’s fear of force as the ‘absolute evil’, the
belief that the ‘first act of force is the original sin’,
and so forth. In his view, on the contrary, force
simply does not have the capacity to initiate evil. It
does however have another ‘role in history, a
revolutionary role’; this role, in Marxist words, is
midwifery:
... it is the midwife of every old society which is
pregnant with the new,... the instrument by the aid of
which social movement forces its way through and
shatters the dead, fossilized, political forms ... (Bums
, p. 278).
The midwifery simile is excellent and conveys
the basic idea extremely well. A midwife cannot
create babies, she can only aid and slightly speed up
their birth, and once the infant is bom the midwife
cannot do much harm either. The most one can say
for her capacity is that she may be necessary for a
successful birth. Engels seems to have no fear that
this sinister midwife might linger after fhe birth and
refuse to go away. He makes this plain by his
comment on fhe possibility of a ‘violent collision’
in Germany which ‘would at least have the advan-
tage of wiping out the servility which has permeated
fhe national consciousness as a result of the humil-
iation of the Thirty Years War’.
There is perhaps an element of truth in the
theory that coercion is and ought to be the slave of
production. The element of truth is this: in pre-
agrarian hunting and gathering societies, surrounded
by a relative abundance of sustenance but lacking
means of storing it, there is no persistent, social,
economic motive for coercion, no sustained
employment for a slave. By contrast, once wealth is
systematically produced and stored, coercion and
violence or the threat thereof acquire an inescapable
function and become endemic. The surplus needs to
be guarded, its socially ‘legitimate’ distribution
enforced. There is some evidence to support the
view that hunting and gathering societies were more
peaceful than the agrarian societies which
succeeded them.
3358
One may put it like this: in societies devoid of a
stored surplus, no surplus needs to be guarded and
the principles governing its distribution do not need
to be enforced. By contrast, societies endowed with
a surplus face the problem of protecting it against
internal and external aggression, and enforcing the
principles of its distribution. Hence they are
doomed to the deployment, overt or indirect, of
violence of the threat thereof. But all of this, true
though it is, does not mean that surplus- less
societies are necessarily free of violence: it only
means that they are not positively obliged to
experience it. Still less does it mean that within the
class of societies endowed with a surplus, violence
on its own may not occasionally or frequently
engender changes, or inhibit them. The argument
does not preclude coercion either from initiating
social change, or from thwarting change which
would otherwise have occurred. The founding
fathers of Marxism directed their invective at those
who raised this possibility, but they never
succeeded in establishing that this possibility is not
genuine. All historic evidence would seem to
suggest that this possibility does indeed often
correspond to reality.
Why is the totally unsubstantiated and indeed
incorrect doctrine of the social unimportance of
violence so central to Marxism?
The essence of Marxism lies in the retention of
the notion of an historical plan, but a re-
specification of its driving force. But the idea of a
purposive historical plan is not upheld merely out of
an intellecUial desire for an elegant conceptual
unification of historical events. There is also a
deeper motive. Marxism is a salvation religion,
guaranteeing not indeed individual salvation, but
the collective salvation of all mankind. Ironically,
its conception of the blessed condition is profoundly
bourgeois. Indeed, it constitutes the ultimate
apotheosis of the bourgeois vision of life. The
bourgeois preference for peacefiil production over
violent predation is elevated into the universal
principle of historical change. The wish is father of
the faith. The work ethic is transformed into the
essence, the very species-definition of man. Work is
our fulfilment, but work patterns arc also the cmcial
determinants of historical change. Spontaneous,
unconstrained work, creativity, is our
Economic Interpretation of History
purpose and our destiny. Work patterns also deter-
mine the course of history and engender patterns of
coercion, and not vice versa. Domination and the
mastery of techniques of the violence is neither a
valid ideal, nor ever decisive in history. All this is
no doubt gratifying to those imbued with the pro-
ducer ethic and hostile to the ethic of domination
and violence: but is it true?
Note that, were it true, Marxism is free to com-
mend spontaneously cooperative production, devoid
of ownership and without any agency of
enforcement, as against production by competition,
with centrally enforced ground rules. It is free to do
it, without needing to consider the argument that
only competition keeps away centralized coercion,
and that the attempt to bring about propertyless and
total cooperation only engenders a new form of
centralized tyranny. If tyranny only emerges as a
protector of basically pathological forms or orga-
nization of work, then a sound work-pattern will on
its own free us for ever from the need for either
authority or checks on authority. Man is held to be
alienated from his true essence as long as he works
for extraneous ends: he finds his true being only
when he indulges in work for the sake of creativity,
and choses his own form of creativity. This is of
course precisely the way in which the middle class
likes to see its own life. It takes pride in productive
activity, and chooses its own form of creativity, and
it understands what it does. Work is not an
unintelligible extraneous imposition for it, but the
deepest fulfilment.
On the Marxist economic interpretation of his-
tory, mankind as a whole is being propelled towards
this very goal, this bourgeois-style fiilfil- ment in
work without coercion. But the guarantee that this
fulfilment will be reached is only possible if the
driving force of history is such as to ensure this
happy outcome. If a whole multimde of factors,
economic, cultural, coercive, could all interact
unpredictably, there could hardly be any historic
plan. But if on the other hand only one factor is
fundamental, and that factor is something which has
a kind of vectorial quality', something which
increases overtime and inevitably points in one
direction only (namely the augmentation of the
productive force of man), then the necessary
historical plan does after all have a film.
Economic Laws
unprecarious base. This is what the theory requires,
and this is what is indeed asserted.
The general problem of the requirement, ulti-
mately, of a single-factor theory, with its well-
directed and persistent factor, is of course related to
the problems which arise from the plan that
Marxists discern in history. According to the above
quotation from Marx, subsequent to primitive
communism, four class-endowed stages arise,
namely the Asiatic, the ancient, the feudal, and the
modem bourgeois, which is said to be the last
‘antagonistic’ stage (peaceful fulfilment follows
thereafter). Marxism has notoriously had trouble
with the ‘Asiatic’ stage because, notwithstanding
what Engels claimed, it does seem to exemplify and
highlight the autonomy of coercion in history, and
the suspension of progress by a stagnant, self-
maintaining social system.
But leaving that aside, in order to be loyal to its
basic underlying intuition of a guaranteed pro-
gression and a final happy outcome, Marxism is not
committed to any particular number or even any
particular sequence of stages. The factual
difficulties which Marxist historiography has had in
finding all the stages and all the historical
sequences, and in the right order, are not by them-
selves necessarily disastrous. A rigid unilinealism is
not absolutely essential to the system. What it does
require (apart from the exclusiveness, in the last
analysis, of that single driving force) is the denial of
the possibility of stagnation, whether in the form of
absolute stagnation and immobility, or in the form
of circular, repetitive developments. If this
possibility is to be excluded, a number of things
need to be true: all exploitative social forms must be
inherently unstable; the number of such forms must
be finite; and circular social developments must not
be possible. If all this is so, then the alienation of
man from his true essence - free fulfilment in
unconstrained work - must eventually be attained.
But if the system can get stuck, or move in circles,
the promise of salvation goes by the board. This
would be so even if the system came to be stuck for
purely economic reasons. It would be doubly
disastrous for it if other factors, such as coercion,
were capable of freezing it. The denial of any
autonomous role for violence in history is the
3359
most important, and most contentious, element in
the Marxian economic theory of history.
So what the Marxist economic interpretation of
history really requires is that no non-economic
factor can ever freeze the development of society,
that the development of society itself be pushed
forward by the continuous (even if on occasion
slow) growth of productive forces, that the social
forms accompanying various stages of the devel-
opment of productive forces should be finite in
number, and that the last one be wholly compatible
with the fullest possible development of productive
forces and of human potentialities.
The profound irony is that a social system
marked by the prominence and pervasiveness of
centralized coercion, should be justified and
brought about by a system of ideas which denies
autonomous historical agency both to coercion and
to ideas. The independent effectiveness both of
coercion and of ideas can best be shown by
considering a society built on a theory, and one
which denies the effectiveness of either.
Bibliography
Bums, E. 1935. A handbook of Marxism. London: Victor
Gollancz.
Engels, F. 1878. Anti-Duhring. In Bums (1935).
Marx, K. 1859. A contribution to ‘the critique of political
economy’. In Bums (1935).
Marx, K., and F. Engels. 1845-1846. The German ideology.
London: Lawrence & Wishart, 1940.
Morishima, M. 1982. Why has Japan 'succeeded’? Western
technology and the Japanese ethos. Cambridge:
Cambridge University Press.
Tucker, R.C. 1961. Philosophy and myth in Karl Marx.
Cambridge: Cambridge University Press.
Economic Laws
Stefano Zamagni
Keywords
Ceteris paribus; Economic laws; Law of
diminishing returns; Law of variable propor-
tions; Marginal revolution; Natural law; Neo-
3360 Economic Laws

Austrian economics; Positive economics; Sub- type, the laws of nature are providentially imposed
stitution; Tendency laws norms of conduct. In such a setting it would have
been pointless to separate means and ends, since the
implementation of natural laws is both an end and a
JEL Classifications means, and even more pointless to think of a tension
BO between 'explaining’ and ‘understanding’ economic
behaviour. Causal and teleological, positive and
The social sciences, and economics in particular, normative, theoretical and practical started being
separated from moral and political philosophy in the seen as separate categories only when the economic
second half of the 18th century when the results of discourse freed itself from the philosophy of natural
the myriad of intentional actions of people were law and all its implications.
perceived to produce regularities resembling the Post-classical economics set out to be a science
laws of a system. Both Physio- cratic thought and of the laws regulating the economic order and of the
Smith's Wealth of Nations reflect this extraordinary conditions allowing these laws to operate. It became
discovery: scientific laws thought to be found only the basis of a theory that, in Jevons’s own terns,
in nature could also be found in society. This proposed to construct a ‘social physics’. The view
extension poses several problems. A serious one of a social world ordered according to transcendent
refers to the tension of combining individuals’ ends was abandoned in favour of an ideal of
freedom of action with the scientists' desire to objective knowledge of economic phenomena
discover the systematic aspects of the unintended gained through a ‘positive’ study of the laws that
and quite often unpredictable consequences of regulate market activities. In so doing, neoclassical
human action, that is, the desire to arrive at laws ‘positive’ economics solves the aforementioned
characterized by a certain degree of generality and tension by extrapolating the theoretical model of
peimanence. namral sciences to economics: economies is to
In the history of economic thought this funda- produce the laws of motion similar to those of
mental tension has been solved in different ways. In physics, chemistry, astronomy.
the 18th century, the mechanistic ideal of the But what is a scientific law and which role do
natural sciences, combined with the natural law idea laws play within the logical positivist's perspective
of a harmonious order of nature, determined the adopted by neoclassical economics? Laws provide
way social phenomena were treated. There was a the foundation of a deductive scientific method of
desire to discover the ‘natural laws’ of economic inquiry. According to the deductive-nomological
life and to formulate the natural precepts which rule conception of explanation, due to C. Hempel, laws
human conduct. are universal statements not requiring reference to
The classical economists upheld the notion that any one particular object or spatio-temporal
natural laws are embedded in the economic process location. To be valid, laws are constrained neither to
as beneficial laws, along with the belief in the finite populations nor to particular times and places;
existence of rales of nature capable of being dis- they are, in effect, expressions of natural
covered. Thus the belief that things could follow the stationarities. This interpretation of the notion of
beneficial ‘natural course’ only in a rationally law provides the so-called covering-law model of
organized society which it was a duty to create explanation with an unquestionably firm inferential
according to the precepts of nature. The economic foundation. Deductive logic is employed to ensure
system is the mechanism by which the individual is the truth status of propositions and, since the
driven to fostering the prosperity of society while deductions are (by hypothesis) predicated on true
pursuing his private interest. Hence the automatic universal statements (laws), the empirical validity
operation of the economic system may be combined of these statements may be ascertained. However,
with freedom of individual action. This is the core what sort of constraints on economic discourse are
of the doctrine of economic harmony. Besides being
causal laws of a mechanical
Economic Laws
imposed by this positivistic structure? On the one
hand this structure constitutes its object; on the
other hand it generates specific economic questions
together with their method of solution. Following
the model of natural sciences and its success in
controlling a natural world made up of objects and
unvarying relations among them expressed in the
form of laws, the neoclassical approach arrives at a
study of regularities conceived of as specifying the
nature of its objects.
To capture the different interpretations of the
notion of law by classical and neoclassical econ-
omists let us refer to one of the most famous of
economic laws: the law of diminishing returns, also
known as the law of variable proportions. Studying
agricultural production, Ricardo had noted that
different quantities of labour, assisted by certain
quantities of other inputs (farm tools, fertilizers, and
so on), could be employed on a given piece of land,
that is, it was possible to vary the proportions in
which land and complex labour (labour assisted by
other inputs) are employed. He accordingly arrived
at the law which states that production increases
resulting from equal increments in the employment
of complex labour, while the quantity of land
farmed remains constant, will initially be increasing
and then decreasing. (To be sure, the first statement
of the law is due to the Physiocratic economist
Turgot. )
Three points deserve attention. First, Ricardo
and classical authors in general offer no formal
demonstration of this law. To them, it is basically
an empirical law, on which no functional associa-
tion between output and variable inputs can be built.
Second, the classics’ use of the law refers to their
theories of distribution and development: as the
supply of land in the whole system is fixed, sooner
or later a point will be reached at which economic
growth will come to a halt, notwithstanding any
countervailing effects due to technical progress.
Finally, the law presupposes a comparative statics
framework: the pattern of the marginal products of
complex labour refers to different observable
equilibrium positions and not to hypothetical or
virtual variations.
With the advent of the marginalist revolution,
two subtle changes in the interpretation of the law
took place, (a) The de facto elimination of the
3361
distinction between the extensive case (the case of
the simultaneous cultivation of pieces of land of
different fertility) and the intensive case (the
application of successive doses of capital and labour
to the same piece of land) with an overevaluation of
the latter. Classical economists, being interested in
the explanation of rent, concentrated on the
extensive case; they took also the intensive case into
consideration but with many qualifications. Indeed,
whereas the various levels of productivity of
E
different qualities of land is a circumstance which
may be directly observed in a given situation, the
marginal productivity of a given input is related to a
virtual increment in output and therefore to a virtual
change in the situation, (b) The change in the
method of analysis - it was preferred to reason in
terms of hypothetical rather than observable
changes - brought about by the shift of interest
towards the intensive margin, supported the thesis
of the symmetrical nature of land and other inputs.
This in turn favoured the extension of the
substitutability between land and complex labour
from agricultural production to all kinds of
production, including those in which land does not
figure as a direct input. It so happened that whereas
in classical economics the substitutability between
land and complex labour presupposes that simple
labour and equipment are strictly complementary, in
neoclassical economics this substitutability is
applied to all inputs indiscriminately.
However, the neoclassical interpretation of the
law poses serious problems. In the first place, there
is the problem of justifying, on empirical grounds,
the general applicability of the substitution
principle. Secondly, and more importantly, in order
to allow the substitution of inputs to take place, a
certain lapse of time is required during which the
required modifications to the productive structure
can be made. (It is certainly true that coal can
replace oil to provide heating, but before this can
happen it will be necessary to change the heating
system.) The well-known distinction between the
short run and the long run is a partial and indirect
way to take the temporal element into consideration.
In the short run the plant is fixed by definition. It is
therefore the fixed input which, in the neoclassical
interpretation of the law, plays the
3362
same role as land in the classical interpretation.
Now, neoclassical theory correctly states the law of
diminishing returns with respect to the short run;
however it is in the long run that the substitutability
of inputs becomes actually feasible. One is therefore
confronted with a dilemma: the neoclassical
interpretation of the law seems to be more plausible
in a long-run framework when there exists the
necessary time to accommodate input adjustments;
on the other hand, fixed inputs cannot, by definition,
exist in the long run so that the law of variable
proportions cannot be stated in such a context.
This dilemma is the price neoclassical theory
has to pay for its interpretation of the law in
accordance with the positivistic statute. Indeed, the
power of deductive, truth-preserving rules of
scientific inference is not purchased without a cost.
A school of economic thought which is not prepared
to sustain such a cost is the neo- Austrian. The neo-
Austrian economists solve what has been called the
fundamental tension by arguing economics cannot
and should not provide general laws since, by its
very nature, it is an idiographic and not a
nomothetical discipline. The general target of
economics is ‘understanding’ grounded in Verstehen
doctrine: by introspection and empathy, the study of
the economic process should aim at explaining
individual occurrences, not abstract classes of
phenomena. It follows that if by a scientific law one
should mean a universal conditional statement of
type ‘for all x, if.r is A, then x is B\ statements
regarding unique events cannot by definition
express any regularity for the simple reason that any
regularity presupposes the recurrence of what is
defined as regular. In the words of L. von Mises,
who shares with F. von Hayek the paternity of the
neo-Austrian school, what assigns economics its
peculiar and unique position in the orbit of pure
knowledge ‘... is the fact that its particular theorems
are not open to any verification or falsification on
the ground of experience ... the ultimate yardstick of
an economic theorem’s correctness or incorrectness
is solely reason unaided by experience’ (von Mises ,
p. 858).
There is indeed a place for economic Taws’ in
the framework of Austrian economics. The
Economic Laws
familiar Taws’ of economics (diminishing marginal
utility, supply and demand, diminishing returns to
factors, Say’s Law and so on) are seen as ‘necessary
truths’ which explain the essential structure of the
economic world but with no predictive worth. In
other words, economic laws are not generalizations
from experience, as it is the case within the
positivistic paradigm, but are theorems which
enable us to understand the economic world. It is
ironic that Mises’ position of radical apriorism
joined to Hayek’s attack on scientism and
methodological monism are completely at variance
with the position taken by the father of the Austrian
school, Carl Menger ( ), who announced that
in economic theories
exact laws are defined which are just as rigorous as
in fact are the laws of nature.
Between the extreme positions of neoclassical
positive economic and neo- Austrian economics are
those who, without denying that economics is in
search for laws in the same sense in which natural
sciences are and that laws perform an explanatory
as well as a predictive function, underline that the
explicative structure of economics, albeit
nomothetical, substantially differs from that of
natural sciences. This intermediate position can be
traced back to Keynes’s ( ) methodol
ogy which considers the conditions of truth and
universality of the positivistic conception of sci-
entific laws as far too rigid for a discipline such as
economics. Two main reasons account for the
different epistemological status of laws in natural
sciences and in economics. First, the knowledge of
economic phenomena is itself an economic variable,
that is, it changes, along with the process of its own
acquisition, the economic situation to which it
refers. The formulation of a new physical law does
not change the course of physical processes; it does
not influence the truth or falsity of the prognosis.
This is not the case in economics where the
prognosis, say, that in two years time there will be a
boom can cause overproduction and a resulting
recession. In turn, this specific aspect is strictly
connected to the fact that the object of study of
economics possesses an historical dimension.
Economics is in time in a way that natural sciences
are not. The ensuing mutability of observed
regularities is well expressed by Keynes
Economic Laws
when he writes, ‘As against Robbins, economics is
essentially a moral science and not a natural
science. That is to say it employs introspection and
judgements of value’ ( , p. 297) to which
he adds, ‘It deals with motives, expectations, psy-
chological uncertainties. One has to be constantly
on guard against treating the material as constant
and homogenous’ (p. 300).
Second, the role played by ceteris paribus
clauses in natural sciences and in economics is
substantially different. The modem economists
appeal to the ‘other things being equal’ clause -
which according to Marshall is invariably attached
to any economic law - in all those cases where the
classical economists were talking of ‘disturbing
causes’. J.S. Mill’s ( ) discussion
of inexact sciences is suggestive here:
When the principles of Political Economy are to be
applied to a particular case then it is necessary to take
into account all the individual circumstances of that
case ... These circumstances have been called
disturbing causes. This constitutes the only uncer-
tainty of Political Economy. ( , p. 300)
Also in natural sciences we find ceteris paribus
clauses. Indeed, a scientific theory that could dis-
pense with them would in effect achieve perfect
closure, which is a rarity. So where lies the differ-
ence? The example of the science of tides used by
Mill is revealing. Physicists know the laws of the
greater causes (the gravitational pull of the moon)
but do not know the laws of the minor causes (the
configuration of the sea bottom). The ‘other things’
which scientists hold equal are the lesser causes. So
could we conclude that just about all generalizations
in both natural sciences and economics express in
fact tendency laws, in the sense that these Taws’
truly capture only the functioning of ‘greater
causes’ within some domain? Certainly not, since
there is a world of difference between the two cases.
Galileo’s law of falling bodies certainly
presupposes a ceteris paribus clause, so much so
that he had to employ the idealization of a ‘perfect
vacuum’ to get rid of the resistance of air. However,
he was able to give estimates of the magnitudes of
the amount of distortion that friction and the other
‘accidents’ would determine and which the law
ignored. In other words, whereas in natural sciences
the ‘disturbing causes’
3363
have their own laws, this is not the case in eco-
nomics where we find tendency statements with
unspecified ceteris paribus clauses or, if specified,
specified only in qualitative terms. In economics it
is generally impossible to list all the conceivable
inferences implied in a lawlike statement and to
replace the ceteris paribus clause with precise
conditions. So, for example, the law that ‘less will
be bought at a higher price’ is not refuted by panic
buying, nor is it confirmed by organized consumer
boycotts. No test is decisive unless ceteris are really
paribus.
These remarks help to understand the role
acknowledged by Keynes to laws in economic
inquiry. Besides general laws, there are also rales
and norms which are significant in the explanation
of economic behaviour. To Keynes, it makes no
sense to reduce all forms of explanation in
economics to that of the covering-law model.
Indeed, whereas to justify a law one has to show
that it is logically derivable from some other more
general statements, often called principles or pos-
tulates, the justification of rales occurs through the
reference to goals and the justification of norms
through the reference to values which are not
general sentences, but rather intended singular
patterns or even ideal entities. Since no scientific
law, in the natural scientific sense, has been
established in economics, on which economists can
base predictions, what are used and have to be used
to explain or to predict are tendencies or patterns
expressed in empirical or historical generalizations
of less than universal validity, restricted by local
and temporal limits. Recently, Arrow has amazed
orthodox economists when raising doubts about the
mechanistically inspired understanding of economic
processes: ‘Is economics a subject like physics, true
for all time or are its laws historically conditioned?’
(Arrow , p. 322).
The list of generally accepted economic laws
seems to be shrinking. The term itself has come to
acquire a somewhat old-fashioned ring and econ-
omists now prefer to present their most cherished
general statements as theorems or propositions
rather than laws. This is no doubt a healthy reac-
tion: for too long economists have been under the
nomological prejudice, of positivistic origin, that
3364
the only route towards explanation and prediction is
the one paved with laws, and laws as forceful as
Newton’s laws. Images in science are never inno-
cent: wrong images can have disastrous effects.
Bibliography
Arrow, K. 11>X5. Economic history: A necessary though not
sufficient condition for an economist. American Economic
Review: Papers and Proceedings 75: 320-323. Keynes, J.M.
107.3. The general theory and after. Part II: Defence and
development. In The collected writings of John Maynard
Keynes, vol. 14. London: Macmillan. Monger, C. 1883.
Untersclnmgen liber die Methode der Sozialwissenschaften.
Leipzig: Duncker & Humblot. Mill, J.S. 1836. On the
definition of political economy and the method of
investigation proper to it. Repr. In: Collected works of John
Stuart Mill, Essays on economy and society, vol. 4, ed. J.M.
Robson. Toronto: University of Toronto Press, 1967.
von Miscs, L. ll)4l). Human action: A treatise on economics.
London: William Hodge.
Economic Man
Shaun Hargreaves-Heap and Colin G. Clark
Abstract
Economic man ‘knows the price of everything
and the value of nothing’, so said because he or
she calculates and then acts so as to satisfy best
his or her preferences. The value of these pref-
erences is immaterial. The hypothesis has nev-
ertheless proved remarkably powerful not only
in economics but across the social sciences
where it has spawned 'rational choice’ accounts
of many aspects of social life. This ambition has
attracted critics both from without and within.
The latter have developed, with insights from
psychology on how people acquire and use
information, a less elegant but arguably more
realistic model.
Keywords
Behavioural economics; Bounded rationality;
Comparative statics; Economic man; Expected
Economic Man
utility maximization; Game theory; Homo
economicus; Hume, D.; Information economics;
Institutional economics; Kant, I.; Law of small
numbers; Learning; Neoclassical economics;
Preferences; Rational behaviour; Rational
expectations; Reference dependence; Risk;
Satisficing; Self-serving biases; Social norms;
Subjective probability; Welfare economics;
White noise
JEL Classifications
B4
Among the many different portrayals of economic
agents, the title of homo economicus is usually
reserved for those who are rational in an instru-
mental sense. For example, this is how agency is
defined in neoclassical economics. In its ideal type
case the agent has complete, fully ordered prefer-
ences (defined over the domain of the consequences
of his or her feasible actions), perfect information
and all the necessary computing power. After
deliberation, he or she chooses the action that
satisfies their preferences better (or at least no
worse) than any other. No questions are raised
about the source or worth of preferences, reason
focuses on the efficient selection of the means to
given ends.
This basic model is then made more sophisti-
cated. The theory of risk allows for the point that an
action may have several possible consequences.
When preferences are represented via the device of
a utility function, the agent assesses his or her
expected utility by discounting the utility of each
consequence by how likely it is to be the actual one.
That requires the agent to have a probability
distribution for the consequences, even if only a
subjective one. Other refinements include allowance
for costs of acquiring information, of processing it
and of action. Then there arc complexities,
illustrated by game theory, when actions of other
agents form part of the environment in which the
person acts. The basic vision remains, however, one
of agents who arc rational in the sense that they
maximize an objective function subject to
constraints (or act ‘as if’ this were the case).
Economic Man
This vision is not unique to neoclassical eco-
nomics. For example, Marx’s profit- maximizing
capitalist fits the same instrumental model of
rationality. Institutionalist accounts of, for instance,
banks or trade unions often conceive economic
bodies as similar unitary rational agents. Nor is the
vision confined to any specific motivating desire in
agents, like a selfish pleasure- maximizing drive.
There is scope for allowing ethical preferences
alongside the symptomatic textbook desires for
apples and oranges. Agents are, however, regarded
as self-interested, in the looser sense that they are
moved to satisfy whatever preferences they happen
to have. Furthermore, granted that de gustibus non
est disputandum, this modest base is enough to
ground a full-blown social theory on a model of
agency which can be exported to other social
sciences.
Such a social theory is individualist and
contractarian, with a pedigree that includes
Hobbes’s Leviathan and Benthamite utilitarianism.
The satisfaction of individual preference, aided by
felicific calculation, is what makes the social world
go round. Social relations become instrumental, in
the sense that they embody exchanges in the service
of individual preferences (see Becker ). For
instance, marriage has been analysed in this spirit as
an arrangement to secure the mutual benefit of
exchange between two agents with different
endowments. Crime has been claimed to occur
because calculation of costs and benefits proves it to
be the action that maximizes expected utility.
Meanwhile, institutions, which feature in
elementary microeconomics as constraints on
individual choice, become deposits left by earlier
transactions, often deliberately so as devices to
prevent preferences being frustrated by situations of
the Prisoner’s Dilemma type. Government policies
are explained on the hypothesis that the political
arena is also peopled by individuals maximizing
expected utility, who form coalitions in support of
policies that will secure reelection (see Downs ). In
short, homo economicus morphs into a universal
homo sapiens.
Such a full-blown social theory may be too
ambitious because assumptions that are plausible
3365
for simple market transactions become suspect
when scaled up. For example, the ideal-type case
makes agents, so to speak, transparent to them-
selves, and does not allow for history occurring
behind their backs. Freudians would object to
transparency of preferences and Marxians would
invoke theories of false consciousness. (Although
Marx’s capitalists are instrumentally rational, their
desire to maximize profit is an alienated one,
‘forced’ on them by a competitive capitalist
E
system.) Many other social theorists would object to
the treatment of norms and social relations as
instrumental, on the grounds that norms are prior to
preferences. For instance, cultural forms like the
rales of orchestral composition are a soiuce of
musical preferences rather than a solution to a priori
problems of maximizing musical enjoyment. Or, to
put this differently, game theory yields too many
instances of indeterminacy for an ambitious
programme of reducing all social practices to the
exercise of instrumental reason by the individual
participating agents.
Such objections, of course, need not affect the
more modest enterprise of explaining economic
transactions within the parameters of social institu-
tions like the market. But even here homo
economicus has critics. Philosophically, it is not
plain that preferences can be taken as given in a
sense which makes them impervious to the agent’s
beliefs about the moral quality of his or her actions.
In supposing that only desires can motivate agents,
the economist is taking sides in a continuing phil-
osophical dispute between Humeans, who regard
reason as the slave of the passions, and Kantians,
who make place for the rational monitoring of
desire. This dispute surfaces plainly in welfare
economics, when it is asked whether all preferences
should count equally or whether ‘capabilities’ are
more appropriate for the evaluation of social states
than degrees ofpreference satisfaction, but bears on
the elementary model of action too (see Sen ).
There are also methodological doubts about the
empirical standing of the model. What would falsify
the claim that economic agents seek the most
effective means to satisfy their preferences?
Apparent counter-examples can always be dealt
with by treating them as evidence that preferences
3366
have changed or been dismissed through a careful
individuation of outcomes. Indeed, since prefer-
ences are unobservable, they can be identified only
if the correctness of the model is presupposed. In
other words, there is room for deeper dispute about
the foundations of orthodox microeconomics than is
always realized.
Even within economics there are critics. The
most substantial attack comes from those who think
that perfect information is not a useful limiting case
of imperfect information. Granted that there is often
no way of calculating the likely marginal costs and
benefits of acquiring extra information (short of
actually acquiring it), how shall the agent decide
rationally when to stop? Simon ( ) uses the
question to argue for
‘satisficing’ models, in place of maximizing ones,
and for ‘procedural’ or ‘bounded’ rationality.
Rationality, he suggests, is a matter of following a
procedure that halts with a good solution, and
should not be defined in terms of best solutions.
While this is a tempting thought, it is not obvious
that searching for a ‘good’ solution is any easier
than the best one if ‘good’ is some kind of second-
best version of the ‘best’. As a result, ‘behavioural
economists’ have been drawn to the large
experimental literature in psychology on how
people actually behave and have produced
economic models of decision-making that incor-
porate a variety of psychological processes such as
‘self-serving biases’, the Taw of small numbers’
and ‘reference dependence’ (see Kahneman ). In
this way, homo economicus has become more
psychologically complex and more of an
institutional or organizational person than an
abstract maximizer.
The rational expectations hypothesis offers a
different approach to the information issue. A
rational agent who is short of information should
not use an information-generating mechanism that
gives rise to systematic errors. If errors are
systematic, the agent should be able to learn how to
eliminate them by amending the mechanism. There
is an incentive to do so, because improved estimates
of future variables will be profitable. On the face of
it this makes rational expectations the natural ally of
the pure economic-man models. Economic Man can
Economic Man
proceed much as before, in the assurance that
inadequate information involves nothing more
systematic than ‘white noise’ and with the benefit of
fresh analytic results that flow from a rational
expectations hypothesis.
But this is to sidestep the informational problem
set earlier, unless one sees how rational agents will
leam to remove systematic errors. When there are
costs to learning then it may not be rational to
expend the effort that achieves a rational expecta-
tion. If we set such costs aside, in some simple
learning situations a Bayesian updating procedure
turns a rational expectations-generating process into
an approximation of adaptive expectations, which
could be construed as a procedural rule of thumb.
But no general rapprochement between maximizing
and procedural models of rationality follows. In
more general learning situations the rational agent is
trying to leam the rational expectations equilibrium
relationship between variables - the one which, if
used by agents to form their expectations, would
reproduce itself in experience (white noise apart).
This sounds easy, in that repeated experience of a
particular relationship should lead to convergence
on accurate parameter estimates. However,
ignorance of the rational expectations equilibrium
values produces behaviour that departs from those
values. So observed values of variables embody a
distortion which agents cannot correct without
knowing the dimensions of their own ignorance. To
know this, however, they would have to know the
rational expectations equilibrium values already. To
put it as the procedural critics might, learning would
be feasible only if there were nothing to leam. The
information question has been begged; and the door
again opens on to psychology and its rich literature
on what people actually do.
Nevertheless, the ideal-type Economic Man
remains a powerful model of action not only in
neoclassical theories, where insights in comparative
statics have been especially notable, but elsewhere
too. How powerful it finally is depends, within
economics, on what becomes of the informational
difficulties and on whether procedural or bounded
models can come up with rival results of equal
scope and elegance. For the wider social sciences, it
offers a tempting analysis of social
Economic Organization and Transaction Costs
behaviour at large both for transactions in other
social arenas and for the emergence of the institu-
tions that govern those arenas. But the greater its
ambitions, the more serious become the unresolved
doubts about the origin of preferences and their
relation to norms and institutions.
See Also
► Altruism, History of the Concept
► National Behavic ::
► R lity, History of the Concept
► Utilitarianism and Economic Theory
Bibliography
Becker, G. 1976. The economic approach to human behav-
iour. Chicago: Chicago University Press.
Downs, A. 1957. An economic theory of democracy. New
York: Harper Row.
Kahneman, D. 2003. Maps of bounded rationality: Psy-
chology for behavioural economics. American Economic
Review 93: 1449-1475.
Sen, A. 1999. Commodities and capabilities. Oxford: Oxford
University Press.
Simon, H.A. 1976. From substantive to procedural ratio-
nality. In Method and appraisal in economics, ed. S.
Latsis. Cambridge: Cambridge University Press.
Economic Organization
and Transaction Costs
Steven N. S. Cheung
One important extension of the Coase Theorem
states that, if all costs of transactions are zero, the
use of resources will be similar no matter how
production and exchange activities are arranged.
This implies that in the absence of transaction costs,
alternative institutional or organizational
arrangements would provide no basis for choice and
hence could not be interpreted by economic theory.
Not only would economic organization be
randomly determined; there actually would not be
any organization to speak of: production and
3367
exchange activities would simply be guided by the
invisible hand of the market.
But organizations or various institutional
arrangements do exist, and to interpret both then-
presence and their variation, they must be treated as
the results of choice subject to the constraints of
transaction costs.
In the broadest sense transaction costs encom-
pass all those costs that cannot be conceived to exist
in a Robinson Crusoe economy where neither
property rights, nor transactions, nor any kind of
economic organization can be found. This breadth
of definition is necessary because it is often
impossible to separate the different types of cost. So
defined, transaction costs may then be viewed as a
spectrum of institutional costs including those of
information, of negotiation, of drawing up and
enforcing contracts, of delineating and policing
property rights, of monitoring performance, and of
changing institutional arrangements. In short, they
comprise all those costs not directly incurred in the
physical process of production. Apparently these
costs are weighty indeed, and to term them
‘transaction costs’ may be misleading because they
may loom laige even in an economy where market
transactions are suppressed, as in a communist state.
By definition, an organization requires someone
to organize it. In the broadest sense, all production
and exchange activities not guided by the invisible
hand of the market are organized activities. Thus,
any arrangement that requires the use of a manager,
a director, a superviser, a clerk, an enforcer, a
lawyer, a judge, an agent, or even a middleman
implies the presence of an organization. These
professions would not exist in the Crusoe economy,
and payments for their employment are transaction
costs.
When transaction costs are defined to include all
costs not found in a Crusoe economy, and economic
organizations are defined equally broadly to include
any arrangement requiring the service of a visible
hand, a corollary appears: all organization costs are
transaction costs, and vice versa. That is why during
the past two decades economists have striven to
interpret the various forms of organizational
arrangements in terms of the varying costs of
transactions.
3368
Some obvious examples will illustrate the point.
A worker in a factory (an organization) may be paid
by a piece rate or by a wage rate. If the costs of
measuring and enforcing performance (one type of
transaction cost) are zero, then either arrangement
will yield the same result. But if these costs are
positive, the piece-rate contract will more likely
prevail if the costs of measuring outputs arc
relatively low, whereas the wage contract will more
likely be chosen if the costs of measuring hours and
enforcing performance arc low relative to the costs
of measuring outputs. As another example, some
restaurants (again an organization ) measure the
quantity of food sold; others serve buffet dinners,
allowing customers to cat as much as they please at
a fixed price per head. The cost of metering and
quantifying food consumption relative to the basic
cost of the food will determine which arrangement
is chosen. In the total absence of transaction costs,
the factory or the restaurant would not exist in the
first place, because consumers would buy directly
from the input owners who produce the goods and
services.
As early as 1937, R.H. Coase interpreted the
emergence of the firm (an organization) in light of
the costs of determining market prices (transaction
costs). When these costs are substantial because of
the difficulties of measuring separate contributions
by workers and of negotiating prices for separate
components of a product, a worker may choose to
work in a factory (a firm); he surrenders the right to
use his labour by contract and voluntarily submits to
direction by a visible hand, instead of personally
selling his services or contributions to customers
through the invisible hand of the market. The film is
therefore said to supersede the market. As the
supersession progresses, the saving in the costs of
determining prices will be countered by the rising
costs of supervision and of management in the firm,
Equi- librium is reached when, at the margin, the
cost saving in the former equals the rising cost in
the latter.
The firm superseding the market may be
regarded as a factor market superseding a product
market. If all costs of transactions were zero, the
two markets would be inseparable in that a payment
made by a customer to the owner of a factor
Economic Organization and Transaction Costs
of production would be the same as payment made
to a product seller. In such a world it would be a
fallacy to speak of the factor market and the product
market as coexisting entities.
The presence of transaction costs is a prelude to
separate the factor market from the product market.
However, in some arrangements, such as the use of
certain piece rates, it may become impossible to
separate the one market from the other. Therefore,
instead of viewing the film as superseding the
market, or the factor market as superseding the
product market, it is more correct to view the
organizational choice as one type of contract
superseding another type. In these terms, the choice
of organizational arrangements is actually the
choice of contractual arrangements.
When organizational choices arc viewed as
contractual choices, it becomes evident that it is
often impossible to draw a clear dividing line
separating one organization from another. Take the
film, for example. It is often the case that the
entrepreneur who holds employment contracts (and
it is not clear whether it is the entrepreneur who
employs the workers or the workers who employ
the entrepreneur) may contract with other films; a
contractor may subcontract; a subcontractor may
sub-subcontract further; and a worker may contract
with a number of ‘employers’ or ‘films’. If the
chain of contracts were allowed to spread, the ‘firm’
might encompass the whole economy. With this
approach the size of the firm becomes indeteiminatc
and unimportant. What arc important are the choice
of contracts and the costs of transactions that
determine this choice.
Traditional economic analysis has been confined
to resource allocation and income distribution.
Contractual arrangements as a class of observations
have been slighted in that tradition. In a world
complicated by transaction costs, this neglect not
only leaves numerous interesting observations
unexplained, but actually obscures the
understanding of resource allocation and income
distribution. The economics of organization or
instimtion or, for that matter, the workings of
various economic systems, were never placed in the
proper perspectives under the traditional approach.
For generations students were told that
Economic Organization and Transaction Costs
various kinds of ‘imperfections’ were the cause of
seemingly mysterious observations: policies were
‘misguided’, or antitrust specialists were barking up
the wrong trees.
The costs of introducing new and more valid
ideas must have been enormous. Even today text-
books still discuss marginal productivity theory
only with reference to fixed wage and rental pay-
ments. Yet economists have known all along that
(for laboin alone) payments may be in the periph-
eral forms of piece rates, bonuses, tips, commis-
sions, or various sharing arrangements; moreover,
even wage rates may assume a number of forms.
Each type of contract implies different costs of
supervision, of measurement, and of negotiation,
and the form of economic organization, along with
the function of the visible hand, changes whenever
a different contractual arrangement is chosen.
The choice of contractual arrangements is not, of
course, confined to the factor markets. In the
product markets, pricing arrangements such as tie-in
sales, full-line forcing, or membership fees
associated with clubs, may similarly be interpreted
in light of transaction costs. Further, business
organizations in mergers, franchises, and various
forms of integration are now beginning to be
viewed as transaction-cost phenomena. Indeed,
close inspection of department stores and shopping
centres reveals pricing and contractual
arrangements between a central agent and individ-
ual sellers, as well as among the sellers themselves,
which could not be explained by textbook
economics.
Transaction costs are often difficult to measure
and, as noted earlier, difficult to separate by type.
However, the measurement problem can be avoided
if only we are able to specify how these costs vary
under different observable circumstances, and their
different types are separable if viewed in terms of
changes at the margin. These two conditions are
requisite in the derivation of testable implications
for the interpretation of organizational behaviour.
The use of transaction costs to analyse institu-
tional (organizational) choice is superior to three
other approaches. One approach would focus on
incentives. However, incentives are not in
3369
principle observable, and we will do better in
deriving testable propositions if the same problem is
viewed in terms of the costs of enforcing per-
formance. A second approach adopts risk. However,
it is difficult to ascertain how risk is altered under
different circumstances. Many risk problems, such
as the uncertainty of whether an agreement will be
honoured, are also problems of transaction costs,
and it is easier to deal directly with those. Finally,
some recent advances in transaction-cost analysis
have called attention to the costs embodied in
dishonesty, cheating, shirking, and opportunistic
behaviour. Yet these are loose terms and, whatever
they describe, to some extent are always to be
found. To the degree that we can identify the
particular costs of transactions that promote
dishonesty, that shadowy explanation is no longer
needed. After all, in what sense can we say a person
is ‘increasingly dishonest’ or ‘increasingly
opportunistic’?
The transaction-cost approach to analysis of
economic organizations can be extended upward
from a few participants to the ‘government’ or even
the nation itself. At the lower level, the owners of
condominium units almost as a rale form
associations with specific by-laws and elect
committees to act on matters of common concern,
the decisions being determined by majority vote.
The transaction costs of ballot voting are less than
those of using prices and dollar votes in certain
circumstances, and trivial matters may even be
delegated to a ‘dictatorial’ manager to further
reduce the cost of voting. Similarly, residents in a
particular location may choose to incorporate into a
city, selecting their own mayor, with a committee
setting up the building codes, hiring firemen and
policemen, and deciding other matters of common
concern.
Private property rights offer the unique advan-
tage of allowing individual property owners the
option of not joining an organization. This choice is
an effective restraint against the adoption of an
organization with higher transaction costs. It is true
that a home-owner in a given region may, by
majority vote, lose his option of not joining in a city
corporation (unlike a worker who, in a free
enterprise economy, always has the option of not
joining a ‘firm’). But with private property
3370
rights the majority vote aims at cost saving, and a
reluctant resident may exercise his own judgement
by selling his house and moving elsewhere.
Private property rights further reduce transaction
costs under competition. An entrepreneur or agent
who wants to recruit other resource owners to join
his organization must, under competition, offer
attractive terms, and this can be achieved only if his
organization can effectively reduce transaction
costs. On the other hand, the resource owner
competing to join an organization will be more
inclined to deliver a good performance when at risk
of losing his job.
The option of not joining an organization and
the cost-reducing function of competition are, of
course, restrained when an organization is extended
to encompass an entire nation. When citizenship is
dictated by birth, the option of not joining is
restrained, and competition among nations to recruit
members is decidedly less than among organizations
within a nation. This relative lack of cost-reducing
mechanisms is all the more evident in a communist
state, where a citizen does not have the option of
choosing an organization within that state.
A communist state may be regarded as a
‘superfirm’ in which comrades lack the option of
not joining. Each worker is assigned to a particular
job supervised and directed by the visible hands of
comrade officials of varying ranks. In this aspect the
communist state is remarkably similar to what
Coase calls a ‘firm’, where workers are told what to
do instead of being directed by market prices. But
the lack of market prices in the communist state is
not due to the costs of determining prices; rather, in
the absence of private property rights market prices
simply do not exist, and visible supervision by a
hierarchy ranking becomes the remaining alternative
to chaos.
The transaction costs of operating an organiza-
tion are necessarily higher in a communist state than
in a free enterprise economy, due to the lack of
option of not joining and the lack of competition
both to recruit members among organizations and to
induce members to perform well.
If the transaction costs of operating organization
were zero, resource allocation and
Economic Organization and Transaction Costs
income distribution would be the same in a
communist state as in a free enterprise state:
consumer preferences would be revealed without
cost; auctioneers and monitors would provide freely
all the services of gathering and collating
information; workers and other factors of
production would be directed free of cost to produce
in perfect accord with consumer preference; each
consumer would receive goods and services in
conformity with his preferences; and the total
income received by each worker, as determined
costlessly by an arbitrator, would equal his marginal
productivity plus a share of the rents of all resources
other than labour, according to any of a number of
criteria costlessly agreed upon. But such an ideal
situation is obviously not to be found.
We therefore conclude that the poor economic
performance of a communist state is attributable to
the high transaction costs of operating that
organization. Under the postulate of constrained
maximization, the communist state survives for the
same reason that any ‘inefficient’ organization
survives: namely, the transaction costs of changing
an organizational (institutional) arrangement are
prohibitive. Such costs include those of obtaining
information about the workings of alternative
institutions, and of using persuasive or coercive
power to alter the status of the privileged groups
whose incomes might be adversely affected by the
institution of a different form of economic
organization.
See Also
Bibliography
Alchian, A.A., and H. Demsetz. 1972. Production, infor-
mation costs, and economic organization. American
Economic Review 62: 777-795.
Barzel, Y. 1982. Measurement costs and the organization of
markets. Journal of Law and Economics 25: 27-48.
Cheung, S.N.S. 1969. Transaction costs, risk aversion, and
the choice of contractual arrangements. Journal of Law
and Economics 12: 23—42.
Economic Sanctions
Cheung, S.N.S. 1982. Will China go ‘capitalist’? Hobart
Paper 94. London: International Economic
Association.
Cheung, S.N.S. 1983. The contractual nature of the firm.
Journal of Law and Economics 26: 1-21.
Coase, R.H. 1937. The nature of the firm. Economica 4: 386-
405.
Jensen, M.C., and W.H. Meckling. 1976. Theory of the firm:
Managerial behavior, agency costs and ownership
structure. Journal of Financial Economics 3: 305-360.
Klein, B., R.G. Crawford, and A. A. Alchian. 1978. Vertical
integration, appropriable rents, and the competitive
contracting process. Journal of Law and Economics
21:297-326.
Knight, F.H. 1921. Risk uncertainty and profit. Boston:
Houghton Mifflin.
McManus, J.C. 1975. The costs of alternative economic
organizations. Canadian Journal of Economics 8: 334.
Williamson, O.E. 1975. Markets and hierarchies: Analysis
and anti-trust implications. Glencoe: Free Press.
Economic Sanctions
Jeffrey J. Schott
Abstract
Economic sanctions are tools of statecraft used
to achieve a broad range of foreign policy goals
by threat or deployment of coercive measures
such as trade embargoes, asset freezes, or with-
holding of development aid. Throughout the
post-war era, the United States and other coun-
tries frequently have imposed economic sanc-
tions, even though they have contributed only
infrequently to foreign policy successes. Glob-
alization has made the exercise of economic
coercion increasingly complex, but has not
obviated the utility of sanctions as part of the
foreign policy arsenal.
Keywords
Economic development; Economic sanctions;
Globalization; Terrorism
JEL Classifications
F51
3371
Economic sanctions are tools of statecraft used to
influence the behaviour of foreign countries by the
threat or actual withdrawal of trade and sources of
finance. Traditional means of coercion include trade
embargoes, withholding development assistance,
and asset freezes. The objective is to confront a
foreign country with a choice: either bear the cost of
lost trade and finance, or change policies to comply
with the demands of those imposing the sanctions
(the sender countries). Projecting power through
economic coercion is deemed more forceful than
diplomatic reproach yet less drastic than military
intervention. In practice, economic measures
generally are deployed as part of a broader
programme of foreign policy responses
encompassing diplomatic entreaties, covert or
quasi-military intrusions, and threat of or
preparation for military action.
Countries impose sanctions in pursuit of a
variety of foreign policy goals. Historically, eco-
nomic sanctions have preceded and then accom-
panied military conflict. The oil embargo of Japan
was a prelude to the Second World War in the
Pacific; so, too, were the United Nations’ sanctions
against Iraq following its invasion of Kuwait in
1990. Obviously, sanctions are part and parcel of
‘hot wars’ that sever economic ties between the
combatants; but they are also prevalent in ‘cold
war’ episodes, where the goal is to impair military
capabilities through denial of weapons and dual- use
technologies (for example, post-war sanctions
against the Soviet Union and its satellites under the
auspices of the Consultative Group and Coor-
dinating Committee for Multilateral Export Con-
trols, or CoCom, and efforts to blunt the
development of nuclear weapons in Iran and North
Korea). In addition, sanctions have sought to
impede or reverse military inclusions across borders
(for example, the League ofNations effort to get
Italy to withdraw from Abyssinia in 1936) and
between warring factions within a country (the sad
recent history of several West African states).
Not all sanctions episodes respond to or presage
military actions. Many post-war cases have been
advanced to counter other types of aberrant
behaviour such as state sponsored terrorism, pro-
liferation of weapons of mass destruction, or
3372
human rights abuses. In these cases, sender coun-
tries impose sanctions in an effort to redress foreign
outrages, to deter emulation by others (the rationale
in most anti-proliferation cases), and to punish the
target regime for its misdeeds (for example, the US
grain embargo alter the Soviet invasion of
Afghanistan in 1989 ). In a number of cases,
sanctions pursue the goal of regime change sotto
voce - whether the target is Moammar Gaddafi in
Libya, Kim Jong-il in North Korea, or the
Afrikaaners in South Africa. Sanctions that portend
regime change obviously meet stauncher resistance
than those that seek narrow changes in governance
by the target government.
Do Sanctions 'Work'?
Foreign policy ventures seldom yield unambiguous
results. Gauging the effectiveness of sanctions
involves a combination of quantitative method and
intuition, and often requires subjective evaluation of
incomplete results. Sanctions alone seldom are
sufficient to change foreign practices, but they can
contribute to the achievement of policy goals in
conjunction with other instruments of statecraft, if
properly designed and implemented. That is easier
said than done.
Sanctions are blunt policy instruments; they are
better at impairing economic performance over time
than at inflicting surgical strikes on target countries.
Senders that expect immediate gratification often
tire of the effort, especially if the sanctions impose
significant costs on their own firms and workers.
Moreover, when sanctions are hard hitting, it is
difficult to avoid innocent victims within the target
country and in neighbouring states; in such cases,
the debilitating effect of sanctions often results in
substantial suffering among the civilian population.
Humanitarian exemptions from the sanctions
designed to soften the blow to the general public
invariably weaken the economic impact of the
sanctions and muddy the policy signal to the target
regime. To be siue, such loopholes in the sanctions
net are important both on moral grounds and to
maintain the cohesion of the coalition of sender
countries, but the loopholes are prone to abuse
(witness the
Economic Sanctions
scandalous operation of the United Nations’ oil-for-
food programme, which was supposed to channel
Iraqi oil export revenues to humanitarian assistance)
and reduce the economic pressiue to comply with
the sender’s demands.
Almost all sanctions leak; targeted countries can
evade the full thrust of the economic restrictions by
redirecting trade and finance to non-sanctioning
states or by engaging in clandestine operations.
Countries seeking economic or political influence
with the target regime often conspire to evade the
sanctions; the Cold War period was replete with
examples of ‘Black Knight’ countries coming to the
rescue of targeted regimes with aid to offset the
impact of sanctions imposed by the United States or
the Soviet Union. Smugglers still outwit even the
most comprehensive embargoes - witness the
billions of dollars earned by Saddam Hussein, the
former Iraqi president, from illicit oil exports diuing
the period of ‘comprehensive’ UN sanctions against
Iraq. For a price, targeted regimes can still prociue
goods, services and technologies; the profit motive
seems to be an irresistible force regardless of region
or culture!
That said, sanctions have contributed to a few
notable successes in the post-war era, including the
collapse of the apartheid regime in South Africa and
the renunciation of terrorism by President Gaddafi
in Libya. Hufbauer et al. ( )
found success - measured by the partial fulfillment
or better of policy goals - in more than a quarter of
the almost 200 sanctions episodes documented in
the 20th century. (The third edition of this
comprehensive study of economic sanctions
contains updated policy analysis and case studies,
and an extensive bibliography. See also Baldwin ,
for an examination of the tools of economic
statecraft, and Martin , for analysis of the use of
multilateral economic sanctions.) Most of these
cases, however, involved relatively modest
demands on the target country. When the stakes are
high, resistance by the target regime stiffens.
Accordingly, most high-profile sanctions cases -
like those seeking to oust President Castro in Cuba
or to deter support for terrorism and the
development of nuclear weapons by the ayatollahs
in Iran - have been abject failures.
Economic Science and Economics 3373

Can Sanctions Be Effective in an Era familiar goals of economic coercion now must be
of Rampant Globalization? pursued through measures adapted to the changing
conditions in global markets. The use of economic
Economic sanctions traditionally have been the sanctions needs to be reconsidered and revamped,
domain of big powers, acting unilaterally or as part but not abandoned.
of a broader international coahtion. Until recently,
the big powers controlled the trade lanes and purse
strings of international commerce, and held a near See Also
monopoly on advanced technologies. Since the
mid-1980s, however, the success of postwar
economic development, spurred in part by the
spread of technological innovation, has eroded the
franchise of the big powers and created alternative
sources of goods, technology and capital for coun-
tries targeted by economic sanctions. Simply put,
Bibliography
globalization has made it much harder to design an
effective sanctions policy. Baldwin, D.A. 1985. Economic statecraft. Princeton:
In addition, global politics are now more com- Princeton University Press.
plex than in the period of East-West rivalry. Former Hufbauer, G.C., K. Elliott, J.J. Schott, and B. Oegg. 2007.
Economic sanctions reconsidered. 3rd ed. Washington, DC:
allies differ regarding strategies and priorities for
Peterson Institute for International Economics. Martin, L.L.
using sanctions to deal with regional trouble spots. 1992. Coercive cooperation: Explaining multilateral
For example, Europe is more vulnerable than the economic sanctions. Princeton: Princeton University Press.
United States to an interruption of energy supplies
from the Middle East, and thus is less willing to
constrain oilfield development and to take actions
that risk political retaliation. Similarly, China and
Japan are highly dependent on imported energy and
thus sensitive to sanctions against Iran and other oil Economic Science and Economics
- producing states.
Henry Sidgwick
Globalization also has contributed to the
decentralization of power, allowing smaller
countries - especially those rich in energy resources
- to provide offsetting assistance to blunt the
economic impact of sanctions. But the influence of The terms ‘economy’ and ‘economic’ or ‘econom-
globalization goes beyond the realm of state-to- ical’, are now used chiefly in two meanings, which
state intervention; terrorism, for example, now it is well to distinguish clearly; since, though diver-
operates in a stateless domain of sleeper cells and gent in their history, they are liable to fusion, and
territories outside of governmental control linked therefore in some degree to confusion.
through informal financial and telecommunications ‘Economy’ originally meant, in Greek, the man-
networks. For that reason, sanctions policies agement of the affairs of a household, especially the
increasingly seek to target individuals and provision and administration of its income. But
corporations as well as governmental bodies, and to since both in the acquisition and in the employment
favour financial measures to interdict inter-bank of wealth it is fundamentally important to avoid
electronic transfers in addition to the more waste either of labour or of its produce, ‘economy’
traditional controls on trade, investment and in modem languages has come to denote generally
development assistance. the principle of seeking to attain, or the method of
In sum, economic sanctions continue to play a attaining, a desired end with the least possible
major role in international relations. However, the expenditure of means; and the words ‘economy’,
‘economic’, ‘economical’, are often used in this
3374
sense, even without any direct relation to the pro-
duction, distribution, or consumption of wealth.
Thus we speak of‘economy of force’ in a mechan-
ical arrangement without regard to its utility, and of
'economy of time’ in any employment whether
productive of wealth or not.
On the other hand, as there is an obvious anal-
ogy between the provision for the needs of a state
and the provision for the needs of a household,
'political economy’, in Greek, came to be recog-
nized as an appropriate teim for the financial branch
of the art or business of government. It is found in
this sense in a treatise translated as Aristotle’s in the
13th century; and so, when, in the transition from
medieval to modem history, the question of ways
and means obtrusively claimed the attention of
statesmen, ‘political economy’ was the name
naturally given to that part of the art of government
which had for its aim the replenishment of the
public treasury, and - as a means to this - the
enrichment of the community by a provident
regulation of industry and trade. And the term
retained this meaning till the latter part of the 18th
centtiry without perceptible change - except that,
towards the end of this period, the enrichment of
the people came to be less exclusively regarded
from the point of view of public finance, and more
sought as a condition of social well-being.
But in the latter part of the 18th century,
under the influence primarily of the leading
French ‘Economistes’ or ‘Physiocrats’ -
Quesnay, De la Riviere, and others - the
conception of political economy underwent a
fundamental change, in consequence of a
fundamental change in the kind of answer which
these thinkers gave to the question ‘how to make
a nation wealthy’. The physiocrats proclaimed to
France, and through France to the world, that a
statesman’s true business was not to make laws for
industry and trade in the hope of increasing
wealth; but merely to ascertain and protect from
encroachment the simple and immutable laws of
nattire, under which the production of wealth
would regulate itself in the best possible way if
governments would abstain from meddling. A
view broadly similar to this, but less extreme,
and, partly for this reason, more directly
influential, was expounded in Adam Smith’s Wealth of
Nations. Instead of showing the
Economic Science and Economics
statesman how to ‘provide a plentiful revenue or
subsistence for the people’ - which was one of the
two main objects of political economy, according to
the traditional view - Adam Smith aims at showing
him how nature, duly left alone, tends in the main to
attain this end better than the statesman can attain it
by governmental interference. Accordingly, so far
as the widespread influence of Adam Smith’s teach-
ing went, that branch of the statesman’s art which
aimed at ‘providing a plentiful revenue for the
people’ tended almost - though not altogether - to
shrink to the simple maxim of laisserfaire: leaving
in its place a scientific sttidy of the processes by
which wealth is produced, distributed, and
exchanged, through the spontaneous and partly
unconscious division of labour among the members
of human society, independently of any govern-
mental interference beyond what is required to
exclude violence or fraud. A part, indeed, of the old
art of political economy - that which aimed at
‘supplying the state with a revenue sufficient for the
public service’ - remained indispensable to the
statesman; but it was held that this traditional art
required to be renovated by being rationally based
on the docttincs of the new-born science just
described. It is, then, this scientific sttidy of the
department of social activity that most writers on
the subject now primarily mean by the tcim 'polit-
ical economy’: such part of the old governmental art
so called, as the doctrine of the new science is held
to admit, being commonly regarded as ‘applied
political economy’. In consequence of this change
the adjective ‘economic’, instead of the too
cumbrous ‘politico-economic’, has come to denote
the matters investigated by the science of political
economy, and the propositions and arguments
relating to them.
By thinkers and duly-instructed students this
distinction between ‘science’ and ‘art’ - between the
study of‘what is’ and the study of‘what ought to be’
- is usually regarded as simple and clear; and
accordingly when such persons speak of the ‘laws
of political economy’ they mean not rules by which
the process of the social production and distribution
of wealth ought to be governed, but general
relations of eo-existenee and sequence among phe-
nomena of this class, ascertained by a scientific
study of this process as it actually takes place.
Economic Sociology
This distinction, however, has been found difficult
to establish in common thought: even well-
educated persons still occasionally speak of the
Taws of political economy’ as being ‘violated’ by
the practice of statesmen, trades-unions, and other
individuals and bodies. It is partly in order to
prevent this confusion that the terms ‘economic
science’ and ‘economics’ have recently come more
and more into use, as a preferable alternative for
political economy, so far as it is the name of a
science. As to the scope of this science - it would be
generally agreed that it is a branch of a larger
science, dealing with man in his social relations;
that it is to an important extent, but not altogether,
capable of being usefully studied in separation from
other branches of this science; and that it is mainly
concerned with the social aspect - as distinct from
the special technical aspect - of such human
activities as are directed towards the production,
appropriation, and application of the material means
of satisfying human desires, so far as such means
are capable of being exchanged. It would also be
generally agreed that the method of economic
science is partly deductive, partly inductive and
historico-statistieal. But to attempt a more precise
determination of its method and scope, and
especially of its relation to the art or system of
practical rules which should guide the action of
governments or private individuals in economic
matters, would require us to enter into questions of
a highly controversial kind; which will be more
conveniently discussed when we come to deal with
the older and wider term Political Economy.
R e p r i n t e d f r o m Palgrave’s Dictionary
of Political Economy.
Economic Sociology
Richard Swedberg
Abstract
The term ‘economic sociology’, used primarily
by sociologists, is defined as the application of reader is referred to the entry for
sociological concepts and methods of analysis
3375
to economic phenomena. Founded by Durk-
heim, Weber, and Simmel, and continued by
Schumpeter and Polanyi, it began to flourish in
the mid- 1980s around the notion that economic
actions are embedded in personal networks. The
concept of networks and other concepts and
perspectives from ‘new economic sociology’
facilitate the analysis of topics like the links
between corporations and between firms, job
search, production markets, finance markets,
insurance markets, industrial markets,
consumption, and ethnic entrepreneurship. Its
long-term impact on economics remains
uncertain.
Keywords
Becker, G; Capitalism; Weber on; Polanyi on;
Coase, R; Division of labour; Smith vs Durk-
heim on; Durkheim, E; Economic sociology;
Emotions; Entrepreneurship; Ethnic; And
immigration; Forced; Granovetter, M; Immi-
gration; And ethnic entrepreneurship; Jevons, S;
Marx, K; Networks; Strong vs weak ties; And
job search; And consumption; And groups of
firms; New Institutional Economics; Parsons, T;
Polanyi, K; On capitalism; Schumpeter, J; On
social economics;; Simmel, G; On bust; Smith,
A; Trust; Simmel on; Weber, M; On capitalism
JEL Classifications
Z1
The first recorded use of the term ‘economic soci-
ology’ is in a 1879 work by Stanley W. Jevons; and
it is clear from the context that Jevons viewed
economic sociology as part of the overall enterprise
of economics rather than as an area belonging to
another social science, such as sociology. Today, in
contrast, the term ‘economic sociology’ is used
primarily by sociologists, and they define it as the
application of sociological concepts and methods
of analysis to economic phenomena. While it is
definitely possible to treat the great concern with
institutions in New Institutional Economics, for
example, as a kind of economic sociology, the
3376
this topic for this type of analysis. Similarly, while
Gary Becker at times has referred to his extension
of the economic model to non-economic topics as
‘economic sociology’, the reader is similarly
referred to the entry for his work.
Here, the first section, on classical economic
sociology, is followed by sections on more recent
economic sociology. This way of proceeding not
only follows the general development of the field of
economic sociology but is often how economic
sociology is taught today, since the classics play a
somewhat different role in economic sociology (as
in sociology itself) to that in economics. In brief,
while sociologists are trained through work with the
classics as well as modem material, today’s
economists read the classics primarily when they
study the history of their discipline.
Classical Economic Sociology
The work of Karl Marx (1818-83) can be seen as a
type of economic sociology, in the sense just
mentioned. More generally, Marx closely linked
classical economic categories, such as value, price
and capital, to distinctly social categories, such as
class, work and relations of production.
Nevertheless, Marx has played a marginal role in
economic sociology as an academic enterprise -
except as a catalyst and inspiration for a number of
scholars, including Max Weber and Joseph
Schumpeter.
Modem academic sociology is generally
regarded as having three founders - Max Weber,
Emi le Durkh cirri and Georg Simmel - all of whom
were interested in the economy. Georg Simmel
(1858-1918), who pioneered sociology in Germany,
wrote on the sociological role of money,
competition and trust in the economy (Simmel ;
). He closely linked different types of
money to different types of social authority, and
also attempted to show how money is linked to the
element of relativism in modem society. Compe-
tition, he argued, releases the energy of all partic-
ipants to the benefit of the public, whereas in a
conflict combatants are pitted against each other and
block each other’s efforts. Trust, finally, is central
to the economy as well as society at
Economic Sociology
large; without bust, the economy as well as society
would collapse.
Emile Durkheim (1858-1917), unlike Simmel,
attempted to institutionalize economic sociology,
partly by encouraging some of his students to
specialize in this field. Durkheim’s own most
important contribution to economic sociology can
be found in his doctoral study of the division of
labour, which contains a sharp critique of the
argument in Adam Smith’s The Wealth of Nations
(Durkheim ). According to Durkheim, while Adam
Smith had seen the significance of division of
labour exclusively from the perspective of the
creation of wealth, he had neglected its importance
for the cohesion of society. More precisely, Smith
had failed to realize that the primary function of the
division of labour in modem society is to tie people
together: people who do very different things need
each other, and this is also what gives cohesion to
modem society.
The most sustained effort to lay a solid theo-
retical foundation for economic sociology and also
to carry out empirical studies can be found in the
work of Max Weber (1864-1920) (Swedberg ).
While Weber is famous for The Protestant Ethic
and the Spirit of Capitalism ( ), it is less
well known that his work is part
of a more general attempt to develop a new aca-
demic field that would complement economic his-
tory and economic theory, namely, economic
sociology.
At first Weber carried out empirical and histor-
ical studies with this goal in mind, and of these The
Protestant Ethic is by far the best known (but see
also Weber ; ). Weber’s thesis,
which holds that a certain type of religion (‘ascetic
Protestantism’) had helped to create the mentality of
modem capitalism in the 16th and 17th centuries
(‘rational capitalism’; Weber ), has led to
a heated debate. Most commentators have found
Weber’s thesis unconvincing, but it should be
emphasized that the debate is still going on with as
much fervour as in the early 20th century (see, for
example, Marshall ).
The heart of Weber’s economic sociology is to
be found in Economy and Society, a work that was
incomplete when Weber died. It is here, for exam-
ple, that Weber set out his well-known typology of
Economic Sociology 3377

capitalism: political capitalism, traditional capital- including trade, money, division of labour and
ism and rational capitalism. While the former two different ways of appropriation.
have existed for thousands of years, rational cap-
italism has emerged only in modem times and in the
West. While traditional capitalism is non-dynamic After the Classics
and centred around small enterprises involving
While the founding fathers of sociology were all
trade and the exchange of money, political
interested in economic sociology and promoted it,
capitalism is profit-making that either takes place
the topic did not become popular among sociolo-
through the state or under its direct protection, as in
gists until the mid- 1980s with the emergence of so-
imperialism. Rational capitalism, in contrast, gets
called ‘new economic sociology’. The reason for
its name from the strong element of conscious and
this is not clear, but may well have been a strong
methodical calculation: the activities of the firm are
sense among sociologists that the economists were
carried out with the help of accountants and a
better equipped to deal with economic topics. In any
trained staff; similarly, the activities of the state
case, very little work on economic sociology was
bureaucracy (including in the legal system) are
produced between 1920 and the mid-1980s.
predictable and rational. All of this makes possible
There were, however, a few exceptions. For one
a truly dynamic and revolutionary form of
thing, sociologists did discuss topics relating to the
capitalism, according to Weber.
economy, even if they did so under labels other than
Economy and Society also contains a serious
‘economic sociology’. One example is industrial
attempt by Weber to develop the central theoretical
sociology, which saw as its main task to analyse
categories of economic sociology (Weber , pp. 63-
situations when people work in groups, in the
211). The basic unit of analysis is ‘economic social
factory as well as the office. An important research
action’, which differs from economic action in
result is that workers develop norms in a number of
economic theory by partly being determined by its
areas, including what is seen as the maximum
social dimension. Economic social action is defined
effort. Those who breach these norms are punished
by Weber as behaviour that is (a) invested with
(for example, Whyte ).
meaning, (b) aimed at utility and (c) oriented to
Three individuals who all made important con-
another actor. Utility is what makes the action
tributions to economic sociology also appeared
‘economic’; and Weber’s definition of ‘social’ is to
during the period after the classics: Joseph
be found in the formula ‘orientation to another
Schumpeter, Karl Polanyi and Talcott Parsons.
actor’. The emphasis on meaning explains why
According to Schumpeter (1885-1950), economics
Weber’s sociology is called an interpretive
should be a broad science (‘social economics’) and
sociology; his economic sociology was to be a form
encompass four areas: economic theory, economic
of interpretive economic sociology.
history, economic statistics and economic sociology
Weber then proceeds to economic relationships
(Schumpeter , pp. 12-24). Schumpeter did work in
in which two actors orient their actions to one
each of these fields, including economic sociology.
another. These relationships can be either open or
According to Schumpeter, economic sociology
closed; and there is a general tendency for open
deals with institutions, while economic theory deals
economic relationships to become closed when
with economic mechanisms. Schumpeter’s three
there are not enough resources to go round.
most famous essays in economic sociology deal
Economic organizations are defined as closed social
with the issues of social class in economic life, the
relationships of a certain type; there also has to be a
role of taxation (‘fiscal sociology’) and imperialism
staff. Economic systems, finally, can be oriented
(Schumpeter ). Schumpeter thought highly of these
either to profit-making (as in capitalism) or to the
essays and they are all considered minor classics
provision for a household (as in socialism or earlier
today.
non-market economies). Weber also discusses a
host of other topics,
3378
But one can also find elements of economic
sociology in some of Schumpeter’s non-sociological
writings. This goes for the famous analysis of
entrepreneurship in Theory of Economic
Development, not least the element of resistance
from the environment that the entrepreneur usually
confronts (Schumpeter ). Similarly in Capitalism,
Socialism and Democracy, we find a sociological
portrait of contemporary capitalism. The US
economy was doing very well, according to
Schumpeter, but its institutions were decaying
(Schumpeter ).
Like Schumpeter, Karl Polanyi (1886-1964)
came from the Austro-Hungarian Empire and ended
his life on the American continent. Like
Schumpeter, he wrote a famous book on capitalism
— The Great Transformation — and contributed to
the economic sociology of his days (Polanyi ). It is
to Polanyi that we owe the term ‘embeddedness’,
even if he used it in his own, very political sense: all
economies had been embedded in politics and
religion before the advent of capitalism, and were
disembedded by the traumatic ‘great
transformation’. The political task of the day, in
other words, was to re-embed the economy into
political and human values.
Polanyi covered historical distances with great
ease and was as much at home in ancient Babylonia
as in 19th-century Britain or 20th-century United
States. The scope of his knowledge about the
economy is also reflected in one of his most useful
sets of categories: the concepts of reciprocity,
redistribution and exchange (for example, Polanyi ).
In a kinship situation, for example, reciprocity may
be used as a way of distributing resources. A
political centre, like the state, would in contrast
redistribute resources; and a market distributes
resources through exchange. Most economic
systems draw on each of these three ways of
distributing resources, with their corporate sectors
(‘exchange’), state sectors (‘redistribution') and
household sectors (‘reciprocity’ ).
Talcott Parsons (1902-1979) had begun his
career as an economist, only to switch to sociology,
since he thought that utilitarian thought was unable
to properly capture the structure of modem society.
Parsons argued for a general systems
Economic Sociology
perspective in social theory, and suggested in
Economy and Society (together with Neil Smelser)
that the economy should be conceptualized as a sub-
system of the general system of society (Parsons
and Smelser ). Just as each society has to have a
distinct goal (‘Polity’) and a value-system (‘Latent-
Pattem-Maintenance’), it also has to adapt to nature
and reality (‘Economy’). While it is part of society,
the economy is also its own society, with a ‘polity’,
Tatent- pattem-maintenance’, and so on.
New Economic Sociology
Around the mid-1980s American sociologists
suddenly started to become interested in economic
sociology, and it is this development that is
generally known as ‘new economic sociology’. One
article in particular operated as a catalyst in this
process, and that is Mark Granovetter’s ‘Economic
action and social structure: the problem of
embeddedness’ ( ). Its central argument is
that all economic actions are embedded in personal
networks, and it is this quality that brings them into
the sociologist’s domain. While this message was
important enough in itself, the article’s implicit or
subliminal message that sociology had neglected a
whole area of social life which lent itself to
sociological analysis, namely, the economy, also
explains its great impact. Since sociological skills
had not been applied to economic problems,
sociologists might also be able to solve a number of
important puzzles that the economists had failed to
do, according to Granovetter.
Since the mid-1980s economic sociology has
advanced steadily, and it is now fully institution-
alized in the United States. It is routinely taught in
sociology departments in all the major universities
and also has a strong presence among the major
journals of the profession. The American Socio-
logical Association has a special section for eco-
nomic sociology; a number of readers have been
published as well as a huge handbook (Smelser and
Swedberg ; ).
Economic sociology is becoming increasingly
popular and accepted in Europe as well, though in
Economic Sociology
a somewhat different form than in the United
States, which is only natural given the various
national traditions in sociology. While interesting
contributions can be found in many European
countries, it is especially in France that one can find
highly original contributions that stand up well to
international competition (for England, see for
example Dodd ; for Scotland, Mac- Kenzie ; for
Germany, Beckert ; for Italy, Trigilia ; and for
Sweden, Aspers )•
The three key figures in French economic soci-
ology are Pierre Bourdieu, Luc Boltanski and
Michel Callon (see also the works of Lebaron , and
Steiner ). Bourdieu (1930-2002) has, among other
things, analysed consumption in an innovative
manner in his celebrated study Distinction ( ); he
has also sketched a whole
programme for economic sociology, drawing on his
three key concepts of habitus, field, and different
types of capitals (Bourdieu ; ).
Luc Boltanski has contributed to the discussion of
modem capitalism through an important study of
class formation and also co-authored a provocative
volume on ‘the new spirit of capitalism’
(Boltanski ; Boltanski and Chiapello ). And Michel
Callon ( ) has introduced the
so-called theory of performativity or the idea that
economic theory may be as successful as an
explanatory approach for the simple reason that it
analyses phenomena that it has helped to create in
the first place.
The number of studies in economic sociology
(books and articles) amounts to several thousand by
now, which makes it hard to summarize its
achievements. One way to convey a sense of this
literature, however, would be to discuss the
methods that are being used to gather and analyse
data as well as some of the most important topics.
That economic sociology indeed has a distinctive
profile that sets it off from mainstream economics
emerges very clearly from a discussion of these two
themes.
The data that is being used in economic soci-
ology has often been put together by the analyst,
and it is considerably less common than in main-
stream economics to draw on official data of the
type that is produced by government agencies.
3379
One example is historical studies in economic
sociology, as illustrated by Brace Carrather’s City
of Capital ( ). The focus in this work is
the emergence of one of the world’s first financial
markets, and the author draws heavily on various
primary and secondary sources. In particular,
Carrathers succeeds in showing that early trade in
shares often followed party lines; that is, sellers
were reluctant to trade with political opponents.
Comparative studies are long-standing in eco-
nomic sociology and have also been popular in new
economic sociology. In one of these, Forging
Industrial Policy, Frank Dobbin ( ) compares
the ways in which the railroad industry developed in
the 19th century in the United States, Britain and
France. The author shows that industrial policy has
largely mirrored the general political culture in its
approach to solving problems in each of these three
countries. In the United States, there has been
scepticism towards the state and reliance on the
corporations; in France, the state has been the
central actor; and in Britain there has been an
attempt to protect the individual firm from com-
petition as well as from interventions from the state.
Dobbin claims to have found that there is no one
best way of doing things. Rather, people generalize
from how they themselves do things and proclaim
this to be the universally rational way to proceed.
Economic sociologists also draw on ethnogra-
phy and participant observation, two methods that
allow the researcher to handle huge amounts of
empirical detail and to approach things from the
perspective of the actors. Michael Burawoy ( ), for
example, worked as a shop steward
in order to better understand how workers interact
and deal with the demands of their work (especially
boredom); and Mitchel Abolafia ( ; )
passed an examination as a stock
broker in order to better understand what goes on in
various stock and bond exchanges.
By far the most significant single method used
by economic sociologists today, however, is that of
networks. This is a very flexible tool, which allows
for quantification and therefore goes well with a
large number of research tasks. It has been used, for
example, to analyse the links that exist between
corporations by virtue of having the same
3380
individual on their boards (so-called interlocks).
Through the resultant system of communication,
various ways of doing things may be diffused. The
so-called poison pill (a measure against hostile
takeovers) has, for example, been shown to diffuse
quickly among corporations linked by common
board members (Davis ). That links between
corporations are not to be understood exclusively in
terms of instrumental actions may be exemplified
by the fact that, when a board member resigns or
dies, he or she is only replaced in something like
half of the cases (Palmer ).
Using networks is also a popular way in eco-
nomic sociology to approach collaboration between
corporations as well as the relationship between
firms and their customers and suppliers (see, for
example, Gulati and Gargiulo ). The area where it
has been most successful, however, may well be the
labour market; and here the classic study is Mark
Granovetter’s Getting a Job (Granovetter ). While
one may have thought that the most important
source of assistance for a person seeking a job is
that’s person’s closest friends and family (‘strong
ties’), in fact it is his or her more casual contacts
(‘weak ties’), whose number depends on how many
jobs a person has had. The reason for this ‘strength
of weak ties’ is simply that, whereas one’s ‘strong
ties’ all share the same information, ‘weak ties’ can
provide access to new and varied information,
including information about job opportunities.
In European economic sociology an attempt has
also been made to expand the notion of networks to
include not only people and organizations in the
category of actors but also objects (so-called actor-
network-theory; see, for example, Law and Hassard
). That objects can be actors in the conventional
sense of this term is no doubt wrong; the weaker
claim that objects can be part of networks is,
however, more interesting. One may, for example,
see a machine as a link between people, some
objects may be used for communication between
people, and so on - and all this can affect the
structure of the network. More generally, the
advocates of actor- network-theory also argue that
the traditional approach of economists and
sociologists tends totally to ignore the role that
objects play in the
Economic Sociology
economy and to focus exclusively on actions, social
relations and the like. The perspective that argues
for including objects in the analysis is usually
referred to as ‘materiality’.
When it comes to the topics that are often
analysed, new economic sociologists have first and
foremost tried to focus on economic institutions as
opposed to phenomena situated at the boundary of,
say, religion and the economy or politics and the
economy. The reason for this has been a desire to
take on truly ‘economic’ topics and go beyond the
old division of labour between economics and
sociology, when the former dealt with the economy
and the latter with society minus the economy. As
examples of this is the interest among contemporary
economic sociologists in markets and corporations,
which have attracted a large number of studies.
One type of study has attempted to develop a
general model for markets that differs sharply from
the standard economic model of the perfect market.
The most prominent example of this is the work of
Harrison White ( ; ) on so-called
production markets, by which he roughly means
industrial markets. Production markets, it is argued,
differ from so-called exchange markets primarily
because their participants have permanent roles as
either sellers or buyers and do not switch between
these two roles as is common in financial markets.
According to White, the typical production
market holds about a dozen actors who closely
follow what the other actors are up to. Markets
come into being, White argues, precisely because
economic actors position themselves in relation to
the products of other actors. Prices are not set
through demand and supply but by producers
relating the revenue of their goods to the volume
that is being sold. Individual markets, finally, are
connected to each other in giant networks, either
‘upstreams’ (suppliers) or ‘downstreams’
(customers).
A number of studies of financial markets have
also been carried out, and here the work of Donald
MacKenzie is outstanding (for example, MacKen-
zie ; MacKenzie and Millo ). MacKenzie has picked
up from Callon the theme of performativity, and he
uses it, for example, in his
Economic Sociology
analysis of trade in options. The pricing of options
was very difficult, the argument goes, until Black,
Scholes and Merton suggested a solution for which
the latter two would win the Nobel Prize in 1997.
While this formula covers most cases with much
precision, according to MacKenzie it does not cover
all - and this was to have important consequences.
Since this fact was not well understood, however,
and since economic reality was mistaken for how it
was portrayed in finance theory (performativity),
there have been cases in which people were
unprepared for what was happening (as in the case
of Long-Term Capital Management). MacKenzie
traces this development and also shows how actors
have tried to protect themselves against exceptional
cases by keeping a margin against the price
predicted according to the Black-Scholes-Merton
formula.
Economic sociologists have suggested several
new ways to approach consumer markets. Viviana
Zelizer ( ), for example, has analysed the
growth of the market in life insurance in the United
States and shown how the idea of putti ng a price on
a human life initially attracted hostility, for reli-
gious reasons. But as people moved into the cities
and religion had to adjust to new circumstances, a
different view of life insurance emerged. Zelizer
has recently also started to look at consumption
among children, both how children are socialized
into becoming consumers and the ways in which
they themselves relate to objects and goods in their
environment (Zelizer ).
DiMaggio and Louch ( ) have attempted to
use networks to analyse consumption. While it is
well known that people will turn to others in their
surroundings to find out where to buy something,
and which merchants, traders and so on are reliable
(‘search embeddedness’), DiMaggio and Louch
examine situations in which people approach
someone in their personal network in order to buy
something (‘within-networks exchange’). As it
happens, this is quite common, especially
infrequent purchases of the type that involve legal
services, home repair maintenance and the buying
of a car or a home.
The number of studies in economic sociology
that deal with corporations is very great, but a few
studies nonetheless stand out. One of these is
3381
Mark Granovetter’s pioneering 1994 article on
business groups. Against R.H. Coase, Granovetter
argues that it is not so much the existence of the
individual firm that needs to be explained but the
common phenomenon of groups of firms. In many
countries, such as India, South Korea and Japan,
these business groups control large parts of the
economy, but have not received the scholarly
attention that they deserve. The impact of business
groups in the United States is not clear from
Granovetter’s work, except that US antitrust leg-
islation has mled out some common forms of this
phenomenon.
The business groups that Granovetter studies
lend themselves to a networks approach, and so do
the corporations that Ronald Burt ( ) has
analysed in his study of US industrial markets. Each
firm, according to Burt, can be conceptualized as
situated at the centre of a network in which there are
a number of competitors, suppliers and customers.
The fewer competitors there are, the more suppliers,
and the more customers, the more the corporation is
characterized by ‘structural independence’. And
with more structural independence comes more
profit, as Burt shows.
The emphasis on corporations in interaction, as
opposed to the single corporation, is also obvious in
another landmark study in economic sociology,
Regional Advantage by AnnaLee Saxenian ( ).
Following Alfred Marshall in analysing
industrial districts, Saxenian carries out a compar-
ative study of the computer industry during the
post-war period in Silicon Valley and the area
around Route 128 in Boston. Silicon Valley has
clearly overtaken Route 128 during recent decades,
and the reason for this, according to Saxenian, has
to do with the nature of the interaction in the two
regions. While in Route 128 the corporations are
loath to cooperate, rely on banks for finance, and
prosecute employees who switch to competitors, in
Silicon Valley there is plenty of cooperation,
finance comes from venture capital firms, and
employees are free to switch as they like. A much
more decentralized and flexible form of
entrepreneurship, in brief, has emerged in Silicon
Valley.
Saxenian’s fascination with entrepreneurship is
shared by many economic sociologists. While
3382
she argues that a radical decentralized industrial
region represents the best conditions for entrepre-
neurship, there exist other perspectives as well.
Granovetter, for example, argues that entrepreneurs
often come from those parts of the social system
which are far away from the controlling centre (for
example, Granovetter ). While this may be termed a
theory of peripheral entrepreneurship, Granovetter
suggests several other situations that are favorable
to entrepreneurship. An entrepreneur may, for
example, be someone who crosses a social
boundary in society and thereby becomes the first to
unite resources from two otherwise separated
regions (for example, Granovetter ). On
immigration, Granovetter also points out that some
ethnic groups that are not entrepreneurial in their
country of origin may be highly entrepreneurial in
their new country because they often leave parts of
the extended family behind (Granovetter ). This
means that they do not have to provide jobs for their
relations or share their wealth with relatives.
Economic sociologists have been very active in
studying ethnic entrepreneurship, (for example,
Light ). Ethnic entrepreneurs, for example, often
have to overcome the fact that their initial market
consists of their countrymen (‘the ethnic market’),
and that they will have to go beyond this market if
they are to expand. In many cases they have become
entrepreneurs simply because they have no other
way of making a living (‘forced entrepreneurship ’).
Economic sociologists have also emphasized the
collective nature of entrepreneurship and attempted
to explode the myth of the creative Schumpeterian
individual. One important example of this can be
found in the research by Rosabeth Moss Kanter (
) on entrepreneur-
ship within the corporation, so-called intra-
preneurship. Through a combination of
ethnographic studies and survey research, Kanter
has attempted to show the conditions under which it
is possible to put together creative and entrepre-
neurial groups in modem corporations. Someone
has to suggest the creation of such groups and
provide them with resources and legitimacy. The
group also has to be defended from outside inter-
vention while it operates, internal conflicts have to
Economic Sociology
be solved, and so on. According to Kanter, this type
of group is common among modem corporations.
While economic sociologists have been unable
to present a general theory of entrepreneurship, it is
nonetheless clear that a number of insights have
been accumulated. Economic sociologists are also
expanding their work into such topics as social
entrepreneurship and the diffusion of courses
among business schools (for example, Swedberg )•
Concluding Remarks
Economic sociology is currently in a very active
phase of its development, and all signs indicate that
this trend will continue. Economic sociologists are
also gradually expanding their range of topics of
study. There has recently, for example, been an
attempt to introduce law into the analysis, and some
economic sociologists are trying to formulate a
position on the relationship between the economy
and technology. Some economic sociologists are
also in the process of investigating the role of
emotions in the economy; and there is a growing
number of studies of gender and the economy. What
all of this adds up to, again, is a steady growth of
studies in economic sociology and a confirmation
that economic sociology is established as a distinct
and accepted area of sociology. But it remains to be
seen whether economic sociology will be able to
make inroads into economics itself and gain respect
from economists, along the lines of, say,
behavioural economics.
See Also
► \k . k>: George Art:: ;o (3 cm J 0)
► A: c\:
► Entrepreneurship
Bibliography
Abolafia, M. 1996. Making markets: Opportunism and
restraint on Wall Street. Cambridge, MA: Harvard Uni-
versity Press.
Economic Sociology
Abolafia, M. 1998. Markets as culture: An ethnographic
approach. In The laws of the markets, cd. M. Callon.
Oxford: Blackwell.
Aspers, P. 2001. A market in vogue: A study of fashion
photography in Sweden. Stockholm: City University
Press.
Beckert, J. 2004. Unverdients Vermogen. Soziologie des
Erbrechtes. Frankfurt: Campus Verlag.
Boltanski, L. 1987. The making of a class: Cadres in French
society. Cambridge: Cambridge University Press.
Boltanski, L., andE. Chiapello. 1999. LeNouvelEsprit du
Capitalisme. Paris: Gallimard.
Bourdieu, P. 1979. Algeria 1960. Cambridge: Cambridge
University Press.
Bourdieu, P. 1986. Distinction: A social critique of the
judgment of taste. London: Routledge.
Bourdieu, P. 2005. Principles of an economic anthropology.
In The handbook of economic sociology, 2nd ed, ed. N.
Smelser and R. Swedberg. Princeton: Princeton
University Press.
Burawoy, M. 1979. Manufacturing consent: Changes in the
labor process under monopoly capitalism. Chicago:
University of Chicago Press.
Burt, R. 1983. Corporate profits and cooptation: Networks of
market constraints and directorate ties in the American
economy. New York: Academic.
Burt, R. 1992. Structural holes: The social structure of
competition. Cambridge, MA: Harvard University Press.
Callon, M. (ed.). 1998. The laws of the markets. Oxford:
Blackwell.
Carruthers, B. 1996. City of capital: Politics and markets in
the English financial revolution. Princeton: Princeton
University Press.
Davis, G. 1991. Agents without principles? The spread of the
poison pill throughout the intercorporate network.
Administrative Science Quarterly 36: 583-613.
DiMaggio, P., and H. Louch. 1998. Socially embedded
consumer transactions: For what kind of purchases do
people most often use networks? American Sociological
Review 63: 619-637.
Dobbin, F. 1994. Forging industrial policy: The United
States, Britain, and France in the railway age. New York:
Cambridge University Press.
Dodd, N. 1994. The sociologs’ of money: Economics, reason
and contemporary society. Cambridge: Polity Press.
Durkheim, E. 1893. The division of labor in society, 1984.
New York: Free Press.
Grano vetter, M. 1974. Getting a job: A study of contacts and
careers. Cambridge, MA: Harvard University Press.
Granovetter, M. 1985. Economic action and social structure:
The problem of embeddedness. American Journal of
Sociology 91:481-510.
Granovetter, M. 1994. Business groups. In The Handbook of
economic sociology, ed. N. Smelser and R. Swedberg.
Princeton: Princeton University Press.
3383
Granovetter, M. 1995. The economic sociology of firms and
entrepreneurship. In The economic sociology of
immigration, ed. A. Portes. New York: Russell Sage
Foundation.
Granovetter, M. 2005. The impact of social structure on
economic outcomes. Journal of Economic Perspectives
19(1): 33-50.
Gulati, R,, and M. Gargiulo. 1999. Where do
interorganizational networks come from? American
Journal of Sociology 104: 1439-1493.
Jevons, S. 1879. The theory of political economy, 5th ed. New
York: Augustus M. Kelley, 1965.
Kanter, R. 1983. The change masters: Innovation and
entrepreneurship in America. New York: Simon and
Schuster.
Law, J., and J. Hassard (eds.). 1999. Actor network theoty and
after. Oxford: Blackwell.
Lebaron, F. 2000. La Crovance Economique: Les
Economistes entre Science et Politique. Paris: Seuil.
Light, I. 2005. The ethnic economy. In The handbook of
economic sociology, 2nd ed, ed. N. Smelser and R.
Swedberg. Princeton: Princeton University Press.
MacKenzie, D. 2003. Long-term capital management and the
sociology of arbitrage. Economy and Society 32: 349-380.
MacKenzie, D., and Y. Millo. 2003. Constructing a market,
performing theory: The historical sociology of a financial
derivatives exchange. American Journal of Sociology
109: 107-145.
Marshall, G. 1982. In search of the spirit of capitalism: An
essay on Max Weber s protestant ethic thesis. London:
Hutchinson.
Palmer, D. 1983. Broken ties: Interlocking directorates and
intercorporate coordination. Administrative Science
Quarterly 28: 40-55.
Parsons, T., and N. Smelser. 1956. Economy and society: A
study in the integration of economic and social theory.
New York: The Free Press.
Polanyi, K. 1957. The great transformation. Boston: Beacon
Hill.
Polanyi, K. 1971. The economy as instituted process. In
Trade and market in the early empires, ed. K. Polanyi, C.
Arensberg, and H. Pearson. Chicago: Henry Regnery.
Saxenian, A. 1994. Regional advantage: Culture and com-
petition in silicon valley and route 128. Cambridge, MA:
Harvard University Press.
Schumpeter, J. 1934. The theory of economic development.
Cambridge, MA: Harvard University Press.
Schumpeter, J. 1954. History of economic analysis. London:
Allen & Unwin.
Schumpeter, J. 1991. In The economics and sociology of
capitalism, ed. R. Swedberg. Princeton: Princeton Uni-
versity Press.
Schumpeter, J. 1942. Capitalism, socialism and democracy.
London: Routledge, 1994.
Simmel, G. 1900. The philosophy of money. London:
Routledge, 1978.
3384 Economic Surplus and the Equimarginal Principle

Simtnel, G. 1908. Competition. In Conflict and the web of

JEL Classifications
group-affiliation. New York: The Free Press, 1955.
D5
Smelser, N., and R. Swedberg (eds.). 1994. The handbook of
economic sociology. Princeton: Princeton University
Press. Marginal analysis is actually only a particular case
Smelser, N., and R. Swedberg (eds.). 2005. The handbook of of a more general theory, the theory of surpluses
economic sociology, 2nd ed. Princeton: Princeton and the economy of markets, which, if considered
University Press.
Steiner, P. 2005. L'Ecole Durkheimienne et I’Economie.
first, facilitates the discussion of the equimarginal
Geneva: Droz. principle.
Swedberg, R. 1998. Max Weber and the idea of economic
sociology. Princeton: Princeton University Press.
Swedberg, R. 2000. The social science view of entrepre- The General Theory of Surpluses and the
neurship. In Entrepreneurship: The social science view,
Economy of Markets: Fundamental
ed. R. Swedberg. Oxford: Oxford University Press.
Trigilia, C. 2002. Economic sociology: State, market and Concepts and Theorems
society in modem capitalism. Oxford: Blackwell.
Weber, M. 1895. The national state and economic policy To simplify the exposition, it is assumed that one
(Freiburg Address). Economy and Society 9: 428-449 good (U), enters all preference and production
(1980).
functions, and that its quantity can vary continu-
Weber, M. 1905. The protestant ethic and the spirit of
capitalism. New York: Charles Scribner’s Sons, 1958. ously. Except for the hypothesis of continuity with
Weber, M. 1909. The Agrarian sociology of ancient civili- respect to this good (U), the discussion in this first
zations. London: New Left Books, 1976. part is free of any restrictive hypothesis of conti-
Weber, M. 1914. Economy and society: An outline of
interpretive sociology. Berkeley: University of California
nuity, differentiability or convexity for the goods
Press, 1978. (V),.. ,,(W) considered, and the preference indexes
White, H. 1981. Where do markets come from? American and production functions. (For an exposition of the
Journal of Sociology 87: 517-547. following theory in the case where no one good
White, H. 2002. Markets from networks: Socioeconomic
models of production. Princeton: Princeton University plays a particular role, see Allais , Section II, pp.
Press. 139-41.)
Whyte, W. 1955. Money and motivation. New York: Harper.
Zelizer, V. 1979. Morals and markets: The development of life Structural Conditions
insurance in the United States. New York: Columbia The needs of every unit of consumption, individual
University Press.
Zelizer, V. 2005. Culture and consumption. In The handbook or collective, can be entirely defined by considering
of economic sociology, 2nd ed, ed. N. Smelser and R. a preference index
Swedberg. Princeton: Princeton University Press.

Ii=fi{Ui,Vi,...,Wi)

(1)
Economic Surplus and the Equimarginal
Principle increasing as it passes from a given situation to one
it finds preferable. Every quantity V, is counted
Maurice Allais positively if it refers to a consumption, negatively if
it refers to a service supplied.
The set of feasible techniques for a unit of
production j can be represented by a condition of
Keywords
the form
Allais, M.; Economic surplus; Economy of
fj(Uj,Vj,...,Wj)> 0
markets; Equimarginal principle; General
equilibrium; Marginalism; Marginal equiva- where every quantity Vj is considered as
lence; Marginal rate of substitution; Maximum representing a consumption or an output depending
efficiency; Preference index; Production func- on whether it is positive or negative.
tions; Surplus
Economic Surplus and the Equimarginal Principle 3385

The extreme points corresponding to the boundary released following this shift (Allais , pp. 610-16).
between possible and impossible situations represent The surplus considered here differs essentially from
states of maximum efficiency for the production unit the concepts of consumer surplus as normally
considered. They may be represented by the considered in the literature (for example,
condition Samuelson , pp. 195-202; Blaug 1985, pp. 355-70;
Allais , pp. 297-8, and , nn. 12-13).
f j ( U j , V j , . . . , W j ) = 0. (2) Let us consider an initial state (<f\) character-
ized by consumption values Ub 1/,...,IF, and Uj,
The function /, may be called the production func- Vj,..., Wj (positive or negative) of the different units
tion. It is defined up to any transformation which of consumption and production. We have
leaves its sign unchanged.
From a technical point of view, maximum effi- ^iUi + XjUj = U0;
ciency implies quite specific conditions. If, for L/VT + 'LjVj = L0;...; + JjWj = W0
instance, one considers a production technique A =
(4)
A (X, K..., /) and if n production units are technically where U0, V0,..W0 designate available
preferable to a single one, we should have (Allais , resources. Let (<5<?i) be a feasible modification of
pp. 187-8; , pp. 319-22) (<%\) characterized by finite variations 8Ub
3Vi, . . . , SWh SUj, SVj,..., SWj, a n d l e t
^ M A • > . . .Z ) > A UCj,
(if 2) = ( < ? i ) + 8(ifi)
'LYj,...,'LZj
(3) represent the new state.
According to ( ) we naturally have
In the opposite case we have

JJ J

>'LA(Xj,Yj,...,Z]). (3*)
E<5V/ + 2«5V,- = 0

for every good (U), (V),..., (W). From ( ) we also

An industry is referred to as differentiated if the use
have for every unit of production j
of distinct production units is technically more
advantageous than the concentration of all pro-
fj (Uj + SUj, Vj + 8Vj,..., Wj +
duction operations into a single production unit. It
is called non-differentiated in the opposite case.
S W j ) = 0.
Conditions ( ) and ( ) are two particular illustra
tions of differentiation (Allais , p. 637).
From inequality ( ) it is possible to show that the According to ( ) the preference indexes become
whole production function of a differentiated
industry is asymptotically homogeneous. In this h + 81 i =fi(Ui + SUj, V; +
case (n 1) there is quasihomogeneity (Allais
, pp. 201-6; ). 8Vh ...,Wt + SWj).
The 81 i can be positive, zero, or negative.
Distributable Surplus Corresponding to a Let us now define a third state (<§3) by the
Given Modification of the Economy condition that by the modification -8aui of just the
The distributable surplus au relative to a good (U) quantities Lj + 8 Lj all the preference indexes
and to a realizable modification of the economy return to their initial values.
which leaves all preference indexes unchanged is We then have the conditions
defined as the quantity of that good which can be
(5)
3386 Economic Surplus and the Equimarginal Principle

The state ( S 3 ) can be termed ‘isohedonous’ with the In such a situation the distributable surplus is
state ( S i ) . In passing from ( S 1) to ( S 3 ) the quantity zero or negative for all possible modifications of the
economy compatible with its structural relations,
<5<T„ = £ 80 ul (6) and it is impossible to find any set of prices that
l would permit effective bilateral or multilateral
of the good (U) is released, as all the units of exchanges (accompanied by the implied production
consumption find themselves again in situations operations) which are advantageous to all the agents
which they consider equivalent, since their pref- concerned.
erence indexes return to the same values (Allais , A situation of maximum efficiency can be
pp. 637-8). defined as a situation in which it is impossible to
The surplus 8au has been released during the improve the situation of some people without
passage from (Si) to ( S 3 ) . It may then be considered undermining that of others, i.e. to increase certain
that in the situation ( S \ ) this surplus was both preference indices without decreasing others. The
realizable and distributable. It may further be set of states of maximum efficiency represents the
considered that in passing from ( S \ ) to ( S 3 ) , it has boundary between the possible and the impossible
in effect been distributed. (Fig- )•
The distributable surplus thus defined covers the From those definitions of the situations of max-
whole economy, but this definition can be used for imum efficiency and stable general economic
any group of agents. It is necessary only to consider equilibrium, it follows, with the greatest generality
the functions f and j) and the resources relating to and without any restrictive hypothesis of continuity,
this group in the preceding relations. differentiability or convexity, except for the
Any exchange system, with the corresponding common good (U), that:
Any state of stable general economic equilibrium is
production operations it implies, is deemed one of maximum efficiency (First theorem of equiv-
‘advantageous’ when a distributable surplus is alence). Any state of maximum efficiency is one of
achieved and distributed, so that the preference stable general economic equilibrium (Second theo-
index of any consumption unit concerned increases. rem of equivalence).

If an exchange and production system is Since there can be no stable general economic
advantageous, there must be at least one system of equilibrium if there is any distributable surplus,
prices which allows it, the prices used by each pair every state of stable general economic equilibrium
of agents being specific to them. The distribution of is a state of maximum efficiency. Conversely, if
the realized surplus between agents is determined there is maximum efficiency, there is no realizable
by the system of prices used in the exchanges surplus which could be used to increase at least one
between them. preference index without decreasing the others, and
Conditions of Equilibrium and Maximum consequently, every state of maximum efficiency is
Efficiency a state of stable general economic equilibrium.
In essence all economic operations of what ever Because of the theorems of equivalence, the
type may be considered as reducing to the search terms ‘conditions of stable general economic
for, the achievement of, and the distribution of equilibrium’ and ‘conditions of maximum effi-
surpluses. Thus stable general economic ciency’ are used interchangeably below.
equilibrium exists if, and only if, in the situation The Dynamic Process of the Economy:
under consideration, there is no realizable surplus, Decentralized Search for Surpluses
which means In their essence all economic operations, whatever
they may be, can be thought of as boiling down to
8gu < 0 (7) the piusuit, realization and allocation of

for all feasible modifications of the economy (Allais distributable surpluses. The corresponding
, pp. 606-12).
Economic Surplus and the Equimarginal Principle
Economic Surplus and the Equimarginal Principle, Fig. 1 Process of dynamic evolution. Illustrative diagram
model is the Allais model of the economy of
markets ( ), defined by the fundamental rule
that every agent tries to find one or several other
agents ready to accept at specific prices a bilateral
or multilateral exchange (accompanied by
corresponding production decisions) which will
release a positive surplus that can be shared out, and
which is realized and distributed once discovered.
Thus the evolution of the market’s economy is
characterized by the condition
81 i > 0
for every consumption unit.
Since in the evolution of an economy of markets
surpluses are constantly being realized and
allocated, the preference indexes of the consump-
tion units are never decreasing, at the same time as
some are increasing. This means that for a given
structure, that is to say, for given preferences,
resources, and technical know-how, the working of
an economy of markets tends to bring it nearer and
nearer to a state of stable general economic
equilibrium, hence a state of maximum efficiency
(Fig. ), which is the third fundamental theorem.
3387
Naturally such evolution takes place only if
sufficient information exists about the actual pos-
sibilities of realizing surpluses.
To any given initial situation whatsoever,
assumed not to be a situation of equilibrium, there
corresponds an infinite number of possible
equilibrium situations, each corresponding to a
particular path and each satisfying the general
condition that no index of preference should take on
a lower value than in the initial situation
(Fig- )•
Economic Loss
The loss a* which is associated with a given
situation is defined as the greatest quantify of the
good (U) which can be released in a transformation
of the economy for which all the preference indexes
remain unchanged (Fig. ) (Allais , pp. 638—49).
It is a well determined function
a* u = F [h , h, . . . Jn , U 0 , V 0 ,. . . , W 0 ] (8)
of the preference indexes 7, and of the resources V0
which characterize this situation. The loss a* is an
3388
indicator of inefficiency, and -cr* an indicator of
the efficiency of the economy as a whole.
The loss is minimum and nil in every state of
maximum efficiency, and positive in every feasible
situation which is not a state of maximum
efficiency. It decreases in any modification of the
economy, whereby some preference indexes
increase, others remaining unchanged, or whereby
some surpluses are released with no decline in some
preference indexes.
Paths to States of Economic Equilibrium
and Maximum Efficiency
Since the preference indices 7, are continuous
functions of the quantities 17, of the common good
(17), the boundary between the possible and the
impossible situations in the hyperspace of
preference indexes is constituted by a continuous
surface. On this surface the loss cr* is nil. This
representation allows an immediate demonstration
by simple topological considerations of propositions
whose proof would otherwise be very difficult. (The
paternity of this representation has been unduly
attributed to P. Samuelson , but it was in fact
published for the first time in Allais , and
systematically used by Allais in later years
especially and ; see Allais , n.ll, p. 385; and , n.18,
pp. 176-7.)
For every feasible situation which is not a state
of maximum efficiency, represented by a point such
as M0, there are an infinity of realizable
displacements MqM enabling a situation of maxi-
mum efficiency M to be approached, such that all
the preference indexes have greater values than in
the initial situation M0.
Figure presents an illustration ofthe process of
dynamic evolution by releasing and sharing out of
surpluses during which the loss cr* is constantly
decreasing (Allais , , and , p. 121).
The Changing Structure of the Economy
As psychological patterns vary, as techniques are
improved, or as new resources are discovered (or
existing resources depleted), the set of situations of
maximum efficiency relative to the indexes of
preference constantly undergoes change over time.
Consequently, situations of equilibrium and
Economic Surplus and the Equimarginal Principle
maximum efficiency are never reached, and what is
really important is to determine the rules of the
game which must be applied to come constantly
closer to them as rapidly as possible. At a given
time t, if information is sufficient and if the adjust-
ments are sufficiently rapid, the point representing
the economy will never be very far from the max-
imum efficiency surface of that time t.
General Comment
An economy of markets can be defined as one in
which the agents - consumption, production, and
arbitrage units - coexist and are free to undertake
any exchange transaction or production operation
which can result in rendering some distributable
surplus available. The principle of the market
economy is that any surplus realized is shared
among the operators involved. How the surpluses
achieved are shared out depends on the specific
systems of prices used in the exchanges between the
agents concerned. The prices used are always
specific to the exchange and production operations
considered and there is never a unique system of
prices used in common by all the agents.
Diagrammatic representation like that of Fig.
reveals clearly three basic facts:
1. There is an infinity of situations of maximum
efficiency corresponding to a given initial sit-
uation characterized by some distribution of
property.
2. To each situation of maximum efficiency there
corresponds a final distribution of property.
3. This final distribution depends on the initial
situation and the distribution of surpluses in the
course of the transition.
Thus there is a very strong interdependence
between the point of view of efficiency
corresponding to the discovery and realization of
surpluses and the ethical point of view
corresponding to their sharing.
In any event, since only what is produced can be
shared, the incentive stemming from the partial or
total appropriation of the surpluses by the various
agents appears as a fundamental factor for the
functioning of the economy of markets.
Economic Surplus and the Equimarginal Principle 3389

On the general theory of surpluses and the Convexity and Concavity

economy of markets in the general case, and on the The local properties of diminishing or increasing
fundamental theorems see Allais ( , marginal returns are related to local conditions of
pp. 112-77; 181-211; 604-56; , § 8-65; convexity or concavity. Convexity is defined as
, vol. 2; , , follows:
pp. 27—48; ). Ordinal fields of preference A field of choice is
said to be convex in the whole space (postulate of
general convexity) if, at all points of the field, the
The Equimarginal Principle condition
Continuity and Differentiability
The preceding definitions and theorems are very 7(M0) < I(M1)
general and do not make any hypothesis of conti-
nuity, derivability or convexity, except the hypoth- entails
esis of continuity for the common good (17).
We now assume in addition only that all the 7(M0) < /(M)
quantities and functions considered are continuous
and that all functions have first and second order with
derivatives, the following developments being
totally independent of any hypothesis of general M = XM0 + (1 - 1)M10 < X < 1.
convexity. There is local convexity at M 0 if this condition is
From the sign conventions adopted earlier it satisfied only for
follows that for any i, j and V
|M0Mi | < e
f'iv = d f i/ d V i > 0, /', = df j / dV j > 0.
where e is a given positive number.
The second partial derivatives are written When differentiability is assumed local con-
vexity implies.
f 'L = d ^ / d Vi S W i , /"w = d 2 f j / dV j dW j .
In the following, the symbol fig represents the
d2fiu < 0 for d/j = 0 •
second differential
---- w Fields of production A field of production is
d 2 g = zg [ : , A v 2 + 2 i : v g [: w dv d w
said to be convex over the whole space (postulate of
general convexity) if, for any two possible points
of a function g(U, V,..., If) when all parameters in M0 and Mi, the centre of gravity defined by the
that function are taken as independent, while the relation
symbol d2gu represents what this second differential
becomes after dn has been replaced by its M = XM0 + (1 - X)M1 is
expression derived from
likewise a possible point for.
w
dg = Sg'dy = 0
0 < X < 1.
(Allais , vol. 2, pp. 77-8; , pp. 151-5; Local convexity obtains at M0 if the preceding
, pp. 688-9). condition is satisfied only for

|M0Mi | < e
3390 Economic Surplus and the Equimarginal Principle

where e is a given positive number. Let 8Vki be the quantity of (V) received by the
When differentiability is assumed, local con- consumption or the production unit k from the
vexity implies consumption or production unit 1. By definition, we
have
d2flu< 0 for d/j = 0 •
8Vk=-L8Vk, (11)
In fact there is no production operation that does
not begin by providing increasing marginal returns, k^\
and it is only beyond a certain threshold that
diminishing marginal returns are observed. That is a 8Vlk = -8Va. (12)
general physical law of nature (Allais , pp. 193-5; ,
vol. 2, pp. 68-96; , Assuming that the displacement (8S) is such that
pp. 362—4; , pp. 153-7). Similarly it can be
considered as an introspective datum that psycho-
'L8Vk = 0,...,'LWt = 0. (13)
logical returns begin by increasing but in the end
always decrease beyond certain threshold values.
k k
That is a general psychological law (Allais , vol. 2,
pp. 109-38; , pp. 360-2; ,
pp. 153-5 ). These are two fundamental properties Let
of fields of choice and production. They rule out the
postulate of general convexity which is generally £
v,u = g'lcv/g'ku- (14)
accepted in the contemporary literature.
The ratio ej is the coefficient of marginal
Generation of Distributable Surplus equivalence (or marginal rate of substitution) of
Consider any economic state (S) and a realizable goods (F) and (U) for agent k (Allais , pp. 609-10,
modification (8S) such that all the preference and 617-21).
indexes /,• remain constant (isohedonous modifi- From ( ) and ( ) we have the relation ( )
cation). Let the conditions of constancy of these
indexes and the conditions corresponding to the dUk = - [EkvudVk + • • • + EkwudWk] ( 1 5 )
production functions be written in the same general
form between the first order differential dUk, dVk,..dWk.
If dt4 is positive, agent k receives a quantity dt4
gk(Uk,Vk,...,Wk) = 0 to within the second order. If dt4 is negative, agent
k supplies a quantity dt4 to within the second order.
(9) From the condition ( ), it follows that the
where Uk, Vk,..., Wk represent the consumption of displacement considered releases a global
both consumption and production units. By con- distributable surplus
vention, any quantity V , if positive, represents
k

consumptions, either by a consumption or a pro- 8au = -Y8Uk

k
duction unit. For any production or consumption
unit, any parameter V , if negative, represents representing the excess of the quantities supplied
k

production of a good or a service. over the quantities received of good (U) whose first
Let dC4, dly,..., dWk, be the first order differ- order differential is
entials of the variations 8Uk, 8 Vk,..., 8Wk of con-
sumptions Uk, V/ ;, . .., Wk in the displacement (8S). 8a„ = - E d Uk.
From ( ), we have k

g'kudUk +g'kvdVk H--------F g'hvAwk — 0. (10) From ( )and( )

Economic Surplus and the Equimarginal Principle 3391

agent k therefore creates an additional positive

value
d Uk X d Vk ,
k,l
k< I d Gukl — (vk ~
Vl)dVld.
and from ( ), we have (Allais 1952c, p. 31; , vol. 2, If in this ‘isohedone’ transformation surpluses are
p. 174; , p. 88) released, all positive, they can be distributed in such
W a way as to increase all preference indexes. In such
dcr„ = X X (4-4,)^Vk,- (16) a modification of the economy, the maximum
v
k,l distributable surplus diminishes, and the point
k<l
representing the economic situation considered
moves closer to the surface of maximum efficiency
According to definitions ( ) and ( ) der„ is the
in the hyperspace of preference indexes. Naturally,
first differential of the global distributable surplus
for this condition to obtain, the corresponding
<5<x„ released in the displacement considered. For
exchanges and the changes of the consumptions and
all economic agents the unit of value is defined by
productions they imply in the production system,
condition uk it 1. The marginal values vk,..., wk of
must effectively occur.
goods (V),..., (W) for unit k are defined with respect
to the uk by the relations
Psychological Values and Marginal
Psychological Values
Sku __ Skv __ _ ____ Skw (17)
Uk Vk Wk Naturally, the vk are only marginal values for the
agents. The psychological values v* of the con-
uk = u = 1. (18) sumption Vi of a subject i is defined by the relation

Under the adopted sign convention, all the vk are flip i + v*Vi,0,...,Wi) =f(Ui,Vi,...,Wi)
positive. We have from ( ) and ( )
where v* V, is the sum he would accept to receive to
4= k v
(19) offset the drop in his consumption Vt to zero. The
unit value v* is generally much higher than the
and relation ( ) is written marginal value v, corresponding to relations ( ),
W ( ) and ( ).
dau = ^2 “ vi)dVki (2°) In any event, a consumption is only advanta-
’ ’ k, I geous when its psychological value is higher than
k<l its marginal value, because, if this were not so, it
would be in the subject’s interest to reduce his
where vk and v; are the marginal values of good (V)
consumption
for units k and /. This summation covers all agents,
both consumption and production units. It can thus
be seen that all the differences between the Conditions of Stable General Economic
marginal values in the situation £ can give rise to Equilibrium and Maximum Efficiency of the
the release of potential surpluses which can be Economy
released and distributed. From condition ( ) it follows that the necessary and
The meaning of relation ( ) is immediate. sufficient condition for a situation (£) to be of stable
Thus if vk > v/ the relative value of good (V) is equilibrium and maximum efficiency is that the
higher for agent A: than for agent /. The transfer of distributable surplus 8au defined by ( )and( ) be
a positive quantity dVk/ of good (V) from agent l to negative or zero for every feasible modification
(8£), that is every modification that is compatible
3392 Economic Surplus and the Equimarginal Principle

with the constraint conditions, that is, the structural could indeed do this because of the existence of
relations of the economy ( ) and ( ) above. other operators who are in a situation of tangential
Condition ( ) implies the two conditions (Allais , equilibrium for good (V).
p. 612) The second order condition ( ) holds whether
or not the d/,: are equal to zero. It is only subject to
der„ = 0 (first order condition) (21) the constraint ( ). If we consider only the mod
ifications of the economy involving units k and /,
d2er„ < 0 (second order condition) (22) condition ( ) is written

for any realizable and reversible modification (SS )

2
in which the expressions of dau and d' a u represent d au = + ~7r^~ < 0 for dauk + daui = 0
J ku J lu.
the first and second differential of 8au.
Thus we have according to ( ) and ( ) using shows that when in a situation of maximum effi-
the above notations ciency consumption or production units consume
(or produce) the same goods, one unit at most is in a
&au = i:&Gui = i:&ii/riu = o (23) situation of local concavity, that is, in a situation of
marginal increasing returns (Allais ,
pp. 196-9; , n.125, p. 184; , p. 65).
d2u„ = E^ + E^<0 for d<T„ = 0. Consequently, when maximum efficiency
1
J iu J J ju
(24) obtains, most operators are in a situation of local
convexity and marginal decreasing returns. How-
Actually, and according to relation ( ), the first ever, this condition cannot be interpreted as mean-
order condition ( ) implies that when the quantities ing that all fields of choice and production are
Vk are not nil, all the marginal values vk are equal to convex everywhere, this hypothesis being totally
a same value v and a same system of prices u, v,..w contradicted by observed data.
then exists for all the agents k concerned, such that. When local convexity obtains for a consumption
unit, its index of preference is effectively at a
§ku Skv _ &k\v (25) maximum, subject to the budgetary constraint,
UV W equilibrium prices being taken as given. Similarly,
if local convexity obtains for a production unit, the
These equalities condense the general equimarginal
unit’s income is effectively at a maximum,
principle into a single formulation. They express the
equilibrium prices again being taken as given.
fact that in a situation of equilibrium and maximum
However, these two principles, winch in any case
efficiency, the psychological (or objective) value vk
could be valid only for a situation of maximum
of the last dollar is the same, for any agent
efficiency, cannot be considered as corresponding
(consumption or production unit), whatever use it is
in all cases to optimum behaviour, and they cannot
put to.
be taken to be of general value. As a matter of fact
For the quantities Vk which are nil (terminal
and for instance, if, in a situation of maximum
equilibria), we necessarily have
efficiency, a production unit is in a situation of local
concavity, its income is minimum, the equilibrium
Vk<V
prices being considered as given.
since, if this were not true, the operator’s interest Conditions ( ) and ( ) show the total sym
would be to increase Vk from the value Vk = 0; he metry of the implications of the psychological and
technical structures of the economy.
Economic Surplus and the Equimarginal Principle 3393

Approximate Value of the Economic Loss foresight: Allais ( , pp. 343-84;

Corresponding to the Non-equality of pp. 23-228; , , , vol. 2). Illustra-
Marginal Values in the Neighbourhood of a tive models: ( , pp. 631-771). Capitalistic
Situation of Maximum Efficiency optimum theory: Allais ( , pp. 179-228;
The integration of Eq. ( ) along a path leading to , ). Demographic optimum theory:
a state of maximum efficiency leads to the follow- Allais ( , pp. 749-85). Case of risk: Allais
ing approximate estimate to within third order ( ). Application of marginal analysis to
accuracy of the global loss involved in the initial transport: Allais ( and ). For a general overview on
situation (relation ) the meaning, limits, generalizations, and history of
the equimarginal analysis see Allais ( ).

zu v
(26)
k,l
k<l General Overview

In this relation, the quantities vk- v; represent the
differences of marginal values in the initial state
considered, and the 5V*kl are the quantities of the
good (V) received by operator k from operator l in
the transition from the initial to the final state.
Relation ( ) is of the broadest generality, and
holds whatever the initial state (Allais , pp.31-2,n.8;
, vol. 2,p. 207; ,p. 110).
Its simplicity is really extraordinary in view of
the complexity of the concept it represents, namely
the maximum of the distributable surplus for all the
modifications which the economy can undergo
while leaving the preference indexes unchanged.
In the neighbourhood of a situation of maximum
efficiency, the (vy. - vj) and S Vu are of the first
order quantities, where as the loss er* is only of the
second order. However, since the SV*kl are of the
first order, the variations SI, of the preference
indexes are also of the first order. As a result, and
for instance, in the neighbourhood of a situation of
maximum efficiency, taxes have major first order
effects on the distribution of income but only
second order effects on the efficiency of the
economy.
On the theoretical foundations of the
equimarginal principle, see Allais ( pp. 604-56,,
, pp. 28-32;
, vol. 2, , Irving Fisher (
, and 1986). Illustrative models: Allais ( ,
Annexe I, pp. 4-24; , pp. 57-69). On
its extension see: cases of perfect and imperfect
Theory of Surpluses and Marginal Analysis
As a matter of fact a single relation, the relation (
) (or the equivalent relation ( )) condenses
the whole marginal approach as it has developed for
over a century. Subject only to the hypotheses of
continuity and derivability implied by any marginal
theory, it applies in all cases, and its simplicity is
really extraordinary'.
It also shows that equilibrium and maximum
efficiency can obtain only when all marginal values
are equal, which is the equimarginal principle.
The equimarginal principle was discovered first
by Gossen ( ), and rediscovered, broad
ened and introduced independently into economics
by Jevons ( ), Menger ( ) and Walras
( -7). In the following years numerous new
developments of the principle have been presented
by their immediate successors, especially by
Edgeworth ( ), Irving Fisher ( ) and
Vilfredo Pareto ( -1911). Particularly striking
illustrations of the role of differences in marginal
equivalences are Ricardo’s theory of comparative
costs ( ) and Dupuit’s theory of economic
losses ( -53).
This principle corresponds to the outcome of the
dynamic process of the economy induced by
differences in marginal equivalences. According to
), with whose judgement
I agree fully, ‘No idea has been more fruitful in the
history of economic science.’ Its applications
3394
and generalizations dominate all economic analysis
in real terms.
From the foregoing a double conclusion
emerges: the classical theory of marginal equiva-
lences is irreplaceable to make understandable the
underlying nature of all economic phenomena; the
general theory of surplus, of which classical mar-
ginal theory is only a special case, allows one to
extend the propositions of marginal analysis to the
most general case of discrete variations and
indivisibilities.
As important as the analysis of the conditions of
general equilibrium and maximum efficiency may
be, the analysis of the dynamic processes which
enable surpluses to be generated from a given
situation is much more important. From this point
of view the analyses by Dupuit, Jevons, Edgeworth,
Pareto, and the marginal school and its predecessors
in general, appear much more realistic than the
contributions which rest only upon the
consideration of Walras’s general model of
equilibrium.
In fact, what is really important is not so much
the knowledge of the properties of a state of max-
imum efficiency as the rales of the game which
have been applied to the economy effectively to
move nearer to a state of maximum efficiency.
The decentralized search for surpluses is truly
the dynamic principle from which a thorough and
yet very simple conception of the operation of the
whole economy can be derived. Whereas in the
market economy model the search for efficiency is
essentially focused on the determination of a certain
set of prices, the analysis of the model of the
economy of markets is based on the search for
potential surpluses and their realization. Not only is
the economy of markets model much more realistic
than the market economy model while lending itself
to much simpler proofs, but also these proofs are
not subordinated to any restrictive assumptions
relating to continuity, differentiability of functions,
or convexity. All of economic dynamics is reduced
to a single principle: the search for and realization
of potential surpluses, which leads to the
minimization of loss for the economy as a whole.
Economic Surplus and the Equimarginal Principle
On all these points see especially: Allais
( and ).
The Tendencies of the Contemporary
Literature
From Walras on, the literature became
progressively - and unduly - concentrated on
equilibrium analysis which, however interesting it
could be, is less so than the analysis of the
processes by which the economy tends at any time
towards situations of equilibrium which in fact are
never reached.
Today there is a tendency to neglect the
dynamic marginal approach based on the consid-
eration of differences in marginal equivalences; and
in the name of a so-called rigour it has been
replaced by new theories. A fortiori, the general
theory of surpluses which generalizes marginal
analysis is simply ignored. This development,
which in reality, and despite the too-widely held
belief to the contrary, represents an immense step
backward, basically stems from the unquestioning
acceptance of ‘established truths’ taught by the
dominant ‘establishments’, whose only real basis is
their incessant repetition.
As a matter of fact the guiding principles of the
contemporary theories descending from Walras: the
adoption of the market economy model; the
hypothesis that a common price system applicable
to all operators prevails at each instant; the
assumption of general convexity; and the exaltation
of mathematical formalism of the theory of sets to
the detriment of conformity with actual facts,
constitute an impediment to any genuine progress in
analysis of the economy in real terms.
The essential difference between the market
economy model and the model of the economy of
markets is that, in the latter, the exchanges leading
to equilibrium take place successively at different
prices, and that, at any given moment, the price sets
used by different operators are not necessarily the
same. Whereas in the first model the final situation
is determined totally by the initial situation, which
correspondingly plays a privileged role without any
real justification, in the second the final situation
depends both on the
Economic Surplus and the Equimarginal Principle
initial situation and the path taken from it to the
final situation (Fig. ).
Whereas the market economy model postulates
perfect competition and a large number, if not an
infinity, of operators, the model of the economy of
markets applies just as well to the cases of monop-
oly as to the cases of competition.
Not only is the market economy model unreal-
istic, but it also gives rise to considerable mathe-
matical difficulties when an attempt is made to
demonstrate the above three fundamental theorems.
Whether differential calculus or set theory is used,
the theorems can only be demonstrated under
extremely restrictive conditions, and the difficulties
they imply are, from an economic standpoint,
completely artificial, for they arise solely from the
unrealistic nature of the model used. Paradoxically,
whereas these restrictive assumptions are totally
unrealistic, most of the theoretical difficulties
encountered disappear, as shown above, once they
are discarded.
The market economy approach leads to impos-
ing on any economic model, for it to be considered
satisfactory, conditions which actually apply to a
particular model, which are generally not fulfilled
in reality, and for which, at all events, no rigorous
justification can be found.
By departing from the great tradition of mar-
ginal theory and by adopting an unrealistic model
and unrealistic assumptions, the contemporary
theories, purely mathematical, have doomed
themselves to sterility as regards the understanding
of reality.
On the contemporary theories see especially:
Samuelson ( ), Arrow ( ), Debreu (
and ), Blaug ( and 1985), Arrow and Hahn ( ),
Hutchison ( , pp. 62-97 and
161-70), Woo ( ), and Allais (
, , , and ).
See Also
► Allais, Maurice (Bom
► Efficient Allocation
► General Equilibrium
3395
► Optimality and Efficiency
► : plus
Bibliography
Allais, M. 1943. A la recherche dime discipline economique.
Premiere partie. I'economie pure,
2 vols. Paris: Ateliers Industria. 2nd ed., Traite
d’economie pure, 5 vols, Paris: Imprimerie Nationale,
Paris, 1952. (The second edition is identical to the first,
except for a new Introduction. )
Allais, M. 1945. Economiepure et rendement social. Paris:
Sirey.
Allais, M. 1947. Economie et interet, 2 vols. Paris:
Imprimerie Nationale and Librairie des Publications
Officielles.
Allais, M. 1952a. Introduction to the 2nd edn of Allais
(1043). Paris: Imprimerie Nationale.
Allais, M. 1952b. L’extension des theories de l’equilibre
economique general et du rendement social au cas du
risque. Colloques Internationaux du Centre National de
la Recherche Scientifique 40, Econometric, 81-120. A
summarized version was published under the same title
in Econometrica, (1953), 269-290.
Allais, M. 1962. The influence of the capital output ratio on
real national income. Econometrica 30: 700-728.
Republished in American Economic Association,
Readings in Welfare Economics, vol. 12, with an addi-
tional Note, 1969.
Allais, M. 1963. The role of capital in economic develop-
ment. In Study work on the econometric approach to
development planning, Pontificiae Academiae
Scientiarum Scripta Varia 28, Pontifica Academia
Scientiarum, Amsterdam: North-Holland, 1963 and
Chicago: Rand McNally, 1965.
Allais, M. 1964. La theorie economique et la tarification
optimum de Tusage des infrastructures de transport. La
Jaune et la Rouge (publication of the Societe Amicale
des Anciens Eleves de l’Ecole Polytechnique), special
issue Les Transports, Paris, 1964.
Allais, M. 1967. Les conditions de Tefficacite dans
Teconomie. Fourth International Seminar, Centro Studi e
Ricerche su Problemi Economico-Sociali, Milan. Italian
translation: ‘Le condizioni dell’
efficienza nelT economia’ in Programmazione E
Progresso Economico. Milan: Franco Angeli, 1969.
Original French text in M. Allais, Les Fondements du
Calcul Economique, vol. 1. Paris: Ecole Nationale
Superieure des Mines de Paris, 1967.
Allais, M. 1968a. Les fondements du calcul economique,
3 vols. Paris: Ecole Nationale Superieure des Mines,
Paris, vol. 1, 1967, and vols 2 and 3, 1968.
Allais, M. 1968b. The conditions of efficiency in the
economy. Economia Internazionale 21 (3): 399-420.
3396
Allais, M. 1968c. Pareto, Vilfredo: Contributions to eco-
nomics. In International encyclopedia of the social
sciences, vol. 2. New York: Macmillan and Free Press.
Allais, M. 1968d. Fisher, Irving. Vn International encyclo-
pedia of the social sciences, vol. 5. New York: Mac-
millan and Free Press.
Allais, M. 1968e. L’economique en tant que science. Revue
d'Economie Politique, January-February, 5-30. Trans, as
‘Economics as a science’, Colliers Vilfredo Pareto,
(1968), 5-24.
Allais, M. 1971. Les theories de l’equilibre economique
general et de l’efficacite maximale - impasse recentes et
nouvelles perspectives. Congres des Economistes de
Langue Francpaise, 2-6 June. Revue d'Economie
Politique 3: 331^409. Spanish translation: ‘Las theorias
del equilibrio economico general y de la eficacia maxima
- recientes callejones sin salida y nuevas perspectivas’. El
Trimestre Economico (Mexico), 39(1972), 557-633;
English translation: see Allais (1974a).
Allais, M. 1973a. La theorie gelnerale des surplus et l’apport
fondamental de Vilfredo Pareto. Revue d'Economie
Politique 6: 1044-1097.
Allais, M. 1973b. The general theory of surplus and Pareto’s
fundamental contribution. Convegno Internationale
Vilfredo Pareto. Roma, 25-27 October, Rome:
Accademia Nazionale dei Lincei, 1975 (English trans. of
Allais, 1973a.)
Allais, M. 1974a. Theories of general economic equilibrium
and maximum efficiency. Vienna Institute for Advanced
Studies. In Equilibrium and disequilibrium in economic
theory, ed. G. Schwodiauer. Dordrecht: Reidel, 1977.
(English version of Allais, 1971, with some additions.)
Allais, M. 1974b. Les implications de rendements croissants
et decroissants sur les conditions de I’equilibre
economique general et d’uned cfficacitc maximale. In
Hommage a Francois Perrons. Grenoble: Presses
Universitaires de Grenoble, 1978.
Allais, M. 1981. La theorie generate des surplus. Economies
et Societes, 2 vols. Montrouge institut de sciences
mathematique et economies appliquees.
Allais, M. 1985. The concepts of surplus and loss and the
reformulation of the theories of stable general economic
equilibrium and maximum efficiency. In Foundations
and dynamics of economic knowledge, ed. M. Baranzini
and R. Scazzieri. Oxford: Basil Blackwell.
Allais, M. 1987. The equimarginal principle, meaning, limits,
and generalisations. Centre d'Analyse Economique.
Revista Internationale di science economiste e
commerciale.
Arrow, K.J. 1968. Economic equilibrium. In International
encyclopedia of the social sciences, vol. 4. New York:
Macmillan/Free Press.
Arrow, K.J., and F.H. Hahn. 1971. General competitive
analysis. San Francisco Edinburgh: Holden-Day/
Oliver.
Blaug, M. 1979. Economic theory in retrospect. 4th ed.
London: Heinemann Educational Books, 1985.
Economic Surplus and the Equimarginal Principle
Debreu, G. 1959. Theoiy of value. New York: Wiley.
Debreu, G. 1985. Theoretic models: Mathematical form and
economic content. Frisch Memorial Lecture, Fifth World
Congress of the Econometric Society, MIT, 17-24
August.
Dupuit, J. 1844. De la mesure de l’utilite des travaux publics.
Annales des Pouts et Chaussees, 2nd series, Memories et
Documents No. 116, vol. 8.
Dupuit, J. 1849. De l’influence des peages sur l’utilite des
voies de communication. Annales des Ponls et
Chaussees, 2nd series.
Dupuit, J. 1853. De l’utilite et de sa mesure. Journal des
Economistes 36 (147): 1-28.
Edgeworth, F.Y. 1881. Mathematical psychics: An essay on
the application of mathematics to the moral si lences.
London: Regan Paul. Reprinted, New York: Kelley,
1953.
Fisher, I. 1892. Mathematical investigations in the theory of
value and prices. New Haven: Yale University Press,
1925.
Gossen, H.H. von. 1854. Entwickelung der Gesetze des
menschlichen Verkehrs und der daraus fliessender
Regelnfur menschliches Handeln. 3rd ed., introduction
by Friedrich Hayek, Berlin: Prager. 1927.
Hutchison, T.W. 1977. Knowledge and ignorance in eco-
nomics. Oxford: Blackwell.
Jevons. W.S. 1871. The theoiy of political economy. London:
Macmillan. 5th ed.. trans. as La theorie de economic
politique. Paris: Giard. 1909.
Menger. C. 1871. Grundsatze der Volkswirtschaftslehre.
Vienna: Braumneller.
Pareto, V. 1896 7. Cours d 'economic politique, 2 vols.
Lausanne: Rouge. Reprinted, Geneva: Droz, 1964.
Pareto, V. 1901. Anwendungen der Mathematik auf
Nationalokonomie. Encvklopadie der Mathematichen
Wissenschaften, vol. 1. Leipzig.
Pareto, V. 1906. Manuale d’economia politico. Milan. Trans,
as Manuel d 'economic politique. Paris: Giard et Briere,
1909, and Geneva: Droz, 1966, and as Manual of
political economy. Reprinted, New York: Kelley, 1971.
Pareto, V. 1911. Economic mathematique. Encyclopedic des
Sciences Mathematiques. Paris: Gauthier Villars, 1911,
also in Statistique et economic mathematique. Geneva:
Droz, 1966, published in English as ‘Mathematical
economics’, International economic papers no. 5. New
York, 1955.
Ricardo, D. 1817. On the principles of political economy and
taxation. Vol. I of The works and correspondence of
David Ricardo, ed. P. Sraffa. Cambridge: Cambridge
University Press, 1951 1955.
Samuelson, PA. 1947. Foundations of economic analysis. 2nd
ed. Cambridge, MA: Harvard University Press, 1948.
Samuelson, PA. 1950. Evaluation of real national income.
Oxford Economic Papers NS 2: 1 29.
Walras, L. 1874 7. Elements d’economic politique pure
theorie de la richesse soeiale. 6th ed. Paris: Guillaumin;
reprinted, Paris: Pichon et Durand-Auzias,
Economic Theory and the Hypothesis of Rationality
1952. English trans. of the 6th ed., as Elements of pure
economics, ed. W. Jaffe, London: Allen & Unwin, 1954.
Woo, H.K.H. 1985. What’s wrong with formalization in
economics? -An epistemological critique. Hong Kong:
Hong Kong Institute of Economic Science.
Economic Theory and the
Hypothesis of Rationality
Kenneth J. Arrow
In this paper, I want to disentangle some of the
senses in which the hypothesis of rationality is used
in economic theory. In particular, I want to stress
that rationality is not a property of the individual
alone, although it is usually presented that way.
Rather, it gathers not only its force but also its very
meaning from the social context in which it is
embedded. It is most plausible under very ideal
conditions. When these conditions cease to hold, the
rationality assumptions become strained and
possibly even self-contradictory. They certainly
imply an ability at information processing and
calculation that is far beyond the feasible and that
cannot well be justified as the result of learning and
adaptation.
Let me dismiss a point of view that is perhaps
not always articulated but seems implicit in many
writings. It seems to be asserted that a theory of the
economy must be based on rationality, as a matter
of principle. Otherwise, there can be no theory. This
position has even been maintained by some who
accept that economic behaviour is not completely
rational. John Stuart Mill ( bk. 2, ch. 4) argued that
custom, not competition, governs much of the
economic world. But he adds that the only possible
theory is that based on competition (which, in his
theories, includes certain elements of rationality,
particularly shifting capital and labour to activities
that yield higher returns); ‘Only through the
principle of competition has political economy any
pretension to the character of science’, ([ ] multiplier and, with it, the explanatory power of the
1909, p. 242).
3397
Certainly, there is no general principle that pre-
vents the creation of an economic theory based on
hypotheses other than that of rationality. There are
indeed some conditions that must be laid down for
an acceptable theoretical analysis of the economy.
Most centrally, it must include a theory of market
interactions, corresponding to market clearing in the
neoclassical general equilibrium theory. But as far
as individual behaviour is concerned, any coherent
theory of reactions to the stimuli appropriate in an
economic context (prices in the simplest case) could
in principle lead to a theory of the economy. In the
case of consumer demand, the budget constraint
must be satisfied but many theories can easily be
devised that are quite different from utility
maximization. For example, habit formation can be
made into a theory; for a given price-income
change, choose the bundle that satisfies the budget
constraint and that requires the least change (in
some suitably defined sense) from the previous
consumption bundle. Though there is an
optimization in this theory, it is different from
utility maximization; for example, if prices and
income return to their initial levels after several
alterations, the final bundle purchased will not be
the same as the initial. This theory would strike
many lay observers as plausible, yet it is not rational
as economists have used that term. Without
belabouring the point, I simply observe that this
theory is not only a logically complete explanation
of behaviour but one that is more powerful than
standard theory and at least as capable of being
tested.
Not only is it possible to devise complete
models of the economy on hypotheses other than
rationality, but in fact virtually every practical
theory of macroeconomics is partly so based. The
price- and wage-rigidity elements of Keynesian
theory are hard to fit into a rational framework,
though some valiant efforts have been made. In the
original form, the multiplier was derived from a
consumption function depending only on current
income. Theories more nearly based on rationality
make consumption depend on lifetime or ‘perma-
nent’ income and reduce the magnitude of the
Keynesian model. But if the Keynesian model
3398
is a natural target of criticism by the upholders of
universal rationality, it must be added that mone-
tarism is no better. I know of no serious derivation
of the demand for money from a rational optimi-
zation. The loose arguments that substitute for a
true derivation, Friedman’s economizing on shoe
leather or Tobin’s transaction demand based on
costs of buying and selling bonds, introduce
assumptions incompatible with the costless markets
otherwise assumed. The use of rationality in these
arguments is ritualistic, not essential. Further, the
arguments used would not suggest a very stable
relation but rather one that would change quickly
with any of the considerable changes in the
structure and technology of finance. Yet the
stability of the demand function for money must be
essential to any form of monetarism, not excluding
those rational expectations models in which the
quantity theory plays a major role.
I believe that similar observations can be made
about a great many other areas of applied econom-
ics. Rationality hypotheses are partial and
frequently, if not always, supplemented by
assumptions of a different character.
So far, I have argued simply that rationality is
not in principle essential to a theory of the econ-
omy, and, in fact, theories with direct application
usually use assumptions of a different nature. This
was simply to clear the ground so that we can
discuss the role of rationality in economic theory.
As remarked earlier, rationality in application is not
merely a property of the individual. Its useful and
powerful implications derive from the conjunction
of individual rationality and the other basic
concepts of neoclassical theory - equilibrium,
competition, and completeness of markets. The
importance of all these assumptions was first made
explicit by Frank Knight ( , pp. 76-79). In the
terms of Knight’s
one-time student, Edward Chamberlin ( pp. 6-7), we
need not merely pure but perfect competition before
the rationality hypotheses have their full power.
It is largely this theme on which I will expand.
When these assumptions fail, the very concept of
rationality becomes threatened, because perceptions
of others and, in particular, of their rationality
become part of one’s own rationality. Even if
Economic Theory and the Hypothesis of Rationality
there is a consistent meaning, it will involve com-
putational and informational demands totally at
variance with the traditional economic theorist’s
view of the decentralized economy.
Let me add one parenthetic remark to this
section. Even if we make all the structural
assumptions needed for perfect competition
(whatever is needed by way of knowledge, con-
cavity in production, absence of sufficient size to
create market power, etc.), a question remains. How
can equilibrium be established? The attainment of
equilibrium requires a disequilibrium process. What
does rational behaviour mean in the presence of
disequilibrium? Do individuals speculate on the
equilibrating process? If they do, can the
disequilibrium be regarded as, in some sense, a
higher-order equilibrium process? Since no one has
market power, no one sets prices; yet they are set
and changed. There are no good answers to these
questions, and I do not pursue them. But they do
illustrate the conceptual difficulties of rationality in
a multiperson world.
Rationality as Maximization in the
History of Economic Thought
Economic theory, since it has been systematic, has
been based on some notion of rationality. Among
the classical economists, such as Smith and
Ricardo, rationality had the limited meaning of
preferring more to less; capitalists choose to invest
in the industry yielding the highest rate of return,
landlords rent their property to the highest bidder,
while no one pays for land more than it is worth in
product. Scattered remarks about technological
substitution, particularly in Ricardo, can be
interpreted as taking for granted that, in a compet-
itive environment, firms choose factor proportions,
when they are variable, so as to minimize unit costs.
To be generous about it, their rationality hypothesis
was the maximization of profits by the firm,
although this formulation was not explicitly
achieved in full generality until the 1880s.
There is no hypothesis of rationality on the side
of consumers among the classicists. Not until John
Stuart Mill did any of the English classical
economists even recognize the idea that demand
Economic Theory and the Hypothesis of Rationality
might depend on price. Cournot had the concept a
bit earlier, but neither Mill nor Cournot noticed -
although it is obvious from the budget constraint
alone - that the demand for any commodity must
depend on the price of all commodities. That insight
remained for the great pioneers of the marginalist
revolution, Jevons, Walras, and Menger
(anticipated, to be sure, by the Gregor Mendel of
economics, H.H. Gossen, whose major work,
completely unnoticed at the time of publication
[1854], has now been translated into English [ ]).
Their rationality hypothesis for
the consumer was the maximization of the utility
under a budget constraint. With this formulation,
the definition of demand as a function of all prices
was an immediate implication, and it became pos-
sible to formulate the general equilibrium of the
economy.
The main points in the further development of
the utility theory of the consumer are well known.
(1) Rational behaviour is an ordinal property. (2 )
The assumption that an individual is behaving
rationally has indeed some observable implications,
the Slutsky relations, but without further
assumptions, they are not very strong. (3) In the
aggregate, the hypothesis of rational behaviour has
in general noimplications; that is, for any set of
aggregate excess demand functions, there is a
choice of preference maps and of initial endow-
ments, one for each individual in the economy,
whose maximization implies the given aggregate
excess demand functions (Sonnenschein ; Mantel ;
Debreu ; for a survey, see Shafer and
Sonnenschein , sec. 4).
The implications of the last two remarks are in
contradiction to the very large bodies of empirical
and theoretical research, which draw powerful
implications from utility maximization for,
respectively, the behaviour of individuals, most
especially in the field of labour supply, and the
performance of the macroeconomy based on ‘new
classical’ or ‘rational expectations’ models. In both
domains, this power is obtained by adding strong
supplementary assumptions to the general model of
rationality. Most prevalent of all is the assumption
that all individuals have the same utility function
(or at least that they differ only in broad categories
based on observable magnitudes,
3399
such as family size). But this postulate leads to
curious and, to my mind, serious difficulties in the
interpretation of evidence. Consider the simplest
models of human capital formation. Cross- sectional
evidence shows an increase of wages with
education or experience, and this is interpreted as a
return on investment in the form of foregone
income and other costs. But if all individuals are
alike, why do they not make the same choice? Why
do we observe a dispersion? In the human capital
model (a particular application of the rationality
hypothesis), the only explanation must be that
individuals are not alike, either in ability or in
tastes. But in that case the cross- sectional evidence
is telling us about an inextricable mixture of
individual differences and productivity effects.
Analogously, in macroeconomic models involving
durable assets, especially securities, the assumption
of homogeneous agents implies that there will never
be any trading, though there will be changes in
prices.
This dilemma is intrinsic. If agents are all alike,
there is really no room for trade. The very basis of
economic analysis, from Smith on, is the existence
of differences in agents. But if agents are different
in unspecifiable ways, then remark (3) above shows
that very few, if any, inferences can be made. This
problem, incidentally, already exists in Smith’s
discussion of wage differences. Smith did not
believe in intrinsic differences in ability; a porter
resembled a philosopher more than a greyhound did
a mastiff. Wage differences then depended on the
disutilities of different kinds of labour, including the
differential riskiness of income. This is fair enough
and insightful. But, if taken seriously, it implies that
individuals are indifferent among occupations, with
wages compensating for other differences. While
there is no logical problem, the contradiction to the
most obvious evidence is too blatant even for a
rough approximation.
I have not carried out a scientific survey of the
uses of the rationality hypothesis in particular
applications. But I have read enough to be con-
vinced that its apparent force comes only from the
addition of supplementary hypotheses. Homoge-
neity across individual agents is not the only aux-
iliary assumption, though it is the deepest. Many
3400 Economic Theory and the Hypothesis of Rationality

assumptions of separability are frequently added. Marshall , pp. 842-3). It was really not until the last
Indeed, it has become a working methodology to 30 years that it has been used systematically as an
start with very strong assumptions of additivity and economic explanation, and indeed its use coincided
separability, together with a very short list of with the first experimental evidence against it (see
relevant variables, to add others only as the original Allais ). The expected-utility hypothesis is an
hypotheses are shown to be inadequate, and to stop interesting transition to the theme of the next
when some kind of satisfactory fit is obtained. A section. It is in fact a stronger hypothesis than mere
failure of the model is attributed to a hither to maximization. As such it is more easily tested, and
overlooked benefit or cost. From a statistical it leads to stronger and more interesting
viewpoint, this stopping rule has obvious biases. I conclusions. So much, however, has already been
was taught as a graduate student that data mining written about this area that I will not pursue it
was a major crime; morality has changed here as further here.
elsewhere in society, but I am not persuaded that all
these changes are for the better.
The lesson is that the rationality hypothesis is by Rationality, Knowledge, and Market
itself weak. To make it useful, the researcher is Power
tempted into some strong assumptions. In particular,
It is noteworthy that the everyday usage of the term
the homogeneity assumption seems to me to be
‘rationality’ does not correspond to the economist’s
especially dangerous. It denies the fundamental
definition as transitivity and completeness, that is,
assumption of the economy, that it is built on gains
maximization of something. The common
from trading arising from individual differences.
understanding is instead the complete exploitation
Further, it takes attention away from a very impor-
of information, sound reasoning, and so forth. This
tant aspect of the economy, namely, the effects of
theme has been systematically explored in
the distribution of income and of other individual
economic analysis, theoretical and empirical, only
characteristics on the workings of the economy. To
in the last 35 years or so. An important but
take a major example, virtually all of the literature
neglected predecessor was Holbrook Working’s
on savings behaviour based on aggregate data
random-walk theory of fluctuations in commodity
assumes homogeneity. Yet there have been repeated
futures and securities prices ( ). It was
studies that suggest that saving is not proportional
based on the hypothesis
to income, from which it would follow that
that individuals would make rational inferences
distributional considerations matter. (In general, as
from data and act on them; specifically, predict-
data have improved, it has become increasingly
ability of future asset prices would be uncovered
difficult to find any simple rationally based model
and used as a basis for current demands, which
that will explain savings, wealth, and bequest data.)
would alter current prices until the opportunity for
The history of economic thought shows some
gain was wiped out.
other examples and difficulties with the application
Actually, the classical view had much to say
of the rationality hypothesis. Smith and the later
about the role of knowledge, but in a very specific
classicists make repeated but unelaborated
way. It emphasized how a complete price system
references to risk as a component in wage differ-
would require individuals to know very little about
ences and in the rate of return on capital (e.g., Mill [
the economy other than their own private domain of
] 1909, pp. 385,406,407, 409). The English
production and consumption. The pro- foundest
marginalists were aware of Bernoulli’s expected-
observation of Smith was that the system works
utility theory of behaviour under uncertainty
behind the backs of the participants; the directing
(probably from Todhunter’s History of the Theoiy
‘hand’ is ‘invisible’. Implicitly, the acquisition of
of Probability) but used it only in a qualitative and
knowledge was taken to be costly.
gingerly way (Jevons [ ] 1965, pp. 159-60;
Even in a competitive world, the individual
agent has to know all (or at least a great many)
Economic Theory and the Hypothesis of Rationality
prices and then perform an optimization based on
that knowledge. All knowledge is costly, even the
knowledge of prices. Search theory, following
Stigler ( ), recognized this problem. But
search theory cannot easily be reconciled with
equilibrium or even with individual rationality by
price setters, for identically situated sellers should
set identical prices, in which case there is nothing to
search for.
The knowledge requirements of the decision
may change radically under monopoly or other
forms of imperfect competition. Consider the sim-
plest case, pure monopoly in a one-commodity
partial equilibrium model, as originally studied by
Cournot in Cournot . The firm has to know not only
prices but a demand curve. Whatever definition is
given to complexity of knowledge, a demand curve
is more complex than a price. It involves knowing
about the behaviour of others. Measuring a demand
curve is usually thought of as a job for an
econometrician. We have the curious situation that
scientific analysis imputes scientific behaviour to its
subjects. This need not be a contradiction, but it
does seem to lead to an infinite regress.
From a general equilibrium point of view, the
difficulties are compounded. The demand curve
relevant to the monopolist must be understood
mutatis mutandis, not ceteris paribus. A change in
the monopolist’s price will in general cause a shift
in the purchaser’s demands for other goods and
therefore in the prices of those commodities. These
price changes will in turn affect by more than one
channel the demand for the monopolist’s produce
and possibly also the factor prices that the
monopolist pays. The monopolist, even in the
simple case where there is just one in the entire
economy, has to understand all these repercussions.
In short, the monopolist has to have a full general
equilibrium model of the economy.
The informational and computational demands
become much stronger in the case of oligopoly or
any other system of economic relations where at
least some agents have power against each other.
There is a qualitatively new aspect to the nature of
knowledge, since each agent is assuming the
rationality of other agents. Indeed, to construct a
rationality-based theory of economic behaviour,
3401
even more must be assumed, namely, that the
rationality of all agents must be common knowl-
edge, to use the term introduced by the philosopher
David Lewis ( ). Each agent must not
only know that the other agents (at least those with
significant power) are rational but know that each
other agent knows every other agent is rational,
know that every other agent knows that every other
agent is rational, and so forth (see also Aumann ). It
is in this sense that rationality and the knowledge of
rationality is a social and not only an individual
phenomenon.
Oligopoly is merely the most conspicuous
example. Logically, the same problem arises if there
are two monopolies in different markets. From a
practical viewpoint, the second case might not offer
such difficulties if the links between the markets
were sufficiently loose and the monopolies
sufficiently small on the scale of the economy that
interaction was negligible; but the interaction can
never be zero and may be important. As usually
presented, bargaining to reach the contract curve
would, in the simplest case, require common
knowledge of the bargainer’s preferences and
production functions. It should be obvious how
vastly these knowledge requirements exceed those
required for the price system. The classic
economists were quite right in emphasizing the
importance of limited knowledge. If every agent has
a complete model of the economy, the hand running
the economy is very visible indeed.
Indeed, under these knowledge conditions, the
superiority of the market over centralized planning
disappears. Each individual agent is in effect using
as much information as would be required for a
central planner. This argument shows the severe
limitations in the argument that property rights
suffice for social rationality even in the absence of a
competitive system (Coase ).
One can, as many writers have, discuss
bargaining when individuals have limited knowl-
edge of each other’s utilities (similarly, we can have
oligopoly theory with limited knowledge of the cost
functions of others: see, e.g., Arrow ). Oddly
enough, it is not clear that limited knowledge means
a smaller quantity of information than complete
knowledge, and optimization
3402
under limited knowledge is certainly computa-
tionally more difficult. If individuals have private
information, the others form some kind of conjec-
ture about it. These conjectures must be common
knowledge for there to be a rationality-based
hypothesis. This seems to have as much informa-
tional content and to be as unlikely as knowing the
private information. Further, the optimization
problem for each individual based on conjectures
(in a rational world, these are probability distribu-
tions) on the private information of others is clearly
a more difficult and therefore computationally more
demanding problem than optimization when there is
no private information.
Rational Knowledge and Incomplete
Markets
It may be supposed from the foregoing that infor-
mational demands are much less in a competitive
world. But now I want to exemplify the theme that
perfect, not merely pure, competition is needed for
that conclusion and that perfect competition is a
stronger criterion than Chamberlin perhaps
intended. A complete general equilibrium system,
as in Debreu ( ), requires markets for all con
tingencies in all future periods. Such a system could
not exist. First, the number of prices would be so
great that search would become an insuperable
obstacle; that is, the value of knowing prices of less
consequence, those of events remote in time or of
low probability, would be less than the cost so that
these markets could not come into being. Second,
markets conditional on privately observed events
cannot exist by definition.
In any case, we certainly know that many - in
fact, most - markets do not exist. When a market
does not exist, there is a gap in the information
relevant to an individual’s decision, and it must be
filled by some kind of conjecture, just as in the case
of market power. Indeed, there turn out to be strong
analogies between market power and incomplete
markets, though they seem to be very different
phenomena.
Let me illustrate with the rational expectations
equilibrium. Because of intertemporal relations in
consumption and production, decisions made
Economic Theory and the Hypothesis of Rationality
today have consequences that are anticipated.
Marshall ( , bk 5, chs 3-5) was perhaps the
first economist to take this issue seriously. He
introduced for this purpose the vague and muddled
concepts of the short and long mns, but at least he
recognized the difficulties involved, namely, that
some of the relevant terms of trade are not
observable on the market. (Almost all other
accounts implicitly or explicitly assumed a
stationary state, in which case the relative prices in
the future and between present and future are in
effect current information. Walras ( , lessons
23-25) claimed to treat a progressive state with net
capital accumulation, but he wound up unwittingly
in a contradiction, as John Eatwell has observed in
an unpublished dissertation. Walras’s arguments
can only be rescued by assuming a stationary state.)
Marshall in effect made current decisions, including
investment and savings, depend on expectations of
the future. But the expectations were not completely
arbitrary; in the absence of disturbances, they would
converge to correct values. Hicks ( , chs 9-10)
made
the dependence of current decisions on expectations
more explicit, but he had less to say about their
ultimate agreement with reality.
As has already been remarked, the Ml com-
petitive model of general equilibrium includes
markets for all future goods and, to take care of
uncertainty, for all future contingencies. Not all of
these markets exist. The new theoretical paradigm
of rational expectations holds that each individual
forms expectations of the future on the basis of a
correct model of the economy, in fact, the same
model that the econometrician is using. In a com-
petitive market-clearing world, the individual agent
needs expectations of prices only, not of quantities.
For a convenient compendium of the basic literature
on rational expectations, see Lucas and Sargent (
). Since the world is uncertain,
the expectations take the form of probability dis-
tributions, and each agent’s expectations are con-
ditional on the information available to him or her.
As can be seen, the knowledge situation is much
the same as with market power. Each agent has to
have a model of the entire economy to preserve
rationality. The cost of knowledge, so emphasized
by the defenders of the price system
Economic Theory and the Hypothesis of Rationality
as against centralized planning, has disappeared;
each agent is engaged in very extensive information
gathering and data processing.
Rational expectations theory is a stochastic form
of perfect foresight. Not only the feasibility but
even the logical consistency of this hypothesis was
attacked long ago by Morgenstem ( ).
Similarly, the sociologist Robert K. Merton ( )
argued that forecasts could be self-denyi ng
or self-fillfilling; that is, the existence of the fore-
cast would alter behaviour so as to cause the
forecast to be false (or possibly to make an other-
wise false forecast true). The logical problems were
addressed by Grunberg and Modigliani ( ) educational level, so the ability of the workers could
and by Simon ( , ch. 5 ). They argued
that, in Merton’s terms, there always existed a self-
fiilfilling prophecy. If behaviour varied
continuously with forecasts and the future reali-
zation were a continuous function of behaviour,
there would exist a forecast that would cause itself
to become true. From this argument, it would
appear that the possibility of rational expectations
cannot be denied. But they require not only
extensive first-order knowledge but also common
knowledge, since predictions of the future depend
on other individuals’ predictions of the future. In
addition to the information requirements, it must be
observed that the computation of fixed points is
intrinsically more complex than optimizing.
Consider now the signalling equilibrium orig-
inally studied by Spence ( ). We have large
numbers of employers and workers with free entry.
There is no market power as usually understood.
The ability of each worker is private information,
known to the worker but not to the employer. Each
worker can acquire education, which is publicly
observable. However, the cost of acquiring the
education is an increasing function of ability. It
appears natural to study a competitive equilibrium.
This takes the form of a wage for each educational
level, taken as given by both employers and
workers. The worker, seeing how wages vary with
education, chooses the optimal level of education.
The employer’s optimization leads to an
‘informational equilibrium’ condition, namely, that
employers leam the average productivity of workers
with a given educational level.
3403
What dynamic process would lead the market to
leam these productivities is not clear, when
employers are assumed unable to observe the
productivity of individual workers. There is more
than one qualitative possibility for the nature of the
equilibrium. One possibility, indeed, is that there is
no education, and each worker receives the average
productivity of all workers (I am assuming for
simplicity that competition among employers
produces a zero-profit equilibrium). Another
possibility, however, is a dispersion of workers
across educational levels; it will be seen that in fact
workers of a given ability all choose the same
be deduced from the educational level ex post.
Attractive as this model is for certain circum-
stances, there are difficulties with its implementa-
tion, and at several different levels. (1) It has
already been noted that the condition that, for each
educational level, wages equal average productivity
of workers is informationally severe. (2) Not only is
the equilibrium not unique, but there is a continuum
of possible equilibria. Roughly speaking, all that
matters for the motivation of workers to buy
education are the relative wages at different
educational levels; hence, different relations
between wages and education are equally self-
fulfilling. As will be seen below, this phenomenon
is not peculiar to this model. On the contrary, the
existence of a continuum of equilibria seems to be
characteristic of many models with incomplete
markets. Extensive non-uniqueness in this sense
means that the theory has relatively little power. (3)
The competitive equilibrium is fragile with respect
to individual actions. That is, even though the data
of the problem do not indicate any market power, at
equilibrium it will frequently be possible for any
firm to profit by departing from the equilibrium.
Specifically, given an equilibrium relation
between wages and education, it can pay a firm to
offer a different schedule and thereby make a
positive profit (Riley ). This is not true in a
competitive equilibrium with complete markets,
where it would never pay a firm to offer any price
or system of prices other than the market’s. So far,
this instability of competitive equilibrium is a
3404
property peculiar to signalling models, but it may be
more general.
As remarked above, the existence of a contin-
uum of equilibria is now understood to be a fairly
common property of models of rational market
behaviour with incomplete information. Thus, if
there were only two commodities involved and
therefore only one price ratio, a continuum of
equilibria would take the form of a whole interval
of price ratios. This multiplicity would be non-
trivial, in that each different possible equilibrium
price ratio would correspond to a different real
allocation.
One very interesting case has been discussed
recently. Suppose that we have some uncertainty
about the future. There are no contingent markets
for commodities; they can be purchased on spot
markets after the uncertainty is resolved. However,
there is a set of financial contingent securities, that
is, insurance policies that pay off in money for each
contingency. Purchasing power can therefore be
reallocated across states of the world. If there are as
many independent contingent securities as possible
states of the world, the equilibrium is the same as
the competitive equilibrium with complete markets,
as already noted in Arrow ( ). Suppose there are
fewer securi
ties than states of the world. Then some recent and
partly still unpublished literature (Duffie ; Wemer ;
Geanakoplos and Mas-Colell ) shows that the prices
of the securities are arbitrary (the spot prices for
commodities adjust accordingly). This is not just a
numeraire problem; the corresponding set of
equilibrium real allocation has a dimensionality
equal to the number of states of nature.
A related model with a similar conclusion of a
continuum of equilibria is the concept of‘sunspot’
equilibria (Cass and Shell ). Suppose there is some
uncertainty about an event that has in fact no impact
on any of the data of the economy. Suppose there is
a market for a complete set of commodity contracts
contingent on the possible outcomes of the event,
and later there are spot markets. However, some of
those who will participate in the spot markets
cannot participate in the contingent
Economic Theory and the Hypothesis of Rationality
commodity markets, perhaps because they have not
yet been bom. Then there is a continuum of
equilibria. One is indeed the equilibrium based on
‘fundamentals,’ in which the contingencies are
ignored. But there are other equilibria that do
depend on the contingency that becomes relevant
merely because everyone believes it is relevant. The
sunspot equilibria illustrate that Merton’s insight
was at least partially valid; we can have situations
where social tmth is essentially a matter of
convention, not of underlying realities.
The Economic Role of Informational
Differences
Let me mention briefly still another and counter-
intuitive implication of thoroughgoing rationality
As I noted earlier, identical individuals do not trade.
Models of the securities markets based on
homogeneity of individuals would imply zero trade;
all changes in information are reflected in price
changes that just induce each trader to continue
holding the same portfolio. It is a natural hypothesis
that one cause of trading is difference of
information. If I leam something that affects the
price of a stock and others do not, it seems rea-
sonable to postulate that I will have an opportunity
to buy or sell it for profit.
A little thought reveals that, if the rationality of
all parties is common knowledge, this cannot occur.
A sale of existing securities is simply a complicated
bet, that is, a zero-sum transaction (between
individuals who are identical apart from
information). If both are risk averters, they would
certainly never bet or, more generally, buy or sell
securities to each other if they had the same infor-
mation. If they have different information, each one
will consider that the other has some information
that he or she does not possess. An offer to buy or
sell itself conveys information. The offer itself says
that the offerer is expecting an advantage to himself
or herself and therefore a loss to the other party, at
least as calculated on the offerer’s information. If
this analysis is somewhat refined, it is easy to see
that no transaction will in fact take
Economic Theory and the Hypothesis of Rationality 3405

place, though there will be some transfer of infor- clear. On the one hand, it may be that recognizing
mation as a result of the offer and rejection. The the limits on rationality will reduce the number of
price will adjust to reflect the information of all equilibria. On the other hand, the problem may lie
parties, though not necessarily all the information. in the concept of equilibrium.
Candidly, this outcome seems most unlikely. It Rationality also seems capable of leading to
leaves as explanation for trade in securities and conclusions flatly contrary to observation. I have
commodity futures only the heterogeneity of the cited the implication that there can be no securities
participants in matters other than information. transactions due to differences of information.
However, the respects in which individuals differ Other similar propositions can be advanced,
change relatively slowly, and the large volume of including the well-known proposition that there
rapid turnover can hardly be explained on this basis.cannot be any money lying in the street, because
More generally, the role of speculators and the someone else would have picked it up already.
volume of resources expended on informational The next step in analysis, I would conjecture, is
services seem to require a subjective belief, at least,
a more consistent assumption of computability in
that buying and selling are based on changes in the formulation of economic hypotheses. This is
information. likely to have its own difficulties because, of
course, not everything is computable, and there will
be in this sense an inherently unpredictable element
Some Concluding Remarks in rational behaviour. Some will be glad of such a
conclusion.
The main implication of this extensive examination
Reprinted from Journal of Business, 1986, vol.
of tiie use of the rationality concept in economic
59, no. 4, pt. 2.
analysis is the extremely severe strain on
information-gathering and computing abilities.
Behaviour of this kind is incompatible with the
limits of the human being, even augmented with
See Also
artificial aids (which, so far, seem to have had a
trivial effect on productivity and the efficiency of ► Models and Theory
decision making). Obviously, I am accepting the ► ^references
insight of Herbert Simon ( , chs 14, 15 ), on ► .ational Behaviou;
tiie importance of recognizing that rationality is ► Rationality, Bounded
bounded. I am simply trying to illustrate that many
of the customary defences that economists use to
Bibliography
argue, in effect, that decision problems are
relatively simple break down as soon as market Allais, M. 1979. The so-called Allais paradox and rational
power and the incompleteness of markets are decisions under uncertainty. In Expected utility hypoth-
recognized. esis and the Allais Paradox, ed. M. Allais and O. Hagen.
Boston: Reidel.
But a few more lessons turned up. For one thing, Arrow, K.J. 1953. Le role des valeurs boursieres dans la
the combination of rationality, incomplete markets, repartition la meilleure des risques. In Econometrie.
and equilibrium in many cases leads to very weak Paris: Centre National de la Recherche Scientifique.
conclusions, in the sense that there are whole Arrow, K.J. 1979. The property rights doctrine and demand
revelation under incomplete information. In Economics
continua of equilibria. This, incidentally, is a
and human welfare, ed. M.J. Boskin. New York: Aca-
conclusion that is being found increasingly in the demic Press.
analysis of games with structures extended over Aumann, R.J. 1976. Agreeing to disagree. Annals of Statistics
time; games are just another example of social A: 1236-1239.
Cass, D., and K. Shell. 1983. Do sunspots matter? Journal of
interaction, so the common element is not
Political Economy 91: 193-227.
surprising. The implications of this result are not
3406 Economic Theory of the State

Chamberlin, E. 1950. The theory of monopolistic competition, Walras, L. 1874. Elements of pure economics. Trans.
6th ed. Cambridge, MA: Harvard University Press. W. Jaffe. London: Allen & Unwin, 1954.
Coase, R. 1960. The problem of social cost. Journal of Law Werner, J. 1985. Equilibrium in economies with incomplete
and Economics 3: 1-44. financial markets. Journal of Economic Theory 36: 110-
Cournot, A.A. 1838. Researches into the mathematical 119.
principles of the theory of wealth. Trans. N.T. Bacon. Working, H. 1953. Futures trading and hedging. American
New York: Macmillan. 1927. Economic Review 43: 314-343.
Debreu, G. 1959. Theory of value. New York: Wiley.
Debreu, G. 1974. Excess demand functions. Journal of
Mathematical Economics 1: 15-23.
Duffie, J.D. 1985. Stochastic equilibria with incomplete
financial markets. Research Paper No. 811. Stanford: Economic Theory of the State
Stanford University, Graduate School of Business.
Geanakoplos, J.. and A. Mas-Colell. 1986. Real indetermi- R. Jessop
nacy with financial assets. Paper No. MSRI 717-86.
Berkeley: Mathematical Science Research Institute.
Gossen, II.II. 1983. The laws of human relations. Cambridge,
MA: MIT Press.
Grunberg, E., and F. Modigliani. 1954. The predictability of
social events. Journal of Political Economy 62: 465M78. The basic forms, social functions, institutional
Hicks, II.R. 1946. Value and capital. 2nd ed. Oxford: boundaries and legitimating principles of states vary
Clarendon. across historical epochs and also differ among
Jevons, W.S. 1871. The theory of political economy. 5th edn;
specific regimes in the same epoch. This makes it
reprinted. New York: Kelley, 1965.
Knight, F. 1921. Risk, uncertainty, and profit. Boston: difficult (some would even say impossible) to
I loughton Mifflin. develop a theory which applies to all states -
Lewis, D. 1969. Convention. Cambridge, MA: Harvard whether in general or simply in their economic
University Press.
aspects. This entry limits itself to some economic
Lucas, R., and 'I. Sargent. 1981 .Rational expectations and
econometric practice. 2 vols. Minneapolis: University of aspects of the capitalist state.
Minnesota Press.
Mantel, R. 1974. On (lie characterization of excess demand.
Journal of Economic Theory 6: 345 354.
Marshall, A. 1920. Principles of economics, 8th edn;
reprinted. New York: Macmillan, ll>48. The Capitalist Type of State
Merton, R.K. 1957. The self-fulfilling prophecy. In Social
theory and social structure, ed. R.K. Merton, revised and Capital accumulation has occurred under the most
enlarged edn. Glencoe: Free Press. divergent state forms, but not all state forms are
Mill, J.S. 1848. Principles of political economy, 1909.
equally supportive of capital accumulation. Various
London: Longmans, Green.
Morgenstem, O. 1935. Vollkommene Voraussicht und attempts have been made to construct theoretically
wirtschaftliches Gleichgewicht. Zeitschrifi fur an ideal type of state which is both possible and
Nationalokonomie 6: 337 357. particularly appropriate under capitalism without
Riley, J.G. 1979. Informational equilibrium. Econometica
claiming, however, that this ‘capitalist type of state’
47: 331-360.
Shafer, W., and H. Sonnenschein. 1982. Market demand and exists always and everywhere in capitalist societies.
excess demand functions. In Handbook of mathematical Among other characteristics of this state form, three
economics, 2nd ed, ed. K.J. Arrow and M. Intriligator. institutional features are worth noting here: it has an
Amsterdam: North-1 lolland. effective monopoly of coercive power, its resources
Simon, H. 1957. Models of man. New York: Wiley.
Sonnenschein, H. 1973. Do Walras’s identity and continuity are purchased with money derived from taxation,
characterize the class of community excess demand and its activities are subject to the rule of law. Each
functions? Journal of Economic Theory 6: 345 354. of these features is not only compatible with but
Spence, A.M. 1974. Market signaling. Cambridge, MA: also potentially supportive of the capitalist
Harvard University Press.
Stigler, G.J. 1961. The economics of information. Journal of economic order.
Political Economy 69: 213-225. Firstly, the state is able to monopolize coercion
because capital appropriates the surplus labour of
Economic Theory of the State 3407

workers through the wage-relation rather than market forces. Whether such problems occur
through extra-economic compulsion. This depends not only on the form of the state and its
monopoly is also functional since it prevents par- integration into the circuit of capital but also on the
ticular economic agents from using direct force to changing balance of political forces.
subvert the free play of market forces. Secondly,
state resources can be purchased because capitalism
involves generalized commodity production and Economic Aspects of the Capitalist State
money mediates the exchange of all commodities
Nowhere are economic systems self-reproducing,
(including labour-power). The state should raise
self-regulating and self-sufficient. They always
monetary taxation because it cannot meet its
depend on other institutional systems and the con-
reproduction needs by selling its own output, and
tingent support of non-economic forces. The cap-
cannot expropriate them forcibly only from those
italist state clearly has a key role in securing such
who happen to produce them without undermining
institutional preconditions; and it is also the nodal
the formal equality and property rights which
site for political support. This does not mean,
underpin capitalism. Thirdly, the rule of law can
however, that one can enumerate a set of essential
exist because capitalism presupposes the formal
economic functions which must be performed by
freedom and equality of all economic agents. Only
the capitalist state. Indeed, paraphrasing Max
if it exists can such agents rely on a stable and
Weber’s more general comment on the modem
impartial legal and political environment for their
state, one could say: there are no economic activ-
long-term economic activities.
ities which capitalist states have not at some time
These three institutional traits of the state facil-
undertaken and none which they undertake invari-
itate capital accumulation. But they are neither
ably and exclusively. In particular the capitalist
logically nor historically necessary; nor, where they
state is neither confined to producing ‘public goods’
occur, do they guarantee accumulation. This is not
nor is it the sole producer of such goods. Instead,
simply because economic factors themselves
even if certain broad developmental tendencies can
engender recurrent crises within capitalism. There
be identified, its precise economic activities are
are also distinct political reasons. These are rooted
always conjunctural. They are always influenced,
in the institutional form of the state and in the
furthermore, by political and ideological as well as
struggles which occur around the nature and
economic factors.
purposes of state power. To take only three exam-
ples. The institutional separation between the state
and economy is crystallized above all in the state’s Economic Periodization of the
legitimate coercive monopoly and its incarnation of Capitalist State
national-popular unity vis-a-vis the antagonistic
private interests of civil society. This means that the The structural relations between state and economy
state has the political and ideological capacities to and the forms of state intervention typically vary
disturb as well as to promote capital accumulation. across time as well as nations. This has encouraged
Nor does the tax form have any self-evident limits. attempts at periodization. Although labels vary,
It can produce fiscal crises and/or disproportions four phases are often identified: mercantilism,
between state expenditure and the requirements of liberal capitalism, (simple) monopoly capitalism
capital accumulation. Thirdly, because the rule of and late (or state monopoly) capitalism. Without
law implies formal neutrality towards particular necessarily endorsing these attempts at
economic agents, it is correspondingly inadequate periodization, the basic features of each stage can
as a steering mechanism. But more purposive, ad be presented as follows.
hoc, discretionary interventions can produce Under mercantilism state power is used to establish
bureaucratic overload and also disrupt the labour the dominance of the capital relation
process and capitalist
3408
and market forces. This is the period of primitive
accumulation and capitalist manufacture and is
associated historically with the absolutist state.
Once this dominance is secured, a liberal phase is
said to follow. This involves the nightwatchman
state which is restricted to securing the general
external conditions of production and has no sig-
nificant directly economic role. The third phase is
linked to the dominance of monopoly capital and
the rise of imperialism. In this stage the state serves
to regulate the economic dominance of monopoly
capital, assumes an active role of managing the
economic and political relations between organized
capital and the labour movement, and also employs
extra-economic coercion abroad in inter-imperialist
competition. Next comes the state monopoly
capitalist stage. State management of the domestic
economy through taxation, state credit, public
enterprise and/or the so-called military-industrial
complex now have an increasingly important role;
the welfare state system and collective consumption
become central to the reproduction of labour-power
and to political management; and international and
transnational state organizations have a key role in
managing the world economy. Not all national
economies have experienced all four stages and
much depends on the timing of their capitalist
development and on their place within the
international division of labour.
Such changes in the state’s economic role also
involve reorganizing its overall institutional form.
Growing state intervention is typically associated
with the strengthening of the executive at the
expense of the legislative branch, the rise of func-
tional (as opposed to territorial) representation
closely tied to the administration, the increased
importance of the state economic apparatus and the
growing dominance of economic criteria within
non-economic departments, and the decline of the
substantive rale of law (as opposed to the simple
maintenance of legal forms) in favour of more
discretionary forms of intervention. Thus the
growth of state economic intervention leaves neither
the economy nor the state unchanged. The circuit of
capitals is socialized through the state and the state
is reorganized to reflect economic needs (cf.
Poulantzas ).
Economic Theory of the State
Explanations for the Economic Role
of the State
Various explanations have been offered for the
state’s assumption of economic functions and for
their general developmental tendencies. Broadly
speaking these comprise two main groups: expla-
nations which focus on the essential structure and
laws of motion of capitalism and explanations
which focus on the social relations which obtain
between class forces. Included among the former
are explanations which emphasize the inability or
failure of individual capitals, market forces or the
law of value to secure all the institutional and
economic conditions needed for capital accumu-
lation. These conditions are frequently said to
include: (a) ‘general external conditions’ such as
bourgeois law or a formally rational monetary
system; (b) public goods such as fire services, sea
walls or statistical services which facilitate
production in all branches; and (c) material factors
productively consumed in all or most branches,
such as labour-power or energy supplies. The more
‘class-theoretical’ explanations focus either on the
state’s instrumentalization by particular (capitalist)
class interests and/or on its relatively autonomous
role in managing the balance of class forces both
within the economic sphere and in society more
generally. In turn the relative strength of class
forces is sometimes attributed to changes in the
mode of production and sometimes to broader
social and political factors ranging from
unionization to wars.
The reasons advanced for increasing state
intervention can be used to illustrate such argu-
ments. Some theorists highlight changes in the
forces of production (e.g their increasing sociali-
zation, growing capital intensity or lengthening
turnover time of capital). Others emphasize changes
in the relations of production (e.g. the shift from
absolute to relative surplus value, growth of
monopoly capital, increased importance of banking
or financial capital, the internationalization of
production, or changing forms of economic crisis).
Yet others have stressed an increased importance of
the tendency of the rate of profit to fall. Whatever
reasons are advanced, however, the same
conclusion is drawn. In the
Economic Theory of the State
course of capital accumulation there is a growing
need for state intervention to socialize the forces of
production (e.g. infrastructural provision, man-
power training, technological innovation) and/or the
relations of production (e.g. state credit, economic
management or collective consumption) to
compensate for the failures of market forces and
competition adequately to coordinate and integrate
the circuit of capital. Whilst such explanations often
identify important structural changes in capitalism,
they do not provide a satisfactory explanation for
the political response to such changes. Nor does an
emphasis on the mediating role of class struggle
help much here unless attention is paid to the full
range and forms of political forces.
The General Limits to State Intervention
There has been considerable interest in the limits as
well as the reasons for state intervention. Again we
find both general explanations and arguments
relating to various stages of intervention. The
following factors are frequently cited here:
(a) the exclusion of the state from the heart of the
production process - which means it must react a
posteriori to events it cannot directly control or
engage in ineffective a priori planning;
(b) its tendency to respond to economic problems
and crises in terms of surface appearances (e.g.
inflation, unemployment, trade deficits) which have
no obvious or consistent relationship to the real
course of capital accumulation - which means that
state policies often have limited or perverse effects;
(c) the inherent limitations of law and money as
steering mechanisms for a constitutional tax-state -
since both mechanisms operate at a distance from
real economic agents and processes; (d) the
contradictions involved in the expansion of non-
commodity forms of provision - they may promote
capital accumulation but they also withdraw money
from the circuit of capital, they can promote fiscal
crises, and they suggest that the commodity form is
neither natural nor necessary; and (e) the sui
generis interests of state managers which can
conflict with the supposed needs of capital. Most of
3409
these difficulties are aggravated by the co-existence
of an effective world economy and a multiplicity of
nation-states.
Political and Ideological Complications
The state’s economic role is always affected by its
other tasks. These include its own organizational
reproduction, maintaining domestic political order
and territorial integrity, and defining and
interpreting national unity. Thus economic policies
are typically inserted into more general political
strategies and influenced by political and
ideological struggles. This affects the inputs,
‘withinputs’ and outputs of the state system.
On the input side economic needs must be
translated into political demands through whatever
organizational and institutional channels are
available; and they must be coupled with political
values and legal norms which are often only indi-
rectly relevant to economic considerations. Within
the state system it is the balance of political forces
which determines how these economic demands are
expressed in economic policies. This will vary with
the individual forms of policy production (e.g.
bureaucratic, purposive programming, participation,
delegation to professionals) and with the manner in
which some basic unity is imposed on the state’s
manifold activities. Each mode of policy-
production contains its own limitations; moreover,
problems of internal unity often preclude the
flexible responses needed for economic
management. All this is aggravated because
political forces are generally most immediately
concerned with other political forces and only
indirectly with the economic sphere. Accordingly, it
is the political repercussions of economic events
and crises which matter more than their inherent
economic form or substance. Finally, the outputs of
the state are generally mediated in and through its
own forms of intervention which operate at one or
more removes from the real economy.
Even the increasingly dominant state economic
apparatus must operate in this environment and it is
also prey to muddling through, administrative
inertia, political pressures and ideological
3410 Economic Theory of the State

thinking. State-owned industries and central banks
typically operate in a political environment which
shapes their economic activities and distinguish
them from private industrial or financial enterprises.
In general, state intervention reflects the balance
among all political forces and these extend well
beyond the classes, fractions and strata defined by
the circuit of capital. This helps to explain the
incoherence of economic policies and the
difficulties of rational economic planning.
Indeed the state’s current expanded role involves
two double-binds: the one economic, the other
political. Firstly, when the state intervenes to
alleviate structural economic crises, it must
substitute its own policies for the purgative effects
of market-mediated reorganization. Thus it typically
changes the forms in which economic crises operate
rather than eliminating them and even internalizes
such crises within the state. Here they can take such
forms as fiscal crises, legitimacy crises,
representational crises, crises of internal unity and
crises of governmental effectiveness or overload.
But, since the state’s role has now become vital for
accumulation, it cannot solve economic crises
simply by withdrawing or refusing to intervene. At
best it can reorganize how it intervenes. Moreover,
in so far as economic crises are seen to follow from
such withdrawal, refusal or reorganization, they can
also precipitate new forms of political crisis.
Secondly, in attempting to resolve crises on behalf
of capital, it faces a political dilemma. If its crisis-
management deliberately favours one fraction of
capital at the expense of others, it is liable to
aggravate economic problems for capital as a whole
nation-states are too varied and because economic
issues are always influenced by non-economic
factors. But a theoretically informed account of the
economic aspects of particular capitalist states is
certainly possible. In this context it would be worth
exploring the following issues. What forms are
taken by the institutional separation of the state
from the economic realm and what do these forms
imply for the nature and limits of state intervention?
How can one identify the collective interests of
capital when these are always overdetermined by
contingent political and ideological factors and
when alternative paths and strategies are followed
in different national economies? What difference do
the various forms of political representation and
intervention make to the economic role of the state
in capitalist societies? What scope is there for
international state organizations to regulate or
manage economic crises? In answering such
questions one must recognize that, despite the
above-mentioned limitations to the state’s capacities
to manage capitalism, some states and regimes are
more successful than others. This suggests the need
for much more detailed historical analyses and for
taking seriously the ‘political’ moment of political
economy.
See Also
► sianvr.:.
► Marx, Karl Heinrich (1818-1883)
►i f z z f ca
►

and to weaken its own legitimacy. But even if it

succeeds in winning support for policies in the
collective interests of capital, it cannot thereby
avoid favouring some capitals more than others.
This will modify the balance of forces and could
disturb the initial alliance which sustained such
policies.
Further Research
A general economic theory of the capitalist state is Kraetke, M. 1985. Kritik der Finanzwissensduift. Frankfurt:
impossible because national economies and
Bibliography
Alford, R.R., and R. Friedland. 1986. Powers of theory: The
state, capitalism, and democracy. Cambridge: Cambridge
University Press.
Badie, B., and P. Bimbaum. 1983. The sociology of the state.
Chicago: Chicago University Press, de Brunhoff, S. 1978.
The state, capital, and economic policy. London: Pluto.
Galbraith, J.K. 1967. The new industrial state. London: Andre
Deutsch.
Jessop, B. 1982. The capitalist state. Oxford/New York:
Martin Robertson/New York University Press.
VSA.
Economic War
Luhmarm, N. 1982. Politische Tlieorie im Wohlfahrstaat.
Munich: Olzog.
O’Connor, J. 1973. The fiscal crisis of the state. London/
New York: Macnrillan/St Martin’s.
Offe, C. 1984. Contradictions of the welfare state. London:
Hutchinson.
Offe, C. 1985. Disorganized capitalism. Oxford: Polity Press.
Poggi, G. 1978. The development of the modern state.
London/Stanford: Hutchinson/Stanford University
Press.
Poulantzas, N. 1978. State, power, socialism. London: New
Left Books.
Economic War
P. J. D. Wiles
Economic war constitutes all economic measures
taken, before, during or instead of a military war, to
harm an enemy. Compare protectionism, which is
all the measines taken to ‘defend’ the national
economy. These latter are often precisely the same
measines. The subjective perception of how they do
defend our own long-run economic interests is very
often incorrect, and always controversial: for free
trade lies at the root of Western economics. By
contrast there is little theory about economic war,
and (or so?) most of the measures taken seem by
common admission well fitted to their time and
place.
In view of the paucity of ‘embargological’
writing this entry must be of a frankly introductory
character. First, it is well to establish some key
definitions:
Embargo - a state’s (or alliance’s) prohibition to all
its (their) citizens to sell to, or buy from, a named
party, even when the price is right. An embargo is
not an act of military war, and one on imports is
little different from protectionism, except that its
motive is to harm the foreign seller not benefit his
domestic competitor. Blockade - the prohibition by
a state upon third states to trade with the second
state, its enemy; a blockade must be enforced by
military means
3411
and so is an act of war, possibly even against
third states.
Both embargoes and blockades normally list
specific goods and services. Note that the
embargo of a sufficiently wide alliance is as
good as a blockade, but is still no act of war.
Boycott - an embargo, usually popular or informal,
on purchases alone. Typically the state machine
is not involved, but some social group.
Contraband-goods on such a list that a third state
tries to smuggle through a blockade.
Sanctions - the League of Nations’ word for its
members’ punishment of an aggressor by a
combined official blockade (the wording of
Article 16 is vague in all original and amended
versions, so the word ‘blockade’ is a little
strong).
Transport strangleholds - when one country’s
transport system monopolizes, or nearly so,
access to another. The great case recently is
Mozambique over Southern Rhodesia (the Beira
railway, see below). A near case used to be the
Arab League’s use of the Suez Canal against
Israel.
Black List - when the state imposing an embargo (or
blockade) seeks to enforce it by a secondary
embargo directly on specific firms within a third
(capitalist) country, that are ‘violating’ the orig-
inal embargo as they are of course entitled by
international law and the law of their own state.
The Arab League runs such a blacklist. The
USA enforces its stricter view of CoCom in the
same way. Communist enterprises are of course
‘imblacklistable’ - apart from their states.
Hostile Planner - the external authority who
intervenes in the market (his, ours or the
world’s) in order to do us harm. This concept is
necessary to remind us that interventions are not
always benevolent. However, just as those of the
friendly (and so mainly internal) planner may be
mistaken and so maleficent, those of the hostile
planner may be mistaken and so beneficient.
Bottleneck effect - when an unsubstitutable import is
successfully embargoed, and some activity
must, at least in the short ran, be shut down.
3412
CoCom - the Co-ordinating Committee of the NATO suspicion - a sign of national weakness even if
powers plus Japan, Australia and New Zealand.
Administers an embargo of militarily significant
industrial products. Is consultative only, each
member remaining sovereign.
Dual Use - services like rail freight and goods like special
steel and aero engines have dual military and civilian
use. Thus an embargo on military goods must hit
some civilian ones. Economic war has a very long
history indeed. Its variety is best appreciated by
considering its first in the Mercantilist era. In the 17th
and 18th centuries the state was broadly proto-
Keynesian. It sought to expand the quantity of
money, in order to increase employment, encourage
development and - above all - collect a gold stock
in case of war. Economic war was the normal
condition of Mercantilist international relations,
interrupted only by military war. Although it
certainly had military implications, it was not, as
in the 20th century, a sign of extreme hostility'
or easily distinguished from mere
protectionism.
But how does a bankless state acquire money?
If, as was normal, it had no gold or silver mines and
could not steal any in its colonial conquests it could
only run a balance of payments surplus. This would
not only bring in money, it would also set off the
foreign-trade multiplier - a concept dimly perceived
but not analysed; i.c. the new money would not be
hoarded. So trade was a zero-sum game - the
international division of labour dates only from
Smith - and indeed a war of all against all in search
of gold. Therefore one embargoed imports and
encouraged exports, for one's own good. In peace
time one did this contra mundum, in wartime one
concentrated on one’s enemy, doing oneself good
and him harm all at once. Exporting to the enemy
(except technology) was very patriotic, since it
harmed him. liven military supplies were allowed
(British cloth for the Grande Armee), though not
actual arms. This was all an essentially monetary,
not an input/out- put, view of economic war.
Banks and paper money added to but did not
modify these policies, notably in the Napoleonic
Wars. Paper money was regarded with extreme
Economic War
convertible, since clearly the authorities had already
failed to gather enough gold for all purposes. So wc
add to our goals the destmetion of the enemy’s
convertibility. When Pitt went off gold it was a
Napoleonic victory, due to France’s superior
exports (of wheat). The nature of this victory was
that it was a blow to the morale of an enemy with a
weak balance of payments. Drained externally of
the means of internal payment, Pitt was faced with
severe unemployment in Yorkshire, and a budget
deficit if he wished to do something about it (in the
absence of an existing and functioning welfare
state). He therefore went off gold and printed the
money - a defeat all in itself.
Let us jump to the 19th century, during which
‘embargology’ declined as free trade doctrine
spread, and the notion spread that war is an cpi-
phenomenon on the real, freely trading, peaceful,
liberal, capitalist, democratic world of the planetary
economy. In such an environment, where also in
practice few wars were fought between major
powers, there was no incentive even to consider
economic measures short of war.
The 20th century has not forgotten the 19th, and
it is only shamefacedly that it has reverted to the
practice of the 17th and 18th cenUiries. Not
accidentally protectionism has grown back too, but
the two are not mixed up as under Mercantilism. All
modem states have banks and paper money, and the
monetary peculiarities of late Mercantilism have
dropped away. With the welfare state and
fiscal/monetary policy the modem state can
sufficiently mitigate external crises to retain
domestic political stability. Inconvertibility and
inflation will not alter its warlike stance. Economic
war has therefore - again very rationally - become
an input-output matter, though the state of our
enemy’s gold reserve continues to be a
preoccupation since gold is fungible into any input.
But the main change, surely due to 19th century
example, is that economic war is no longer waged
for economic ends (make him economically weaker
so that I can be economically stronger, trade being a
zero-sum game), but only for ‘political’ ends (make
him economically weaker
Economic War
so that I can be militarily stronger). We may even
infer that civilization has advanced: dirtly tricks are
no longer played by states merely for civilian gain.
Let us examine a few examples of the new, more
purely military economic war.
In the simplest case a specific export is
embargoed to the enemy. If it is not a finished good,
like a weapon, but an input (e.g. special steel) or
both (e.g. refined petrol), our enemy must shut
down some activity because of the bottleneck effect.
This is the main weapon of modem economic war.
If, however, his gold reserve is low and his balance
of payments strained we may also embargo his
exports, quite in general. This will force him to cut
an import of his choice, and so suffer a mild
bottleneck.
The practical complications are illuminating.
Should the USA embargo the sale of wheat to the
USSR, or should France embargo the purchase of
Soviet gas? Provided that France does not become
dependent on this gas (e.g. above five per cent of all
fuel consumption) she clearly has a better economic
case for doing what she prefers. For the USA, wheat
relieves a serious and immediate bottleneck: that of
fodder, leading to the immediate slaughter of Soviet
livestock.
By a simple and well tried ‘iteration’, the live-
stock slaughter first raises, then lowers the supply
of meat, the great crucial consumer good shortage
that has already lead to very serious rioting and
many deaths (Novocherkassk, 1962), not to mention
a huge consumer subsidy. For comparison, in 1801
Britain imported French wheat to avoid a serious
food shortage and despite Mercantilist doctrine. The
mad Tsar Paul suggested a wheat embargo, this
being his period of alliance with France. But that
would have been to embargo an export, so everyone
pointed out that he was only the mad Tsar Paul.
Napoleon, of course, supported by current doctrine,
had no qualms about his export. Anyway had not
low farm prices contributed to the Vendee?
Similarly Reagan fears, or feared, low farm prices,
and brought Carter’s wheat embargo to an end.
Yet again, wheat is a perfectly competitive
commodity, and so much less suitable to be
embargoed (though sometimes easy enough to
blockade). In fact under President Carter the
3413
USSR bought wheat from Argentina instead. But
the price was higher, the docking facilities worse
and the delay considerable. All this imposed
external costs the USSR, while USA, selling else-
where in the world, had very minor external losses.
Her losses were internal, indeed mainly only
transfers, embarrassing the government but not
much impoverishing the people: price support
outlays, storage costs and electoral shifts.
Nevertheless it is part of the conventional
wisdom of modern ‘embargology’ to count as far as
possible in physical terms. The embargo deprived
USSR of scarcely any bushels of wheat, so it is
accounted a failure. The notion of a discriminatory
export tax, of depressing the enemy’s terms of
trade, has achieved no recognition: the intellectual
world of modem economic war is one of input-
output and, seemingly, fixed co-efficients
Mercantilism knew better. Even the export of
money itself (long-term loans) is not taxed, but
simply subjected to administrative control. But it
has eventually been agreed, among the NATO
powers, no longer to subsidize loans to Warsaw
Pact countries; i.e. not to operate export credit
guarantees in favour of even non-embargoed
exports. At least, like machinery, large long-term
loans are not perfectly competitive and so much
easier to control.
If Mercantilism knew little about foreign lend-
ing, it knew as well as we do about technology
transfer. Technology, like gold itself, was an
exception: it must never be exported. For with better
technology ‘we’ beat ‘them’, both in war and in the
exportation of ordinary goods and services. In
modem times technological levels differ much
more, and the subject has become more important.
Although no one country has a monopoly, the
advanced have become very advanced, and it has
become much more difficult to absorb their output;
their active help is needed. There has also grown up
an unduly sharp distinction between civilian and
military technology - as if dual use were
inconceivable. Moreover, military R&D bulks much
larger in the total.
It was the beginning of the end of Mercantilism
when David Hume declared that, In opposition to
his narrow and malignant opinion, I will venture to
assert, that the increase of riches and commerce
3414
in any one nation, instead of hurting, commonly
promote (sic) the riches and commerce of all its
neighbours; and that a state can scarcely carry its
trade and industry very far, where all the sur-
rounding states are buried in ignorance, sloth and
barbarism (Three Essays... II: On the Jealousy of
Trade, Josiah Tucker’s edn, London 1787, first
page).
Economic war contributes much less than
nothing if we only want to prosper. In the circum-
stances of the Cold War, the sole long-term eco-
nomic war that the world now knows, this is clearly
still true, but irrelevant. The great question is
purely, will this new political system - opposed to
'us’ on principle, and both expecting and working
for ‘our’ total defeat - become more friendly just
because it is richer? Or will it spend the extra
resources on yet more arms?
The political aims of an economic war are
seldom clear. Do wc want (i) to incapacitate our
enemy, (ii) to dissuade him, or (iii) much more
ambitiously, to change his policy and aims? And
with which economic instruments should we pro-
ceed in each ease? In the absence of good theory
modem political leaders enter upon economic war
in permanent ignorance and temporary passion;
their Mercantilist predecessors were far better
served.
Case (iii) is bimodal. It includes, as a valid
‘offensive’ tactic, bringing the enemy into our
group, transferring technology to him, lending him
money at a discount and so enriching him: ‘stab
with a sausage’. In a basically economic analysis
we need only say, this is absolutely correct, and the
best policy by far, but only if it is sure to work, and
within reasonable time. If not, ease (iii) means,
bimodally, that veiy severe measures indeed are
appropriate: conversion through fear.
Case (ii) implies short slaps on the wrist, with
valid threats of worse to come. It implies that we
have some ability to change policy, at least in small
matters, and are therefore prepared to ‘fine-tune’
our measures and to agree with each other on
tactics.
Case (i) implies despair- over ultimate friend-
ship, and accepts a 'peace that is no peace’ as a
Economic War
long-term goal: the establishment of military
superiority by permanently slowing up the enemy's
economic growth, without fine-tuning. One cannot
after all fine-tune so diverse and fractious a
coalition as the CoCom.
Modem economic war concerns mainly military
and dual-use goods. This is an unnecessary
restraint: if our enemy can make wheat with diffi-
culty and rifles with ease wc should deprive him of
wheat. Tire logic is irrefragable in Case (iii) strate-
gies, indeed hard to beat in Case (ii). Lipstick,
therefore, is a highly strategic commodity if our
enemy taxes it heavily and his comparative cost
situation makes its production for any reason
expensive for him. The concentration of embargoes
in military goods selves however a good electoral
purpose; ordinary people do not understand the
lipstick argument but do agree that wc should not
deliver weapons (a not wholly correct proposition!).
Do the initiators of economic warfare always
fail in their aims? This is often stated these days, by
those who wish to end the CoCom and widen
embargoes and with it (unilaterally) the Cold War
against USSR. There is, however, no truth in
‘always’; at most one can say, politicians initiate
military war with far more thought, and it is not the
fault of economic war, but of those who wage it,
that its record is so spotty.
We list the main disputed or forgotten incidents
since 1919:
1935-6. League of Nations sanctions against Italy,
on the occasion of her Abyssinian aggression.
Excessive moderation shown: neither oil
imports nor use of the Suez Canal embargoed,
but these were the only two serious bottlenecks.
Reasons: fear of war in Mediterranean, and of
Fascist-Nazi alliance.
1940. Anglo-American partial embargo on oil for
Japan. Japanese general staff estimate military
action will shortly become impossible. Pearl
Habor results. This catastrophe for the initiators
shows, at any rate, the effectiveness of their
threat.
1976. Ian Smith, leader of the illegal white gov-
ernment of Southern Rhodesia, was forced to
Economic War
go to the negotiating table with his black ene-
mies by Samora Machel’s closure of the Beira
railway. In power since 1974, Machel had hes-
itated because of the huge loss of invisible
earnings. The effect of this was to divert all
traffic to the South African network, which is
about five times as far to the sea, and so very
expensive; overloading it was also very unpop-
ular with the South African government (but to
South African pressure was added greater
guerrilla activity). So the success of the
Mozambican embargo redeemed the failure of
the British. The latter was of course grossly mis-
conceived. Even if better administered it could
not have worked before the Portuguese
Revolution.
The beginning of East-West Detente in 1970
merits longer treatment. First Brezhnev offered the
German Treaty, then the Helsinki Declaration and
then, more informally, the emigration of Jews.
These were, in their original form, substantial
concessions, and the quid pro quo was to be
technology transfer, and access to Western capital
markets. In 1972 the deal was in place: the frontiers
of West Berlin were recognized, the European
Security Conference had begun (to end in 1975 with
the Helsinki Declaration on human rights,
communications, etc. ), and the Jews were coming
out.
But in the same year, 1972, Senator Jackson
boasted during elections too much of how he had I
iterally bargained the loans against the emigration.
Sheer pride forced Brezhnev to hold back his emi-
grant Jews and the deal turned sour. This however
does not alter the fact that the original detente was
made possible by the US embargoes on technology
and capital: the very Soviet political concessions
basic to the earlier Detente, which the Western
enemies of CoCom and the renewed Cold War wish
to bring back, were themselves the product of the
relaxation of the still earlier embargoes.
Economic war against South Africa since about
1946 has been, until September 1985, mainly a
matter of private boycott; except that the
Communist powers have embargoed her
3415
(save Mozambique, which is much more dependent
than upon Southern Rhodesia; and the USSR which
has co-operated in the international diamond
duopoly). Ideologically motivated private groups in
the advanced capitalist democracies have refused to
buy this or that export; but since they have never
fully controlled any enterprises this has affected
only consumer goods. States have embargoed the
sale of weapons and police equipment (except Israel
and Brazil). All this is standard stuff - and was very
ineffective.
Much more novel was the ‘extra-territorial’ use
of shareholder power. Much as the US government
forces its firms to boycott Swedish firms that have
been blacklisted for ignoring CoCom, so have
ideological groups of shareholders forced
enterprises with branches in South Africa to raise
black wages above the market level, recognize
black unions and even to evade local laws. This has
been achieved more by bad publicity than by
serious voting blocks at shareholders’ meetings. The
role of the churches, both as shareholders and as
propagandists, has been considerable. Such
interference is known as extraterritoriality: the state
(Sweden or South Africa) on whose territory the
enterprise produces or sells, or the trade union
organizes, loses the degree of control over events
that is normal in a capitalist state owing to foreign
bodies with their own political will. This is not the
case if it has merely to deal with a profit seeking
headquarters abroad. Non-profit seekers are much
more formidable, once in full control.
Disinvestment runs clean contrary to this. Anti-
Apartheid campaigners have divided into
pragmatists wishing to use such little powers as
extra-territoriality confers, and extremists wishing
to keep, above all, their hands clean. Disinvestment
is no weapon at all against a company that does not
want to borrow more, and the refusal to recognize
this simple fact shows us again at what a low
intellectual level economic war is ordinarily
discussed. But disinvestment has a corollary of very
great potency indeed: the refusal to buy new issues.
This refusal mbs off on the bonds and bills of the
South African government. It
3416
was of course the disinvestment controversy, and
the spreading of the consciousness of what Apart-
heid really means, that made conservative West-
ern banking circles refuse to ‘put together a
package’ during the debt crisis of September
1985, turning them into a sort of moralized
IMF. It will be observed that the more monetary,
Mercantilist view of economic war has lost little
validity.
Let us conclude with a mixed bag of applica-
tions of economic theory, for war and trade have
many parallels:
(a) small countries are seldom in a position to
make economic war, but are ideal victims of it:
(b) even large ones are not often well placed.
Countries should form alliances, or coalitions
as one says in oligopoly theory.
(c) even before size comes factor endowment. To
be the monopolist of a raw material is great,
but to possess an irreplaceable transport artery
is still greater. And factor endowment is
always largely historical chance.
(d) trade unions make economic war and throw
up many parallels.
(e) to a most curious extent there is little notion of
compensation for the losses caused to one’s
side by economic war. Once’s image is of rich
corporations losing small sums by not selling,
or delaying the sale of, high technology. So
the issue only arises domestically when small
enterprises (e.g. farmers) are hit. As to inter-
national burden sharing, say with CoCom, the
diplomacy of it would be horrendously com-
plicated and divisive. But could not Britain
have subsidized Mozambique, already in
1975, to close the Beira railway?
See Also
► “ggai
► 7;v an
►
► Trade
► ernativ
Economics in Belgium
Economics in Belgium
Guido Erreygers
Abstract
There is no single Belgian school of economics,
but a number of Belgian economists have made
significant contributions, including Adolphe
Quetelet, Ernest Solvay, Leon Dupriez, Paul
Van Zeeland, Gaston Eyskens, Etienne Sadi
Kirschen, Robert Triffin and Jacques Dreze.
Keywords
Belgium; De Man Plan; De Molinari, Gustave;
Dreze, Jacques; European Economic Associa-
tion; Free trade; Quetelet, Adolphe; Triffin,
Robert
JEL Classifications
BOO; BIO; B20
Although no genuine Belgian school of economics
has ever emerged, except perhaps in the second half
of the 20th century, Belgian economists have made
original and significant contributions to the
discipline and played a major role in the creation of
a European community of economists.
Before the First World War
When Belgium gained independence in 1830,
economics as a scientific discipline virtually did not
exist in the country. By the middle of the 19th
century, however, most universities were offering
courses on economic subjects, economists began
forming associations, and international cooperation
was actively pursued. Throughout the 19th century
French economic thought was undoubtedly the
main source of inspiration for economists working
in Belgium, but British, German and Dutch
economic schools were also influential. Another
characteristic of that period was the
Economics in Belgium
strong separation along ideological lines, with little
interaction between Catholic economists, mainly
associated to the Catholic University of Louvain, on
one side, and liberal and socialist economists,
associated to the Free University of Brussels and
the State Universities of Ghent and Liege, on the
other. The ideological divide is clearly visible in the
rather biased account of the history of economic
thought in 19th-century Belgium published by
Michotte ( ).
Although not in the first place known as an
economist, the polymath Adolphe Quetelet (1796-
1874) needs to be mentioned for his pathbreaking
contributions with regard to the use of statistics in
the social sciences. He introduced the notion of the
‘average man’, which in economics influenced the
work of both the German Historical School and
William Stanley Jevons (Mosselmans ). A fine
example of pioneering statistical research is
provided by the household surveys of Edouard
Ducpetiaux (1804-1868), whose data were used by
Ernst Engel to derive relationships between
consumption and income.
Not surprisingly, many Belgian economists
considered themselves to be part of the liberal
family. The first generation of liberal economists is
probably best represented by Charles De Brouckere
(1796-1860), who combined careers in politics,
academics and business. Together with Adolphe Le
Hardy de Beaulieu (1814-1894), the Italian emigre
Giovanni Arrivabene (1787-1881) and others, he
founded a Belgian association of free-traders
(Erreygers ). The main accomplishment of the
association was the organization of the Congres des
Economistes in September 1847 in Brussels, the
very first international conference of economists
attended predominantly by ardent free-traders, and
also by Karl Marx and Friedrich Engels. The next
generation of liberal economists was headed by
Gustave De Molinari (1819-1912), who advocated
an extreme libertarian form of liberalism, opposing
virtually any form of government intervention.
Some consider him to have laid the foundations of
free-market anarchism, also known as
anarchocapitalism (Hart -2). Although he spent
much of his time in France, where he was for a long
time editor of the Journal des
3417
Economistes, he played a very active role in Bel-
gium. He founded and edited EEconomiste Beige,
animated the Societe Beige d’Economie Politique,
and managed to breathe new life into the free-trade
movement, both nationally and internationally (Van
Dijck ). In many of these initiatives he found a
fellow-traveller in Charles Le Hardy de Beaulieu
(1816-1871), who published several textbooks on
economics. It must be added, however, that few
Belgian economists shared De Molinari’s extreme
view of liberalism. E
In the second half of the 19th century Emile De
Laveleye (1822-1892) was the country’s most
prominent economist. This prolific writer covering a
wide area of topics had been strongly influenced by
the French philosopher and Christian socialist
Framjois Huet, who taught at the University of
Ghent. De Laveleye’s economic publications
included work on the origins and varieties of
property rights, on bimetallism and on socialist
doctrines. He was professor of political economy at
the University of Liege, and authored an often
reprinted textbook on economics. He considered his
views to be close to those of John Stuart Mill and
the German Historical School. Although very much
appreciated by his contemporaries - he built up an
impressive international network of colleagues and
correspondents - his contributions lost most of their
influence soon after he died.
At that time the wealthy industrialist Ernest
Solvay (1838-1922), founder of the chemical firm
Solvay & Cie, started to turn his attention to social
and economic issues. He was convinced that a
change of the monetary system (replacing the
system based on metallic money by a pure- credit
system which he called ‘social comptabilism’)
combined with a sweeping reform of taxation
(replacing all existing taxes by taxes on gifts and
bequests, with rates increasing with the number of
transfers) would provide the clue to solving
society’s problems (Erreygers ; Boianovsky and
Erreygers ). Solvay, a prominent liberal, worked on
these issues in close collaboration with leading
socialist economists such as Hector Denis (1845-
1913) and Emile Vandervelde (1866-1938). Their
monetary
3418 Economics in Belgium

propositions led to a debate with Leon Walras, who Princeton and Columbia. As a result Belgian eco-
saw a great similarity with his own views. On a nomics gradually obtained a more American char-
more practical level, Solvay influenced economics acter and the University of Louvain became much
in Belgium through his generous funding of various more prominent in economic research (Maes and
institutions associated to the University of Brussels,
Buy st ).
the most important of which are the Institut de Dupriez introduced to Belgium the statistical
Sociologie and the Ecole de Commerce Solvay business cycle techniques used by Harvard Uni-
(now Solvay Business School). These institutions versity. He became the driving force of the Institut
allowed economists such as fimile Waxweiler des Sciences Economiques (later renamed Institut
(1867-1916), Maurice Ansiaux (1869-1943 ) and de Recherches Economiques et Sociales, IRES),
Boris Chlepner (1890-1964) to do research. founded in 1928 at the University of Louvain. Van
Socialist doctrines found responsive audiences Zeeland, who had joined the National Bank of
in Belgium, partly as the result of the rapid indus- Belgium as head of its research department, made a
trialization, but also because of the presence of swift career in the bank and was soon considered as
exiles such as Karl Marx and Joseph Proudhon. The one of the country’s leading economic experts. In
Saint-Simonians and Fourierists attracted scores of the troubled political and economic climate of the
young intellectuals. This created a fertile ground for
1930s he was twice prime minister of governments
such figures as Hippolyte Colins de Ham (1783- of ‘national unity’. With the scientific backing of
1859), who proposed to provide all adults with a IRES, Van Zeeland successfully devalued the
capital endowment, Joseph Charlier (1816-1896), Belgian currency in 1936. The Van Zeeland
who launched the idea of an unconditional basic governments also included the socialist Hendrik De
income, and Cesar De Paepe (1842-1890), who Man (1885-1951), who in 1933 had proposed an
tried to bridge the gap between the Marxists and the ambitious ‘Labour Plan’ (also known as the ‘De
anarchists (Cunliffe and Erreygers ). Man Plan’) as a way out of the economic
At the University of Louvain economics had a depression. His project, which was intensively
decidedly Catholic profile in the 19th century. discussed and had broad support in Belgium and in
Charles Perin (1815-1905) and Victor Brants (1856- other countries, involved state planning of the
1917) both aimed at developing a social economics economy and a technocratic way of governing.
in accordance with the doctrine of the Church, A few isolated attempts were made to introduce
along the lines of Le Play. a more mathematical approach in economics. The
most ambitious project was that of Bernard Chait
(1893-1957), an engineer and businessman who was
The Interwar Period in close contact with Jan Tinbergen and Fran go is
Divisia. In his 1938 Ph.D. thesis he constructed a
After the end of the First World War fellowships general mathematical theory capable of explaining
offered by the Commission for Relief in Belgium business cycle movements, but he failed to convince
and the Belgian American Educational Foundation economists of its usefulness (Erreygers and Jolink ).
gave many talented economics students the
opportunity to spend at least one year in the United
States. This was the case with Leon Dupriez (1901- After the Second World War
1986), Paul Van Zeeland (1893-1973) and Gaston
Eyskens (1905-1988), who combined their studies In the first half of the 20th century economics
at the University of Louvain with stays in gained increasing recognition as a mature scientific
respectively Harvard, discipline. Universities created special
Economics in Belgium
schools and separate faculties for economic sci-
ences, and several economics journals were
launched. In the northern part of the country Dutch
gradually replaced French as the language of
instruction. Both the Universities of Bmssels and
Louvain were eventually split into Dutchspeaking
and French-speaking universities.
As early as the 1930s Gaston Eyskens had
become the leading figure of the Dutch-speaking
section of economists at the University of Louvain.
In the years after the war he seemed to be more
receptive to the ideas of Keynes than the leading
figure on the French-speaking side, Leon Dupriez,
who favoined a laissez-faire approach and resisted
to the introduction of Keynesian macroeconomic
policies. The tension between the Dutch-speaking
and French-speaking section at the university led to
the foundation, in 1955, of a separate Dutch-
speaking research institute at the University of
Louvain, the Center for Economic Studies. The
Center provided scientific backing for various eco-
nomic reforms adopted under Eyskens’s period as
prime minister of Belgium (1958-61, 1968-72).
Eyskens also took the initiative to create a govern-
ment planning bureau (Buyst et al. ).
At the University of Bmssels Etienne Sadi
Kirschen (1913-2000) was the main agent of change
in the 1950s. After having worked at the Office of
European Economic Cooperation in Paris, he
organized a team which estimated the first national
accounts for Belgium. In 1957 Kirschen and his
collaborators founded the Departement d’Economie
Appliquee, better known under its acronym
DULBEA, the economic research institute of the
University of Bmssels. It constructed the first input-
output table for Belgium. In close cooperation with
Tinbergen, who lectured at the University of
Bmssels in the 1960s, Kirschen emphasized policy-
oriented research based on quantitative methods, as
exemplified in the ambitious threevolume
Economic Policy in our lime (1964) by an
international team of economists led by Kirschen
(Shjacobs ).
A few Belgian economists emigrated and made a
career abroad. The most striking case is that of
Robert Triffin (1911-1993), who initially studied
3419
at the University of Louvain. Having obtained his
Ph.D. at Harvard, he worked for the Fed and the
International Monetary Fund (IMF) - under its first
director, the former Belgian Finance Minister
Camille Gutt (1884-1971) - and in 1951 became
professor of economics at Yale University. He made
himself a reputation by pointing out a crucial
weakness of the Bretton Woods system (later
known as the Triffin dilemma) and arguing for a
fundamental reform of the international monetary
system. He was an influential economic adviser to
key players in the European economic and mone-
tary integration process; he returned to Belgium in
the 1970s (Maes and Buyst ). A less well- known
emigre is Raymond De Roover (1904-1972), who
after his studies at the Antwerp Catholic business
school decided to specialize in economic history.
He earned his Ph.D. at the University of Chicago,
and was later appointed as professor in Boston and
New York. His work on early banking in Bmges
and Florence made him a leading specialist on late
medieval economic history and thought.
Probably the most important development in
Belgian economics after the Second World War was
initiated by a man who decided not to stay in the
United States after completing his Ph.D. thesis. In
1966 the Center for Operations Research and
Econometrics (CORE) was founded in Louvain.
This was very much the achievement of Jacques
Dreze (b. 1929), a student of the University of
Liege who thanks to a fellowship of the Commis-
sion for Relief in Belgium went to the United States
and obtained a Ph.D. at Columbia. After his return
to Belgium he was appointed at the University of
Louvain, where he rapidly replaced Dupriez as the
dominant figure of the economics faculty, but he
kept close contacts with his American colleagues,
especially at Northwestern and Chicago, where he
was visiting professor in the 1960s. The creation of
CORE marked the adoption of a decidedly
American style of doing research. From the outset
CORE was meant as an interdisciplinary research
institute, bringing together specialists in
econometrics, statistics, operations research, game
theory and
3420
(mathematical) economics. Thanks to grants from
the Ford Foundation and other institutions, it cre-
ated a stimulating environment for research and
offered fellowships to both Ph.D. students and
established researchers. CORE was unique not only
because it brought together Dutch-speaking and
French-speaking economists from the University of
Louvain, but also because Dreze managed to get the
econometricians of the University of Brussels
involved in the project. Moreover, from the very
beginning CORE opened itself to the world: it hired
the Dutch econometrician Anton Barten (b. 1930),
established strong links with other European
institutions focusing on quantitative economics, and
welcomed American and other foreign scholars as
fellows (Maes and Buyst ).
The University of Brussels and CORE played a
major role in the creation of the European Eco-
nomic Review and the European Economic Asso-
ciation. The European Economic Review was
founded in 1969 by the European Scientific Asso-
ciation for Medium and Long Term Economic
Forecasting (ASEPELT), of which Kirschen was
the driving force. His younger colleagues Jean
Waelbroeck (b. 1927) and Herbert Glejser (b.
1938), both of the University of Brussels, were the
founding editors. In 1985 Waelbroeck and three
other Belgian economists, Jean Jaskold Gabszewicz
(b. 1936), Louis Phlips (b. 1933) and Jacqucs-
Frangois Thisse (b. 1946), all affiliated to CORE,
took the initiative to launch the European Economic
Association. Dreze was elected as the first
president.
Dreze’s contributions to economics are exten-
sive, covering uncertainty, general equilibrium
theory, macroeconomics, econometrics and much
more (Dehez and Licandro ). He is by far the most
influential Belgian economist of the second half of
the 20th century. Besides those already mentioned,
other important Belgian economists include Claude
d’Aspremont (b. 1946), Paul De Grauwe (b. 1946),
Mathias Dewatripont (b. 1959), Pierre Pestieau(b.
1943), Jean-Philippe Platteau (b. 1947) and Gerard
Roland (b. 1954). It is remarkable that the top of the
economics profession remains very much
dominated by French- speaking economists.
Economics in Belgium
See Also
► Catholic
► Tinbcrgi
►'
Bibliography
Boianovsky, M., and G. Erreygers. 2005. Social comptabilism
and pure credit systems: Solvay and Wicksell on
monetary reform. In The experiment in the history of
economics, ed. P. Fontaine and R. Leonard, 98-134.
London: Routledge.
Buyst, E., I. Maes, H.W. Plasmeijer, and E. Schoorl. 2005.
Comparing the development of economics during the
twentieth century in Belgium and the Netherlands.
History of Political Economy 37: 61-78.
Cunliffe, J., and G. Erreygers. 2001. The enigmatic legacy of
Fourier: Joseph Charlier and basic income. History of
Political Economy 33: 459-484.
Dehez, P., and O. Licandro. 2005. From uncertainty to
macroeconomics and back: An interview with Jacques
Dreze. Macroeconomic Dynamics 9: 429-461.
Erreygers, G. 1998. The economic theories and social reform
proposals of Ernest Solvay (1838-1922). In European
economists of the early 20th century. Volume One:
Studies of neglected thinkers of Belgium, France, The
Netherlands and Scandinavia, ed. W. Samuels, 220-262.
Cheltenham: Edward Elgar.
Erreygers, G. 2001. Economic associations in Belgium in the
19th century. In The spread of political economy and the
professionalisation of economists: Economic societies in
Europe, America and Japan in the nineteenth century, ed.
M. Augello and M. Guidi, 91-108. London: Routledge.
Erreygers, G., and A. Jolink. 2007. Perturbation, networks
and business cycles: Bernard Chait’s pioneering work in
econometrics. European Journal of the Histoty of
Economic Thought 14: 543-571.
Hart, D.M. 1981-2. Gustave de Molinari and the antistatist
liberal tradition. Journal of Libertarian Studies 5: 263-
290, 399-434; 6: 83-104.
Maes, I., and E. Buyst. 2005. Migration and Americanization:
The special case of Belgian economics. European
Journal of the Histoty of Economic Thought 12: 73-88.
Michotte, P. 1904. Etudes sur les theories economiques qui
dominerent en Belgique de 1830 a 1886. Louvain:
Charles Peeters.
Mosselmans, B. 2005. Adolphe Quetelet, the average man and
the development of economic methodology. European
Journal of the Histoty of Economic Thought 12: 565-582.
Siijacobs, I. 1997. Leconomiste dans le temps. Bruxelles:
Archives de l’ULB.
Van Dijck, M. 2008. From science to popularization, and
back: The science and journalism of the Belgian
economist Gustave de Molinari. Science in Context 21:
377-402.
Economics Libraries and Documentation
Economics Libraries
and Documentation
P. Stages
Libraries for a discipline are formed and charac-
terized by that discipline. It is quite easy for a visitor
to recognize that a library is for scientists or for
humanists or for social scientists just by a glance at
the types of books, periodicals and other material on
the shelves. Libraries for economists reflect the
distinctive and changing soiuces and documentation
of economics, which are in turn products of changes
in the discipline itself. As economists have
successively widened the scope of their enquiries
and added weapons to their methodological
armoury, so the types of material they have needed
to consult have multiplied. As the immediate
communication of their results has become more
and more pressing, so the types of publication they
have favoured have evolved in response. The library
providing effective service to an econometrician
today would have been as irrelevant to a 17th-
century mercantilist as the literature of econometrics
would have been incomprehensible. This article will
deal with economics libraries in their natural context
of economics documentation.
The first economics collections were in the
private libraries of 17th- and 18th-century scholars.
Adam Smith, renowned for his forgetfulness and
carelessness in dress, was able to defend himself
with the claim that at least ‘I am a beau in my
books’. In fact the majority of his 3000 titles were in
the miscellaneous topics a gentleman scholar might
have been expected to cultivate, with only about 100
directly on economic topics. Nevertheless, his
economic method was a book-based one, using
material from his predecessors and building it into
his own system. About 100 authors are quoted in the
Wealth of Nations (1776), though not always by
name. By Smith’s day there was already a large
monograph literature the economist could draw on
in several languages: pamphlets advocating
trading
3421
schemes, tracts on farm management, alarming
arguments on the growth or decline of population,
accounts of particular industries, countries, cities,
etc. The first economists wrote in monograph form
too, either the pamphlet directed at some particular
case or instance, or, from the 18th century onwards,
the full length treatise summing up a whole theory
of economics. Such was to remain the pattern of
economics publishing until the later 19th century.
Not all economists relied on published sources;
Malthus and others travelled widely in search of
facts which they blended with information derived
from extensive reading. There were other
economists like Ricardo, whose theorizing worked
outwards from his own inner store of business
experience and personal insights, for whom access
to a library was of less significance. An economist
whose method did rely on books was usually forced
to be a collector himself since libraries well-stocked
with suitable publications were few. The 19th-
century British economist, Richard Jones, in the
laborious progress of his treatise on Rent (1831)
made frequent calls on the kindness of his friend
William Whewell to provide him with books from
the ‘Public Library’ (in fact the predecessor of
Cambridge University Library). As a country
clergyman with little money for book buying and in
comparative isolation from access to libraries, it
was a difficult assignment for him to develop a
theory of rent based on worldwide evidence.
Indeed, as the complexity and specialization of the
economics discipline grew during the 19th century
it became less and less possible for an economist to
function away from well-stocked libraries.
One reason for the increased dependence was
the question of priority. As the literature of eco-
nomics grew, so did the possibility of the prior
publication of a supposed original idea. W.S.
Jevons, for instance, developed and published his
marginal utility theory in virtual isolation from
other ideas on the topic. His subsequent realization
that Menger and Walras had arrived at the same
theoretical point in the same year of 1871 was
disturbing, but as a voracious collector of
economics literature he began to appreciate that
there were also predecessors,
3422
most notably Hermann Heinrich Gossen, whose
work published in 1854, needed to be acknowl-
edged. It was becoming clear in various inescapable
ways that the economist needed a library. Great
general libraries such as that of the British Museum,
could and still can provide much for the economist.
Karl Marx, for instance, laboured there to great
effect for years, and was able to display an
encyclopedic knowledge of the relevant literature as
a result. Nevertheless, by the end of the 19th
century, the discipline was ready for specialized
economies libraries.
Estimates of the numbers of practising econo-
mists at a given time are even today fraught with
problems of definition. However, it is fairly safe to
say that until the end of the 19th century and the
early 20th century brought the creation of schools
and faculties of economics in the universities of
Europe and North America, and governments and
business began to hire people with the degrees
awarded by these institutions, practising economists
were very few in number and more significantly
were thinly scattered geographically. A few, most
notably Smith himself but others such as Jean-
Baptiste Say, were professors in faculties of law or
other marginally related disciplines. Most were first
and foremost otherwise employed: as a businessman
like Ricardo, an official like J.S. Mill, a clergyman
like the above- mentioned Richard Jones, an
engineer like Dupuit, or a landowner like von
Thiinen. Libraries to serve such a scattered and
heterogenous group were not practical. With the rise
of Cambridge as a centre of economics excellence
in the second half of the 14th century, the creation
of the London School of Economics in 1895, and
the emergence of great faculties of economics at
Harvard, Columbia and Chicago by the beginning
of the 20th century in the USA, there were for the
first time concentrations of economics teachers,
researchers and students, whose needs gave rise to
specialized economics libraries.
The source materials that such libraries might
stock were multiplying fast. The business world was
the chief generator of publications of use to
economists. Journals from many countries in spe-
cific fields such as mining, insurance and banking,
prospectuses and annual reports of railway
Economics Libraries and Documentation
companies, histories of films or industries, biog-
raphies of businessmen, documents from interna-
tional fairs and exhibitions, all burgeoned during the
19th century. For the first time, governments
became significant publishers of economics- related
literature: British Parliamentary Papers and United
States Congressional Documents began to appear
with increased volume and regularity almost
immediately the 19th century began. In other
countries the quantity, if not the regularity, of
government publications also increased swiftly as
the century progressed. Much of their content was
statistical; and the US Census of 1790 and the
British Census of 1801 were milestones in the
practice of number gathering. Statistical societies,
such as that of London (founded 1825, now the
Royal Statistical Society) arose to take advantage of
this material and in the process created a new layer
of publication which economists could exploit.
The growth of economics as a discipline not
only brought about the multiplication of materials
but also created a need for more immediate chan-
nels of communication. The monograph which had
been the chief avenue of publication for so long,
began to lose some of its importance to the
periodical. During the 19th century, economists had
written for the great literary reviews, for general
interest magazines, for newspapers, and for the one
or two specialized publications such as the
Economist. It was only after 1886, however, when
the Quarterly Journal of Economics was founded
by Harvard University, followed quite swiftly by
the Royal Economic Society’s Economic Journal
(1841) and Chicago’s Journal of Political Economy
(1892 ), that there was a serious rival medium for
economic writings. When the American Economic
Review joined them in 1911, amain core of
prestigious journals was taking shape, and the
nature of the communication of economic ideas was
completely altered. The need to convince a
publisher of the validity and interest of one’s ideas,
or alternatively to publish at one’s own expense,
was replaced by interaction with an editor or
editorial board drawn from one’s peers. The
possibility of quite swift and direct communication
of one’s ideas, even on highly technical matters,
was opened up. Economics publication
Economics Libraries and Documentation
became more focused and more isolated from
public debate.
This rich growth of sources permitted econo-
mists to build more and more elaborate structures of
argument and evidence. W.S. Jevon’s perilous but
gallant attempts to link cyclical commercial
fluctuations back through the business statistics to
agricultural yields, to weather data, and finally to
sunspot activity in a chain of causality, was just a
particularly bold exploitation of the wealth that was
becoming available. Even a collector of books as
enthusiastic as Jevons could not reasonably hope to
acquire materials of the number and type required
for work of this kind. It is no coincidence that
Jevons was an enthusiastic member of the Library
Association (UK) and published papers on the topic
of libraries. Whilst the 18th- century or early 19th-
century economist urgently needed to be a collector
of books, an economist of the late 19th and early
20th century like F.Y. Edgeworth needed to own
hardly any books and could rely on libraries
entirely.
The economics libraries which were created to
cope with the growing need of economic
researchers for specialized materials were, in the
first place, historical collections designed to allow
the reconstruction ofpast theory and the recovery of
past knowledge. Two of the greatest have their roots
in the personal collecting of one man. H.S. Foxwell
(1849-1936) bridged the age of the great amateur
collectors and that of specialised professional librar-
ies. He amassed two great collections during a
lifetime of acquisition carried on at a level of bib-
liomania. His first collection, sold in 1901 to the
Goldsmiths Company, is now the basis of London
University Library’s Goldsmiths’ Collection. His
second collection, begun immediately the Gold-
smiths sale had put his finances to rights, was sold
in 1929 to Harvard Business School, which took
possession on Foxwell’s death to form the Kress
Library. Kress is in turn now only part of the Baker
Library of Harvard Business School.
Other great academic libraries also follow this
pattern of a core historical collection alongside a
fast-growing and fast-changing current collection.
Columbia University, with its Seligman Collection,
Johns Hopkins with the Hutzler Collection
(assembled by Jacob Hollander) and the
3423
University of Illinois with its Hollander Collection,
show a similar pattern. In Britain, the historical
riches of Goldsmiths’ are complemented within the
London University system by the London School of
Economics’ British Library of Political and
Economic Science, founded in 1896. Other parts of
the world also have libraries of similar scope -
Japan, for instance, where the Menger, Schumpeter
and Burt Franklin collections of Hitotsubashi
University provide a basis of old economics
material to underpin its modem collections. With
business history now an accepted discipline in
graduate business schools in North America, the
older titles in these libraries’ core collections are
attracting renewed attention from researchers.
Other types of economic research library have
grown up in the 20th century to supplement the
provisions of the academic libraries. Departments of
Government and their associated agencies are the
most common alternative source of economics
materials. In the USA in particular, excellent
libraries of this type are numerous; the US Depart-
ment of Commerce Library, founded in 1913, for
instance, has extensive collections including much
statistical material, and agricultural economics
material is one of the chief strengths of the National
Library of Agriculture, originally established as the
Department of Agriculture Library in 1862. The
Federal Trade Commission, the Treasury, the
Federal Reserve System Board of Governors, and
other US Government agencies have fine economics
libraries. The Library of Congress too, in its Social
Science Reading Room, gives access to an
enormous wealth of economics material. Other
countries also have government economics libraries,
the UK Board of Trade Library being one
particularly fine example. International
organizations also support economics collections;
the International Labour Organisation (ILO) in
Geneva, maintains an extremely large library,
accessible via a computerized retrieval system. The
Chambre de Commerce et d’lndustrie Library in
Paris is an example of a large, modem economics
library supported by a trade association. The
greatest economics collection in the public library
sector is that of the New York Public Library,
whose
3424
Economic Division has since 1919 produced the
weekly Bulletin of the Public Affairs Information
Service, the foremost index to the world’s English
language literature in the related fields of econom-
ics, finance, business, labour and public affairs.
Only a comparatively few libraries have the
resources to devote the necessary attention to the
frill potential range of material required by econ-
omists and the techniques that will permit its
efficient acquisition and retrieval. The Library of
the Instituts fur Weltwirtschaft in Kiel, Federal
Republic of Germany, regarded by many as the
premier economics library in the world, has been
able to do this consistently since its foundation in
1914. Its catalogue of persons gives access to
material not only via authors of books, articles and
chapters, editors of books, symposia and journals,
writers of prefaces and introductions, but also to
material whose subject is a person or persons
connected with economics. Its title catalogue
includes not merely books, but also annual reports,
serials, newspapers, collections, and the corporate
bodies, congresses and conferences which publish
material. The subject catalogue has geographical
entries in addition to conventional subject headings.
Cross reference cards are hardly ever needed, for as
many copies of the full entry card as are necessary
are entered in the relevant places in the catalogue.
The work of the library is largely in the hands of
professional economists whose subject expertise
ensures the accuracy of subject cataloguing.
Because the provision of good catalogues is never
enough to ensure that the user obtains documents
that relate to his interests, Kiel like many other
special libraries alerts users to pertinent new
acquisitions. Mark Perlman ( ) has justly
called its methods ‘the
acme of the traditional approach to economics
literature retrieval’.
In addition to dealing with the complexities of
the multiplying sources for economic research,
libraries have had to come to terms with yet another
shift in the forms of economic communication. The
urgency to establish the priority of ideas or to
publish research before it becomes obsolete has
placed strains on book and journal publishing with
which they have been unable to cope. The average
lag between submission of an
Economics Libraries and Documentation
article and its publication in a journal has increased
over the years, and the maximum wait may be in
excess of two years. This is despite the introduction
of submission fees intended to reduce the number of
submissions to some journals, the growth in the
number of specialized journals and the introduction
of a journal, Economics Letters, specifically
designed to ensure swift publication of material.
Academic monograph publishing is currently under
the severest financial strains, with spiralling costs
leading to higher price per copy to the consumer,
with consequently reduced numbers of sales driving
the unit cost up still further. In their editorial deci-
sions, therefore, publishers are putting increasing
weight on the market value of proposed titles.
The solution to this problem has been the
increasing use of semi-published forms usually
referred to as working papers. This form of distri-
bution for an individual paper, reproduced by some
inexpensive method and circulated via the writer’s
own institution’s mailing list, causes confusion and
distress to some librarians. Some libraries only
acquire working papers if they are free, many do not
catalogue them, some bind them in series, others do
not. Some libraries avoid them altogether because of
the difficulties they cause, and in the conviction that
anything worthwhile which appears as a working
paper will eventually be published in more
permanent form. This is a serious disservice to
economic scholarship. Roy Harrod ( ) said
‘Mimeographed essays issued
in advance of publication, if any, by the research
unit of one university to the professors of other
universities all over the world have come to con-
stitute the main matter for reading, at least among
theoretical economists.’ Indeed, some of the diffi-
culty of knowing all the writings of an economist as
distinguished as the Norwegian Ragnar Frisch stems
from the fact that he published so frequently in
working paper form. The collection of working
papers at Warwick University Library. (UK), their
published Economics Working Papers Bibliography
and their microform service are a major contribution
to this problem area.
In the last quarter of the 20th century, the forms
of source material for the economist have begun to
challenge the service capacities of traditional
Economics Libraries and Documentation
librarianship. Most particularly, published statistics
are no longer adequate for the purposes of many
economists: they are out-of-date when pubhshed;
they are in summary rather than comprehensive
form; and they require recording in computerized
form so that they can be arranged, shuffled and
manoeuvred into revealing forms by the researcher.
The electronic publication of statistics, making
them available originally in computerized form, is
growing rapidly. Increasingly, economic research
depends on access to suitable computer hardware,
availability of programmes which will perform the
required tasks, and tapes of the data, rather than on
books or other paper formats. To some extent this
trend tends to exclude the library from the research
process, but that is not necessarily always the case.
The issue of whether an economics library
should merely confine itself to searching biblio-
graphic databases on behalf of its economist clients
or whether it should go further in identifying and
making available numeric databases, has already
begun to be explored, for instance in the Economics
Library of the Ministry of Agriculture Fisheries and
Food in the UK (O’Sulhvan ). The Baker Library at
Harvard has long been involved with the acquisition
of computerized data bases, but now has
professional staff members who are also actively
involved in reviewing, publicising, and
manipulating numerical databases. What is more, to
add the exploitation of numerical databases,
whether commercial, such as the many provided by
Evans Economics, Inc. (EEI), or government
created, such as the US Bureau of Labor Statistics’
LABSTAT, or the various services of the Bureau of
Economic Analysis, to the functions of a library
which is already managed by computerized systems
often does not seem like a major problem. The
availability of statistics in published or database
form, the relative costs, types of series available in
the alternative forms, the compatibility of
computerized data with the systems available to the
economist, are undeniably difficult issues.
However, librarians in their new roles as
information specialists are proving themselves
capable of lending invaluable assistance to their
clients with this type of material.
3425
The potential of computerized systems for the
swift transmission of information is likely to be
further exploited. Electronic mail, for instance,
already permits the flexible exchange of messages,
long or short, amongst individuals or groups, by
users of computer systems linked by telephone
lines. The potential of electronic mail for the almost
instant communication of research findings amongst
a group of interested experts, an ‘invisible college’,
is obvious. The electronic journal, which is at
present being developed at a number of centres,
seems likely to be an answer to the problem of the
chronic pressure on economics journals. The
electronic journal exists originally as a database
controlled by an editor. Authors send their
‘manuscripts’ to the editor on-line from wherever
their computer is located, the text can be refereed,
edited and amended on-line, and then the
subscribers read the journal on-line, printing out text
on demand. The journal can be altered daily if new
material is received and old material can be
relegated to storage files. Within a group of
specialists, united by the necessary machinery and
associated financial arrangements, accurate, easily
modified information can be available in a form
more swift and convenient than any previously
used. This new communication medium need not
necessarily circumvent specialized libraries, which
in future might actually provide facilities for such
operations, for instance in the archiving of material
from the electronic journal.
With their computer expertise, their staffs of
specialists, and their vested interest in the storage
and dissemination of information, there are very
obviously roles which libraries are developing in the
management of electronic information systems.
Though there seems little indication of the large-
scale return to conventional library-based
scholarship in economics prophesied by Harry
Johnson ( ), a total disregard of the value of
historic and rare books collections already existing
should not be contemplated. No discipline can
safely neglect its past and it is to be hoped that
economics libraries will cherish their treasures for
the use of the minority of scholars whose approach
is more reflective and retrospective. Nonetheless,
the cutting edge of economics librarianship in the
21st century seems likely to be
3426
as different from that of the 20th or the 19th as will
be its forms of documentation, and indeed the
economics discipline which generates them.
Bibliography
Cole. A.H. 1957. The historical development of economic and
business literature. Kress Library Publication no. 12.
Boston: Baker Library, Harvard Business School.
Fletcher. J. 1972. A view of the literature of economics.
Journal of Documentation 28(4): 283-295.
Harrod. R. 1969. How can economists communicate? Times
Literary Supplement. 24 July. 805-806.
Johnson. H.G. 1977. Methodologies of economics. In The
organisation and retrieval of economic knowledge, ed. M.
Perlman. 496-509. London: Macmillan.
Kindleberger. C. 1977. The use of libraries by economists: A
personal view. In The organisation and retrieval of
economic knowledge, ed. M. Perlman. London:
Macmillan.
Koch, J.E., and J.M. Pask. 1980. Working papers in academic
business libraries. College and Research Libraries 41(6):
517-523.
O’Sullivan, S.D.A. 1982. Numeric and bibliographic data-
bases for agricultural statistics: Conflict or
co-operation? In Sixth international online information
meeting. Oxford: Learned Information.
Perlman, M. ll>72. Economic libraries and collections. In
Encyclopedia of library and information science, vol. 7.
New York: Marcel Dekker.
Perlman, M. 1973. Editor’s comment [on Riel Instituts fur
Weltwirtschaft Library], Journal of Economic Literature
11(1): 56 58.
Ruokonen, K. 1981. BILD-integrated online system for
economic and business literature. Lidskrift fur
Dokumentation 37(3): 62-67.
Yohe, G.W. 1980. Current publication lags in economics
journals. Journal of Economic Literature 18(3): 1050-
1055.
Economics of Franchises
Roger D. Blair and Jessica S. Haynes
Abst r act
Franchising, which is common in many Rationale for Franchising advanced
economies, is a contractual form of
vertical integration. This article examines the Brands provide information to consumers in a economic
rationale for choosing franchising mobile society', and so reduce search costs.
Economics of Franchises
over vertical integration. It also examines the
influence of the franchisor's ability to maxi-
mize its own profit.
Keywor ds
Distribution systems; Franchisee; Franchising;
Franchisor; Royalties; Vertical integration
JEL Cl assi f i cat i ons
L14; L24; M55
Franchising is a contractual form of vertical inte-
gration. A manufacturer, for example, produces a
product that must be distributed to consumers. The
manufacturer can perform the distribution function
itself through a chain of retail outlets it owns and
operates. When a manufacturer both produces and
distributes its product, the firm is vertically
integrated by ownership. The manufac- turer can
then control the retail promotion, customer service,
pricing, product availability, delivery and other
relevant decisions at the distribution stage. It does
this through internal managerial decisions designed
to maximize the overall profit of the firm. But this is
not the only way to organize the production and
distribution of the firm’s output. Instead of having a
network of its own distributors, it can license
independent firms to perform the distribution
function. These licensees are termed franchisees,
and the distribution system is called a franchise
system. The manufacturer then engages in a
contractual form of vertical integration. The
franchise contract gives the franchisee the right to
distribute the manufacturer’s product, but also
allows the manufacturer to control retail promotion,
customer service, resale prices (minimum or
maximum), and a host of other things that are
important to the manufacturer, who is now a
franchisor. But if franchising is just a contractual
form of vertical integration, what is the rationale for
franchising?
Economics of Franchises
When a consumer visits a branded outlet, there are
supposed to be no surprises - pleasant or otherwise.
Thus, local residents go to Roger’s Ribs and
Jessica’s Java while visitors go to Sonny’s Barbe-
cue and Starbucks. The consumer’s increased reli-
ance on brand names has resulted in the
development of chains of retail outlets. Chains
benefit from lower costs as a result of economies of
bulk purchasing and economies of scale in
production, new product development, and pro-
motion. In principle, all of the retail outlets could be
corporately owned and managed, but many of the
resulting chains are organized as franchises.
Franchising permits specialization that may lead
to higher overall profits. The franchisor is the
innovator who specializes in developing the brand,
exploiting economies of scale in production and
promotion, and negotiating with vendors on behalf
of the chain. The franchisees bring their
entrepreneurial spirit to their locations. They also
contribute their knowledge of the local market. The
result is a synergistic effect which increases
potential profits (Caves and Murphy ). Franchising
succeeds because it allows each party to do what it
does best. Further, franchise contracts deliberately
organize this relationship to give the franchisor and
franchisee incentives to work in tandem to increase
revenues.
Types of Franchise
Although the lines drawn are somewhat arbitrary,
franchising can be divided into traditional fran-
chising and business format franchising. Traditional
franchising is used by a manufacturer to distribute
its product through distributors (the franchisees)
that are specifically licensed to do so. Traditional
franchising is found in several industries:
automobiles, beer, gasoline, ice cream, and soft-
drink bottling to name a few. Business format
franchising involves the use of the franchisor’s
brand, trademark and trade dress, and distinctive
way of supplying goods and services to the
consumer. In this case, the franchisor develops and
promotes the concept while the franchisees
implement the concept and carry out local
production and distribution. The ‘quick
3427
serve restaurant’ sector is a good example, with
familiar names such as the US-based chains
McDonald’s, Pizza Hut, Subway and Taco Bell.
There are many other sectors in business format
franchising including accounting services, auto-
mobile servicing, health and fitness, hotels and
motels, and real estate.
Franchisor Compensation in Traditional
Franchising
Traditional franchisors sell products to their fran-
chisees which are resold to consumers. The fran-
chisor earns its profit on these sales to franchisees.
The extent to which franchising is a good substitute
for vertical integration depends on two critical
factors: the market structure in distribution and the
relative efficiency of franchisees versus employee-
managers. Assuming that franchisees are neither
more nor less efficient than employee-managers, a
manufacturer will earn the same profit whether it
performs the distribution function itself or
franchises the distribution. Its profit will be the
same because the cost of performing the distribution
function will be in the same in either case. If
franchisees are more efficient than employee-
managers, then distribution costs will be lower,
retail demand will be higher, or both. This will
improve the manufacturer’s profit, due to increased
sales to the franchisees.
If there is monopoly at the distribution stage, the
market structure is one of successive monopoly.
Assuming equal efficiency, the manufacturer will
prefer vertical integration rather than franchising.
Compared with the case of competitive distribution,
successive monopoly results in double
marginalization, which leads to lower output and
higher price. The manufacturer’s profits are reduced
below the level that would result from vertical
integration because double marginalization reduces
the derived demand for the product.
These ill effects can be offset with other con-
tractual provisions such as maximum resale price
constraints, minimum quantity standards and price
advertising (Blair and Esquibel ). If these are
effective, they should eliminate the
3428
exercise of downstream monopoly power and
improve the manufacturer’s profits.
Things are more complicated when franchisees
are more efficient than employee-managers. The
effect of the increased efficiency is to shift the
derived demand to the right. This, of course, tends
to improve the manufacturer’s profits. The exercise
of monopoly power by the franchise, however,
tends to decrease the manufacturer’s profits. The net
effect cannot be determined on a priori grounds.
Presumably, when a manufacturer expects the net
effect to be positive, franchising is selected.
Franchisor Compensation in Business
Format Franchising
Business format franchisors do not sell products to
their franchisees for resale. Instead, they provide a
concept, a brand, and a distinctive way of doing
business. Business format franchisors have many
ways of charging their franchisees for using the
licence: initial franchise fees, sales revenue
royalties, output royalties, rent and sales of nec-
essary inputs, to name a few.
Nearly all franchisors include franchise fees in
their contracts. These fees are structured as either
initial or periodic lump-sum payments, with the
initial form being far more prevalent. Blair and
Lafontaine ( , p. 61) foimd that 99.2 percent
of all franchisors use initial franchise fees. In
principle, both forms allow the franchisor to capture
the maximum profit available under vertical
integration, provided that the franchisee can obtain
all inputs at competitive prices. To realize this level
of profit, the franchisor sets the initial franchise fee
or the present value of the periodic payments equal
to the present value of the stream of future
operating profits that will be generated by the
franchisee. This, in turn, will allow the franchisee to
earn only a competitive return on its investment.
As in traditional franchising, the efficiency of
the franchisees relative to employee-managers
contributes to the level of profits the franchisor can
obtain. If franchisees are more efficient, this will
increase the franchisor’s profit because
Economics of Franchises
franchisees will bid up the franchise fee until it is
equal to the stream of future operating profits.
Although theoretically feasible, franchise fees
generally are not set at such a high level. This is
predominantly a result of the uncertainty surround-
ing future market conditions, future interactions
between the parties, as well as franchisees’ wealth
constraints. Often franchise fees just cover the start-
up costs of opening a new franchise location.
Franchisors can also extract revenue by charg-
ing franchisees royalties, based on either a per-
centage of their sales revenue or a fixed fee on
output sold. Sales revenue royalties are the more
prominent of the two forms and the second most
utilized charge by franchisors, next to initial fran-
chise fees (Blair and Lafontaine , p. 66).
As long as the franchisee can acquire all inputs
at competitive prices and a competitive market
exists among local franchisees, the franchisor can
extract the optimum level of profits with royalties.
In the case of sales revenue royalties, the franchisor
achieves this outcome by setting the royalty rate
equal to the ratio of the difference between the
profit maximizing price and marginal cost to the
profit maximizing price. This rate will result in a
post-royalty price equal to the franchisees’ marginal
cost. The royalties collected will be precisely equal
to the maximum profits that vertical integration
would yield.
Again, franchisees’ efficiency relative to
employee-managers impacts these profits and the
decision to franchise. If the franchisees are more
efficient, then the franchisor will earn more profit
due to the increase in sales revenue. If instead the
employee-manager is more efficient, then the
company will choose vertical integration.
If the franchisee has local monopoly power,
however, the franchisor will be unable to attain the
maximum level of profit through a sales royalty.
The ability to exert local market power makes price
endogenous for franchisees and allows them to
factor sales revenue royalties into output decisions.
Therefore, no sales revenue royalty will allow the
franchisor to capture the amount of profit available
through vertical integration. An equivalent analysis
holds for output-based royalties (Blair and
Kaserman ). Output-based royalties, however, make
little sense in business
Economics of Online Recruitment
format franchises because there are often too many
products to make tracking outputs feasible.
Another way in which franchisors obtain rev-
enue is through input requirements. In many busi-
ness format franchise systems, the franchisees are
required to buy inputs from the franchisor. For
example, a pizza franchisee may be required to buy
pizza dough and sauces from the franchisor. If the
franchisor charges its franchisees prices above the
competitive level, these tying arrangements provide
an alternative way of extracting the profit that
vertical integration would provide (Blair and
Kaserman ). Again, however, this result holds only
when there is competition among the franchisees;
otherwise, the higher input prices cause distortions
that reduce overall profits.
Choosing to Franchise
The choice of franchising ultimately depends on Catherine Thomas
whether it is more profitable than vertical integra-
tion. This, in turn, depends on the structure of the
particular market and the efficiency of franchisees
relative to employee-managers. The less compet-
itive the downstream market structure, the less
attractive franchising becomes. But the more effi-
cient franchisees are, the more attractive franchising
becomes. The choice of whether and how to
franchise will therefore vary from chain to chain.
See Also
► A: : . . v
► Vertical Integration
Bibliography
Blair, R.D., and A. Esquibel. 1996. Maximum resale price
restraints in franchising. Antitrust Law Journal 65: 157- of the work process and payment systems.
180.
Blair, R.D., and D.L. Kaserman. 1978. Vertical integration,
tying, and antitrust policy. American Economic Review
68: 397-402.
Blair, R.D., and D.L. Kaserman. 1980. Vertical control with
variable proportions: Ownership integration and
contractual equivalents. Southern Economic Journal 47:
1118-1128.
3429
Blair, R.D., and D.L. Kaserman. 1983. Law and economics of
vertical integration and control. Orlando: Academic
Press.
Blair, R.D., and F. Lafontaine. 2005. The economics of
franchising. New York: Cambridge University Press.
Caves, R.E., and W.F. Murphy. 1976. Franchising: Firms,
markets, and intangible assets. Southern Economic
Journal 42: 572-586.
Coase, R.H. 1937. The nature of the firm. Economica 4: 386-
405.
Kaufmann, P.J., and F. Lafontaine. 1994. Cost of control: The
source of economic rents for McDonald’s franchises.
Journal of Law and Economics 57: 417—454.
Rubin, P. 1976. The theory of the firm and the structure of the
franchise contract. Journal of Law and Economics
21:223-233.
Spengler, J.J. 1950. Vertical integration and antitrust policy.
Journal of Political Economy 58: 347-352.
Economics of Online Recruitment
Abst r act
Online recruitment describes hiring workers who
were initially selected via the Internet. The
practice is now widespread, with a majority of
US jobseekers undertaking some online job
search activity. The online intermediaries that
allow employers and employees to connect with
each other, leading to online recruitment, vary in
scope. They range from websites that provide
information about workers or openings, thereby
facilitating search (‘job boards’ or ‘job search
engines’) to websites that both provide
information and enable employers and
employees to interact online during the hiring
process (‘online labour markets’). A subset of
online labour markets also provides an
infrastructure that allows for online management
Keywor ds
Adverse selection; Communications technology;
Globalization; Hiring; Intermediation; Job
boards; Labour; Moral hazard; Online
3430 Economics of Online Recruitment

labour markets; Offshoring; Organizational Early job boards ranged from being quite gen-
economics; Outsourcing; Personnel economics; eralist in the type of work posted to being very
Search models; Search engines; Services specialised. Large corporations were some of the
industries; Trade frictions; Websites first recruiters to adopt online technology in their
hiring processes, purchasing technology from
specialised software providers but performing the
JEL Classifications intermediation functions of managing applications
J6; L86 in-house. As early as 2001, 90% of large US
corporations were recruiting via the internet
This article describes the development of interme- (Cappelli ), with the corporate homepage often
diaries that govern online recruitment and the com- providing a ‘careers’ link for prospective applicants.
munications technology that has shaped the course Large newspapers also put their classified section’s
of that development. It presents data collected from job listings online, and non-profit organisations
various sources about the extent and growth of such as the US military began to operate online
online recruitment. It then discusses some of the recruitment services. During the late 1990s, in
features of online recruitment that are of interest for tandem with the Internet boom, a handful of online
general economic research. These features include job-search engines grew to prominence in this
online labour market search and matching pro- industry. These include Indeed, CareerBuilder,
cesses; supply and demand in online labour mar- Monster.com, Hotfobs and SimplyHired. Some of
kets; the implications of online hiring for remote these job boards (such as LinkUp, Indeed and
work where output can be delivered electronically; SimplyHired) are metasearch engines, which means
and the increased availability of data about labour that they aggregate job postings from other boards,
market transactions in online settings. an activity known as ‘scraping’ or ‘wrapping’.
While these large websites are broad in scope, there
are also many industry-specific sites. One example
Job Boards and Job-Search Engines is the IT industry site DICE.com.
Many Internet job-search engines also provide
The Internet allows employers and employees to
additional services that assist in employer and
communicate prior to an employment contract in a
employee search. Most of the large sites are con-
variety of ways. The common feature of online
tinually evolving, offering job-seekers and potential
recruitment intermediaries is that they provide the
employers increasingly sophisticated tools. For
institutional framework, or infrastructure, that
example, Monster.com - now the world’s largest
permits this communication. The first main way in
recruitment website, with over 1.3 million resumes
which they facilitate communication is to make
being posted globally each month (Monster.com ) -
employers aware of potential employees, and vice
offers workers tools including skills analysis, cover
versa. In the early 1990s, along with the develop-
letter and resume-writing tips, and career
ment of the Internet came organisations that allowed
management services such as employer
employers to post online descriptions of job
comparisons and salary and benefits analysis. In
vacancies and allowed prospective employers to
2012, Monster launched BeKnown, a professional
search those listings. Websites that perform this
networking application on Facebook, to allow
function are commonly referred to as ‘job boards ’.
employers to manage their online recruitment
Some sites also allow workers to post their resumes
activities.
online to bring them to the attention of recruiters.
Online search dominates job-seeking activity in
While recruitment via these sites is initiated online,
the USA and in other developed countries. Jansen et
the jobs that are filled via this recruitment channel
al. ( ) estimated that by 2005, more
span many types of work and a broad range of
than 52 million Americans had performed online
sectors in the traditional offline economy.
job searches. By 2010, a group of the largest
Economics of Online Recruitment
online recruiters had 51.1 million visitors per month
among them (comScore ). In 2012, 99% of the
Fortune 1000 firms had purchased services from or
utilised job postings on Mon- ster.com. The
Monster Employment Index (MEI), a survey
measure available on the Mon- ster.com website,
tracks changes in online recruitment activity over
time and across different industry sectors in the
USA, Canada, Europe and India, by gathering and
summarising data from a large representative
selection of career web sites and online job listings.
Online Labour Market Platforms
Intermediaries have also arisen to facilitate the
online recruitment of workers whose output can be
delivered electronically and they look very different
from the search engines described above. When the
work can be done remotely, the ability to recruit
workers online has an even greater transformational
effect on recruitment activities because, while
traditional labour markets are segmented across
geographical boundaries and by distance, online
labour markets for remote work can match
employers and employees across the world. Blinder
and Krueger ( )
estimate that up to 25% of US jobs can be
offshored, made possible, in part, by electronic
product delivery.
Websites that assist employers in finding
workers to deliver electronic output began to
emerge in the late 1990s and early 2000s. While
these websites function as platforms that facilitate
search, they also provide infrastructure that allows
employers to oversee the different stages of the
employment process. Website functions include
hiring, project/work management, payment man-
agement and API (application programming inter-
face). These services reduce the trade frictions
associated with hiring and managing a single
remote worker, as well as the barriers to the coor-
dination of multiple remote workers. Horton ( )
refers to these platform websites as ‘Online
Labour Markets’ (OLMs) and describes them as
having three common features: (1) Labour is
exchanged for money within the platform. As
3431
such, there are two parties involved in each trans-
action, and these platforms differ from online tools
that deliver content without payment, such as
Wikipedia and Linux. (2) Output is delivered
electronically. (3) The allocation of labour is
determined and payments for services are conducted
within the platform. Horton then divides the OLMs
that have these three features into two categories:
‘spot’ OLMs and
‘contest’ OLMs.
Spot OLMs allow employers and employees to
contract with each other regarding a specific piece
of work, at a given price, or for a specified length of
time. The work contracted ranges from data entry,
to website optimisation, to discrete programming
tasks. Large intermediaries in this category include
oDesk.com, ELance, RentACoder and Guru.
Amazon’s Mechanical Turk (AMT) is an example
of a spot OLM that offers a slightly different set of
services, providing employers with the opportunity
to hire a large temporary workforce online to
perform wide-ranging tasks. This platform is often
used both as an experimental site for research
(where the human response is of interest in itself)
and for commercial tasks in which humans are more
effective than computers (for example, identifying
objects in a photograph or transcribing audio
recordings).
Data on the total number of employees and
postings on each of these sites are relatively sparse.
A 2009 report estimates that ten of the largest
platforms had over 2.3 million registered workers at
that time, although these workers are likely to
overlap across sites (SmartSheet ). The types of
tasks performed via spot online labour market
platforms range from small, automated tasks with
low pay, to simple projects, to relatively complex
integrated tasks. Examples include: finding a set of
email addresses or prices; writing a product review;
designing a website or a presentation; programming
software; or developing an algorithm. The tasks
span services and solutions that are built on top of
Micro Tasks platforms (such as often found on
AMT) or technology-assisted relationship
management on top of project management (as
offered by oDesk).
The second category of online labour market
platforms comprises ‘contest’ websites. The
3432 Economics of Online Recruitment

employer-employee relationship on contest OLMs how labour markets work (Rogerson

functions differently from that in spot markets, in
et al. ). The growing prevalence of recruitment in
that firms requiring certain electronic output post a
online labour markets is, hence, of great interest to
competition for that output and invite submissions
researchers because the longstanding and important
on the website. Workers who wish to enter the questions about all labour markets are paralleled in
competition undertake the work, and buyers either
online settings. In addition, the contrast between
select a winner or opt not to do so. Intermediaries of
offline and online labour markets is, in itself,
this type include CrowdSpring, 99Designs, revealing about frictions that exist in all labour
GeniusRocket and LogoToumament. markets.
Technological and organisational advances In particular, economists study labour market
within these platforms have led to the globalisation
efficiency, motivated by several key empirical facts.
of the labour market for this type of work. TheOne critical question is why it is that unemployment
employers demanding these types of labour services
often exists at the same time as unfilled job
are often located in different countries from where
vacancies. Another central question involves the
the most efficient potential employees can be found,
way in which wage levels are set for individual
and employers can recmit large numbers of foreign
workers in employment: why is it that similar
workers with the required skills within these sites.
workers earn persistently different wage rates? For
As an illustration, in June 2010, over 60% of the
example, Mortensen ( ) finds that observ
revenues paid for online work in oDesk.com cameable characteristics typically explain no more than
from employers located in the USA, while workers
30% of observed wage variation in traditional
located in the USA accounted for less than 15% of
labour markets. These two facts each suggest that
the revenues. The country accounting for the largest
trade frictions in labour markets prevent these
share of revenues earned was India, at almost 30%.
markets from clearing at an equilibrium wage rate
A majority of transactions on this website spanthat reflects the interaction of competitive demand
international borders, and employers regularly and supply. As noted by Rogerson et al. ( , p.
employ workers in different countries
960), the empirical evidence sug
simultaneously. gests that ‘there is simply no such thing as a
In addition, the flexibility provided in these
centralised market where buyers and sellers of
virtual workplaces allows for labour market par-
labour meet and trade at a single price, as assumed
ticipation among groups of domestic and foreignin classical equilibrium theory’.
workers that previously had limited employment Labour market economists have shown that
opportunities. The discrete nature of jobs, theintroducing employer and employee search to
variation in job duration, and the worker’s ability to
labour market models can generate theoretical
perform independent work, perhaps at irregular predictions that mirror observed outcomes in these
hours, make part-time online work possible for markets. Much of this research has its origins in
them. The online workforce includes large groups
work done by the winners of the 2010 Nobel Prize
of students, for example. in economic science, Peter Diamond, Dale
Mortensen and Christopher Pissarides. These
models typically consider a forward- looking
Employer-Employee Matching and individual’s job-search problem - the decision of
Wage Determination in Online whether to look for a job based on expected future
Labour Markets wages and the likelihood of finding a job, as well as
job-search costs and any benefits from remaining
Because labour market outcomes are one key unemployed. Similarly, an employer’s problem
determinant of individuals’ economic activity, about whether to post a job and search for a worker
economists devote much time to understanding can be modelled as a function of the probability of
finding a worker, the value of finding a worker to
accept the job, the
Economics of Online Recruitment 3433

expected wage that it would cost the employer, and The overall consequences for the wage determi-
any search costs that would be incurred. Petrongolo nation process are likely to depend on the nature of
and Pissarides ( ) survey early the job posting and the set of skills required in the
matching work in labour markets, and discuss how job, as well as the matching function in the online
various models differ in the wage determination labour market in question.
process, matching function, and in the other
assumptions made.
Online recruitment is generally thought to lower Online Labour Supply and Demand
the costs associated with job search, in comparison
While lower search costs in online recruitment are
to offline laboin markets. When employers can
predicted to increase the efficiency of labour market
browse the resumes of hundreds or thousands of
outcomes for a given set of employers and
potential hires that have all been screened as
employees, it is also likely that lower costs will
possessing relevant qualifications and experience
affect the number and type of workers applying for
for the job, it is much easier to become informed
jobs. As the costs of applying for jobs fall, the
about candidates. In addition, many intermediaries
number of applicants for any one job is predicted to
lower the costs of contacting the workers, and some
increase, and employers may find that having many
even facilitate employer- employee contracting.
more applicants makes it harder and more costly to
This can all be done in front of computers from
find a good fit among the candidates. In particular,
remote locations. Most models of search and
online applicant profiles tend to contain lots of
matching predict that lower search costs increase
information described by Autor ( ) as
match efficiency and, depending on the structure of
Tow bandwidth’, meaning that the information is
the model, often also increase the productivity of
not helpful to employers in distinguishing between
the resulting employment matches. Both these
candidates. In online markets, employers have less
effects increase economic output and have the
of the ‘high bandwidth’ information that can be
potential to transform a whole range of laboin
more easily gathered during inperson meetings and
market outcomes, including wages, job dination,
that allows employers to determine the quality of
aggregate unemployment and aggregate
the match for each potential candidate.
productivity.
Perhaps more importantly, in addition to
Online recruitment, especially in online laboin
increasing the number of applicants, when
market platforms, is also changing the nature of the
employers solicit applications, the composition of
bargaining game between employers and workers
workers applying for jobs may change. When it is
that determines workers’ wages. In traditional
costly to apply, workers with private information
models of laboin markets, wages are often modelled
about their own suitability for the job will apply
as the outcome of Nash Bargaining over the surplus
only if the probability of their being hired, which is
generated by the transaction, relative to each party’s
increasing in their fit, is sufficiently high. When
outside option and bargaining power. One
application costs fall, candidates who judge
implication of the increased ease of search and
themselves to be less well suited to the job will
matching in online markets could well be that the
choose to apply, increasing the problems associated
outside options for employers and employees, as
with adverse selection (Akerlof ). Autor ( )
well as their relative bargaining powers, are
forecasts that this adverse selection of
different than in offline markets. It could be that a
lower-quality - or worse-fit - applicants into the set
given employer, when searching online rather than
of candidates for any given job could work to offset
offline, has a choice between larger numbers of
efficiency gains from lower search costs. There is
potential hires that are closer substitutes and who
some empirical evidence that is consistent with the
have fewer valuable outside options. For a given
presence of adverse selection in several online
worker, on the other hand, the ability to search
markets. Among the largest OLM platforms, a large
online may increase his or her access to potential
share of posted jobs are never
employment opportunities.
3434
filled despite attracting large numbers of applica-
tions, suggesting that distinguishing between can-
didates remains a costly undertaking for recruiters
in these markets.
However, low search costs and more competi-
tion among workers (leading to lower wages) are
also likely to bring employers into the market and
affect the number and type of jobs that employers
post. A 2009 report by Smartsheet.com comments
that many buyers of online labour services are
individual employers or small companies with one-
off projects and limited in-house resources. Often,
the work done online via online recruitment would
have been infeasible to staff through traditional
labour markets. This includes tasks such as image
tagging, transcription and other large volume tasks
that can be completed online at relatively low cost.
Additional Features of Interest in Online
Recruitment for Electronically Delivered
Work
In markets where adverse selection threatens to lead
to market unravelling, efficiency can be increased
through the development of costly signalling
mechanisms (Spence ). In online labour markets, it
is possible that good-quality workers might find it
worthwhile to agree to work at a very low wage
until their type is revealed. High future wages
would compensate for the cost of sending this
signal only for those workers who are subsequently
revealed, on the job, to be high quality. Employers,
then, would be able to distinguish good-quality
from bad-quality workers by offering low initial
wages. In practice, however, there is some evidence
that current online labour market institutions are
unable to facilitate the costly signalling required to
separate good from bad workers. While workers
with no established reputation often bid to work at
very low hourly wages, there is still unemployment
among these workers and, at the same time, job
postings go unfilled. One potential explanation is
that hiring online workers incurs costs for
employers in addition to the wages paid in any
transaction, and the wages
Economics of Online Recruitment
paid to workers cannot fall to levels that are low
enough to induce employers to hire a worker of
unknown quality. For instance, workers typically
cannot bid to work for negative wages on OLMs.
Some online labour market platforms are devel-
oping alternative mechanisms that allow the quality
of inexperienced workers to be credibly revealed.
For example, Stanton and Thomas ( ) show that
inexperienced workers affiliated
with the small organisations in oDesk.com called
outsourcing agencies are high-quality workers.
Affiliation is associated with greater employment
success on the site, which suggests that affiliation
acts as a credible signal of quality.
When work is undertaken remotely by workers
recruited online, employers might be particularly
concerned about moral hazard. For example, when
the work undertaken requires a worker to have
access to proprietary systems and data, an employer
takes on the risk that this information will be
expropriated. There are also the more standard
concerns about the possibility of worker shirking on
the job. The largest online labour markets have been
able to develop an institutional context that
discourages worker moral hazard or, at least,
insures employers against it. Most OLMs operate
dispute resolution systems - tending to favour the
employer - and monitoring systems that reduce
contractual incompleteness. Many have
wellfunctioning employer feedback systems that
allow workers to build valuable reputations in the
marketplace and increase incentives for good
performance on the job.
The low observed wages, especially in online
labour market transactions between employers in
developed economies and workers in low-wage
countries, have led to public debate about whether
these marketplaces constitute ‘digital sweatshops’
(Zittrain ). There is concern that firms hire remote
workers in online markets to engage in cross-border
institutional arbitrage, allowing employers to avoid
their local labour market regulations. It is, however,
not clear that this view is held by the workers on
these sites. In his 2011 paper, Horton surveys
workers on AMT and finds that they perceive
online employers to be slightly fairer and more
honest than offline employers.
Economics of Online Recruitment 3435

One final issue that arises as a consequence of
the growth in online recruitment for work that can
be delivered electronically is that the need for
physical concentration of the workforce is greatly
reduced. A distributed workforce that spans the
confines of a traditional workplace may also render
traditional organisational hierarchies redundant.
Some managerial roles that involve coordinating
teams and processing information may be replaced
by online communications technology. By
facilitating online remote work, it is likely that
developments in online recruitment have
consequences for how firms are structured and for
the way in which economic activity is undertaken,
in general, in the global economy.
workers cannot generally be easily tracked across
different markets. With access to all these data, it
becomes feasible to observe the impact of policy
experiments at a level of detail that is not practical
in offline labour markets. Nonetheless, the extent to
which any inferences drawn are externally valid, for
example, in offline labour markets, would depend,
however, on the experiment in question.
Researchers should bear in mind that the type of
jobs (and the type of workers) on online sites are a
selected sample from the overall labour market.
See Also

Data Availability About Market ► k

Transactions >,:>■
►! . nic
Overall, then, online recruitment opens up new C c
channels of communication that can be
characterised as reducing search costs and affecting
the bargaining power of each party, perhaps making
labour supply more competitive. In labour
economics, both the matching function and the ►
bargaining function are central in determining
labour market outcomes (and, thus, individual and
Bibliography
economy-wide economic outcomes). But, in
addition to changing the nature of these functions, Akerlof, G. 1970. The market for ‘lemons’, quality uncer-
online recruitment offers unprecedented access to tainty and the market mechanism. The Quarterly Jour-
micro-level data about the recruitment process that nal of Economics 84(3): 488-500.
will allow the nature of these functions to be studied Autor, D. 2001. Wiring the labor market. Journal of Eco-
nomic Perspectives 15(1): 25-40.
in much greater detail. Precisely because Blinder, A., and A. Krueger. 2009. Alternative measures of
interactions happen electronically, at arm’s length, offshorability: A survey approach. NBER Working
data on the nature of these interactions can be stored Papers 15287, National Bureau of Economic Research,
and retrieved by researchers ex post. Inc.
Cappelli, P. 2001. Making the most of on-line recruiting.
As an example, it is now possible to determine Harvard Business Review, March.
the profiles that employers choose to view, how comScore. 2010. comScore media metrix ranks top-growing
long they look at each profile, and whom they properties and site categories for January 2010. Avail-
choose to contact. In OLMs, researchers can able at:
observe entire employment histories for workers
and entire hiring histories for employers. In some . Accessed 15 Oct 2012.
cases, these websites collect data on the bargaining Horton, J. 2010. Online labor markets. Working paper.
process over wages, along with both employers’ Horton, J. 2011. The condition of the Turking class: Are
and employees’ outside options, as revealed by online employers fair and honest? Economics Letters
111(1): 10-12.
other job offers or potential hires. One remaining
Jansen, B., K. Jansen, and A. Spink. 2005. Using the web to
drawback is that firms and look for work. Implications for online job seeking and
recruiting. Internet Research 15(1): 49-66.
3436
Monster.com. 2012. Monster Internal Data, monthly average
Q3.
Mortensen, D. 2003. Wage dispersion. Cambridge, MA: MIT
Press.
Petrongolo, B., and C. Pissarides. 2001. Looking into the
black box: A survey of the matching function. Journal of
Economic Literature 39(2): 390-431.
Rogerson, R., R. Shimer, and R. Wright. 2005. Search-
theoretic models of the labor market: A survey. Journal of
Economic Literature 43(4): 959-988. Smartsheet.com. 2009.
Paid crowdsourcing. Current state and progress toward
mainstream business use. Available at:
C:: ,/dsou c: f/OZZ-y;c/=202009%20-%20Rs:s.
. Accessed 15 Oct
2012.
Spence, M. 1973. Job market signaling. Quarterly Journal of
Economics 87(3): 355-374.
Stanton, C., and C. Thomas. 2012. Landing the first job: The
value of intermediaries in online hiring. Working paper.
Zittrain, J. 2009. Work the new digital sweatshops.
Newsweek, December. Available at:
J serf svre vsek/2009/12/07/work-' s-:sv-
Economics, Definition of
Roger E. Backhouse and Steven Medema
Abstract
Economics is difficult to define unambiguously,
many definitions having been proposed as the
subject has evolved. Definitions are ex post
constructions, even rationalizations, but they can
nonetheless influence what economist do and
how they set about doing it. This article
considers the main definitions from the late 18th
century to the present, pointing out some of the
ways in which changing views reflect and have
influenced changes in the subject.
Keywords
Cameralism; Economic science; Economics
imperialism; Economics, definition of; Institu-
tional design; Marshall, As; Mill, J. S; Opera-
tions research; Political economy; Rational
Economics, Definition of
choice; Robbins, L. C; Samuelson, P. A; Scar-
city; Steuart, Sir J; Value judgements
JEL Classifications
B0
The definition of economics has evolved signifi-
cantly over time, influenced by and influencing the
focus of economic study. The definition often
attributed to Jacob Viner, ‘economics is what
economists do’, reflects the difficulty of providing
an unambiguous definition. The problem, of course,
is that definitions of the field are proposed ex post in
an attempt to impose order upon a body of work that
has grown up as economists have sought to tackle
diverse practical and intellectual problems. Viner’s
statement suggests that there is no need for a tight,
specific definition of the subject, which may explain
the tendency of economists blithely to ignore
definitions, and hence to not analyse them in detail,
except sporadically. However, definitions of the
subject do have effects through influencing what
economists choose to study and the methods they
think legitimate for analysing them.
The root of the word ‘economics’ lies in the
Greek oiKovopia, meaning the management of a
household, as in Xenophon’s OiKovopiKoq, written
around 400 BC. In the 18th century, the idea of
efficiently providing for the wants of a household
was extended to the nation as a whole, under the
heading ‘political economy’, the term first used for
the discipline that later became economics. The first
systematic English-language book on the subject
was James Steuart’s An Inquiry into the Principles
of Political Oeconomy ( ,
p. 16). Though Steuart made an analogy between
‘providing for all the wants of a family, with
prudence and frugality’ and doing the same for the
state, there was a difference, for the ruler of the state
could not direct people in the way that the head of a
household was able to do. This had the consequence
that,
The great art therefore of political oeconomy is, first
to adapt the different operations of it [the state] to
the spirit, manners, habits and customs of the
Economics, Definition of
people; and afterwards to model these circumstances so,
as to be able to introduce a set of new
and more useful institutions. (Steuart , p. 16)
No doubt influenced by German Cameralism,
Steuart saw institutional design as lying at the heart
of political economy. This usage was followed by
Adam Smith, who saw political economy as ‘a
branch of the science of a statesman or legislator’
with two objects: providing the people with
‘plentiful revenue or subsistence’ and providing the
state with enough revenue to provide public services
(Smith , p. 428).
Many of the classical economists, however,
disagreed with the focus on policy, arguing that
political economy was concerned with the laws that
govern the production, distribution and consumption
of wealth, the clearest example of this being Jean
Baptiste Say, whose major work ( ) human character was of crucial importance: it was
is Traite d’economie politique, ou simple exposition
de la maniere dont seforment, se distribuent et se
consomment les richesses (A treatise on political
economy, or a simple account of the way in which
wealth is formed, distributed and consumed). This
definition formed the basis for Nassau Senior’s
Outline of the Science of Political Economy ( ) in
which he argued that the science was
based on four propositions, the first and most
important of which was ‘That every man desires to
obtain additional Wealth with as little sacrifice as
possible’ (Senior , p. 26).
Neither of these definitions was acceptable to
John Stuart Mill, whose ‘On the Definition of
Political Economy; and the Method of Investigation
Proper to It’, first published in 1836, was the last of
his Essays on Some Unsettled Questions of Political
Economy (1844). To define political economy as the
rules for making a nation rich was to confuse ‘art’
and ‘science’. However, it was not enough to define
it as the laws relating to the production and use of
wealth, for these included many physical laws that
lay outside its remit. He thus favoured a more
limited definition: ‘The science which treats of the
production and distribution of wealth, so far as they
depend upon the laws of human nature’ or ‘The
science relating to the moral or psychological laws
of the production and distribution of wealth’ (Mill ,
3437
p. 318). Mill went on to argue that even this
definition was too broad, for political economy
related only to man in society.
The most significant challenge to this definition
of political economy as, loosely, the science of
wealth, came from Alfred Marshall, who offered the
well-known definition:
Political Economy or Economics is a study of man-
kind in the ordinary business of bfe; it examines that
part of individual and social action which is most
closely connected with the attainment and with the
use of the material requisites of wellbeing. (Marshall
, p. 1)
This definition is significant not so much for
changing the name of the discipline to economics as
for its focus on the study of mankind. For Marshall,
as for many of his generation, the evolution of
important to study actual human behaviour, but it
was important, especially in the longer run, to
consider how activities and consumption served to
influence character and hence behaviour. Wants
could not be taken as given but depended on
activities.
In these discussions there was, as Neville
Keynes pointed out, an ambiguity in the use of the
word ‘economic’. On the one hand it referred to
attaining an end ‘with the least possible expenditure
of money, time and effort’ (Keynes , pp. 1-2) whilst
on the other hand it was used as an adjective
corresponding to the noun, wealth. The economists
who laid most emphasis on the first of these were
the Austrians - Carl Menger and his successors -
who focused on economizing behaviour. It was his
familiarity with this literature that led Lionel
Robbins to deny originality for his much-quoted
definition, ‘Economics is the science which studies
human behaviour as a relationship between ends
and scarce means which have alternative uses’
(Robbins , p. 16). Robbins’s definition put scarcity
and choice at the centre of economic analysis. He
emphasized that ‘any kind of human behaviour’ that
demonstrates the scarcity aspect falls within the
scope of economics, and that there are ‘no
limitations on the subject-matter of Economic
Science’ beyond involving ‘the relinquishment of
other desired alternatives’ (choice) ( , p. 17).
3438
The significance of this definition lies in its
analytical nature: instead of defining economics in
terms of its subject matter, it defines it as an aspect
of behaviour. In spite of Robbins’s claim that he
was simply describing professional practice, the
initial reaction of the profession to his definition of
economics, at least as it surfaced in academic
journal articles and introductory textbooks (where
the definition of economics was primarily
discussed), was negative (for a detailed discussion,
see Backhouse and Medema ). Throughout the
1930s and 1940s, textbook writers continued to
define economics in terms more reminiscent of Mill
and Marshall than Robbins, in that, even where
reference was made to scarcity, this was frequently
qualified: economics was described as a social
science concerned with the study of wealth, of
earning a living or a study of the system of free
enterprise. Robbins’s choice- based definition was
seen as too wide, and needed to be restricted so as
to rule out matters that did not come within the
‘traditional’ boundaries of economics. The
acceptance of the Robbins definition came
piecemeal. First, scarcity came to be stressed as
important to the subject. The first edition of Paul
Samuelson’s Economics ( ),
undoubtedly the leading textbook in the post-war
period, captures well the qualified attitude with
which the Robbins definition was approached.
Samuelson explained that economics was about
scarcity, for ‘the American way of life’ required
more resources than were available, but he chose to
define the subject in terms of ‘what’, ‘how’ and ‘for
whom’ - that is, as concerning the production and
consumption of goods and services. There is
nothing here that is inconsistent with Robbins, but
this approach was equally consistent with a more
traditional approach. Books such as George
Stigler’s Theory of Price ( ), which adopted
the Robbins definition, laid great stress on both
scarcity and choice, but others carefully refrained
from doing so.
It was only in the late 1950s and 1960s that the
use of Robbins’s definition became widespread. By
the late 1960s, Samuelson’s Economics was
claiming that economists agreed on ‘a general
definition something like the following’:
Economics, Definition of
Economics is the study of how men and society
choose, with or without the use of money, to employ
scarce productive resources, which could have
alternative uses, to produce various commodities
over time and distribute them for consumption, now
and in the future, among various people and groups
in society. (Samuelson , p. 5).
However, support for this was still not universal.
For Richard Lipsey, whose Introduction to Positive
Economics was one of the most success- fill rivals
to Samuelson’s Economics, scarcity was ‘one of the
basic problems encountered in most aspects of
economics’, not the entire subject (Lipsey /71, p.
50). Economics also dealt with questions related to
failure to achieve a point on the production
possibility frontier, such as explaining
unemployment, which could not be reduced to
problems of scarcity.
The move by Robbins to define economics as an
aspect of behaviour made it just a short step to
defining economics in terms of a method - that of
rational choice-which could be applied not simply
to production and consumption choices, but to all of
human behaviour. This move was encouraged by
the tendency, in the aftermath of the Second World
War, to see economics though the lens of operations
research, as social engineering, in which
optimization techniques were central and game
theory played a significant role. It has also been
argued that this move towards emphasising rational
choice had ideological attractions during the Cold
War. During the 1960s, economics became
increasingly conceived as the ‘science of choice’,
without reference to a particular social domain,
even, at times, without reference to scarcity: the
subject could encompass non-market as well as
market activities. The work of Theodore Schulz and
Gary Becker on human capital, James Buchanan,
Anthony Downs and Gordon Tullock on political
processes, and Becker on discrimination and on
crime and punishment laid foundation for what
came to be called ‘economics imperialism’, the
application of economics to fields including politics,
law, history, and sociology. These theoretical moves
were reinforced by advances on the empirical side,
where the techniques developed by, for example,
James Heckman and Daniel McFadden for
analysing
Economics, Definition of
cross-section data sets on individuals and house-
holds were used to investigate phenomena, such as
non-marital fertility, that lie outside the traditional
domain of economics as concerned with market
behaviour.
Robbins’s definition of economics in terms of
the allocation of scarce resources remains the most
widely cited definition of the subject, but it has
never commanded universal assent Though scarcity
can be defined in such a way as to make it true,
there have always been significant numbers of
economists who have considered that it does not
encompass all aspects of their discipline and that
qualifications or extensions are required. These
result in definitions closer to those found in the
19th-century literature, focusing on phenomena
such as the production and distribution of wealth. At
the other end of the spectrum, there are economists
for whom rational choice is more fundamental than
scarcity. To this extent, then, there is no universally
agreed upon definition of the subject.
The reason this does not present a problem is
that economists can proceed with their work
irrespective of how their subject is defined. Defi-
nitions of fields generally come only after the field
is established; as fields change, so definitions
change. Despite this, however, definitions can
matter. As Mill recognized, questions of method
and definition are linked. The clearest example of
this is Robbins, who sought to derive all the main
propositions of economics from the premise of
scarcity. His definition, therefore, was the basis for
claiming that economic theory was central to
economics - that it was far more important than
Marshall had believed it to be. Also significant was
his reference to economic science, for the word
science is far from neutral. Robbins had argued that
value judgements, including those necessary to
make interpersonal welfare comparisons, did not
come within the scope of economic science, but
belonged instead to the realm of ‘political
economy’. In claiming this, he was arguably
attempting to clarify the status of economists’
arguments, for, as he later made very clear, offering
any advice on economic policy requires such value
judgements. Thus if economics includes
3439
policy advice it must encompass more than eco-
nomic science as Robbins defines it. However, such
is the prestige of ‘science’ that Robbins’s definition
caused many economists to try to dispense with
value judgements altogether, even in welfare
economics. An exercise in clarification (and no
doubt a critique of certain views of the subject) thus
had the effect of significantly narrowing the subject.
Attempting to define economics thus was not and is
not simply a descriptive exercise; it has
consequences for what economists do, and how
they go about doing it.
See Also
► Acbbins, Lionel Charles (1
► ICIMTI, ul An
► J : States, Economics in (194;
Bibliography
Backhouse, R.E., and, S.G. Medema. 2007. Defining eco-
nomics: Robbins’s definition in theory and practice.
SSRN working paper. Abstract online. Available at
. Accessed 19 May
2007.
Keynes, J.N. 1891. The scope and method of political
economy. London: Macmillan.
Lipsey, R.G. 1963. An introduction to positive economics, 3rd
ed. London: Weidenfeld & Nicolson, 1971.
Marshall, A. 1890. Principles of economics. London:
Macmillan, 1949.
Mill, J.S. 1844. Essays on some unsettled questions of
political economy. In Collected works of John Stuart
Mill, ed. J.M. Robson. Toronto: University of Toronto
Press, 1967.
Robbins, L.C. 1932. An essay on the nature and significance
of economic science, 2nd ed. London: Macmillan, 1935.
Samuelson, P.A. 1948. Economics. New York: McGraw Hill.
Samuelson, P.A. 1967. Economics, 7th ed. New York:
McGraw Hill.
Say, J.-B. 1803. Traite d’economie politique, ou simple
exposition de la maniere dontseforment, se distribuent et
se consomment les richesses. Paris: Deterville.
Senior, N. 1836. An outline of the science of political
economy. New York: Augustus Kelley, 1965.
3440 Economies and Diseconomies of Scale

Smith, A. 1776. An inquiry into the nature and causes of the
wealth of nations, 2 vols. Indianapolis: Liberty Press,
1976.
Steuart, J. 1767. An inquiry into the principles of political
oeconomy, 2 vols. Edinburgh: Oliver & Boyd, 1966.
Stigler, G.J. 1946. The theory of price. New York: Macmillan.
Iffis a strictly concave function and/(0) > 0, or
if/is homogeneous of degree less than one, then
decreasing returns to scale prevail. Conversely,
homogeneity of degree greater than one is a suffi-
cient condition for increasing returns to scale.

Internal and External Economies and

Diseconomies
It is sometimes useful (see, e.g., section, “
” below)
Economies and Diseconomies of to consider economies of scale that appear only at
Scale the aggregate level and not at the level of the
individual firm. For example (see Chipman ) let
Joaquim Silvestre
there be two firms with cost functions C/yJ) = kjyj, j
= 1, 2. Firm j treats kj as a parameter, and in this
sense its technology displays constant returns to
scale. But suppose that kj actually depends on the
Conceptual Issues amount of output of the other firm, say kj = \yif.
Then the aggregate cost is lyffyi + \yifyi' We have
Definitions external economies if /? < 0 and external
We consider the unit costs of producing a (single or diseconomies if /i > 0.
composite) output under a given technology (no
technical change). We say that there are economies Explaining Diseconomies and
(or diseconomies) of scale in some interval of Economies of Scale
output if the average cost is decreasing (or We consider diseconomies first. Decreasing returns
increasing) there. This definition focuses on imply that duplicating all inputs yields less than
economies and diseconomies of a technical char- twice the amount of output. But an exact clone of a
acter. It is sometimes extended to cover business production process that exhaustively lists all factors
activities other than production (such as marketing, of production should give exactly the same output.
financing, training: see Scherer ). The failure to double the output suggests the
Note that, in the case of a composite output, the presence of an extra input, not listed among the
proportions among the goods produced are kept arguments in the production function, that cannot be
constant. (A different notion, that of ‘economies of duplicated. This idea goes back to Ricardo’s rent as
scope’ contemplates variations in cost as the output based on the impossibility of duplicating agricultural
mix varies.) The definition of cost may, on the other land of a given quality. Alternatively, the extra input
hand, imply that the input proportions are adjusted can be interpreted as managerial skill.
in order to minimize expenditures. A related idea is Consider (see McKenzie ) a strictly concave
that of returns to scale: here both the output and production function fix), where x is an L-
input proportions are kept fixed, and one compares dimensional input vector. One can associate to it a
the amount of (the simple or composite) output y(x) constant returns to scale technology with L + 1
produced by a given input vector x with the amount inputs F(x; z) and a fixed level of the extra input,
produced by vector Xx, for X > 1. Increasing (or say z = 1, such that fix) = F(x; 1), i.e., / describes the
decreasing, or constant) returns to scale are said to amounts of output obtainable by varying the first L
prevail if fiXx) is greater than (or smaller than, or inputs when the ‘managerial skill’ is kept at the
equal to) Xfix). Under some conditions (see, e.g., constant level z = 1. To this end, define F(x; z) =
Fuss and McFadden , p. 48) increasing (or zfix/z). It is easy to check that F is quasiconcave and
decreasing) returns to scale are equivalent to homogeneous of
economies (or diseconomies) of scale.
Economies and Diseconomies of Scale
degree one, i.e., constant returns to scale. Moreover,
competitive profits can be viewed as the
competitive reward to the ‘managerial skill’ (at z =
1, z(dF/dz) = fix) - V/(x) • x).
A similar notion can be applied to the case of
external diseconomies of scale: the extra input can
then be identified with a common pool resource
(say, clean water), available in a limited amount.
Conversely, the extra public input may be created
by the activity of the industry (say, information or
specific training of the labour pool): this will
generate external economies of scale.
We turn now to internal economies of scale.
Koopmans ( ) reviews some controversies on
this issue and remarks (p. 152 fn.3), ‘I have not
found one example of increasing returns to scale
where there is not some indivisible commodity in
the surrounding circumstances.’ The following
ideas have appeared in the literature.
(a) Indivisible input. Assume for instance that the
only input is some specific capital good (a
machine, plant, ship or pipeline) which is
indivisible in the sense that it becomes useless
if physically divided. It has a given maximal
capacity y, but it can be underutilized to pro-
duce amounts of output less than y. Then C(y)
looks like Fig. , and there are economies of
scale in each of the intervals [0, }’ ], IJ. 2y ],K, (d) Adam Smith’s division of labour. The Wealth
[(/? - 1 )y, ny ], K.
(b) Set-up cost. Take the only input to be labour
time and assume that a certain amount of time
has to be spent in preparation for the task (the
Economies and c(y)
Diseconomies of
Scale, Fig. 1
3441
set-up cost can be given several interpretations,
as time spent in: (1) concentrating and getting
psychologically ready for the task; (2) learning
how to do it; (3) preparing the tools needed).
Once the set-up cost is paid, the amount of
output is proportional to the extra labour spent.
This looks like Fig. , where increasing returns
to scale prevail. Set-up costs can here be
viewed as a form of indivisibility: ‘readiness’
(or ‘information’ or ‘preparation’) is indi-
visible: a ‘half-ready’ worker is useless.
(c) The above examples can be extended to more
than one capital good (or type of set-up cost).
Consider, for instance, pipelines ten miles long.
Only metal sheet is used in their production:
the amount needed is proportional to the radius
of the pipeline. Output y (flow of oil between
two points ten miles apart) is proportional to
the section area, a quadratic function of the
radius. A pipeline of a given radius is
indivisible, but one can build pipelines of any
radius. The cost function looks like Fig. . The
vertical coordinate can be interpreted as the
minimal dollar outlay of a firm that buys pipe-
lines and sells y, or as the amount of the input
‘square yards of metal sheet’ used by a verti-
cally integrated firm that produces its own
pipelines and sells output y.
of Nations attributes to the ‘division of labour’
the increase in output per worker. The main
argument seems to be based on the set-up costs
of (b) above. Smith’s notion is related
3442 Economies and Diseconomies of Scale

Economies and C(y)

Diseconomies of
Scale, Fig. 2

Economies and C(y)

Diseconomies of
Scale, Fig. 3

to another fundamental idea: the Ricardian be at equilibrium at a level of output at which the
gains from specialization and trade. But, in average cost faced by the firm is decreasing.
Arrow’s ( ) words, ‘the Ricardian idea of This in particular implies that competition can-
specialization lacks some characteristics of the not prevail under the presence of internal econo-
Smithian; in Ricardo’s system the abilities to mies of scale at all levels of output. The argument
produce are given. In Smith’s view, special- allows for economies of scale which are external to
ization is more a matter of deliberate choice.’ the competitive firm. (But the laissez- faire equi-
librium will then typically be suboptimal.) The idea
of constant (or increasing) average cost at the firm
Economies of Scale and Market Structure level but of decreasing average cost at the aggregate
level (industry economies of scale) did play an
Perfect Competition as Price Taking Behaviour important role in the controversies on the
A perfectly competitive firm is often defined as one compatibility between economies of scale and
that faces a horizontal demand curve. It is clear that, competition (see Chipman ). This idea is illustrated
as long as C(0) = 0, no such firm can in the example of section, “
Economies and Diseconomies of Scale
Economies and P
Diseconomies of
Scale, Fig. 4
above: the reader is referred to Chipman ( ) none will exist if the cost curves are U-shaped. The
for a rigorous study.
We now focus on internal economies. Let the
market demand curve be as in Fig. where the
average and marginal curves of a typical firm are
also drawn. This situation does not per se violate
the price taking rule: one could have, for instance, a
single price taking firm operating aty 1. But such a
combination of demand and cost does not fit well
with the idea of perfect competition.
First, it is graphically clear that at most two
firms may operate in this market. But it is then
unrealistic to assume that each firm will take the
market price (or the price charged by the other
firm) as given.
Second, with two firms the aggregate supply
curve would look like the discontinuous curve in
Fig. , where supply at no price equals demand.
Difficulties with the existence of competitive
equilibrium will in general appear as soon as the
average cost is somewhere decreasing.
These difficulties become more severe when
considering unrestricted entry (with identical cost
curves for all incumbents and entrants), a natural
attribute of perfect competition: it is then required
that at a ‘long run’ equilibrium no potential entrant
have incentives to enter. But if potential entrants
are themselves price takers, a long-mn equilibrium
implies zero profits, and typically
3443
entry model of Baumol et al. ( ) faces the
same existence difficulties.
Explaining the Number of Firms
The previous discussion suggests that imperfect
competition will prevail under economies of scale.
On the other hand, one would expect the number of
firms in an industry to be inversely related to the
degree of scale economies relative to the extent of
the market. Novshek’s ( )
approach yields a rigorous version of this idea.
Novshek considers a model of Cournot oligopoly
with free entry where potential entrants adopt
themselves the Cournot conjecture that active firms
will keep their output constant (as proposed by Bain
; Sylos-Labini ).
An example will illustrate Novshek’s method.
Consider a market where the aggregate (inverse)
demand is given by p = a — b Y. The number of
firms in the industry is not given, but all firms,
incumbent or potential, have access to the same cost
function: C(y) - g + cy. The positive parameter g is a
set-up cost: the larger g, the stronger the economies
of scale. Similarly, the larger a or the smaller b, the
larger the extent of the market.
Write y(n) = (a — c)/(n + 1 )b. This is the out-
put of a firm at the (unique and symmetric) Cournot
equilibrium for n active firms. For this
3444
Economies and P
Diseconomies of
Scale, Fig. 5
to be a Novshek (or Long-run Cournot) Equilibrium
we require that: (a) the price be not less than
average cost; this is a no exit condition, (b) No
potential entrant has incentives to enter, i.e., a —
b(n x (n) - y) < c - g/y for all y > 0: this is a no entry
condition.
It can be checked that (a) and (b) impose lower
and upper bounds on the number of firms n that can
prevail at a Novshek Equilibrium. The no exit
condition implies that n + 1 < (a - c)M(bg), and the
no entry condition implies that n + 1 > (1/2) (a —
c)/xj(fyg). The expression (a — c)/xj(bg) can be
viewed as an index of the ‘extent of the market
relative to the scale economies’, since it is
increasing in a and decreasing in b andg. Both
bounds increase with this index, and in this sense
the number of firms in an industry increases with it.
Perfect Competition as a Limit
Section, “
” above discussed the existence diffi-
culties that appear when production functions are
not concave. These difficulties, serious for com-
petitive equilibrium, are attenuated when consid-
ering noncompetitive equilibria: existence results
for general equilibrium models can be found in
Arrow and Hahn ( ) and Silvestre ( , out). But Guesnerie and Hart also show that
Economies and Diseconomies of Scale
). We focus now on partial equilibrium mar-
kets of the Novshek type (as in section, “ ”
above, but
perhaps with U-shaped costs). Consider a sequence
of such markets each with the same technology but
with increasing size of the consumer sector (say,
the parameter b of section, “ ” above tends
to zero). Equilibria turn out to exist at least for all
but a finite number of markets in such a sequence.
Moreover, the equilibria of such a sequence
converge to an optimal state (zero welfare loss)
where the price equals the marginal cost. Such a
limiting state can motivate an alternative definition
of a (long-run) competitive equilibrium: this
approach has the virtue of providing a justification
(from the noncooperative, Cournot viewpoint) of
the price taking postulate (see Mas-Colell , ).
The convergence to long-mn competitive equi-
librium obtains both in the case of U-shaped cost
curves and in the case of everywhere decreasing
average costs (see Guesnerie and Hart ). This
suggests a certain degree of compatibility between
increasing returns and competition, in the sense that
if the economy is sufficiently large, the price will
be approximately equal to marginal cost in either
case (even when price taking behaviour is ruled
Economies and Diseconomies of Scale
the per capita welfare loss tends to zero much more
rapidly in the U-cost case. Thus, in their words,
‘there remains a sense in which everywhere
increasing returns do cause greater problems for the
competitive model than do increasing returns which
are eventually exhausted’ (p. 541).
Normative Analysis
Consider a commodity that is produced with econ-
omies of scale and that is sold in a market at a
uniform price. Efficiency requires that price be
equal to marginal cost. But the marginal cost is
lower than the average cost. Hence, efficient pricing
requires that the producing firm suffer losses. This
is a basic obstacle to efficiency under increasing
returns to scale. (When the commodity is not easily
transferable among buyers efficiency can
sometimes be achieved by means of price
discrimination or nonlinear pricing schedules.)
One institutional arrangement that can in prin-
ciple resolve the conflict is the public ownership of
the firm. The firm can then be instructed to set
prices equal to marginal costs and be subsidized for
the resulting losses (see Hotelling ). This motivates
the concept of Marginal Cost Pricing Equilibrium
(see Guesnerie ; Beato ), an extension of the notion
of general competitive equilibrium where firms
with increasing returns to scale must follow the
marginal cost pricing rale instead of profit
maximization. Any efficient allocation can be
attained as a marginal cost pricing equilibrium for
some redistribution of income. But setting prices
equal to marginal cost only guarantees the first
order conditions for efficiency, not sufficient here.
Thus, one should not expect that all marginal cost
pricing equilibria will be efficient. A weaker
desideratum is the existence of at least one efficient
marginal cost pricing equilibrium compatible with a
given income distribution. This turns out to obtain
in some special cases (e.g., when there is a
representative consumer) but not in general (see
Guesnerie ; Brown and Heal , ; Beato and
Mas-Colell ).
There are, on the other hand, practical obstacles
to the achievement of efficiency by a publicly
3445
owned firm. First, the absence of a profit maximiza-
tion target may reduce the incentives for cost mini-
mization. Second, the implied redistribution from
taxpayers to buyers may be ethically objectionable
if, say, only the wealthy are buyers. Or the public
firm may find itself legally or politically
constrained to break even, because it may in
practice be hard to distinguish the losses mandated
by marginal cost pricing from those caused by
mismanagement. First best efficiency is then
unattainable.
An interesting second best problem for a pub-
licly owned or regulated firm with economies of
scale is the following one (see Ramsey ; Boiteux ).
Let the firm be constrained to break even, and let it
sell the same commodity in two separate markets.
Efficiency would require setting the prices in both
markets equal to the (common) marginal cost, but
this violates the break-even constraint. The second
best solution requires charging different prices: the
market with less elastic demand is charged a higher
price.
See Also
Bibliography
Arrow, K J. 1979. The division of labor in the economy, the
polity and society. In Adam Smith and modem political
economy, ed. Gerald P. O’Driscoll Jr. Ames: Iowa State
University Press.
Arrow, K.J., and F.H. Hahn. 1971. General competitive
analysis. San Francisco: Holden-Day.
Bain, J.S. 1956. Barriers to new competition. Cambridge,
MA: Harvard University Press.
Baumol, W.J., J.C. Panzar, and R.D. Willig. 1982. Con-
testable markets and the theory ofindustiy structure. New
York: Harcourt Brace Jovanovich.
Beato, P. 1982. The existence of marginal cost pricing with
increasing returns. Quarterly Journal of Economics 97:
669-689.
3446
Beato, R, and A. Mas-Colell. 1985. On marginal cost pricing
with given tax-subsidy rules. .Journal of Economic
Theory 37: 356 365.
Boitcux, M. 1956. Sur la gestion des monopoles publics
astreints a Tequilibre budgetaire. Econometrica 24: 22
40.
Brown, D.J., and G. Heal. 1979. Equity, efficiency and
increasing returns. Review of Economic Studies 46: 571
585.
Brown, D.J., and G. Heal. 1980. Two-part tariffs, marginal
cost pricing and increasing returns in a general
equilibrium model. .Journal of Public Economics 13: 25
49.
Chip man, ,I.S. 1965. A survey of the theory of international
trade: Part 2, the neo-classical theory. Econometrica 33:
685 760.
Chip man, J.S. 1970. External economies of scale and com-
petitive equilibrium. Quarterly Journal of Economics
84(3): 347-385.
Fuss, M., and It. McFadden (eds.). 1978. Production eco-
nomics: .1 dual approach to theory and applications.
Amsterdam: North-1 lolland.
Guesncric, R. 1975. Pareto optimality in non-convex econ-
omics. Econometrica 43: 1-29.
Guesncric, IT, and O. Hart. 1985. Welfare losses due to
imperfect competition: Asymptotic results for
Cournot Nash equilibria with and without free entry.
Internationa! Economic Review 26(3): 525-545.
Hotelling, II. 1938. The general welfare in relation to
problems of taxation and of railway and utility rates.
Econometrica 6: 242-269.
Koopmans, T.C. 1957. Three essays on the state of economic
science. Now York: McGraw-Hill.
McKenzie, L. 1959. On the existence of general equilibrium
for a competitive market. Econometrica 27: 54-71.
Mas-Colell, A. (ed.). 1780. Noncooperative approat lies to
the theory of perfect competition. Symposium issue,
Journal of Economic Theoty. Reprinted, Now York:
Academic Press, 1782.
Mas-Colell, A. 1781. Coumotian foundations of Walrasian
equilibrium theory: An exposition of recent theory. In
Advances in economic theory’, ed. W. Hildenbrand.
Cambridge: Cambridge University Press.
Novshek, W. 178(1. Cournot equilibrium with free entry.
Review of Economic Studies 47: 473-486.
Ramsey, F. 1727. A contribution to the theory of taxation.
Economic Journal 37: 47-61.
Scherer, F.M. 178(1. Industrial market structure and eco-
nomic performance, 2nd ed. Boston: Houghton Mifflin.
Silvestre, .1. 1777. General monopolistic equilibrium under
non-convexities. International Economic Review 18: 425
434.
Silvestre, .1. 1778. Increasing returns in general
non-competitive analysis. Econometrica 46(2): 397 4(12.
Sylos-Labini, P. 1962. Oligopoly and technical progress.
Trans. Elizabeth I lenderson, Cambridge, MA: I Harvard
University Press.
Economy as a Complex System
Economy as a Complex System
Alan Kirman
Abstract
Complex systems are composed of particles or
agents which interact directly with each other.
The rules for this interaction may be very
simple and may not reflect the sort of rationality
associated with standard economic models.
Interaction is not through some exogenously
given market, nor does it depend on the
complicated reasoning involved in game theory.
A complex system exhibits emergent aggregate
properties as it organizes itself and these can
explain important phenomena such as bubbles,
herding behaviour, and segregation. In each
case the aggregate state of the economy or
market could not be predicted from the average
behaviour of the individuals.
Keywords
Agent-based models; Bubbles; Chaos; Clusters;
Complex systems; Complexity; Computational
complexity; Coordination; Deterministic chaos;
Economy as a complex system; Econophysics;
Emergence; Equilibrium; Ergodicity; Evolution;
Excess volatility; Financial market contagion;
Forecasting; Game theory; Herding; Imperfect
competition; Law of large numbers; Minority
game; Neighbourhood effects; Prisoner’s
Dilemma; Punctuated equilibria; Residential
segregation; Social interaction; Social networks;
Statistical mechanics; Statistical physics; Steady
state; l it for tat; Tournaments
JEL Classifications
D85
Introduction
The term ‘complex system’ has been widely used in
science and many different definitions have
Economy as a Complex System
been given. Frequently, rather than give a definition
of such a system, scientists have fallen back on
certain characteristics that these systems exhibit.
For example, emergence, selforganization,
synergetics, collective behaviour, and non-
equilibrium have all been cited in this regard. It is
useful at the outset to make the distinction between
‘complexity’ and ‘complex system’. The former
involves a number of ideas which are important in
economics and which are inherited from computer
science but which will not be dealt with here. In
particular, the notion of computational complexity,
as it applies to decision-making or to the
computation of equilibria or of dynamic
programming problems is central to certain aspects
of economic theory. However, here the discussion
will turn on the idea of the economy as a complex,
adaptive, evolving system. For economists the first
real incarnation of this approach was with the intro-
duction of deterministic chaos. The idea of complex
dynamic behaviour, which would not explode or
cycle or converge to a steady state, was fascinating
for a science long dominated by the ideas of
convergence to a static equilibrium or to a steady
state. Jean-Michel Grandmont ( ) that agents react to signals from some central
developed a simple model of ‘business cycles’
involving the ‘tent map’, which gave rise to such
chaotic behaviour. Apart from the idea of the
complicated dynamics involved, it was clear that the
fact that a small perturbation in the initial conditions
governing such a process could produce radically
different trajectories was also of great intellectual
interest. Two important innovations were involved.
Firstly, there was the idea that the economy should
be thought of as a truly dynamic system and that the
initial conditions of such a system might play a key
role. Secondly, there was the idea that there might
be no continuity in the dependence on those initial
conditions and that small changes might radically
influence the trajectory of the system; hence the
famous allusion to the influence of the fluttering of
a butterfly’s wing on the world’s weather. These
two aspects led economists to focus their attention
on deterministic chaos. Yet in making such a close
link between complexity and chaos, economists
may have lost sight of the
3447
broader implications of complexity for the analysis
of economic systems.
To see why this is so, consider what sort of
systems are referred to as ‘complex’ in other dis-
ciplines. Typically they have some, or all, of the
following characteristics:
• The agents are heterogeneous and interact
directly with each other.
• The interaction and the information of agents are
‘local’.
• The agents’ behaviour is governed by simple
‘rules of thumb’.
• The aggregate behaviour of the system is not that
of an ‘average’ or representative agent.
• This aggregate behaviour ‘emerges’ from the
complicated interaction between the individuals.
To someone who has not studied theoretical
economics, all of these characteristics might seem
rather intuitive as features of an economy. Yet they
are very different from the traditional view. In that
view, the economy is a system in which the only
interaction is through the market. By this it is meant
authority such as an auctioneer. In some way the
central prices adjust so as to coordinate the
activities of the agents. The system adjusts in this
way until the activities are coordinated - for
example, in a market economy, until aggregate
demand for all products is equal to the aggregate
supply of those products. Once this is achieved, the
signals will not change and no agent has an
incentive to modify his behaviour and to deviate
from this ‘equilibrium state’.
This description reveals another important fea-
ture of the collective model. No agent takes account
of any influence that he might have on the outcome
of the system. Many economists will react to this
description by arguing that models of ‘imperfect
competition’ abound, and in these models agents
take into account the impact of their actions on the
state of the system and know that other agents do
the same.
This brings us to a second view of the economy,
that based on game theory. Here all agents take
account of the reciprocal impacts of their
3448
actions and know that all the other agents do the
same. This view is very different from the basic
model of the economy. However, it is also very
different from that of a complex system, since it
attributes unlimited calculating capacity and depth
of reasoning to the agents.
The vision of the economy as a complex system
falls between these two approaches. It requires
neither the central coordinating mechanism of the
competitive market, nor the analytically
sophisticated players of game theory.
A good comparison might be between, on the
one hand, an economy organized as a set of markets
that are open simultaneously, each with an
auctioneer and, on the other, an ants’ nest. In the
former there are structured central price-giving
mechanisms, and the actors gradually reveal their
willingness or non-willingness to pay until the
goods are allocated efficiently. In the latter, the
individuals pursue their own different activities and
react to each other and to outside stimuli. The
system organizes itself but there is no central
mechanism for achieving such organization. No one
would think of trying to describe the activity of an
ants’ nest by examining the behaviour of the
‘representative ant’ yet many would describe the
allocation of effort and resources as ‘efficient’.
The sort of system that could be described as
complex in the sense outlined above can be physical
or biological or social. A typical reaction to the use
of physical or biological analogies in economics or
other social sciences is that social systems are
populated by individuals that have intentions and
undertake purposeful activity, while the other
systems are composed of purposeless molecules or
particles. It is therefore argued that the sort of
analysis that can be applied to the other systems is
not pertinent to the analysis of economic systems.
This reasoning does not stand up to close
inspection. If individuals follow well- defined mles
and their interaction is well specified, the simple
models that are used in physical and biological
models can be applied. Precisely why the
individuals should follow these mles is a different
question.
Why is the complex systems approach of par-
ticular interest currently to economists? Economic
theory has recently been attacked on two fronts.
Economy as a Complex System
The first is the problem of aggregation: how is the
behaviour of the economic system related to that of
the individuals that make it up? The second is the
question of why individuals behave as they do. The
answer to the first question is simple but
undermines much of modem macroeconomics that
is based on the idea that the behaviour of the
aggregate can be treated as the behaviour of an
individual. Yet what is known is that the standard
model of a system composed of isolated individuals
each solving his own maximizing problem does not
allow one to treat the system as an individual. (This
is not the place to enter into the details of this
assertion but the basic argument is given in
Kirman , and stems from the results of
Sonnenschein, Mantel and Debreu). The second
question is that posed by behavioural economics
that questions the idea of the isolated maximizing
individual. Ideas from Simon ( ) onwards
have suggested that individuals reason in a limited
and local way. Experiments, observation, and
examination of the neural processes utilized in
making decisions all suggest that homo economicus
is not an accurate or adequate description of human
decision making. (For a good survey of the relevant
literature, see Rabin ).
All of this suggests that one might want to take a
very different view of how the economy functions.
In particular, the notion of a complex system as
used in many parts of science seems to correspond
well to an intuitive vision of the economy. Just as in
an ants’ nest individuals perform tasks without
having any idea of the behaviour of the system,
individuals in an economy go about their business
and achieve a remarkable degree of coordination.
Take a simple example that of bees in a hive. The
tasks for house bees are varied but temperature
control is one of the important duties. When the
temperature is low, bees cluster to generate heat for
themselves, but when it is high some of them fan
their wings to circulate air throughout the hive. The
general hive temperature required is between 33°
and 36°C, while the brood chamber requires a
constant heat of 35°. Honey has to be cured in order
to ripen, and this also requires the help of
circulating air. According to Crane ( ),
12 fanning bees positioned across a hive entrance
25 cm wide can produce an air flow amounting to
Economy as a Complex System
50-60 litres per minute. This fanning can go on day
and night during the honey-flow season. Hon-
eybees’ wings beat 11,400 times per minute, thus
making their distinctive buzz.
What is the lesson here for us? The typical
economist’s response to this phenomenon would be
to consider a representative bee and then study how
its behaviour responds to the ambient temperature.
This would be a smooth function of temperature,
wing beats going up or down with the temperature.
Yet this is not what happens at all. Bees have
different threshold temperatures and they are either
on (beating at 11,400 beats per minute) or off. As
the temperature rises more bees join in. Thus
collectively with very simple 1, 0 mles the bees
produce a smooth response. This sort of
coordination, with each agent doing something
simple, can only be explained by having a
distribution of temperature thresholds across bees.
Aggregation of individuals with specific local and
differentiated behaviour produces smooth and
sophisticated aggregate behaviour.
Nobody would argue that, in social systems, all
coordination is achieved by simple interaction.
Markets make a powerful contribution to economic
coordination. Yet the important question is not
whether such mechanisms exist, but how they come
into being and develop and modify their mles. As
already explained, the idea that the existence of
such markets facilitates the allocation of resoinces is
clear and generally accepted. What is not so clear is
that the abstract idea of a market governed by
centralized prices which are adjusted to equilibrate
the market has any descriptive value. The idea of
markets and networks of communication and
transactions as emergent and changing phenomena
is much more persuasive.
Considering the economy in this light is far from
a new idea. When Adam Smith discusses the
‘invisible hand’ some of these notions are apparent,
Pareto’s work contains some of these ideas and
Hayek is perhaps he who was closest to this vision.
Schelling in his Micromotives and Macrobehavior (
) clearly foresaw the role of
self-organization. A recent development of these
ideas had an introduction on the formal level by
Foellmer ( ), who adopted the basic Ising
model. He posited a system in which individuals
3449
were situated in space and whose preferences were
dependent on those around them. This dependence
was stochastic, that is, the probability of having
certain preferences depended on the preferences of
an individual’s neighbours. If all the preferences are
independently drawn, then one can determine the
expected values of the equilibrium prices. However,
if the interdependence of the individuals is too
strong, this is no longer hue. The Taw of large
numbers’ no longer applies. There is no easy
transition from the micro to the macro level by
simple averaging.
Foellmer’s contribution was left to one side for
a long time. However, the complexity approach to
economics took on new life with the work at the
Santa Fe Institute of a number of economists,
physicists and other scientists such as Arthur, Bak,
Blume, Durlauf, Geanakoplos, and Holland. A good
picture of this sort of work can be found in The
Economy as an Evolving Complex System
(Anderson et al. ) and the two additional volumes
that followed it (Arthur et al. ; Blume and Durlauf ).
The emphasis on the increasing ‘socialization’
of economics, which is intrinsic to models of
interacting agents, permits one to introduce the
influence of neighbours and groups on individual
behaviour. Such an approach is standard in soci-
ology and anthropology but has remained a very
thinly populated field in economics. A good survey
of this work is to be found in Durlauf and Young (
).
One important part of the research on complex
systems in economics has been that on agent- based
models. Here the idea is to look at a set of linked
individuals whose behaviour is influenced by and
which influences their neighbours, and to simulate
the dynamics of that interaction. Perhaps the best-
known early example of this was Axelrod’s work on
the Prisoner’s Dilemma, which is summarized in
Axelrod ( ). He
started from a series of tournaments. The strategies
used for these were those that individuals proposed
for a repeated Prisoner’s Dilemma game. These
strategies were then played against each other in a
series of tournaments and the winning strategy
tinned out to be ‘tit for tat’, which is basically
cooperative.
3450 Economy as a Complex System

Axelrod was concerned that those who had
entered his tournament had already anticipated the
strategies that would be proposed by others. To
overcome this he ran simulations in which new
strategies were introduced into the pool of existing
strategies. To do this he assigned existing strategies
randomly to his artificial agents and then modified
them using a ‘genetic algorithm’. (For an
introduction to the theory and use of genetic
algorithms see Mitchell ). The set of strategies thus
evolved in two ways. After the strategies had played
against each other a new generation with more of
the successful strategies was created. To these were
also added new strategies generated by mutations
and crossovers from the current population. After a
while reciprocating strategies - that is, strategies
which respond to cooperation with cooperation but
which defect in the face of defection - took over,
giving high payoffs. Here we have a selection
process working on strategies that evolved rather
than were consciously chosen. The behaviour of
this basic but complex system - indeed, Axelrod
refers to himself as a complexity scientist - led to
the evolution of interesting aggregate
characteristics. In this context it is also interesting
to look at the work of Lindgren ( ), who also
allowed the
evolution of the strategy pool and generated periods
of stability in which one strategy dominated,
followed by periods of instability as the population
was invaded by another strategy. This corresponds
to the idea of ‘punctuated equilibria’ introduced into
evolutionary theory by Eldredge and Gould ( ).
The notion of evolution, which can also be
interpreted in the human or social context as
adaptive learning, is important here. We can think
of selection among a population of automata
endowed with single strategies or of the idea that
individuals learn to use more successful strategies.
In complex systems governed by local interac-
tions, it may be the case that as a result of some
perturbation there is a major change in aggregate
behaviour. This is an important idea which is central
to statistical mechanics. The idea here is that local
interaction can generate a rapid transition from one
‘phase’ to another of an economic system and, more
importantly, that one cannot simply apply the Taw
of large numbers’ to evaluate the impact of
stochastic shocks. An example of this is provided by
Bak et al. ( ), who consider
a model of ‘self-organized criticality’ to describe an
economy composed of a large number of productive
units, each supplying a limited number of customers
and, in turn, each supplied by a limited number of
suppliers; both customers and suppliers are located
near the productive unit.
The graph outlining the location of productive
units is a cylindrical lattice. In other words, each
production unit is supplied by the firms above it on
a vertical line and supplies the customers next to it
on a horizontal line. The demand for each final
good producer is characterized by stochastic
fluctuations, which affects the variability of orders
received by the suppliers. Such orders (and shocks)
are locally and vertically correlated, as every final
producer is supplied by the two upstream firms
situated a line up along the network representing the
productive system. In such a context, characterized
by local interaction, Bak et al. ( ) prove that, if
individual costs are
non-convex, the aggregation of small independent
individual shocks may lead to large aggregate
fluctuations in the productive system, breaking
therefore the law of large numbers. These small
shocks do not cancel each other out but are ampli-
fied by their interaction. Thus fluctuations at the
aggregate level cannot be explained by reducing the
whole model to one of an individual.

Phase Transitions Coordination: The Schelling Model

Recalling the characterization of complex systems Now let us pursue the discussion of the relationship
given above, it is worth considering a few between aggregate and individual behaviour. One of
examples. the important features of complex systems is that
the system can coordinate on a solution which could
not be predicted from a careful analysis of
Economy as a Complex System
the average or typical individual. In other words,
patterns at the aggregate level can emerge as the
individuals in an economy or market interact with
each other. The emergence of such aggregate pat-
terns cannot be forecast from the specification of
the individual characteristics. A good example of
this was provided by Tom Schelling at the end of
the 1960s (for a summary see Schelling ). He
introduced a model of segregation involving local
interaction, in the sense that peoples’ utility depends
on the race of their neighbours. He showed that,
even if people have only a very mild preference for
living with neighbours of their own colour, as they
move to satisfy their preferences complete
segregation will occur.
The basic model is very simple. Take a large
chess board, and place a certain number of black
and white counters on the board, leaving some free
places. A counter prefers to be on a square where
half or more of the counters in his Moore
neighbourhood, (the eight squares around him) are
of its own colour (utility 1) to the opposite situation
(utility 0). From the counters with utility zero, one
is chosen at random and moves to a preferred
location. This model, when simulated, yields
complete segregation even though people’s
preferences for being with then own colour are not
strong. Indeed, the result holds when individuals are
happy even when more than half of their neighbours
are of a colour different from their own. This result
was greeted with surprise and has generated a large
literature.
In fact, this result is not surprising and some
simple physical theory (see Vinkovic and Kirman ),
can explain the segregation phenomenon. Numerous
variants on Schelling’s original model have been
developed. In particular, the form of the utility
function used by Schelling, the size of
neighboiuhoods, the rules for moving, and the
amount of unoccupied space have all been studied
(see Panes and Vriend , for a survey). The physical
model encompasses all of these variants.
An attempt to provide a formal structure has
been made by Pollicott and Weiss ( ). They
however, examine the limit of a Laplacian process
in which individuals’ preferences are strictly
increasing in the number of like neighbours. In this
situation it is intuitively clear that there is a
3451
strong tendency to segregation. Yet Schelling’s
result has become famous because the preferences
of individuals for segregation were not particularly
strong. The model is of interest because it illustrates
the emergence of an aggregate phenomenon which
is not directly foreseen from individual behaviour
and because it concerns an important economic
problem, that of segregation.
The physical analogue to Schelling’s model,
developed in Vinkovic and Kirman ( ),
exhibits three features of the resultant segregation.
The first is the organization of the system into
‘regions’ or clusters, each containing individuals of
only one colour. Second, it explains the shape of the
frontier between the regions. Lastly, in the case
where several clusters of one colour may form it
allows one to analyse the size distribution of the
clusters.
The basic idea is simple. Think of utility as the
negative of energy. Particles with high energy in the
physical system correspond to individuals with low
utility in the social system. Where are the unhappy
or high-energy individuals to be found? Clearly
they are individuals on the frontiers of clusters.
Those within clusters of their own colour are happy
and have no possibility of increasing their utility by
moving. Those on the frontier, on the other hand,
are in contact with those of the other colour and
there may be too many of the latter. In this case
these individuals correspond to particles with high
energy. A physical system with these characteristics
will seek to minimize its energy. The energy is
highest on the frontier between clusters. Thus the
way for the system to minimize its energy is to
reduce the length of these frontiers. It will achieve
this by organizing itself into clusters, and the shape
and size of these clusters will depend on the precise
variant of the model. In the original model the
system will organize itself into two giant clusters,
each composed of individuals of one colour. If we
only allow people to move to currently free places,
then the number of these will be important for the
outcome. If there are not enough, the system will
‘freeze’ with many small clusters. If, on the other
hand, individuals can swap places the system will
segregate, but there will be perpetual movement
within it. Thus, a simple physical
3452
model generates the result obtained by Schelling
and, furthermore, shows how the form of the
segregation depends on the exact version of the
model. (For a discussion of the emergent properties
of the Schelling model see emergence ).
The 'El Farol Bar'
Another interesting example of emergent coordi-
nation is that provided by Brian Arthur ( ) in
his ‘El Farol Bar’ problem. The simple model that
he develops and which has been taken up by many
physicists under the name of ‘the minority game’
shows how individuals using mles -of thumb can
come to coordinate in a way which yields a satis-
factory social outcome even though no individual
had any such intention. The idea is that the bar can
hold 100 people. Being at the bar with fewer than
60 people is, by common consent, better than
staying at home. However once attendance goes
over 60 the bar becomes too crowded and home is
the preferable alternative. The question then is how
people will decide whether to go to the bar.
Suppose that they all reason strategically. In this
case they must decide in function of what their
neighbours will decide. Thus, to anticipate whether
there will be more than 60 people at the bar they
must reflect on the strategies employed by the
others. However, they must also take into account
that the others are doing the same and know that the
others know that they know that they are behaving
in this way. This leads to an infinite regress that
poses logical problems for the foundations of such
game-theoretic reasoning. Rather than attribute such
calculating capacities to his agents, Brian Arthur
imagined that each was endowed with a set of
forecasting rules based on previous attendance at
the bar. Given his set of rules the individual chooses
that rule which has forecast best up to the present,
‘best’ meaning the forecast that has the smallest
sum of squared prediction errors, for example.
Now, each agent uses, as information, just the
attendance observed at the bar, and updates in
consequence. There is no coordinating mechanism,
yet the model quickly settles to the ‘equilibrium’
solution with 60 people at the bar with occasional
small deviations.
Economy as a Complex System
Furthermore, each agent receives a fixed number of
forecasting rales, some of which may be rather
stupid. Nevertheless, coordination is achieved at the
aggregate level.
Some things about this model are worth noting.
It is not guaranteed that all agents will learn to
forecast correctly; some may persist in erroneous
forecasts. The way in which the model is set up
means that whenever attendance goes to 61 many
people are unhappy, which is not the case when it
goes to 59. This asymmetry does not prevent the
achievement of collective coordination, however.
Thus, the relation between satisfactory performance
at the aggregate level and satisfaction at the
individual level is tenuous. While many may find
this example intriguing, one might enquire as to
how it can be directly applied to economic
problems. An interesting answer is to be found in a
book by some Oxford physicists who specialize in
complex systems and who apply the model to
financial markets (see Johnson et al. pp. 81-136).
Financial Markets
This brings us to another important example, that of
financial markets. Models of economies with
interacting agents in the spirit of complex systems
may, as we have just seen, be able to show how
certain aggregate coordination may emerge. They
may also help us to analyse some of the observed
features of markets which normal economic anal-
ysis has difficulty explaining. For example, one of
the major problems with the standard model of
financial markets is that they do not reproduce
certain well-established stylized facts about
empirical price series. In standard models, where
there is uncertainty about the evolution of prices,
the usual way of achieving consistency is to assume
that agents have common and ‘rational’
expectations. Yet, if agents have such common
expectations, how can there be trade? Indeed there
are many ‘no trade’ theorems for such markets.
How, then, do we deal with the fact that the volume
of trade on financial markets is very important and
that agents do, in fact, differ in their opinions and
forecasts and that this is one
Economy as a Complex System
of the main sources of such trade? There is also an
old problem of ‘excess volatility,’ that is, prices
have a higher variance than the returns on the assets
on which they are based. One answer is to allow for
direct interaction between agents other than through
the market mechanism. Models reminiscent of the
Ising model from physics have been used to doing
this. For example, one might suggest that
individuals may change their opinions or forecasts
as a function of those of other agents. In simple
models of financial markets such changes may be
self-reinforcing. If agents forecast an increase in the
price of an asset and others are persuaded by their
view, the resultant demand will drive the price up,
thereby confirming the prediction. However, the
market will not necessarily ‘lock on’ to one view
for ever. Indeed, under certain rather reasonable
assumptions, if agents make stochastic rather than
deterministic choices, then it is certain that the
system will swing back to a situation in which
another opinion dominates. The stochastic choices
are not irrational, however. The better the results
obtained when following one opinion, the higher is
the probability of continuing to hold that opinion.
Such models will generate swings in opinions,
regime changes and Tong memory’, all of which
are hard to explain with standard analysis. An
essential feature of these models is that agents are
wrong for some of the time, but whenever they are
in the majority they are essentially right. Thus they
are not systematically irrational. (For examples of
this sort of model see, Lux and Mar- chesi ; Brock
and Hommes ; and Kir- man and Teyssiere , and for
a recent survey, De Grauwe and Grimaldi ). Thus
the behaviour of the agents in the market cannot
correctly be described as ‘irrational exuberance’, in
the well- known words of Alan Greenspan,
Chairman of the Board of Governors of the Federal
Reserve from 1987 to 2006.
Economists faced with this sort of model are
often troubled by the lack of any equilibrium
notion. The process is always moving; agents are
neither fully rational nor systematically mistaken.
Worse, the process never settles down to a
particular price even without exogenous shocks.
Suppose that we accept this kind of model: can
3453
we say anything analytic about the time series that
result? If we consider some of these models, for
certain configurations of parameters they could
become explosive. There are two possible reactions
to this. Since we will never observe more than a
finite sample, it could well be that the underlying
stochastic process is actually explosive, but this will
not prevent us from trying to infer something about
the data that we observe. Suppose, however, that we
are interested in being able, from a theoretical point
of view, to characterize the long-run behaviour of
the system. In particular, if we treat the process as
being stochastic and do not make a deterministic
approximation, then we have to decide what, if
anything, constitutes an appropriate long-run
equilibrium notion. Such a concept provides an
answer to those who consider that complex systems,
by their nature, are not amenable to formal analysis.
Foellmer et al. ( ), examined the
sort of price process discussed here and produced
some analytical results characterizing the process.
Furthermore, they provided a long-run equilibrium
notion that is not the convergence to a particular
price vector.
If prices change all the time, as they will do in
an evolving complex system, how may one speak of
‘equilibrium’? The idea is to look at the evolving
distribution of prices and to try to characterize its
long-run behaviour. Foellmer et al. ( )
examined the process governing the evolution of
asset prices and the profits made by traders, and
gave conditions under which it is ergodic, that is,
the proportion of time that the price takes on each
possible value converges over time and that the
limit distribution is unique. (For a discussion of the
mathematical background, see ergodicity and
nonergodicity in economics). This means that,
unlike the ‘anything can happen’ often associated
with deterministic chaos, in the long run the price
and profits process does have a well-defined
structure.
Conclusion
To view the economy as a complex system implies
a fundamental rethinking of theoretical
3454
economics. The basic idea is that of a decentralized
system with no central source of signals, whose
aggregate behaviour cannot be reduced to that of an
individual. Furthermore, the individuals are
endowed with local information and interact
directly with each other, and their behaviour can be
characterized by simple mles. Such a vision is far
from new in economics. Its origins can be traced
back at least to Adam Smith and a long chain of
economists leads from him to Hayek and Simon,
who preceded the developments described here. The
most recent contributions borrow heavily from other
disciplines such as statistical physics and the
appearance of ‘econophysics’ represents a shift
from the path that led from classical mechanics to
axiomatic mathematical models as the basic par-
adigm of economic theory. This sort of approach
has already allowed economists to analyse problems
such as contagion, neighbourhood effects, financial
bubbles, and herding behaviour, none of which fits
well into the standard economic framework. In
addition, many of the features that are imposed on
standard models emerge as a result of the
interaction between agents (see emergence).
Perhaps, most importantly, looking at economies
in this way provides a very different and more
intuitive vision of the economy as a vast interactive
system whose aggregate properties reflect the self-
organization of the system and its continual
adaptation. However, entrenched ideas die hard and
it remains to be seen whether Steven Hawking’s
prediction that the 21 st century will be the ‘age of
complexity’ will hold true for economics.
See Also
► 3::.:..‘-,::,-jence
► city and N
► Interacting Agenl
► Social Interactic
► Social Interaction
► ctworks.
► Statistical Mecha
Economy as a Complex System
Bibliography
Anderson, P.W., K.J. Arrow, and D. Pines. 1988. The
economy as an evolving complex system. Redwood City:
Addison-Wesley.
Arthur, W.B. 1994. Inductive reasoning and bounded ratio-
nality. American Economic Review 84: 406-411.
Arthur, W.B., S.N. Durlauf, and D.A. Lane, ed. 1997. The
economy as an evolving complex system II. Reading:
Addison-Wesley.
Axelrod, R. 1997. The complexity of cooperation: Agent-
based models of competition and collaboration.
Princeton: Princeton University Press.
Bak, P., K. Chen, J. Scheinkman, andM. Woodford. 1993.
Aggregate fluctuations from independent sectoral shocks:
Self-organized criticality in a model of production and
inventory dynamics. Ricerche Economiche 47: 3-30.
Blume, L., and S.N. Durlauf. 2006. The economy as an
evolving complex system III. Oxford: Oxford University
Press.
Brock, W.A., and C. Hommes. 1997. A rational route to
randomness. Econometrica 65: 1059-1095.
Crane, E. 1999. The world history of beekeeping and honey
hunting. London: Duckworth.
De Grauwe, P., and M. Grimaldi. 2006. The exchange rate in
a behavioral finance framework. Princeton: Princeton
University Press.
Durlauf, S.N., and H.P. Young. 2001. Social dynamics:
Economic learning and social evolution. London: MIT
Press.
Eldredge, N., and S.J. Gould. 1972. Punctuated equilibria: An
alternative to phyletic gradualism. In Models in
paleobiology, ed. T.J.M. Schopf. San Francisco: Free-
man, Cooper and Co.
Foellmer, H. 1974. Random economies with many interacting
agents. Journal of Mathematical Economics 1: 51-62.
Foellmer, H., U. Florst, and A. Kirman. 2005. Equilibrium in
financial markets with heterogeneous agents: A new
perspective. Journal of Mathematical Economics 41:
123-155.
Grandmont, J.-M. 1985. On endogenous competitive business
cycles. Econometrica 53: 995-1046.
Johnson, N., P. Jeffries, and P.M. Hui. 2003. Financial
market complexity. Oxford: Oxford University Press.
Kirman, A. 1992. What or whom does the representative
individual represent? Journal of Economic Perspectives
6(2): 117-136.
Kirman, A., and G. Teyssiere. 2005. Testing for bubbles and
change points. Journal of Economic Dynamics and
Control 29: 765-799.
Lindgren, K. 1991. Evolutionary phenomena in simple
dynamics. In Artificial life II, ed. C.G. Langton, C.
Taylor, J.D. Farmer, and S. Rasmussen. Redwood City:
Addison Wesley.
Lux, T., and M. Marchesi. 1999. Scaling and criticality in a
stochastic multi-agent model of a financial market.
Nature 397: 498^150.
Econophysics
Mitchell, M. 1996. An introduction to genetic algorithms.
Cambridge, MA: MIT Press.
Panes, R., andN.J. Vriend. 2007. Schelling’s spatial proximity
model of segregation revisited. Journal of Public
Economics 91: 1-24.
Pollicott, M., and H. Weiss. 2001. The dynamics of Schelling-
type segregation models and a nonlinear graph Laplacian
variational problem. Advances in Applied Mathematics
27: 17-40.
Rabin, M. 1998. Psychology and economics. Journal of
Economic Literature 36: 11-46.
Schelling, T. 1978. Micromotives and macrobehavior.
New York: W.W. Norton and Co.
Simon, H.A. 1957. Models of man: Social and rational. New
York: Wiley.
Vinkovic, D., and A. Kirman. 2006. A physical analogue of
the Schelling model. Proceedings of the National
Academy of Sciences 103: 19261-19265.
Econophysics
J. Barkley Rosser Jr.
Abstract
Econophysics, a term neologized only in 1995,
refers to physicists studying economics problems
using conceptual approaches from physics.
Certain ideas are emphasized, especially the
ubiquity of scaling laws in distributions of
financial returns, income and wealth, firm sizes,
city sizes, and other economic phenomena.
However, economists have been using many of
these techniques since much earlier, and the
influence of ideas from physics on economics
dates as far back as 1801 at least. Arguably, if
economics successfully absorbs the most useful
of this work, ‘econophysics’ may cease to exist.
Keywords
Bachelier, L.; Black-Scholes formula; Bounded
rationality; Brownian motion; Canard, N.-F.;
Econobiology; Econochemistry; Econophysics;
Levy distribution; Lognormal distribution;
Pareto distribution; Pareto, V.; Random walk;
Scaling laws; Statistical mechanics
3455
JEL Classifications
C6;D5
According to Bikas Chakrabarti ( , p. 225),
the term ‘econophysics’ was neologized in 1995 at
the second Statphys-Kolkata conference in Kolkata
(formerly Calcutta), India, by the physicist H.
Eugene Stanley, who was also the first to use it in
print (Stanley 1996). Mantegna and Stanley ( , pp.
viii-ix) define ‘the multi
disciplinary field of econophysics’ as ‘a neologism
that denotes the activities of physicists who are
working on economics problems to test a variety of
new conceptual approaches deriving from the
physical sciences’.
The list of such problems has included distri-
butions of returns in financial markets (Mantegna ;
Levy and Solomon ; Bouchaud and Cont ;
Gopakrishnan et al. ; Somette and Johansen ;
Farmer and Joshi ), the distribution of income and
wealth (Dragulescu and Yakovenko ; Bouchaud and
Mezard ; Chatterjee et al. ), the distribution of
economic shocks and growth rate variations (Bak et
al. ; Canning et al. ), the distribution of firm sizes
and growth rates (Stanley et al. , ; Takayasu and
Okuyama ; Botazzi and Secchi ), the distribution of
city sizes (Rosser ; Gabaix ), and the distribution of
scientific discoveries (Plerou et al. ; Somette and
Zajdenweber ), among other problems, all of which
are seen at times not to follow normal or Gaussian
patterns that can be described fully by mean and
variance. The main sources of conceptual
approaches from physics used by the
econophysicists have been from models of
statistical mechanics (Spitzer ), geophysical models
of earthquakes (Somette ), and ‘sandpile’ models of
avalanches, the latter involving self-organized
criticality (Bak ). An early physicist to assert the
essential identity of statistical methods used in
physics and the social sciences was Majorana ( ).
A common theme among those who identify
themselves as econophysicists is that standard
economic theory has been inadequate or insufficient
to explain the non-Gaussian distributions
empirically observed for various of these
3456
phenomena, such as ‘excessive’ skewness and
leptokurtotic ‘fat tails’ (McCauley, ). With their
sense of creating and developing a new science
based on physics that is superior to the older
conventional economics, many of the
econophysicists have focused their publishing
efforts in physics journals, notably Physica A,
Physical Review E, and European Physical Journal
B, to name some of the most frequently used ones,
along with the general science journal Nature and
some more clearly multidisciplinary journals such
as Quantitative Finance. However, increasingly
some of the econophysicists have begun to publish
jointly with economists, with some of these papers
appearing in economics journals as well. This
should not be surprising in that the emergence of
econophysics followed fairly shortly after the
influential interactions and discussions that
occurred between groups of physicists and
economists at the Santa Fe Institute (Anderson et al.
; Arthur et al. ), with some of the physicists
involved in these discussions also becoming
involved in the econophysics movement.
Now we come to a great curiosity and irony in this
matter: some of the main techniques used by
econophysicists were initially developed by economists
(with many others developed by mathematicians), and
some of the ideas associated with economists were
developed by physicists. Thus, in a sense, these efforts by
physicists resemble carrying coals to Newcastle, except
that it must be admitted that many economists
either forgot or never knew of these issues or
methods. This is true of the most canonical of
such models, the Pareto distribution.
The Empirical Focus on Scaling Laws
(Power Laws)
If there is a single issue that unites the
econophysicists it is the insistence that many eco-
nomic phenomena occur according to distributions
that obey scaling laws rather than Gaussian
normality. Whether symmetric or skewed, the tails
are fatter or longer than they would be if Gaussian,
and they appear to be linear in figures with the
Econophysics
logarithm of a variable plotted against its cumula-
tive probability distribution. They search for
physics processes, most frequently from statistical
mechanics, that can generate these non-Gaussian
distributions that obey scaling laws.
The canonical (and original) version of such a
distribution was discovered by the mathematical
economist and sociologist, Vilfredo Pareto, in 1897.
Let N be the number of observations of a variable
that exceed a value x with A and a positive
constants. Then
N = Ax-a.
(1)
This exhibits the scaling property in that
In (N) = \nA — aln(jc). (2)
This can be generalized to a more clearly sto-
chastic form by replacing N with the probability
that an observation will exceed x. Pareto formulated
this to explain the distribution of income and
wealth, and believed that there was a universally
true value for a that equalled about 1.5. More
recent studies (Clementi and Gallegati ) suggest
that it is only the upper end of income and wealth
distributions that follow such a scaling property,
with the lower ends following the lognormal form
of the Gaussian distribution that is associated with
the random walk, originally argued for the whole of
the income distribution by Gibrat ( ).
The random walk and its associated lognormal
distribution is the great rival to the Pareto distri-
bution and its relatives in explaining stochastic
economic phenomena. It was only a few years after
Pareto did his work that the random walk was
discovered in a Ph.D. thesis about speculative
markets by the mathematician Louis Bachelier
(1900), five years prior to Einstein using it to model
Brownian motion, its first use in physics (Einstein ).
Although the Paretian distribution would have its
advocates for explaining stochastic price dynamics
(Mandelbrot ), the random walk would become the
standard model for explaining asset price dynamics
for many decades, although it would be asset
returns that would be so modelled rather than asset
prices
Econophysics
themselves directly as Bachelier did originally. As a
further irony, it was a physicist, M. F. M. Osborne (
), who was among the
influential advocates of using the random walk to
model asset returns. It was the Gaussian random
walk that would be assumed to underlie asset price
dynamics when such basic financial economics
concepts as the Black-Scholes formula would be
developed (Black and Scholes ). If we let p be price,
R be the return due to a price increase, B be debt,
and a be the standard deviation of the Gaussian
distribution, then Osborne characterized the
dynamic price process by
dp = Rpdt + apdB.
(3)
Meanwhile, a variety of efforts were made over
a long time by physicists, mathematicians and
economists to model a variety of phenomena using
either the Pareto distribution or one its relatives or
generalizations, such as the stable Levy ( ) these specifically generates scaling law outcomes,
distribution, prior to the clear emergence
of econophysics. Alfred Lotka ( ) saw scien
tific discoveries as following this pattern. George
Zipf ( ) would see city sizes as doing
so. Benoit Mandelbrot ( ) saw cotton prices
doing so and was inspired to discover fractal
geometry from studying the mathematics of the
scaling property (Mandelbrot , ). Ijiri
and Simon ( ) saw firm sizes also following
this pattern, a result more recently confirmed by
Axtell ( ).
Economists Doing Econophysics?
Also, economists would move to use statistical
mechanics models to study a broader variety of
economic dynamics prior to the emergence of
econophysics as such. Those doing so included
Hans Follmer ( ), Lawrence Blume ( ),
Steven Durlauf ( ), William Brock ( ),
Duncan Foley ( ) and Michael Stutzer
( ), with Durlauf ( ) providing an over
view of an even broader set of applications. How-
ever, by 1993 the econophysicists were fully active
even if they had not yet identified themselves by
this term.
3457
While little of this work explicitly focuses on
generating outcomes consistent with scaling laws, it
is certainly reasonable to expect that many of them
could. It is tme that the more traditional view of
efficient markets with all agents possessing full
information rational expectations about a single
stable equilibrium is not maintained in these models,
and therefore the econophysics critique carries some
weight. However, many of these models do make
assumptions of at least forms of bounded rationality
and learning, with the possibility that some agents
may even conform to the more traditional
assumptions. Stutzer’s ( )
reconciles the maximum entropy formulation of
Gibbsian statistical mechanics with a relatively
conventional financial economics formulation of the
Black-Scholes options formula, based on Arrow-
Debreu contingent claims (Arrow ).
Brock and Durlauf ( ) formalize heteroge
neous agents socially interacting within a utility
maximizing, discrete choice framework. Neither of
but there is nothing preventing them from doing so
potentially.
While some econophysicists seek to integrate
their findings with economic theory, as noted above
many seek to replace conventional economic theory,
seeing it as useless and limited. An irony in this
effort is that it has been argued that conventional
neoclassical economic theory itself was
substantially a result of importing 19th-century
physics conceptions into economics, with not all
observers approving of this (Mirowski ). The
culmination of this effort is seen by many as being
Paul Samuelson’s Foundations of Economic
Analysis (1947), whose undergraduate degree was in
physics at the University of Clue ago. Samuelson
himself noted approvingly that Irving Fisher’s 1892
dissertation (1926) was partly supervised by the
pioneer of statistical mechanics, J. Willard Gibbs (
), and as far
back as 1801 Nicholas-Franco is Canard conceived
of supply and demand ontologically being
contradicting ‘forces’ in a physics sense. So the
interplay between economics and physics has been
going on for far longer and is considerably more
complicated than is usually conceived.
3458
Related Trans-disciplinary Movements
Curiously but unsurprisingly given the tremendous
attention given to the new econophysics movement,
it has spawned imitators since 2000 in the form of
econochemistry and econobiology, although these
have not had nearly the same degree of
development. The former term is the title of a
course of study established at the University of Ulm
by Barbara Mez-Starke, and was used to describe
the work of Hartmann and Rossler ( ) at
conference in 2002 in Urbino,
Italy (see also Padgett et al. , for a more recent
effort). The latter term first appeared in Hens ( ),
although McCauley ( , economy as an evolving complex system. Redwood City,
pp. 196-9) dismisses it as not a worthy competitor
for econophysics. Nevertheless, there has long been
a tradition among economists of advocating
drawing more from biology for inspiration than
from physics (Hodgson ), going back at least as far
as Alfred Marshall’s famous declaration that
economics is ‘a branch of biology broadly
interpreted’ (Marshall , p. 637), even as Marshall’s
actual analytical apparatus arguably drew more
from physics than from biolog} .
In any case, one trend we can expect for some
time is an increase in coauthoring between econ-
omists and physicists within the area of
econophysics (Lux and Marchesi ; Li and Rosser ).
Very likely we shall eventually see the more useful
ideas of econophysics coming to be absorbed into
economics proper. As that comes to pass, it may
also come to pass that the separate and distinct
movement we now know as econophysics will
cease to exist and will be forgotten, just as most
economists do not think about the physics roots of
standard neoclassical economic theory today.
See Also
► iconomy as a Cc fjZEZ..
► Evolutionary Economics
► Gibrat, Robert Pierre Louis (1904-1980)
► Tibrat’s Law
► Inequality (Global)
► Inequality (Measurement)
Econophysics
► Lcgi
► Non
► Pi
► Pi
► Pow
► Redi
► S<
► s:
► Ti
► Jrb;
a
Bibliography
Anderson, P., K. Arrow, and D. Pines, eds. 1988. The
CA: Addison-Wesley.
Arrow, K. 1974. Essays in the theory of risk bearing.
Amsterdam: North-Holland.
Arthur, W., S. Durlauf, and D. Lane, eds. 1997. The economy
as an evolving complex system II. Redwood City, CA:
Addison-Wesley.
Axtell, R. 2001. Zipf distribution of firm sizes. Science
293:1818-1820.
Bachelier, L. 1900. Theorie de la speculation. Annales
Scientifique de I'EcoIe Normale Superieure III-17, 21-86.
(English translation). In The random character of stock
market prices, ed. P. Cootner, 1964. Cambridge, MA:
MIT Press.
Bak, Per. 1996. How nature eorks: The science of self-
organized criticality. New York: Copernicus Press for
Springer-Verlag.
Bak, P., K. Chen, J. Scheinkman, andM. Woodford. 1993.
Aggregate fluctuations from independent sectoral shocks:
self-organized criticality in a model of production and
inventory dynamics. Ricerche Economiche 47: 3-30.
Black, F., andM. Scholes. 1973. The pricing of options and
corporate liabilities. Journal of Political Economy 81:
637-654.
Blume, L. 1993. The statistical mechanics of strategic
interaction. Games and Economic Behavior 5: 387^424.
Botazzi, G., and A. Secchi. 2003. A stochastic model of firm
growth. Physica A 324: 213-219.
Bouchaud, J.-P., and R. Cont. 1998. A Langevin approach to
stock market fluctuations and crashes. European
Physical Journal B 6: 543-550.
Bouchaud, J.-P., and M. Mezard. 2000. Wealth condensation
in a simple model of economy. Physica A 282: 536-545.
Brock, W. 1993. Pathways to randomness in the economy:
emergent nonlinearity and chaos in economics and
finance. Estudios Econdmicos 8: 3-55.
Brock, W., and S. Durlauf. 2001. Discrete choice with social
interactions. Review of Economic Studies 68: 235-260.
Econophysics
Canard, N.-F. 1801. Principes d Economic Politique, 1969. Rome: Edizioni
Bizzarri.
Canning, D., L. Amaral, Y. Lee, M. Meyer, and H.
Stanley. 1998. A power law for scaling the
volatility of GDP growth rates with country size.
Economics Letters 60: 335-341.
Chakrabarti, B. 2005. Econphys-Kolkata: a short story.
In Econophysics of wealth distributions, ed. A. Chatteijee, S.
Yarlagadda, and B. Charkrabarti. Milan: Springer.
Chatteijee, A., S. Yarlagadda, and B. Charkrabarti,
eds. 2005. Econophysics of wealth distributions. Milan: Springer.
Clementi, F., and M. Gallegati. 2005. Power law tails
in the Italian personal income distribution. Physica A
350: 427-438.
Dragulescu, A., and V. Yakovenko. 2001. Exponential and
power-law probability distributions of wealth and
income in the United Kingdom and the United States.
Physica A 299: 213-221."
Durlauf, S. 1993. Nonergodic economic growth. Review of
Economic Studies 60: 349-366.
Durlauf, S. 1997. Statistical mechanics approaches to
socioeconomic behavior. In The Economy as a complex evolving system II,
ed. W. Arthur, S. Durlauf, and D. Lane. Redwood
City, CA: Addison-Wesley.
Einstein, A. 1905. Uber die von der
molekularkinetischen Theorie der Warme geforderte
Bewegung von der ruhenden Fliissigkeiten
suspendierten Teichen. Annalen der Physik 17: 549-560.
Farmer, J., and S. Joshi. 2002. The price dynamics of
common trading strategies. Journal of Economic- Behavior and Organization
49: 149-171.
Fisher, I. 1926. Mathematical investigations into the
theory ofvalue and prices. New Haven: Yale University
Press.
Foley, D. 1994. A statistical equilibrium theory of
markets. Journal of Economic Theory 62: 321-345.
Follmer, H. 1974. Random economies with many
interacting agents. Journal of Mathematical Economics 1:51-62.
Gabaix, X. 1999. Zipf’s law for cities: an explanation.
Quarterly Journal of Economics 114: 739-767.
Gibbs, J. 1902. Elementary principles in statistical mechanics. New Haven: Yale
University Press.
Gibrat, R. 1931. Les Inegalites Economiques. Paris: Sirey.
Gopakrishnan, P., V. Plerou, L. Amaral, M. Meyer, and
H. Stanley. 1999. Scaling of the distributions of
fluctuations of financial market indices. Physical Review E
60:5305-5316.
Hartmann, G., and O. Rossler. 1998. Coupled flare
attractors - a discrete prototype for economic
modelling. Discrete Dynamics in Nature and Society 2: 153-159.
Hens, T. 2002. Evolutionary portfolio theory. Asset allocation
almanac: special report #4. Merrill Lynch. Online. Available at
. Accessed
22 May 2006.
Hodgson, G. 1993. Economics and evolution: bringing life
back into economics. Ann Arbor: University of
Michigan Press.
3459
Ijiri, Y., and H. Simon. 1977. Skew distributions and the sizes of business firms.
Amsterdam: North- Holland.
Levy, M., and S. Solomon. 1997. New evidence for the
power-law distribution of wealth. Physica A 242: 90-94.
Levy, P. 1925. Calcul des Probabilites. Paris: Gauthier- Villars.
Li, H., and J. Rosser Jr. 2004. Market dynamics and
stock price volatility. European Physical Journal B 39: 409-413.
Lotka, A. 1926. The frequency distribution of
scientific productivity. Journal of the Washington Academy of Sciences 12:
317-323.
Lux, T., and M. Marches! 1999. Scaling and criticality
in a stochastic multi-agent model of a financial
market. Nature 397: 498-5"oO.
Majorana, E. 1942. II valore delle leggi statistiche
nelle fisica e nelle scienze sociali. Scientia 36: 58-
66.
Mandelbrot, B. 1963. The variation of certain
speculative prices. Journal of Business 36: 394-419.
Mandelbrot, B. 1983. Die fractal geometry of nature. San Francisco: W.H.
Freeman.
Mandelbrot, B. 1997. Fractals and scaling in finance. New York: Springer-
Verlag.
Mantegna, R. 1991. Levy walks and enhanced diffusion in
Milan stock exchange. Physica A 179: 232-242.
Mantegna, R., and H. Stanley. 2000. An introduction to econophysics:
correlations and complexity in finance. Cambridge: Cambridge University
Press.
Marshall, A. 1920. Principles of economics. 8 ed. London: Macmillan.
McCauley, J. 2004. Dynamics of markets: econophysics and finance. Cambridge:
Cambridge University Press.
Mirowski, P. 1989. More heat than light: economics as
social physics, physics as nature s economics. Cambridge:
Cambridge University Press.
Osborne, M. 1959. Brownian motion in stock markets.
Operations Research 7: 145-173.
Padgett, J., D. Lee, and N. Collier. 2003. Economic
production as chemistry. Industrial and Corporate Change 12: 843-
877.
Pareto, V. 1897. Cours d' Economic Politique. Paris and Lausanne.
Trans. In Manual of Political Economy, ed. A. Schwier, 1971. New
York: Kelly.
Plerou, V., L. Amaral, P Gopakrishnan, M. Meyer, and H.
Stanley. 1999. Similarities between the growth
dynamics of university research and competitive
economic activities. Nature 400: 433-437.
Rosser, J. Jr. 1994. Dynamics of emergent urban
hierarchy. Chaos, Solitons & Fractals 4: 553-562.
Samuelson, P. 1947. Foundations of economic analysis. Cambridge, MA:
Harvard University Press.
Somette, D. 2003. Why stock markets crash: critical events
in complex financial systems. Princeton: Princeton
University Press.
Somette, D., and A. Johansen. 2001. Significance of
log-periodic precursors to financial crashes. Quantita-
tive Finance 1: 452-471.
3460 Eden, Frederick Morton (1766-1809)

Somette, D.,nndD. Zajdenweber. 1999. Economic returns with poverty and pauperism. In the years that
of research: The Pareto law and its implications. followed Eden wrote a number of pamphlets on
European Physical Journal B X: 653 664.

Spitzer, F. 1971. Random fields and interacting particle systems. Providence:

related issues.
American Mathematical Society. The greater part of The State of the Poor records
Stanley, H., V. Afanasyev, L. Aamaral, S. Buldyrev. A. Eden’s findings relating to the acntal conditions
Goldberger, S. Havlin, H. Leschhom, P. Maass, R. prevailing in the parishes of England. Stimulated by
Mantegna, C.-K. Peng, P. Prince, M. Salinger, M.
Stanley, and G. Viswanathan. 1996a. Anomalous
the high prices prevailing in 1794-5, Eden initially
fluctuations in the dynamics of complex systems: set out to study the condition of the poor, but later
from DNA and physiology to econophysics. Physica .1 224: extended this to the labouring classes. He
302 321. encountered at times great resistance from local
Stanley, M., L. Amaral, S. Buldyrev, S. Havlin, H.
Leschhom, P. Maass, M. Salinger, and H. Stanley.
parish authorities, but despite this he was able to
1996b. Scaling behavior in the growth of gather a considerable amount of information on
companies. Nature 379: 804 806. wage levels, diet and prices. This was linked to an
Stutzer, M. 1994. The statistical mechanics of asset appraisal of the nutritional value of available
prices. In Differential equations, dynamical systems, and control science: .1 festschrift in honor of
foodstuffs, such that it was possible to arrive at
lawrence markus, ed. Is.. Elworthy, W. Everitt, and E.

Lee, vol. 152. New York: Marcel Dekker. some kind of comparative assessment of levels of
Takayasu, H., and Is. Okuyama. 1998. Country depen- poverty and want. It emerged from his empirical
dence on company size distributions and a findings that the actual conditions and treatment of
numerical model based on competition and
the poor varied greatly from parish to parish, this in
cooperation. Fractals 6: 67 79.
Zip f, G. 1941. National unity and disunity. Bloomington, IN: part reflecting the patchwork of legislation that had
Principia Press. grown up over the years in relation to the pauper
and the workless. He argued however that existing
legislation implied a policy of support for the
indigent, and that in general a civilized society' had
an obligation to make such provision.
Eden, Frederick Morton (1766-1809)

K. Tribe

Selected Works

1797. F. M. Eden, The state of the poor: Or an

Keywords history of the labouring classes in England , 3 vols.
Eden, F. M.; Poverty London.

JEL Classifications
B31
Edgeworth as a Statistician
The son of Sir Robert Eden, F.M. Eden was edu-
cated at Oxford, gaining a Master’s degree in 1789. Stephen M. Stigler
A co-founder of the Globe Insurance Company, he
published in 1797 the three volumes of his
investigation into the conditions of the labouring
poor. The State of the Poor. This work was perhaps
Francis Edgeworth was the leading theorist of
the most detailed appraisal of social legislation and
mathematical statistics of the latter half of the 19th
its actual workings that had appeared, and the
century, though his influence was diminished by the
findings provided ample material for ensuing debate
difficulty of his exposition. He is most frequently
on the best form of dealing
remembered today for his work on the
Edgeworth as a Statistician
Edgeworth Series, but in fact he touched on nearly
every sphere of modem statistics, from the analysis
of variance to stochastic models, to multivariate
analysis, to the asymptotic theory of maximum
likelihood estimates, to inventory theory. In some
areas such as correlation, his work was decisive in
the development of all that followed.
Edgeworth’s first purely statistical work was
published in 1883, when he began a series of papers
examining the methods, rationale and philosophical
foundations of probability and its application to the
analysis of observational data. Most of this work
appeared in the Philosophical Magazine, Mind, or
the Journal of the London (later the Royal)
Statistical Society. Between 1883 and 1890 he
published over 30 separate papers on a wide
selection of statistical topics; these works are best
viewed as the tracks left by a first-rate mind as it
took an excursion through territory that had already
been explored. He found much that was new, but
his principal occupation re-examining past works,
particularly those of Laplace, to see how they might
be used in social science. A major (and under
appreciated) accomplishment of this period was
Edgeworth’s explanation of how simple
significance tests could be used to compare
averages. The mathematical technique was not new,
but the conceptual framework was subtly different
from that of the early astronomers, and while
Edgeworth’s ( ) explanation may
today seem elementary, it had a lasting widespread
impact. In subsequent work (Edgeworth ; Stigler )
he developed what might now be viewed as an
analysis for an additive effects model for a two-way
classification, and he was sensitive to the effect
non-normality or serial dependence could have
upon the procedures.
Edgeworth’s main orientation in his inferential
work was Bayesian, and he presented both philo-
sophical and mathematical investigations of this
approach. To Edgeworth, a prior distribution was
based in a rough way upon experience. A uniform
prior was often justified because, Edgeworth
observed, we do not find a pattern in nature that
tends to favour one set of values for its constants
over another set. Edgeworth tempered this with a
realization that inferences would frequently not
3461
be very sensitive to the prior specification
(Edgeworth ). When evaluating the significance of
differences, however, Edgeworth reverted to a
sampling theory viewpoint. One of his 1883 works
includes a derivation of Student’s /-distribution as
the posterior distribution for a normal mean. From
1890 to 1893 Edgeworth, reacting to work by
Galton, gave the first fully developed mathematical
examination of correlation and its relation to the
multivariate normal distribution (Edgeworth , ).
Edgeworth
showed how the constants of a multivariate normal
distribution could be expressed in terms of pairwise
correlation coefficients (and hence how the
conditional expectation of one variable given others
could be expressed in terms of correlation
coefficients), and he investigated how a correlation
coefficient could be estimated from data. His work
gave what may be the earliest version of what has
come to be called Pearson’s product moment
estimate (or Pearson’s r). Incidentally, it was
Edgeworth who coined the term ‘coefficient of
correlation’, as Galton had used ‘index’.
Edgeworth’s work on correlation had an
immense influence upon Karl Pearson, and through
him upon all 20th-century work on this topic. In the
1890s, Edgeworth’s statistical work became
increasingly occupied by a competition with Karl
Pearson as to who could best model skew data.
Pearson, with his family of skew curves that
included gamma distributions and a scheme (the
method of moments) for selecting a curve within
this family, is generally conceded to have won the
contest. Edgeworth at one time or another tried
three different approaches. One of these (the
‘method of translation’, or fitting a normal curve to
transformed data) has become popular in more
recent times. Another (fitting separate half-normal
curves to the left and right sides of the distribution)
has been largely forgotten. The third was based
upon what we now call the Edgeworth Series. The
essence of Edgeworth’s approach was to generalize
the central limit theorem by the inclusion of
correction terms, terms that appeared in the
derivation of the distribution of sums but which
became negligible if the number of terms in the sum
was large. The idea was that skew distributions
found in nature
3462
were skew because they were aggregates of rela-
tively small numbers of non-normal components.
Edgeworth was thus taking a theoretical approach,
one that he felt was more appealing than Pearson’s
more ad hoc approach. The Edgeworth Series was
foreshadowed in his work as early as 1883 (when he
found it as a series solution to the heat equation),
but the full development came later (Edgeworth ),
and the labour he put into it after 1895 was
immense, and largely unrewarded. His attempts to
provide a methodology for fitting the series to data
attracted few followers, Arthur Bowley being the
only important one. Bowley’s brave attempt to
explain the method in his assessment of
Edgeworth’s work (Bowley ) was only marginally
more readable than Edgeworth’s own many efforts
on this. Ironically, later statisticians (notably Harold
Cramer, see Cramer ) have found that Edgeworth’s
mode of arranging correction terms was far superior
to alternatives proposed by Bruns, Gram and
Charlier, and the Edgeworth Series has become an
important technique for approximating sampling
distributions (rather than data distributions, as
Edgeworth had intended).
In addition to these major themes, Edgeworth’s
work abounds in minor nuggets. The largest of
these may be a series of papers in 1908-9 that we
can now recognize as containing the germ of a
proof of the asymptotic efficiency of maximum
likelihood estimates. In a contentious 1935 meeting
of the Royal Statistical Society this work was
pointed out to R.A. Fisher by Bowley as an
unacknowledged predecessor, although it seems
doubtful that it had any influence on Fisher (see
Pratt ). Ofmore importance was Edgeworth’s work
on index numbers and on the theory of banking.
While his work on index numbers is more properly
treated with his economic work, it is worth noting
here that he was a pioneer in the application of
probability to the analysis and choice of index
numbers. In regard to banking, based upon
statistical considerations, he promulgated in 1888
the rule that the reserves of a bank need only be
proportional to the square root of its liabilities
(Edgeworth ).
In all Edgeworth’s work one is constantly com-
ing upon minor, often paradoxical observations
Edgeworth as a Statistician
(see for example, Stigler ) that reveal the depth of
his understanding, the subtlety of his thoughts, and a
grasp of mathematics that seems quite at odds with
his lack of formal training in the subject. Edgeworth
was an independent thinker upon statistical matters,
though he was perhaps the earliest to appreciate and
follow up on Galton’s innovative concepts of
regression and correlation. Edgeworth’s most
important influence was upon Karl Pearson, though
Pearson was chary in his recognition of this
influence. Taken together, Galton, Edgeworth and
Pearson shaped modem statistics to a greater degree
than any other individual or group before R.A.
Fisher. Edgeworth’s works on statistics number at
least 75, and it is rare to find one that is self-
contained. Bowley ( ) made an attempt to
summarize all
of Edgeworth’s statistical work, and he gave a
bibliography of most of it. Stigler ( , )
gives a more recent assessment, and comments upon
different aspects of Edgeworth’s work can be found
in papers by Kendall ( , ) and
Pratt ( ).
Bibliography
Bowley, A.L. 1928. F.Y. Edgeworth's contributions to mathematical statistics. London:
Royal Statistical Society. Reprinted, New York:
Augustus M. Kelley, 1972.
Cramer, H. 1972. On the history of certain expansions
used in mathematical statistics. Biometrika 59: 205-207.
Edgeworth, F.Y. 1885a. Observations and statistics. An
essay on the theory of errors of observation and
the first principles of statistics. Transactions of the Cambridge
Philosophical Society 14: 138-169.
. 1885b. Methods of statistics. Jubilee Volume of
the Statistical Society: 181-217.
. 1885c. On methods of ascertaining variations in
the rate of births, deaths, and marriages. Journal of the
Royal Statistical Society 48: 628-649.
. 1888. The mathematical theory of banking. Journal of the
Royal Statistical Society 51:113-127.
. 1892a. Correlated averages. Philosophical Magazine 34(Fifth
Series): 190-204.
. 1892b. The law of error and correlated averages.
Philosophical Magazine 34(Fifth Series): 429-438, 518-26
. 1905. The law of error. Transactions of the
Cambridge Philosophical Society 20: 36-65, 113-41
Edgeworth Price Cycles 3463

. 1908. On the probable errors of frequency- cycles continue to generate public concern with
constants. Journal of the Royal Statistical Society 71: 381-397, 499-512,
claims of collusion often raised, the current
651-678; 72, 81-90
Kendall, M.G. 1968. Francis Ysidro Edgeworth, 1845- evidence favours Edgeworth price cycles bci ng
1926. Biometrika 55: 269-275. the result of stronger competition and the source
. 1969. The early history of index numbers. Review of lower retail gasoline prices.
of the International Statistical Institute 37: 1-12.
Pratt, J. 1976. F.Y. Edgeworth and R.A. Fisher on the
efficiency of maximum likelihood estimation. Annals of Keywords
Statistics 4: 501-514. Markov strategies; Markov perfect equilibria;
Stigler, S.M. 1978. Francis Ysidro Edgeworth, Cournot model; Retail petrol markets
statistician (with discussion). Journal of the Royal Statistical Society,
Series A 141: 287-322.

. 1980. An Edgeworth curiosum. Annals of Statistics 8: 931-

934. JEL Classifications
—. 1986. The history of statistics: The measurement of uncertainty before 1900. Cambridge, L13; L81; L92
MA: Belknap Press of the Harvard University Press.

Introduction

Edgeworth price cycles refer to an asymmetric

pattern of prices that is generated by a dynamic
Edgeworth Price Cycles pricing equilibrium among competing oligopolists
under certain simple assumptions. Most notably, the
Michael D. Noel
oligopolists are assumed to compete in prices,
follow Markov strategies, and face relatively high
price elasticities for their good. The time series of
market prices under this equi librium takes on a
Abstract sawtooth shape, with many small price decreases
Edgeworth price cycles refer to an asymmetric intermpted only by occasional large price increases.
pattern of prices that result from a dynamic The asymmetric price pattern is repeated over and
pricing equilibrium among competing over, even in the absence of any supply or demand
oligopolists. The resulting time series takes on a shocks.
sawtooth shape - many small price decreases Edgeworth price cycles are the leading theory
interrupted only by occasional large price behind the asymmetric price cycles that appear in
increases. Maskin and Tirole (Econometrica many retail gasoline markets around the world.
56(3): 571-599, 1988) formalized the theory, First observed in some US cities in the 1960s, they
and later extensions were provided by Eckert have become commonplace in many US, Canadian,
(International Journal of Industrial Organization Australian and European retail gasoline markets
21(3): 151-170, (e.g. Noel ( ), Eckert ( ), Doyle
2003) and Noel (Journal of Economics and et al. ( ), Wang ( )) and visually are
Management Strategy 17(2): 345-377, 2008). very similar to the theoretical cycles. A single cycle
Edgeworth price cycles are the leading theory is often a week or two long with amplitude up to
for explaining the asymmetric price cycles that about 10% of the price.
appear in many US, Canadian, Australian and Two waves of literature examine these cycles
European retail gasoline markets (e.g. Noel empirically. The first investigates the cause of the
(Review of Economics and Statistics 89(2): 324- retail gasoline price cycles. The near consensus in
334, 2007a), Eckert (Canadian Journal of the literature is that the cycles are Edgeworth price
Economics 35(1): 52-77, 2002), Doyle et al. cycles. The location and shape of the cycles and the
(Energy Economics 32(3): 651-660, 2010), behaviour of different types of firms along the
Wang (Journal of Political Economy 117(6): cycles support the Edgeworth price cycles theory.
987-1030,2009b)). While the gasoline
3464
The second wave of literature examines the welfare
effects of the cycles relative to a stable price
equilibrium. The literature is young, but the results
currently favour the conclusion that the price cycles
are indicative of stronger competition and the
source of lower prices for consumers.
The Theory of Edgeworth Price Cycles
The notion of a competitively driven, dynamic,
asymmetric price cycle dates back to Edgeworth (
). Edgeworth was a strong critic of the
Cournot model and argued that when marginal costs
were increasing (or firms were capacity constrained
in the extreme case), prices in oligopolistic
competition would not be stable as in Cournot’s
model. Instead, they would change continually
along an asymmetric price cycle.
Firms would undercut one another to gain
market share until prices were low enough that one
firm could profitably raise the price and serve the
residual demand left over from the capacity
constrained firm. Interestingly, Edge- worth
considered his cycle a disequilibrium, as the notion
of equilibrium was then equated to stable prices.
The seminal theory paper on Edgeworth price
cycles is by Maskin and Tirole ( ), who gave
the cycles their name. Maskin and Tirole assume
two identical and infinitely lived firms with high
Edgeworth Price Cycles,
Fig* 1 Theoretical
Edgeworth cycle
-a—Firm 1’s Price
Edgeworth Price Cycles
discount factors that sell homogeneous goods and
compete in prices. Firms are restricted to using
Markov strategies, which in this context means that
a firm’s pricing decision depends only on the price
of the other firm currently in effect. Demand is
constant and costs are zero. Maskin and Tirole show
that in this setting two possible types of Markov
perfect equilibria could result. The first generates
stable prices overtime, while the second results in
asymmetric price cycles - that is, Edge- worth price
cycles - in equilibrium.
Figure shows firm prices in a cycling equi-
librium. The mechanism operates as follows.
Starting from prices relatively high above marginal
cost, firms alternately and repeatedly undercut one
another’s price by the smallest possible amount.
Because the goods are identical, this is sufficient to
frilly steal total market demand. Undercutting
continues until prices fall all the way to marginal
cost (zero in this example). At that point, there is no
gain from lowering prices further, but there is a gain
from raising them. If one firm were to raise its price
to a much higher level, the other firm would surely
respond with a higher price too, just slightly
undercutting that of the first firm. As a result, when
prices are at marginal cost, firms play a war of
attrition, each mixing between a higher price and
maintaining the marginal cost price. Eventually, one
firm relents by restoring its price to a high level, the
other follows, and a new round of undercutting
Time Series of:
—♦—Firm 2’s Price
Edgeworth Price Cycles
begins. The cycle repeats over and over even in the
absence of cost and demand shocks.
The theory of Edgeworth price cycles is the
leading theory for explaining the asymmetric price
cycles found in retail gasoline markets. However,
not all of its assumptions are well suited to retail
gasoline. Retail gasoline is not perfectly
homogeneous, marginal costs are not constant,
there are more than two firms, and so on. To
address this, Noel ( ) extends the model to
allow for fluctuating costs, capacity constraints,
product differentiation, triopoly situations, and
other extensions. Noel shows that Edgeworth price
cycles are robust when product differentiation or
capacity constraints are not too strong. Noel also
shows that cycles are robust to triopoly, but now
with challenges in the form of false starts. These
occur when the first firm to increase price abandons
its high price altogether, after waiting too long for
others to follow. Since false starts make it more
costly for a firm to be the first to raise its price, the
cycle peak and trough prices move lower and
average prices fall from increased competition. In
another important extension, Eckert ( ) showed that
if firms share the mar
ket very unequally at equal prices (interpretable as
the case of asymmetrically sized firms), Edge-
worth price cycles are more likely.
Empirical Evidence of Edgeworth
Price Cycles
An early criticism of the theory, dating back to at
least Nichol (1935), was that the model failed to
predict the experience of any known real-world
markets. There has also been very limited success
in generating Edgeworth cycles in laboratory
experiments (Leufkens and Peeters ).
This all changed when - with the availability of
new high-frequency and station-specific datasets -
asymmetric retail price cycles were discovered in
gasoline markets in many countries in the 1990s
and 2000s. As of 2010, cycles have been observed
in many retail gasoline markets in the USA,
Canada, Australia, Norway, Germany and Belgium,
with other discoveries sure to come. (Cycles had
been detected in a few US cities in the
3465
late 1960s and early 1970s, but they received little
attention from economists at the time and then
disappeared from view for thirty more years.)
Recently, inverted asymmetric price cycles have
also been found in keyword advertising auctions at
leading search engines (inverted because com-
petition is among buyers instead of sellers)
(Zhang ).
The first generation of studies in the literature
sought a cause for the empirical asymmetric cycles,
and the results of that literature strongly point to
Edgeworth price cycles (to name a few: Castanias
and Johnson ( ), Lewis and Noel (
), Lewis ( , ), Doyle et al. ( ), in
US retail gasoline markets; Noel ( ,, ,
), Eckert ( , ), Eckert and West
( ), Atkinson ( ) in Canada; Wang
( , ) in Australia; and Zhang ( ) in
Internet auction markets). One opposing view is that
of Foros and Steen ( ), who examine
Norway and argue that the cycles there, although
similar to those in other countries, are possibly a
form of pure collusion.
The first known publication that noted asym-
metric price cycles in retail gasoline was Allvine
and Patterson ( ), who showed cycles in sev
eral US cities in the late 1960s and early 1970s.
Castanias and Johnson ( ) reported summary
statistics on the cycles in Los Angeles area and
noted the resemblance between those cycles and the
then newly published Maskin and Tirole ( )
theory article. In Los Angeles, a single
cycle lasted about one to two months with an
amplitude of roughly 25% of the price.
The first full-length scientific papers were writ-
ten about Edgeworth cycles in Canada, where most
large cities experienced price cycles beginning from
at least the late 1980s. The cycles ranged from
weekly up to bimonthly in duration and longer.
Eckert ( ) showed that price vol
atility in retail gasoline prices in Canadian cities
were consistent with the general predictions of his
extension of the Edgeworth cycle model.
Noel ( ) specifically modelled the asym
metric price movements in Canadian cities and
found that a greater market share of aggressive
independent firms resulted in cycles that were
faster, taller and less asymmetric, consistent
3466
with the theory. Noel ( ) examined station-
specific data in Toronto, Canada, and showed that
behaviours of differently sized firms were also
consistent with the theory. Large refiner-retailers
tended to lead price increases, and smaller inde-
pendents tended to lead price decreases. The dura-
tion of cycles in Toronto averaged a week and the
amplitude was about 8% of the price. Atkinson (
) found similar results with high-frequency
price data in Guelph, Canada.
Retail gasoline price cycles were rediscovered in
dozens of US cities in the early 2000s, although it is
uncertain how long they existed prior to that. The
cycles were weekly in most cities but biweekly in a
few. Lewis ( ,)
support that the cycles in the Midwest US are
Edgeworth price cycles. The amplitude of the cycles
averaged 7% of the retail price. Doyle et al. ( )
present further evidence that cycles
were more likely with more large, price- aggressive
independents, consistent with then- extension of the
theory.
In major Australian cities and many smaller
ones, retail gasoline price cycles were weekly in
duration, except in Perth where they were some-
times biweekly. Wang ( ) estimates espe
cially high cross price elasticities in Perth,
consistent with the theory, and Wang (
shows that the pattern of price increases among the
leading firms is consistent with the use of mixed
strategies, as predicted in a symmetric Edgeworth
price cycles model.
Foros and Steen ( ) examine the cycles in
Norway which share the characteristics of cycles
elsewhere. They offer the opposing view that the
Norwegian cycles may be a pure collusion story on
the basis that the large price increases occur in a
short window on late mornings and early afternoons
on Mondays. While there is no direct evidence of
widespread collusion in Norway or elsewhere,
isolated instances of individual dealers colluding
with one another can occur in cycling markets, just
as they can in non-cycling markets (see Wang (
) for the case of Ballarat, Australia,
and Erutku and Hildebrand ( ) for the case
of four towns in Quebec, Canada).
Edgeworth price cycles have not been well
understood in many circles and the large sweeping
Edgeworth Price Cycles
price increases understandably raise antitrust
scrutiny. The second wave of the Edgeworth price
cycles empirical literature examines the impact that
cycles have on price and welfare.
Most important and difficult is the question of
how markups rise and fall with the presence of
Edgeworth cycles. Noel ( ) shows that in
Canada markups within the same city in a nearby
time period (i.e. controlling for time-city effects)
were one cent lower just after cycles began com-
pared with before. The consumer gain of cycling in
his sample cities is CDN$48 million per year. Doyle
et al. ( ) show that in the USA, markups
in cycling cities are one to two US cents per gallon
lower than in non-cycling cities, controlling for
market structure and other observables. Wang (
), however, finds that prices were 1.8
cents lower when cycles temporarily ceased in the
four months after the passing of the 24 hour price
change pre-notification law in Perth, Australia. Noel
( ) notes it would not be unusual for
prices to temporarily fall in such a situation, as
firms jostle for position and generate a string of
false starts as they adjust to the new notification
requirements.
Lewis ( ) and Lewis and Noel ( )
argue for another pro-competitive aspect to Edge-
) worth price cycles - that they more effectively
anchor prices to wholesale costs over the long run.
In non-cycling markets, the well-known rockets and
feathers phenomenon is that prices rise quickly after
a cost increase but fall slowly after a cost decrease
(Borenstein et al. ). Noel ( ) and Eckert ( )
even show that in
cycling markets, the presence of the asymmetric
cycle can generate or magnify such an effect.
However, compared to non-cycling markets, in
cycling markets the cycles are effective in returning
prices to costs at every trough. Lewis ( )
shows that retail gasoline prices in US
markets with Edgeworth price cycles fell much
more quickly in the months after Hurricane Katrina
than those cities without, for a relative gain of
US$1.33 million per 100,000 people. Lewis and
Noel ( ) look more comprehen
sively at 72 US cities and show that cost shocks are
passed through two to three times faster in cities
with Edgeworth price cycles than in cities
Edgeworth Price Cycles
without, substantially reducing the rockets and
feathers welfare loss relative to non-cycling cities.
Noel ( ) argues for another hidden benefit
of Edgeworth price cycles. Very simple purchase
timing strategies can allow price elastic consumers
in Toronto to easily reduce their gasoline
expenditures by 4% relative to purchasing at ran-
dom times. The gain would be even greater in other
markets (e.g. Norway or Australia) where cycle
troughs are even more easily predictable. Firm price
reoptimization in response to large numbers of
consumers timing the cycles could limit the gain, of
course, but as the Australian experience shows, the
cycles remain active and strong even when there is
significant consumer awareness of them.
The greater implication of Noel ( ) is that
to the extent that consumers use purchase timing
strategies that are not observed by the researcher,
the benefit of Edgeworth price cycles is understated.
This is because when consumers can time purchases
to periods of lower prices during cycles, the more
economically relevant measure of quantity-
weighted prices is likely to be lower than the
unweighted average price in markets with cycles,
but the same in markets without. A comparison of
average quantity-weighted prices under periods of
cycling and non-cycling would then reveal a greater
price advantage to consumers in markets with
Edgeworth price cycles.
Conclusion
Edgeworth price cycles are asymmetric price cycles
generated from equilibrium behaviour in a game of
oligopolistic price competition. Firms repeatedly
steal market demand from one another by
undercutting down to marginal cost. Firms then
sequentially increase prices back to the top of the
cycle and begin undercutting again.
The empirical literature strongly favors the
asymmetric price cycles in retail gasoline markets
being generated by an Edgeworth price cycles
process. While research continues, the weight of the
current evidence also points to the conclusion that
Edgeworth price cycles are indicative of
3467
stronger competition. They benefit consumers with
lower and more efficient prices relative to the less
controversial stable price equilibrium.
The literature on Edgeworth price cycles
continues to grow. An obvious direction for future
work is to search for and uncover additional
examples of Edgeworth price cycles outside of
retail gasoline. Further extensions to the theory can
help guide this search, and conversely the findings
of the search can suggest new extensions to the
theory to help identify the factors most critical to
cycle generation. Finally, the welfare effects of
Edgeworth price cycles have only recently begun to
be understood. An important direction of future
research would be to study and quantify these
effects further, with obvious and important antitrust
and policy implications.
See Also
► idgeworth, Frantic J sidro (1845-1926)
► iasolinc Mark
► Markov Equilibr in Macroeconomics
► Markov Proce;
Bibliography
Allvine, F., and J. Patterson. 1974. Highway robbery: An analysis of the
gasoline crisis. Bloomington: Indiana University Press.
Atkinson, B. 2009. Retail gasoline price cycles:
Evidence from Guelph, Ontario using bi-hourly
station-specific retail price data. Energy Journal 30(1):
85-110.
Borenstein, S., A. Cameron, and R. Gilbert. 1997. Do
gasoline markets respond asymmetrically to crude
oil price change. Quarterly Journal of Economics 112: 305-339.
Castanias, R., and H. Johnson. 1993. Gas wars: Retail
gasoline price fluctuations. Review of Economics and Statistics
75(1): 171-174.
Doyle, J., E. Muehlegger, and K. Samphantharak. 2010.
Edgeworth cycles revisited. Energy Economics 32(3): 65F-
660.
Eckert, A. 2002. Retail price cycles and response
asymmetry. Canadian Journal of Economics 35(1): 52-77.
Eckert, A. 2003. Retail price cycles and presence of
small firms. International Journal of Industrial Organization 21(3): 151-170.
3468 Edgeworth, Francis Ysidro (1845-1926)

Eckert, A., and D. West. 2004. Retail gasoline price

cycles across spatially dispersed gasoline
Edgeworth, Francis Ysidro (1845-1926)
stations. Journal of Law and Economics 22: 997-1015.
Edgeworth, F.Y. 1925. The pure theory of monopoly. In
Papers relating to political economy, vol. I, ed. F.Y. Edgeworth, 111-
John Creedy
142. London: Macmillan.
Erutku, C.. and V. Hildebrand. 2010. Conspiracy at the
pump. Journal of Law and Economics 53( 1): 223 237.
Foros, O.. and F. Steen. 2008. Gasoline prices jump up
on Mondays? An outcome of aggressive competition?
CEPR Working Paper DP61S3.
Abstract
Leufkens, K., and R. Peeters. 2008. Focal prices and
Edgeworth was a major figure in the develop-
price cycles in an alternating price duopoly
experiment. METEOR Working Paper RM 08.021. ment of neoclassical economics, and one of its
Lewis, M. 2009a. Temporary wholesale gasoline price most original theorists, making a wide range of
spikes have long lasting retail effects: The lasting contributions. After describing his
aftermath of Hurricane Rita. Journal of Law and Economics 52(3):
approach to economics, this article discusses his
581-606.
Lewis, M. 2009b. Price leadership and coordination in early work in moral philosophy, which had a
retail gasoline markets with price cycles. The Ohio State strong influence on his economics. His
University Working Paper. important contribution to the theory of
Lewis, M., and M. Noel. In press. The speed of gasoline
exchange, focusing on indeterminacy and the
price response in markets with and without
Edgeworth cycles. Review of Economics and Statistics. role of the number of traders, is examined. His
Maskin. E.. and J. Thole. 1988. A theory of dynamic later work on monopoly, international trade and
oligopoly U: Price competition, kinked demand taxation arc then briefly discussed.
curves and Edgeworth cycles. Econometrica 56(3): 571
599.
Noel, M. 2002. Edgeworth price cycles in retail gasoline
Keywords
markets. PhD dissertation. MIT.
Noel, M. 2007a. Edgeworth price cycles, cost-based Arbitration; Assumptions; Barter; Bentham, J.;
pricing and sticky pricing in retail gasoline Bickerdike, C; Bilateral monopoly; Biological
retail markets. Review of Economics and Statistics 89(2): 324-334. analogies; Bootstrap; Butler, J.; Calculus of
Noel, M. 2007b. Edgeworth price cycles: Evidence from variations; Coalitions; Collusion; Combinations;
the Toronto retail gasoline market. Journal of Industrial Economics
55( 1): 69 l>2.
Competitive (price-taking) equilibrium;
Noel, M. 2008. Edgeworth price cycles and focal prices: Complements: see substitutes and complements;
Computational dynamic Markov equilibria. Journal of Economics Conjectural variations; Contract curve;
and Management Strategy
Correlation coefficient; Cournot, A.; Darwin, C.;
17(2): 345-377.
Noel, M. 2009. Do gasoline prices respond
Deductive method; Determinacy and inde-
asymmetrically to cost shocks? The effect of terminacy of exchange; Distributive justice;
Edgeworth cycles. RAND Journal of Economics 40(3): 582 595. Duopoly; Edgeworth box; Edgeworth, F.; Ego-
Noel, M. 2010a. Edgeworth cycles and intertemporal ism; Equilibrium in exchange; Experimental
price discrimination. UCSD Working Paper.
Noel, M. 2010b. Edgeworth cycles, competition, and
psychology; First fundamental theorem of wel-
antitrust. UCSD Working Paper. fare economies; Giffengood; Harsanyi, .1.; His-
Wang, /,. 2008. Collusive communication and pricing torical School; Hotelling, H.; Idealism;
coordination in a retail gasoline market. Review of Industrial Immiserizing growth; Indeterminacy of contract;
Organization 32( 1): 35-52.

Wang, /,. 2009a. Station level gasoline demand in an

Australian market witli regular price cycles. Australian
Journal of. Igricultural and Resources Economics 53: 467 483.
Wang, /,. 2009b. Mixed strategies in oligopoly pricing:
Evidence from gasoline price cycles before and
under a timing regulation. Journal of Political Economy 117(6):
987-1030.
Zhang, M. 2005. Finding Edgeworth cycles in online
advertising auctions. MITmimeo.
Indifference map; Inference; International trade;
International values, theory of; Intuitionism;
Jevons, W.; Keynes, J. M.; Lagrange
multipliers; Laplace, P.; Launhardt, C.; Law of
indifference; Marshall, A.; Mathematical
economics; Mechanical analogies; Mill, .1. S.;
Monotonieity; Moral philosophy; Negative
income tax; Neoclassical; No-profit
Edgeworth, Francis Ysidro (1845-1926)
entrepreneur; Offer curve; Optimal distribution;
Optimal tariffs; Paley A. and M.; Palgrave’s
Dictionary of Political Economy, Partial
equilibrium theory; Pearson distributions;
Physical sciences; Pigou, A.; Probability;
Progressive and regressive taxation; Rate of
exchange; Reciprocal demand curve;
Recontracting; Royal Statistical Society; Rules
of conduct; Sacrifice theory of tax incidence;
Saddle point; Schumpeter, J.; Sidgwick, H.;
Social contract; Social welfare function;
Statistical inference; Substitutes and comple-
ments; Tax incidence; Taxation, theory of;
Transformations; Utilitarianism; Utility func-
tions; Utility maximization; Vickrey, W.;
Walras, L.
JEL Classifications
B31
Biographia
Francis Ysidro Edgeworth (1845-1926) was bom in
Edgeworthstown in County Longford, Ireland. The
background into which he was bom was dominated
by the ‘larger than life’ figure of his grandfather
Richard Lovell Edgeworth (1744-1817), whose life
was documented in a two-volume memoir (1820)
by his oldest daughter, the famous novelist Maria
Edgeworth (1767-1849). Richard Lovell’s many
scientific and mechanical experiments were helped
by his strong association with the Lunar Society of
Birmingham, whose members included Watt,
Bolton, Wedgwood, Priestley, Darwin and Galton.
In addition, Maria’s scientific acquaintances
included Davy, Humboldt, Herschel, Babbage,
Hooker and Faraday. The marriage of F. Y.
Edgeworth’s cousin Harriet Jessie Edgeworth
(daughter of Richard Lovell’s seventh and youngest
son Michael Pakenham, 1812-1881) to Arthur Gray
Butler provided links with another large and
eminent academic family. These connections
extend even further since A. G. Butler’s sister,
Louisa Butler, married Francis Galton, a cousin of
Charles Darwin.
3469
Richard Lovell’s sixth son, and 17th surviving
child, was Francis Beaufort Edgeworth (1809-
1846), who met his wife, Rosa Florentina Eroles,
the daughter of a Spanish refugee from Catalonia
and then aged 16, while on the way to Germany to
study philosophy; they married within three weeks
in 1831. F. Y. Edgeworth was their fifth son. With
his family background and his knowledge of
French, German, Spanish and Italian, Edgeworth
had wide international sympathies. On the family
background, see Butler and Butler ( ) and for a
full-length treatment of
Edgeworth’s work, see Creedy ( ).
Edgeworth was educated by tutors in
Edgeworthstown until the age of 17, when in 1862
he entered Trinity College Dublin to study
languages. In 1867 Edgeworth entered Exeter
College, Oxford, but after one term transferred to
Magdalen Hall. He transferred to Balliol in 1868,
where in Michaelmas 1869 he obtained a first in
Literae Humaniares. He was called to the bar in
1877, the same year in which his first book, New
and Old Methods of Ethics, was published.
Edgeworth applied unsuccessfully for a professor-
ship of Greek at Bedford College, London, in 1875,
but later lectured there on English language and
literature for a brief period from late 1877 to mid-
1878. He had earlier lectured on logic, mental and
moral sciences and metaphysics to prospective
Indian civil servants, at a private institution run by a
Mr. Walter Wren. In 1880 he applied for a chair of
philosophy, also unsuccessfully, but began lecturing
on logic to evening classes at King’s College
London. Soon after the publication of his second
book, Mathematical Psychics, in 1881, he applied
for a professorship of logic, mental and moral
philosophy and political economy at Liverpool.
Testimonials for two of Edgeworth’s applications
were given by Jevons (see Black , V, pp. 98, 145)
and Marshall.
Edgeworth had to wait until 1890 until he
obtained a professorial appointment: this was at
King’s College London, where he succeeded
Thorold Rogers in the Tooke Chair of Economic
Science and Statistics. In the next year, 1891, he
again succeeded Rogers, this time to become
Dmmmond Professor and Fellow of All Souls’
College, Oxford, a position he held until his
3470 Edgeworth, Francis Ysidro (1845-1926)

retirement in 1922. Edgeworth therefore finally characterized by an original use of techniques,

settled in Oxford at the age of 46 in what was to although he does not appear to have received a
become one of the most illustrious British chairs in formal training in mathematics. However, he came
economics. At the same time he became the first to economics from moral philosophy. The central
editor of the Economic Journal. He was editor or question of distributive justice, rather than simply
co-editor from its first issue until his death. He was the application of mathematics, dominated his
supported by Henry Higgs from 1892 to 1905, when attitude towards economics. His main argument was
the latter became the Prime Minister’s Private that mathematics provided powerful assistance to
Secretary, with further assistance provided at a later ‘unaided’ reason, and could check the conclusions
stage by Alfred Hoare. Keynes was a co-editor for reached by other methods. Thus:
15 years. After a tremendously creative period of He that will not verify his conclusions as far
as possible by mathematics, as it were
the late 1870s and 1880s, Edge- worth had become
bringing the ingots of common sense to be
firmly established as the leading economist, after assayed and coined at the mint of the
Marshall, in Britain. sovereign science, will hardly realise the
In addition to his work in economics, Edge- full value of what he holds, will want a
measure of what it will be worth in however
worth began a series of statistical papers in 1883.
slightly altered circumstances, a means of
He was President of section F of the British Asso- conveying and making it current. (1881, p. 3)
ciation in 1889, a position he held again in 1922.
Edgeworth’s approach contrasts sharply with that of
Edgeworth’s work on mathematical statistics played
Marshall. The contrast between Edgeworth and
an increasingly important role. Indeed, of about 170
Marshall was neatly summarized by Pigou as
papers which he published, approximately three-
quarters were concerned with statistical theory. He follows:
During some thirty years until their recent
became a Guy Medalist (Gold) of the Royal deaths in honoured age, the two outstanding
Statistical Society in 1907 and was President of the names in English economics were Marshall ...
Society during 1912-1914. His main contributions and Edge- worth... Edgeworth, the tool-maker,
to statistics concern work on inference and the law gloried in his tools ... Marshall, on the
other hand, had what almost amounted to an
of error, the correlation coefficient, transformations obsession for hiding his tools away. (Pigou
(what he called ‘methods of translation’), and the and Robertson, , p. 3)
‘Edgeworth expansion’. The latter, a series
Although both men turned to economics from
expansion which provides an alternative to the
mathematics and moral philosophy, Marshall gen-
Pearson family of distributions, has been widely
erally used biological analogies, and was concerned
used (particularly since the work of Sargan ) to
with developing maxims. In contrast, Edgeworth
improve on the central limit theorem in
generally used mechanical analogies, and was more
approximating sampling distributions. It has also
concerned with developing theorems.
been used to provide support for the bootstrap in
In the 1880s and 1890s the deductive method
providing an Edge- worth correction. Edgeworth’s
encountered a great deal of criticism, especially
work in probability and statistics has been collected
from the ‘Historical School’ of economists.
by McCann ( ). His third and final book was
Edgeworth’s defence of the deductive method often
Metretike: or
involved showing how other economists had
the Method of Measuring Probability and Utility
advocated its use. His interest in the natural
(1887). These contributions are not examined here;
sciences often led him to make comparisons with
see Bowley ( ) and Stigler ( ).
scientific laws, and especially to show that the
physical sciences also relied on abstraction and
approximation.
Approach to Economics
Edgeworth argued carefully that the assump-
A dominant characteristic of Edgeworth’s approach tions used in economics are often untestable, and he
therefore took precautions against the
to economics is that it is mathematical,
Edgeworth, Francis Ysidro (1845-1926)
accusation of‘plucking assumptions from the air’.
He was conscious of the fact that the difficulty is in
making the crucial abstractions which make the
particular problem under consideration tractable,
but which are not question begging. His attitude to
many a priori assumptions was directly related to
his approach to statistical inference. In Mathemat-
ical Psychics, for example, he referred to ‘the first
principle of probabilities, according to which cases
about which we are equally undecided ... count as
equal’ (1881, p. 99). This was then transferred to
economics. The appropriate assumption was that all
feasible values, say, of elasticities, were equally
likely, until evidence is obtained. Hence, ‘There is
required, I think ... in order to override the a priori
probability, either very definite specific evidence, or
the consensus of high authorities’ (1925, ii, pp. 390-
391). This also illustrates Edgeworth’s attitude to
authority and his many allusions to the views of
other leading economists. Price ( , p. 38)
referred to his
frequent ‘reference to authority for ... support of
tentative opinion waveringly advanced’.
Edgeworth was also prone to stress negative
results. For example, in discussing taxation, where
the criterion of minimum sacrifice does not alone
provide a simple tax formula, he stated:
Yet the premises, however inadequate to the deduc-
tion of a definite formula, may suffice for a certain
negative conclusion. The ground which will not
serve as the foundation of the elaborate edifice
designed may yet be solid enough to support a
battering-ram capable of being directed against sim-
pler edifices in the neighbourhood. (1925, p. 261)
Edgeworth’s position as editor of the Economic
Journal enabled him to combine both his critical
attitude and his appetite for a wide range of reading.
He contributed 32 book reviews, and in sending
books to other reviewers he would include ‘apposite
remarks on particular points in the text’ (Bowley ,
p. 123). These reviews should also be placed beside
his 17 reviews in the Academy, and 131 articles in
the original Palgrave's Dictionary of Political
Economy. Furthermore, Edgeworth’s later articles
in the Economic Journal, such as those on
international trade and on taxation, took the form of
extended commentaries on contemporary work.
3471
Early Work in Moral Philosophy
Before turning to economics, Edgeworth published
a brief note in Mind in 1876, and his first (privately
printed) book on New and Old Methods of Ethics in
1877. The description by Keynes of Edgeworth’s
first book could just as well be applied to his other
two books:
Edgeworth’s peculiarities of style, his
brilliance of phrasing, his obscurity of
connection, his inconclusiveness of aim, his
restlessness of direction, his courtesy, his
caution, his shrewdness, his wit, his subtlety,
his learning, his reserve -all are there full-
grown. Quotations from the Greek tread on the
heels of the differential calculus. (Keynes, p.
257)
The main focus of this early work, strongly
influenced by the great Cambridge philosopher
Henry Sidgwick (1838-1900), was to examine in
detail the implications of utilitarianism for the
optimal distribution of resources. Edgeworth’s
special and original contribution was to apply
advanced mathematics to this problem.
Edgeworth’s approach was dominated by his
utilitarianism, but the influence of contemporary
psychological research and the impact of evolu-
tionary ideas can also be traced. Both aspects led to
explicit consideration of differences between
individuals and changes which take place over time.
Edgeworth was also influenced by the major
fierce debates in the last half of the 19th century
between egoism, evolutionism, idealism, intui-
tionism, and of course utilitarianism. His brand of
utilitarianism became extremely eclectic, and
embraced the majority of the above principles
(except for those of the Hegelian idealists) while
regarding utilitarianism as the ‘sovereign principle’.
His note in Mind discussed Matthew Arnold’s
views of Joseph Butler, who had examined egoism
at great length. Arnold had argued that Butler’s
term ‘self love’ should be interpreted to mean ‘the
pursuit of our temporal good’. However, Edgeworth
argued that egoism and utilitarianism could be
subsumed under the same principle. He believed
Butler to be saying, ‘duty and interest are perfectly
coincident; for the most part in this world, but
entirely and in every
3472
instance, if we take in the future and the whole’
(1876, p. 571).
Edgeworth generally distinguished between
‘impure’ and ‘pure’ utilitarianism. In the latter case
individuals are assumed to be concerned with the
welfare of society as a whole. The former case in
fact corresponds more closely with a ‘short term’
version of egoism. Economic exchange can usefully
be analysed in terms of ‘jostling egoists’, but he
believed that ultimately individuals would evolve to
become pure utilitarians. A reason for believing that
individuals would make such a transition was later
to be developed by Edgeworth in the form of his
contractarian justification of utilitarianism as the
appropriate principle of distributive justice.
Edgeworth’s early utilitarianism was influenced
by his wide knowledge of work in experimental
psychology. In his books of 1877 and 1881 there are
many references to the work of Delboeuf, Fechner,
Elelmholtz, Weber and Wundt. These references
occur in the context of discussing the nature of
utility functions and, although Edgeworth at this
time was not aware of the earlier work of Jevons,
the same range of psychological work was also
important to Jevons. Edgeworth in 1877 explicitly
suggested, in connection with Fechner, that an
additive form would not be appropriate.
A further aspect, of Edgeworth’s utilitarianism
is his attitude towards authority. An important issue
for early utilitarians involved the nature of inductive
evidence about the consequences of acts. Most
people cannot know the full consequences of their
acts, so that rules of moral conduct must be
followed (in contrast with intuitionism where
individuals are assumed to have immediate
consciousness of moral rules). In arriving at such
rules, the opinions of highly regarded individuals
are taken to be credible though it may not be
possible to show conclusively that they are
‘correct’. Edgeworth argued, for example, that ‘we
ought to defer even to the undemonstrated dicta and
opinions of the wise, who have a power of mental
vision acquired by experience’ (1925, ii, p. 149).
Edgeworth defined the problem of determining
the optimal utilitarian distribution as follows:
Edgeworth, Francis Ysidro (1845-1926)
‘given a certain quantity of stimulus to be distrib-
uted among a given set of sentients ... to find the
law of distribution productive of the greatest
quantity of pleasure’ (1877, p. 43). In treating this
problem mathematically Edgeworth used Lagrange
multipliers, without any explanation, and concluded
that, ‘unto him that hath greater capacity for
pleasure shall be added more of the means of
pleasure’ (1877, p. 43). In using Lagrange
multipliers Edgeworth was also careful to discuss
possible complications, referring to the possibility
of multiple solutions and explicitly discussing
comer solutions and inequality constraints.
Further complexities were then examined,
where Edgeworth emphasized that utilitarianism
implies equality of the ‘means of pleasure’ only
under a special set of assumptions, and in the
general case the prescribed solution will be some
form of inequality. In dealing with the distribution
of effort, he argued not surprisingly that most work
should be provided by those most capable of
providing it. In a yet more general treatment of the
problem, Edgeworth used the calculus of variations,
but again provided the reader with virtually no help
in following his mathematical argument.
Edgeworth’s analysis of the utilitarian optimal
distribution was continued in his paper on ‘The
Hedonical Calculus’ (1879), which was later
reprinted as the third part of Mathematical
Psychics.
Early Work in Economics
The turning point in Edgeworth’s work was his
introduction to Jevons in 1879 by a mutual friend
James Sully, who in 1878 moved to Hampstead,
where Edgeworth had lodgings in Mount Vernon
and where Jevons also lived; see Sully ( ,
pp. 180, 223). His first knowledge of Marshall came
from Jevons, who ‘highly praised the then recently
published Economics of Industry’ (in Pigou , p. 66).
Edgeworth became interested in the problem of the
indeterminacy of the rate of exchange, arising from
the existence of only a small number of transactors.
This led rapidly to Edgeworth’s second and most
important
Edgeworth, Francis Ysidro (1845-1926) 3473

book Mathematical Psychics: An Essay on the ‘Mecanique Sociale’ may one day take her place
Application of Mathematics to the Moral Sciences along with ‘Mecanique Celeste’ [sic], throned
each upon the double-sided height of one
(1881), which was clearly written in a state of maximum principle, the supreme pinnacle of
considerable enthusiasm for his new subject. This moral as of physical science ... the movements
slim volume of 150 pages was known only to a of each soul, whether selfishly isolated or
small group of experts. Marshall’s review began, linked sympathetically, may continually be
realising the maximum energy of pleasure, the
‘this book shows clear signs of genius, and is a Divine love of the universe. (1881,
promise of great things to come’ (Whitaker , p. p. 12)
265). Jevons began by stating that ‘whatever else
Jevons’s work in the Theoiy of Political Econ-
readers of this book may think about it, they would
omy involved the application of very basic math-
probably all agree that it is a very remarkable one’
ematics and of psychological research to the
(1881, p. 581). It was not until the middle of the
analysis of exchange in competitive markets. In
20th century that many of its central ideas began to
addition to this direct stimulus, Edgeworth was also
be more fully appreciated.
influenced by an anonymous review of Jevons’s
Part 1 of Mathematical Psychics (1881, pp. 1-
book in the Saturday Review ( ).
15) was devoted mainly to a justification of the use
The cmcial development following Edgeworth’s
of mathematics in economics where precise data are
contact with Jevons was not simply the realization
not available. There is probably no other ‘apology’
that mathematics could be used to examine
in the whole of economic literature which compares
equilibrium in exchange. Rather, it was that in his
with Edgeworth’s plea for the application of
analysis Jevons explicitly assumed, through his Taw
mathematics. For example, when considering
of indifference’, that all individuals take the
individual utility maximization:
Atoms of pleasure are not easy to distinguish equilibrium prices as given, that is, outside their
and discern; more continuous than sand, more control. In using this law as ‘one of the central
discrete than liquid; as it were nuclei of the pivots of the theory’, Jevons stated that, ‘there can
just-perceivable, embedded in circumambient
only be one ratio of exchange of one uniform
semi-consciousness.
We cannot count the golden sands of life; we commodity at any moment’ ( ,
cannot number the ‘innumerable smile’ of seas p. 87). His theory was explicitly limited to the static
of love; but we seem to be capable of observing equilibrium conditions. He deliberately excluded
that there is here a greater, there a less, the role of the number of competitors from his
multitude of pleasure-units; mass of happiness;
and that is enough. (1881, pp. 8-9)
analysis via the awkward notion of the ‘trading
body’, following correspondence with Fleeming
Great stress was placed on comparison with
Jenkin (1833-1885), who raised the question of
Lagrange’s ‘principle of least action’ in examining
indeterminacy with just two traders; see Black (
the overall effects produced by the interactions
, iii, pp. 166-78). Jenkin could not
among many particles. The connection with
see why two isolated individuals should accept the
Edgeworth’s analysis of competition, involving
price-taking equilibrium, whereas Jevons wished to
interaction among a large number of competitors to
consider the behaviour of two typical individuals in
produce a determinate rate of exchange, is central
a large market.
here. The fact that in the natural sciences so much
In a section on ‘Failure of the Laws of
could be derived from a single principle was
Exchange’, Jevons discussed cases in which some
important for both Jevons and Edgeworth. But
indeterminacy would result. His most notable
Edgeworth took this to its ultimate limit in arguing
example was of house sales, where it was suggested
that the comparable single principle in social
that indeterminacy would result from the discrete
sciences, that of maximum utility, would produce
nature of the good being exchanged. The Saturday
results of comparable value. Referring to Laplace’s
Review article took exception to this, suggesting
massive work, Mecanique Celeste, he suggested
that indeterminacy ‘is really owing in our opinion to
that:
the assumed absence of competition’ (see Black , p.
157). The stress
3474
on indeterminacy was also influenced by Marshall’s
discussion of wage bargaining: Edgeworth (1881, p.
48 n, 1) referred to Thornton's comparison of the
determination of prices in Dutch and English
auctions, and cited Alfred and Mary Paley
Marshall’s joint book on the Economics of Industry
( 1879).'
It was this gap in Jevons’s analysis that Edge-
worth set out to fill. His achievement was to show
the conditions under which competition between
buyers and sellers, through a barter process, leads to
a ‘final settlement’ which is equivalent to one in
which all individuals act independently as price
takers. As he later stated (1925, p. 453), 'the
existence of a uniform rate of exchange between
any two commodities is perhaps not so much
axiomatic as deducible from the process of com-
petition in a perfect market’.
Exchange and Contract
Having argued that ‘the conception of Man as a
pleasure machine may justify and facilitate the
employment of mechanical terms and Mathematical
reasoning in social science’ (1881, p. 15).
Edgeworth moved on to the analysis of the ‘eco-
nomical calculus’, the starting point of which was
the assumption that ‘every agent is actuated only by
self-interest’ (1881, p. 16).
In modern economic analysis the analytical
tools invented by Edgeworth in 1881, such as the
indifference map and the contract curve, are now
used in a vast range of contexts. They were
introduced by Edgeworth to examine the nature of
barter among individuals. He wanted to see if a
determinate rate of exchange would be likely to
result in barter situations where it is assumed only
that individuals wish to maximize their own utility,
considered solely as a function of their own
consumption. With full knowledge of individuals’
utility functions, and their initial endowments of
goods, would it be possible to work out a
‘determinate’ rate of exchange at which trade would
take place? Edgeworth’s direct statement of the
problem is as follows:
Edgeworth, Francis Ysidro (1845-1926)
The PROBLEM to which attention is specially
directed in this introductory summary is: How far
contract is indeterminate - an inquiry of more than
theoretical importance, if it show not only that inde-
terminateness tends to [he present] widely, hut also
in what direction an escape from its evils is to be
sought. (1881. p. 20)
Edgeworth began his analysis of this problem by
taking the simplest case of two individuals
exchanging fixed quantities of two goods. The basic
framework is that described by Jevons, where the
first individual holds all of the initial stocks of the
first good, and the second individual holds all the
stocks of the second good. He wrote the utility
functions of each individual in terms of the amounts
exchanged rather than consumed, using the general
utility function (‘utility is regarded as a function of
the two variables, not the sum of two functions of
each’, 1881, p. 104). He then immediately defined
the contract curve and indifference curves, in that
order.
In the sentence which follows Edgeworth’s
introduction of the general utility function, he raised
the question of the equilibrium which may be
reached with 'one or both refusing to move further’.
In barter the conditions of exchange must be
reached by voluntary agreement, or contract,
between the two parties, and of course it is funda-
mental that no egoist would agree to a contract
which would make him worse off than before the
exchange. The question thus concerns the naUtre of
the settlement reached by two contracting parties.
He immediately answered that contract supplies
only part of the answer so that 'supplementary
conditions . .. supplied by competition or ethical
motives’ are required, and then wrote the equation
of his famous contract curve (1881, pp. 20-T).
The problem of obtaining the equilibrium values
ofx andy which, 'cannot be varied without the
consent of the parties to it' was stated as follows: 'It
is required to find a point (x, y) such that, in
whatever direction we take an infinitely small step,
\Uf\ and [Ufl\ do not increase together, but that,
while one increases, the other decreases’ (1881, p.
21). The locus of such points 'it is here proposed to
call the contract-curve’. Edgeworth’s alternative
derivations of the contract curve
Edgeworth, Francis Ysidro (1845-1926)
involved the movement, from an arbitrary position,
along one person’s indifference curve; ‘motion is
possible so long as, one party not losing, the other
gains’ (1881, p. 23). He thus used the Lagrange
multiplier method of maximizing one person’s
utility subject to the condition that the other
person’s utility remains constant.
In the diagram drawn by Edgeworth (1881, p.
28) he did not use a box constmction. Furthermore
the only indifference curves shown fully were those
which each individual is able to reach in isolation,
and which therefore specify the limits beyond
which each is not prepared to move. Also part of the
offer or reciprocal demand curves of each
individual were drawn on the same diagram,
although they were not defined until ten pages later.
After presenting the results for the two-person
two-good case, Edgeworth (1881, p. 26) examined
the contract curve in the case where three
individuals exchange three goods, stated that it is
given by the ‘eliminant’, and then gave three lines
of three sets of partial derivatives. In fact, the
au,
contract curve in this context is defined by dij -
where —is the marginal utility of person i with
respect to good j, but Edgeworth did not use the
modem notation for determinants and did not set
the Jacobian equal to zero. This early use of deter-
minants in economics would probably have con-
fused many of his readers.
The Problem of Indeterminacy
The concepts of indifference curves and the contract
curve therefore help to specify a range of ‘efficient
exchanges’ of goods between individuals. The
essential feature of the analysis from Edgeworth’s
point of view is precisely that there is a range rather
than a unique point: ‘the settlements are represented
by an indefinite number of points’ (1881, p. 29). At
any particular settlement, the rate of exchange is
expressed simply in terms of the amount of one
good which is given up in order to obtain a
specified amount of the other good. Hence the
existence of a range of efficient
3475
contracts means that the rate of exchange is ‘inde-
terminate’. The rate of exchange achieved in prac-
tice will thus depend to a large extent on bargaining
strength. It was this result which led Edgeworth to
make his often quoted remark that ‘an accessory
evil of indeterminate contract is the tendency,
greater than in a full market, towards dissimulation
and objectionable arts of higgling’ (1881, p. 30).
Edgeworth argued that his analysis of indeter-
minacy in contract between two traders could be
applied to a very wide variety of contexts. In
particular, the tendency of large groups to form
‘combinations’, as in the case of trade unions and
employers’ associations, would serve to increase
the extent of indeterminacy. The general applica-
bility of his analysis of contract and indeterminacy
was summarized by Edgeworth as follows:
What it has been sought to bring clearly into view is
the essential identity (in the midst of diversity of
fields and articles) of contract; a sort of unification
likely to be distasteful to those excellent persons
who are always dividing the One into the Many, but
do not appear very ready to subsume the Many under
the One. (1881, p. 146; Plato’s expression ‘the one in
the many’ was later used by Marshall as the motto
for his 1919 book on Industry and Trade)
Having shown the possibilities of indetermi-
nacy, Edgeworth then went on to show how ‘the
escape from its evils’ requires either competition or
arbitration.
Competition and the Number of Traders
The central question which Edgeworth was trying to
resolve in the second part of Mathematical Psychics
was that of the conditions necessary to remove the
indeterminacy which exists in the case of barter
between two traders. The question naturally arises
as to the extent to which this indeterminacy is the
result of the absence of competition in the simple
two-person market. Edgeworth thus quickly moved
on to the introduction of further traders.
In Edgeworth’s earlier problem of two traders
exchanging two goods, the definition of a range of
3476
efficient exchanges (along the contract curve) is of
course analytically separate from the question of
whether or not two isolated traders would actually
reach a settlement on the contract curve. However,
these two aspects were not clearly separated by
Edgeworth because at the beginning of his analysis
he introduced his stylized description of the process
of barter: this is the famous ‘recontracting’ process.
Edgeworth did not wish to assume that individuals
initially have perfect knowledge. Instead, he
supposed that, ‘There is free communication
throughout a normal competitive field. You might
suppose the constituent individuals collected at a
point, or connected by telephones - an ideal
supposition, but sufficiently approximate to
existence or tendency for the purposes of abstract
science’ (1881, p. 18). The knowledge of the other
traders’ dispositions and resources could be
obtained by the formation of tentative contracts
which are not assumed to involve actual transfers,
and can be broken when further information is
obtained. Edgeworth introduced this in typical style:
‘Is it peace or war?’ asks the lover of ‘Maud’, of
economic competition, and answers hastily: it is
both, pax or pact between contractors during con-
tract, war, when some of the contractors without the
consent of others recontract. (1881, p. 17; the allu-
sion here is to Alfred Tennyson’s poem Maud: A
Monodrama, part 1, verse VII.)
An important role of the recontracting process is
thus to disseminate information among traders. It
allows individuals who initially agree to a contract,
which is not on the contract curve, to discover that
an opportunity exists for making an improved
contract according to which at least one person
gains without another suffering.
However, the real importance of the recon-
tracting process lies in the fact that it allows for
Edgeworth’s analysis of the role of the number of
individuals in a market. With numerous individuals,
the recontracting process makes it possible to
analyse the use of collusion among some of the
traders. Individuals are allowed to form coalitions in
order to improve bargaining strength. Recontracting
enables the coalitions to be broken up by outsiders
who may attract members of a group away with
more favourable terms of exchange.
Edgeworth, Francis Ysidro (1845-1926)
Edgeworth’s analysis was extremely terse and
the following discussion does not therefore follow
his own presentation. The analysis begins by
introducing a second person A and a second person
B. The new traders are assumed to be exact replicas
of the initial pair, with the same tastes and
endowments. This simplification is useful because
the dimensions of the Edgeworth box and the utility
curves are identical for each pair of traders. Hence,
it enables the same diagram to be used as in the
case when only two traders are considered in
isolation. Two basic points can be stated imme-
diately. First, in the final settlement all individuals
will be at a common point in the Edgeworth box.
Second, the settlement must be on the contract
curve. The first point arises because if two indi-
viduals have identical tastes then their total utility is
maximized by sharing their resources equally. It is
useful to consider other types of contract which will
eventually be broken, in order to illustrate the way
in which the introduction of additional traders
provides a role for some kind of competitive
process.
The major question at issue is whether the range
of indeterminacy along the contract curve is
reduced by the addition of these traders. Consider
Fig. and suppose that when A i and B i are trading
independently of A2 and B2, trader i?! has all the
bargaining power and is able to appropriate all the
gains from trade by pushing A! to the limit of the
contract curve at point C. Suppose also that the
same applies to A2 and B2. If the two pairs of traders
are then able to communicate with each other, A2
can now simply refuse to trade with B2 at C. With
no transaction costs, A2 was previously indifferent
between trading at C and consuming at the
endowment point, E. This endowment position is
effectively the ‘threat point’ of the As: it is the
position in which they would find themselves if the
bargaining process were to break down. But A2 no
longer needs to remain in isolation after refusing to
trade with B2, and instead can trade with A\, after A\
has traded with B\ at C and has therefore obtained
some of good Y. The two As can share their stocks
ofX and /equally, arriving at point P; such an equal
division maximizes their total utility.
Edgeworth, Francis Ysidro (1845-1926)
Edgeworth, Francis
Ysidro (1845-1926),
Fig. 1 Two pairs of traders
By reaching point P, halfway between C and E,
the convexity of the indifference curves implies that
they are both better off than anywhere on the no-
trade indifference curve. The two As would be on a
higher common indifference curve, and thus better-
off, if they could consume at a point along the CPE
which is to the north-west of point P However, they
do not have enough resources to move beyond the
halfway point P.
Trader B2, who has been isolated, cannot pre-
vent such a bargain. Thus B i is at C, both A s arc at
P and B2 is at the initial endowment point E. In this
situation Bx has no incentive to change, but B2 has a
strong incentive to offer a better deal to one of the
As than the one offered by trader B:. So long as B2
offers one of the As, say A2, a trade on the contract
curve which allows A2 to reach a higher
indifference curve than U'A, the initial agreement
with B | will be broken and recontracting will take
place.
The implication is that the ability of the As to
turn to someone else, rather than deal with a single
trader, means that the Bs now compete against each
other. However, trader B\, who cannot prevent the
recontracting, has an incentive to make yet a better
offer. Hence, the recontracting process continues.
The stylized process of recontracting with the two
Bs competing against each other will produce a
final settlement at the point C* in Fig. . This has the
property that the indifference curve
3477
U"l passes through C* and P*, where P* is halfway
between C* and E. This means that the two As are
indifferent between C* and P*, and since they
cannot both reach any point between C* and P*
along the line C* E, they are unable to improve on
C*. Hence there is no need to leave one of the Bs in
isolation and the two Bs will trade with the two As
at point C*.
This argument has shown that at the final set-
tlement all traders are at a common point on the
contract curve and the limit has moved inwards
along the old contract curve. The analysis can be
repeated by starting with an alternative situation
whereby the As are initially assumed to be able to
appropriate all the gains from trade. The point C
would then no longer qualify as a point on the new
contract curve. The introduction of the additional
pair of traders means that the contract curve
shrinks, and the range of indeterminacy involved in
barter is correspondingly reduced.
The extent to which the contract curve shrinks
when the additional pair of traders is introduced is
influenced by the fact that the As cannot get further
than halfway along a ray from a point on the
contract curve to the endowment position.
However, if there are three pairs of As and three
pairs of Bs, the repetition of the above analysis
involves two of the As dealing with two of the Bs at
a point on the contract curve. The two As then share
their resources equally with the remaining
3478
Edgeworth, Francis
Ysidro (1845-1926),
Fig. 2 The new limit to the
contract curve
A while the third B is isolated. The As are able to
consume together at a point which is two-thirds of
the way along the ray from the initial endowment
position to the point on the contract curve where the
trade involving the two As and two Bs takes place.
With N pairs, the As can reach a proportion 'AA 1
of the way from the endowment point to the
contract curve. Thus as TV increases, the values of
k approaches unity. This means that the As can
reach all the way from E to the contract curve, so
that the final settlement must be such that the
indifference curve is tangential to the ray from the
origin. A final settlement with many traders is
therefore shown in Fig. as point P on the contract
curve. The effect of working in from the point C
would lead to an equivalent result for an
indifference curve of the Bs, shown as U*B.
The result is that the final settlement looks just
like a price-taking equilibrium. The figure illustrates
the case where there is a single price-taking
equilibrium. If there are multiple equilibria, the
reconfiacting process causes the number of final
settlements, with sufficiently large N, to shrink to
the number of price-taking equilibria. (For discus-
sion of utility functions involving multiple equi-
libria, and comparison of bargaining, competitive
and utilitarian solutions, see Creedy .) This
argument relating to the shrinking contract curve,
first established by Edgeworth, is often referred to
as the limit theorem.
Edgeworth, Francis Ysidro (1845-1926)
OB
GoodX
After Edgeworth’s terse discussion, he stated:
If this reasoning does not seem satisfactory, it would
be possible to give a more formal proof; bringing out
the important result that the common tangent to both
indifference curves ... is the vector from the origin.
(1881, p. 38)
The price-taking solution is necessarily on the
contract curve. This gives rise to what is now
referred to as the ‘first fundamental theorem’ of
welfare economics -that a price-taking equilibrium
is Pareto efficient. Furthermore, the use of price-
taking provides a considerable reduction in the
amount of information required by traders when
compared with the reconfiacting process. Given an
equilibrium set, individuals need to know only the
prices of goods, whereas in the reconfiacting
process they have to leam a considerable amount of
information about other individuals’ preferences
and endowments. But Edgeworth placed more stress
on the equivalence of the competitive price-taking
solution with a reconfiacting barter process
involving large numbers.
Given that coalitions among traders are allowed
in the reconfiacting process, a price-taking
equilibrium cannot be blocked by a coalition of
traders. In this sense the competitive equilibrium is
robust. The argument that a complex process of
bargaining among a large number of individuals
produces a result which replicates a price-taking
equilibrium, allowing for the free flow of
Edgeworth, Francis Ysidro (1845-1926)
Edgeworth, Francis
Ysidro (1845-1926),
Fig. 3 F i n a l
settlement with
many traders
information using recontracting and enabling coa-
litions of traders to form and break up, is an
important result that is far from intuitively obvious.
The recontracting process can be said to represent a
competitive process, and the contract curve shrinks
essentially because of the competition between
suppliers of the same good, although it is carried out
in a barter framework in which explicit prices are
not used (although rates of exchange are equivalent
to price ratios).
The price-taking equilibrium, in contrast, does
not actually involve a competitive process. Indi-
viduals simply believe that they must take market
prices as given and outside their control. They
respond to those prices without any reference to
other individuals. But the result is that the price-
taking equilibrium looks just like a situation in
which all activity is perfectly coordinated.
Edgeworth suggested that similar results apply
when some of the assumptions are relaxed. Thus,
‘when we suppose plurality of natures as well as
persons, we have to suppose a plurality of contract-
curves ... Then, by considerations analogous to
those already employed, it may appear that the
quantity of final settlements is diminished as the
number of competitors is increased’ (1881, p. 40).
He then briefly considered different numbers of As
and Bs, concluding that ‘the theorem admits of
being extended to the general case of
3479
0B
GoodX
unequal numbers and natures’ (1881, p. 43). How-
ever, some of the results do not hold in the general
case; for example, equality within the group of As
no longer holds when there are unequal numbers of
As and Bs. A considerable number of articles have
been written, since the late 1950s, examining
various aspects of the Edgeworth recontract model
under different assumptions.
Reciprocal Demand Curves
It has been mentioned that Edgeworth included in
his diagram (1881, p. 28) the reciprocal demand
curve, or offer curve, of each individual, although
such curves were then called ‘demand-and-supply
curves’. Edgeworth mentioned them only briefly in
the text (1881, p. 39), but the lack of emphasis is
understandable since in imperfect competition they
are not relevant. Edgeworth’s contribution was to
provide the basic ‘analytics’ of the offer curve in
terms of indifference curves, whereby it is ‘the
locus of the point where lines from the origin touch
curves of indifference’ (1881, p. 113).
When there is a lack of competition, giving rise
to indeterminacy, there is nothing to ensure that
individuals will trade on their offer curves and, as
Edgeworth argued, ‘the conceptions of demand and
supply at a price are no longer appropriate’
3480
(1881, p. 31). It is this general preference, in favour
of the analysis of barter in noncompetitive
situations, to which Marshall objected and which
led to the controversy discussed below.
The Utilitarian Calculus
Having shown how indeterminacy can be removed
by increasing the number of traders, Edgeworth
turned to consider the role of arbitration in resolving
the conflict between traders, in a ‘world weary of
strife’ (1881, p. 51). The principle of arbitration
examined was, not surprisingly, the utilitarian
principle, which Edgeworth had earlier used to
examine the optimal distribution. However, the new
context of indeterminacy led him to a deeper
justification of utilitarianism as a principle of
distributive justice. Having arrived at this new link
between ‘impure’ (egoistic) and ‘pure’
utilitarianism, Edgeworth had only to reorientate his
earlier analysis of optimal distribution, contained in
his paper in Mind of 1879.
The need for arbitration with indeterminacy had
been stated by Jevons as follows:
The dispositions and force of character of the
parties ... will influence the decision. These are
motives more or less extraneous to a theory of
economics, and yet they appear necessary
considerations in this problem. It may be that
indeterminate bargains of this kind are best arranged
by an arbitrator or third party.) 171, pp. 124-5)
Edgeworth’s statement of the same point was as
usual rather less prosaic: ‘The whole creation
groans and yearns, desiderating a principle of
arbitration, and end of strifes’ (1881, p. 51). Edge-
worth argument involved two steps. First, he
showed that the principle of utility maximization
places individuals on the contract curve, because the
first-order conditions are equivalent to the tangency
of indifference curves.
It is a circumstance of momentous interest that one
of the in general indefinitely numerous settlements
between contractors is the utilitarian arrangement ...
the contract tending to the greatest possible total
utility of the contractors. (1881, p. 53)
Edgeworth recognized that this result was not
sufficient to justify the use of utilitarianism as a
Edgeworth, Francis Ysidro (1845-1926)
principle of arbitration. It is only a necessary
condition of a principle of arbitration that it should
place the parties somewhere on the contract curve.
Edgeworth’s justification for utilitarianism as a
principle of justice, comparing points along the
contract curve, was as follows:
Now these positions lie in a reverse order of desir-
ability for each party; and it may seem to each that as
he cannot have his own way, in the absence of any
definite principle of selection, he has about as good a
chance of one of the arrangements as another... both
parties may agree to commute their chance of any of
the arrangements for ... the utilitarian arrangement.
(1881, p. 55)
The important point to stress about this statement is
that Edgeworth clearly viewed distributive justice
in terms of choice under uncertainty. He argued that
the contractors, faced with uncertainty about their
prospects, would choose to accept an arrangement
along utilitarian lines. A cmcial component of this
argument, also clearly stated by Edgeworth in this
quotation, is the use of equal a priori probabilities.
The importance to him of this new justification
of utilitarianism cannot be exaggerated. Indeed the
whole of Mathematical Psychics seems to be
imbued with a feeling of excitement generated by
his discovery of a justification based on a ‘social
contract’. This provided the cmcial link between
‘impure’ and ‘pure’ utilitarianism in a more satis-
factory way than his earlier appeal to evolutionary
forces.
Edgeworth believed that he had provided an
answer to an age-old question, stating ‘by what
mechanism the force of self-love can be applied so
as to support the structure of utilitarian politics,
neither Helvetius, nor Bentham, nor any deductive
egoist has made clear’ (1881, p. 128). Nevertheless
this argument was neglected until restatements
along similar lines were made by Harsanyi ( , )
and Vickrey ( ). The
maximization of expected utility, with each indi-
vidual taking the a priori view that any outcome is
equally likely, was shown to lead to the use of a
social welfare function which maximizes the sum of
individual utilities. This approach is now usually
described as ‘contractarian neo-utilitarianism'.
Edgeworth, Francis Ysidro (1845-1926)
In discussing the utilitarian solution as a prin-
ciple of arbitration in indeterminate contract,
Edgeworth did not clearly indicate in 1881 that the
utilitarian solution of maximum total utility could
specify a position which makes one of the parties
worse off than in the no-trade situation. This was
nevertheless later made explicit when, alter
proposing arbitration along utilitarian lines, he
added ‘subject to the condition that neither should
lose by the contract’ (1925, ii, p. 102). This
possibility of course depends largely on the initial
endowments of the individuals.
Later Work in Economics
After the publication of Mathematical Psychics,
Edgeworth concentrated increasingly on mathe-
matical statistics, in particular on the problem of
statistical inference, but, following his appointment
to the Drummond Chair at Oxford, Edge- worth
again made important contributions to economics,
although this work mainly involved reactions to,
and discussions arising from, the later work of other
authors.
Demand and Exchange
In the Principles of Economics (1890, Appendix F)
Marshall included a brief discussion of Edgeworth’s
analysis of barter, and produced a figure showing
the contract curve. During the following year, in the
course of a review written in Italian (translated in
Edgeworth, 1925, ii, pp. 315-19), Edgeworth
criticized Marshall for not having dealt sufficiently
with the problem of indeterminacy. The basic
problem was that Marshall, using a model in which
a series of trades are allowed to take place at
disequilibrium prices, believed he had shown that
prices will eventually settle at the price-taking
equilibrium. However, the argument was not
transparent.
The adjustment process involves moving from
the initial endowment point in a series of trades,
where trading at ‘false’ prices is allowed at each
step. The process must conclude with both indi-
viduals at a point on the contract curve. A feature of
the process is the assumption that each stage or
iteration of the sequence involves Pareto
3481
improvements: individuals trade only if it makes
them better off. Furthermore, it involves trading at
the ‘short end’ of the market, that is, the minimum
of supply and demand. This arises from the
impossibility of forcing any individual either to buy
or sell more than desired at any price.
An example of two disequilibrium trades is
shown in Fig. , where the endowment moves from
E to Ei, and then to Ex- With a price line
represented by EP, there is an excess supply of
good X as person A tries to reach the indifference
curve U'A and person B wishes to reach U'B. Trade
takes place at £), the short end of the market. Point
Ei then becomes the new endowment point. At the
second trading stage, the price of X must be lowered
to induce person B to purchase more. At a price
represented by the line ExPi through the new
endowment point, the excess supply is lower than
formerly and trade takes place at E2. Comparing U'A
and U'B with U"A and U"B with UA and UB
respectively, it can be seen that E2 is a Pareto
improvement relative to £). It is also clear that
person A is better off the slower the fall is in the
price of X relative to Tat each stage.
The combination of Pareto-efficient moves at
each stage and an adjustment process such that an
excess supply leads to a price reduction, and vice
versa, produces a stable process that converges to an
equilibrium somewhere on the contract curve. (This
type of sequence of disequilibrium trades was later
used by Launhardt; see Creedy .)
The basic problem was that Marshall believed
that his assumption of an additive utility function,
combined with the assumption that the marginal
utility of one good is constant for both individuals,
guaranteed a determinate price, if the good having
constant marginal utility was money. Indeed, this
case was mentioned by Edgeworth (see 1925, ii, p.
317 n.l). The contract curve is a straight line parallel
to the y axis (where this good is the one with
constant marginal utility), along which the rate of
exchange is constant. So the equilibrium price does
not depend on the sequence of trades. However,
Edgeworth’s point was that the total amount spent
on good x remains indeterminate.
There was a later, though much milder, dis-
agreement between Marshall and Edgeworth
3482
Edgeworth, Francis °B
Ysidro (1845-1926),
Fig. 4
Disequilibrium
trades
over the so-called Giffen good. In a book review,
Edgeworth argued that, ‘even the milder statement
that the elasticity of demand for wheat may be
positive, though I know it is countenanced by high
authority, appears to me so contrary to a priori
probability as to require very strong evidence (1909,
p. 104). The ‘authority’ was of course Marshall (
, p. 132), who replied
directly to Edgeworth that, ‘I don’t want to argue ...
But... the matter has not been taken quite at random’
(Pigou , p. 438). Marshall gave a numerical
example involving a journey travelled by two
methods, where the distance travelled by the
cheaper and slower method must increase when its
price increases. For further details, see Greedy (
).
It has been mentioned that Edgeworth intro-
duced the generalized utility function. An impli-
cation is that it allows for complementarity,
although Edgeworth did not explicitly consider this
in 1881. The first formal definition of com-
plementarity is attributed to Auspitz and Lieben,
and it was used by Edgeworth in his paper on the
pure theory of monopoly, and also by Pareto: this
amounts to what is now called ‘gross’ comple-
mentarity, defined in terms of cross-price elastic-
ities. It is also sometimes referred to, using the
initials of the four people mentioned above, as
ALEP complementarity.
Edgeworth, Francis Ysidro (1845-1926)
Good X
The first major criticism came from Johnson (
), who pointed out that the criterion was not
invariant with respect to monotonic transformations
of the utility function. His treatment was extended
by Hicks and Allen ( ), so that the
modem definition involves ‘net’ complements in
terms of compensated price changes. There is no
symmetry between gross substitutes and comple-
ments as only the matrix of (compensated) substi-
tution elasticities is assumed to be symmetric.
Monopoly and Oligopoly
In a paper first published in Italian in 1897, and not
translated until the collected Papers (1925),
Edgeworth examined several problems relating to
monopoly. He began his discussion with Cournot’s
( ) example of the ‘source
minerale’ in which there are ‘two monopolists’
(that is, duopolists ), each owning a spring of min-
eral water. It would be natural for Edgeworth to
expect an indeterminate price in this ‘small num-
bers’ context. Cournot had arrived at a determinate
solution for price and output, but Edgeworth
showed that ‘when two or more monopolists are
dealing with competitive groups, economic equi-
librium is indeterminate’ (1925, p. 116). The daily
output from each spring was assumed to be limited
to identical fixed amounts, delivery costs were zero
and all consumers had the same demand
Edgeworth, Francis Ysidro (1845-1926) 3483

curve (purchasing one unit only of output). Hence homogeneous units of inputs of organization and
demand is n( 1 - p) where n is the number of management services are subsumed in the costs of
customers and p is the price. Cournot’s solution was the firm.
that the price would be p 1/4, but Edge- worth Edgeworth’s criticisms of this concept of the no-
argued that one of the ‘monopolists’ had an profit entrepreneur, reproduced in his Papers
incentive to raise the price back to p 1/2, which is (1925), recognized that with Walras’s assumptions
the revenue maximizing price, so that there is not a there was nothing illogical about the argument. The
determinate price. He argued that: theory simply means that nothing remains ‘after the
entrepreneur has paid a normal salary to himself
at every stage ... it is competent to each monopolist (1925, pp. 26,30). Furthermore, ‘if [the general
to deliberate whether it will pay him better to lower
his price against his rival as already described, or expenses] are taken into account, the argument
rather to raise it to a higher ... for that remainder of becomes a fortiori. For why should not a substantial
customers of which he cannot be deprived by his remuneration for the entrepreneur be included in the
rival. ... Long before the lowest point has been general expenses of the business’ (1925, ii, p. 469).
reached, that alternative will have become more
advantageous than the course first described’ (1925,
Edgeworth’s difference with Walras was to some
p. 120) extent ‘only verbal’, but he was also unhappy with
the idea that entrepreneurship is homogeneous and
Edgeworth went on to say ‘the matter may be put in divisible.
a clearer light’, and he then defined what are now
called the reaction curve and isoprofit lines (in that The Theory of Taxation
order) for variations in prices. However, it was not In the 1890s Edgeworth produced two surveys of
until Bowley’s ( ) discussion that considerable importance. These surveys, of the pure
these matters began to be presented in a more theory of taxation and of the pure theory of
transparent manner. international values, were both published in the
Edgeworth then considered the case of com- Economic Journal and subsequently reproduced
plementary demand within the context of ‘bilateral (with alterations) in his Papers (1925, vol. ii). Each
monopoly’, where the two goods are demanded in survey consisted of three separate parts, and
fixed proportions for use in the production of a displayed a staggering breadth of knowledge and
further article. An interesting feature is that he command of the subject. They represent his most
wrote the equations of the reaction curves and serious attempts to produce any kind of synthesis of
explicitly dealt with what are now called conjectural a branch of economic literature. Edgeworth began
variations, reflecting the extent to which one his survey with the rather strong statement that ‘the
duopolist is expected to change price in response to science of taxation comprises two subjects to which
changes made by the second duopolist. In the character of pure theory may be ascribed; the
discussing this problem Edgeworth also introduced laws of incidence, and the principle of equal
the further important concept of the ‘saddle point’, sacrifice’ (1925, p.64). He then considered a variety
which he called the ‘hog’s back’, clearly indicating of special cases and contexts of tax incidence. The
its importance for stability. basic framework for incidence analysis was the
simple partial equilibrium approach, still used in
The No-Profit Entrepreneur many basic textbooks, in which the incidence
Walras ( , p. 225) had introduced the concept depends on the relative values of supply and
of the entrepreneur who neither gains nor loses. demand elasticities.
This result applied only to the competitive equi- The basic approach to incidence analysis actu-
librium, where there are no incentives for entre- ally stemmed from the important paper by Jenkin (
preneurs to enter any industry. This does not of ). It suggests that in general the price of the
course mean that there are no profits, in the taxed good will either remain constant (in the
accounting sense, since the returns to extreme case of inelastic supply ) or will increase.
However, this result ignores interrelationships
3484
among commodities. Edgeworth showed that, when
such interrelationships are explicitly allowed, there
are some circumstances in which the price of the
taxed good will actually fall. When discussing this
‘paradox’, Edgeworth reproduced his argument
which had in fact been explored in more detail in
his paper on monopoly, published in Italian in the
same year (translated in Edge- worth, 1925, i, pp.
111-42). Edgeworth first stated his ‘tax paradox’ in
the following terms:
when the supply of two or more correlated
commodities - such as the carriage of passengers by
rail first class or third class is in the hands of a single
monopolist, a tax on one of the articles - e.g. a
percentage of first class fares - may prove
advantageous to the consumers as a whole. ... The
fares for all the classes might be reduced. (1925, p.
139)
Edgeworth regarded this result as an example of a
situation where, ‘the abstract reasoning serves as a
corrective to what has been called the “metaphys-
ical incumbus” of dogmatic laisser faire’ (1925, i, p.
139; see also 1925, ii, pp. 93-4). Essentially the two
commodities must be substitutes in consumption
and production, and the result is partly brought
about by the fact that the monopolist has an
incentive to increase the supply of the untaxed
commodity. Edgeworth also recognized that the
result could occur in competitive markets (see 1925,
p. 63 ). As with many of Edgeworth’s original
results, this tax paradox was not a subject of
continuous development. Its main practical
importance perhaps arises from the fact that in the
early 1930s it attracted the attention of Hotelling (
). For further discussion of the
paradox, see Creedy ( ).
The section of the taxation survey which
attracted most immediate attention was
Edgeworth’s discussion of the various ‘sacrifice’
theories of the distribution of the tax burden, and his
qualified support for progressive taxation.
Edgeworth’s attitude to taxation was similar to that
of the major classical economists in that he rejected
a benefit approach, on the argument that taxation is
not an economic bargain governed by competition.
Thus in his view the problem was to determine ‘the
distribution of those taxes which are applied to
common purposes, the benefits
Edgeworth, Francis Ysidro (1845-1926)
whereof cannot be allocated to particular classes of
citizens’ (1925, p. 103). A principle of justice is
thus required. His approach can be seen as marking
a crucial stage in the transition towards a ‘welfare
economics’ view ofpublic finance, rather than using
a special set of ‘tax maxims’ such as the famous
criteria laid down by Adam Smith.
Not surprisingly, Edgeworth (1925, p. 102)
argued along neo-contractarian lines set down in
Mathematical Psychics that the utilitarian
arrangement would be accepted by individuals
uncertain of their own prospects and taking an equal
a priori view of the probabilities. He suggested that
each party may reflect that, in the long run of various
cases ... of all the principles of distribution which
would afford him now a greater, now a smaller
proportion of the sum-total utility obtainable ... the
principle that the collective utility should be on each
occasion a maximum is most likely to afford the
greatest utility in the long run to him individually
Having established the use of utilitarianism as a
principle of distributive justice, Edgeworth then
succinctly stated the main argument:
The condition that the total net utility procured by
taxation should be a maximum then reduces to the
condition that the total disutility should be a mini-
mum ... it follows in general that the marginal
disutility incurred by each taxpayer should be the
same. (1925, p. 103)
The implication is that, if all individuals have the
same cardinal utility function, after-tax incomes
would be equalized. Edgeworth also clearly rec-
ognized that, if there is considerable dispersion of
pre-tax incomes relative to the total amount of tax to
be raised, where there is ‘not enough tax to go
around’ (1925, ii, p. 103), the equimarginal con-
dition cannot be fully satisfied unless there is a
‘negative income tax’ which raises the incomes of
the poorest individuals to a common level. Thus,
‘the acme of socialism is for a moment sighted’
(1925, p. 104). But Edgeworth immediately con-
sidered the practical limitations to such high pro-
gressive taxation. The following quotation
illustrates one of Edgeworth’s favourite metaphors,
his respect for Sidgwick, his attitude to authority,
his views on utilitarianism and the
Edgeworth, Francis Ysidro (1845-1926) 3485

applicability of pure theory, and of course his One of Edgeworth’s criticisms of Mill ( )
unmistakable style: was that the latter took as his measure of the gain
In this misty and precipitous region let us take from trade the change in the ratio of exchange of
Professor Sidgwick as our chief guide. He best has exports against imports. Thus Mill in this case
contemplated the crowning height of the utilitarian
first principle, from which the steps of a sublime ‘confounds “final” with integral utility’ (1925, p.
deduction lead to the high tableland of equality; but 22). The same point had in fact been made by
he also discerns the enormous interposing chasms Jevons ( , pp. 154-6). However, Edgeworth,
which deter practical wisdom from moving directly while preferring total utility, admitted that Mill was
towards that ideal. (1925, p. 104)
not otherwise led to serious error in using his own
Among the various limitations, Edgeworth noted measure.
differences in individual utility functions, popula- Edgeworth’s survey was, as always, extremely
tion effects, the disincentives to work, growth of wide-ranging, though for later developments the
culture and knowledge, savings, and of course the most interesting parts are concerned with his elu-
problem of evasion. cidation of Mill’s ‘recognition of the case in which
an impediment may be beneficial - or an improve-
International Trade ment prejudicial - to one of the countries’ (1925, p.
Edgeworth’s survey of the pure theory of interna- 9). These cases would now be discussed under the
tional values was in some ways responsible for a headings of the ‘optimal tariff’ and ‘immiserizing
change of emphasis in the approach to trade theory, growth’. In the case of an optimal tariff, a country
despite the fact that it contained few original ana- acts as monopolist and imposes a price which
lytical contributions. Indeed, he said that, ‘Mill’s enables that country to attain its highest indifference
exposition of the general theory is still unsurpassed’ curve, subject to the other country’s offer curve.
(1925, p. 20), and acknowledged further that, ‘what However, this position is not on the contract curve.
is written... after a perusal of [Marshall’s] privately The detailed specification of the optimum tariff in
circulated chapters ... can make no claim to origi- terms of elasticities had to wait until Bickerdike (
nality’ (1925, p. 46). Edgeworth saw trade theory as ) and Pigou
an application of the general theory of exchange: ( ) and the later revivals of interest in the
The fundamental principle of international trade is 1940s. Edgeworth’s judgement of Bickerdike was
that general theory ... the Theory of Exchange ... that he had ‘accomplished a wonderful feat. He has
which... constitutes the ‘kernel’ ofmost of the chief
problems in economics. It is a corollary of the
said something new about protection’ (1925, ii, p.
general theory that all the parties to a bargain look to 344).
gain by it... This is the generalised statement of the Edgeworth could not of course be expected to
theory of comparative cost. (1925, p. 6) support the use of such tariffs in practice. He
Thus the gains from trade are analogous to the gains acknowledged the possibility of retaliation, but also:
For one nation to benefit itself at the expense of...
from exchange in simple barter and ‘It is useful... to others is contrary to the highest morality ... But in an
contemplate the theory of distribution as analogous abstract study upon the motion of projectiles in
to that of international trade proper’ (1925, p. 19). vacuo, I do not think it necessary to enlarge upon the
Hence trade theory is to Edgeworth simply one horrors of war. (1925, p. 17 n. 5)
more application of the general method of The ‘highest morality’ was, of course, the principal
Mathematical Psychics. In directly applying the of utilitarianism.
theory of exchange to that of trade, Edgeworth was
quite content to use community indifference curves
without clearly specifying how aggregation might Conclusions
be carried out. He said only that ‘by combining
properly the utility curves for all the individuals, we It has been seen that Edgeworth did not begin
obtain what may be called a collective utility curve’ working and writing in economics until his
(1925, p. 293).
3486
mid-30s, but in common with the majority of
neoclassical economists he soon pursued an aca-
demic career as a professor of economics. Indeed, in
a period which saw the rapid and widespread
professionalization of the subject Edgeworth held
an academic position in England that was regarded
as second only to that of Alfred Marshall. In spite of
his wide range of reading and sympathies,
Edgeworth’s work was characterized by the fact
that it was virtually all addressed to his fellow
professional economists. So uncompromising was
he in his view that economics is a very difficult
subject offering only remote and nearly always
negative policy advice that it may fairly be said that
his work was addressed to just a small number of
‘fellow travellers’ in the rarefied atmosphere of the
‘higher regions’ of pure theory. However,
Edgeworth imposed no geographical limitations,
and with his considerable linguistic skills and
international sympathies was in contact with the
majority of leading economists around the world.
The distinguishing feature of the neoclassical
‘revolution’ was its emphasis on exchange as the
central economic problem. The success of this shift
of focus from production and distribution to
exchange was closely associated with the fact that it
had as its foundation a model based on utility
maximization. This allowed for a deeper treatment
of the gains from exchange and the wider
considerations of economic welfare. Schumpeter
summarized the point by stating that utility analysis
must be understood in terms of exchange as the
central ‘pivot’ and ‘the whole of the organism of
pure economics thus finds itself unified in the light
of a single principle’ ( , p. 913). This is indeed
the
context in which Edgeworth’s work in economics
must be seen. Schumpeter’s remark is merely a
more prosaic expression of Edgeworth’s view
quoted above that “‘Mechanique Sociale” may one
day take her place along with “Mechanique Celeste”
[sic], throned each upon the doublesided height of
one maximum principle’. The central theme of
Edgeworth’s work is also clear in his revealing
statement, taken from his presidential address to
Section F of the Royal Society, that:
Edgeworth, Francis Ysidro (1845-1926)
It may be said that in pure economics there is only
one fundamental theorem, but that is a very difficult
one: the theory of bargain in a wide sense. (1925, ii,
p. 288)
This perspective helps the major thread which runs
through all Edgeworth’s work in economics to be
seen. His earlier mathematical analysis of the
implications of utilitarianism for the optimal dis-
tribution, written before he turned to economics,
was not only highly original (and esoteric) but laid
the foundation for his work in economics. Thus, the
transition from New and Old Methods of Ethics to
Mathematical Psychics was not a shift in major
preoccupations but rather a change of emphasis.
Distribution was then seen as an important
concomitant of exchange, so that the analysis of
contract became central for Edgeworth. Edgeworth’s
emphasis on the indeterminacy (the inability of
utility maximization alone to determine the rate of
exchange, only a range of efficient exchanges)
which results from the existence of a small number
of traders led him to his path- breaking analysis of
the role of numbers in competition, along with the
efficiency properties of competitive equilibria.
The analysis of the utilitarian objective as an
arbitration rule led Edgeworth directly to his new
‘social contract’ argument in explaining the
acceptance of utilitarianism as a principle of social
justice. It was the realization of this new justifica-
tion of utilitarianism, using his newly developed
analytical tools, which generated the excitement that
is clearly evident in his first work in economics.
While Mathematical Psychics developed the
techniques of indifference curves and the contract
curve within the ‘Edgeworth box’ - tools which are
now ubiquitous in economic analysis - Edge- worth
himself was clearly driven mainly by his ability to
link the analysis of private contracts in markets to
that of a social contract in which utilitarianism is the
‘sovereign principle’. The integration of his analysis
of barter, and the effects of the introduction of
additional traders into the market, with the
demonstration that the utilitarian arrangement
prescribes a point on the contract curve of efficient
exchanges and is acceptable to risk-averse traders,
was to Edgeworth nothing short of ‘momentous’.
Edgeworth, Francis Ysidro (1845-1926)
The results are of course highly abstract. In
discussing their ultimate value suggested that:
Considerations so abstract it would of course be
ridiculous to fling upon the flood-tide of practical
politics ... it is at a height of abstraction in the
rarefied atmosphere of speculation that the secret
springs of action take their rise, and a direction is
imparted to the pure foundation of youthful enthu-
siasm whose influence will ultimately affect the
broad current of events. (1881, p. 128)
The intellectual pleasure derived from being able to
draw together so many different subjects of
analysis, and strands of his enormous range of
learning, is clearly evident. However, it is precisely
this wide field of vision, combined with the
technical level and idiosyncratic style of writing,
which made Mathematical Psychics so difficult for
his contemporaries, and which continue to make the
book seem so strange and yet so rewarding to the
modem reader.
Selected Works
1876. Mr. Matthew Arnold on Bishop Butler’s
doctrine of self love. Mind 1: 570-571.
1877. New and old methods of ethics: Or ‘physical
ethics ’ and ‘methods of ethics ’ . O x f o r d :
Parker.
1879. The hedonical calculus.
Mind 4 : 3 9 4 - 4 0 8 . 1 8 8 1 . Mathematical
psychics: An essay on the application of
mathematics to the moral sciences. L o n d o n :
Kegan Paul.
1 8 8 7 . Metretike, or the method of measuring
probability and utility. London:
Temple.
1909. Review of free trade in being. Economic
Journal 19: 104-105.
1925. Papers relating to political economy, 3 vols.
London: Macmillan, for the Royal Economic
Society.
Acknowledgment I am grateful to Denis O’Brien and Steven
Durlauf for comments on an earlier draft of this article.
Bibliography
Bickerdike, C. 1906. The theory of incipient taxes. Economic
Journal 16: 529-535.
3487
Black, R., ed 1977. Papers and correspondence of William
Stanley Jevons. London: Macmillan, for the Royal
Economic Society.
Black, R., ed. 1981. Papers and correspondence of William
Stanley Jevons, Papers on political economy. Vol. VII.
London: Macmillan, for the Royal Economic Society.
Bowley, A. 1924. The mathematical groundwork of eco-
nomics. Oxford: Clarendon Press.
Bowley, A. 1928. Edgeworth's contribution to mathematical
statistics. London: Royal Statistical Society.
Bowley, A. 1934. Francis Ysidro Edgeworth. Econometrica
1: 113-124.
Butler, J., and H. Butler. 1927. The Black Book of
Edgeworthtown and other Edgeworth Memories 1585-
1817. London: Faber and Gwyer.
Cournot, A. 1838. Researches into the mathematical
principles of the theory of wealth, trans. N. Bacon, ed. I.
Fisher. London: Stechert-Hafner, 1927.
Creedy, J. 1986. Edgeworth and the development of neo-
classical economics. Oxford: Basil Blackwell.
Creedy, J. 1988. Wicksell on Edgeworth’s tax paradox.
Scandinavian Journal of Economics 90: 101-112.
Creedy, J. 1990. Marshall and Edgeworth. Scottish Journal of
Political Economy 37: 18-39.
Creedy, J. 1994a. Exchange equilibria: Bargaining, utilitarian
and competitive solutions. Australian Economic Papers
33: 34-52.
Creedy, J. 1994b. Launhardt’s model of exchange. Journal of
the History of Economic Thought 16: 40-60.
Edgeworth, M. 1820. Memories of Richard Lovell Edge-
worth Esq., begun by himself and concluded by his
daughter, Maria Edgeworth, 2 vols. London: Hunter.
Harsanyi, J. 1953. Cardinal utility in welfare economics and
in the theory of risk taking. Journal of Political Economy
61:434—435.
Harsanyi, J. 1955. Cardinal welfare, individualistic ethics, and
interpersonal comparisons of utility. Journal of Political
Economy 63: 309-321.
Hicks, J., and R. Allen. 1934. A reconsideration of the theory
of value. Economica 14 (52-76): 196-219.
Hotelling, H. 1932. Edgeworth’s taxation paradox and the
nature of demand and supply functions. Journal of
Political Economy 40: 577-616.
Jcnkin, F. 1871. On the principles which regulate the
incidence of taxes. Reproduced in Readings in the
economics of taxation, ed. R. Musgrave and C. Shoup.
London: Allen and Unwin, 1959.
Jevons, W. 1871. The theory of political economy. 5th ed., ed.
H. Jevons. New York: Augustus Kelly, 1957.
Jevons, W. 1881. Review of Mathematical Psychics. Mind 6:
581-583.
Johnson, W. 1913. The pure theory of utility curves.
Economic Journal 23: 483-513.
Keynes, J.M. 1972. Essays in biography. Vol. X of the
collected writings of Keynes. London: Macmillan, for the
Royal Economic Society.
Marshall, A. 1890. Principles of economics, ed C. Guillebaud,
2 vols. (variorum ed.). London: Macmillan, 1961.
3488
Marshall, A., and M. Marshall. 1879. Economics of industry.
London: Macmillan.
McCann, C. 1996. EY. Edgeworth: Writings in probability,
statistics and economics. 3 vols. Cheltenham: Edward
Elgar.
Mill, J.S. 1848. Principles of political economy. Reprinted
with editorial material by W. Ashley. London: Longmans,
Green, 1920.
Pigou, A. 1908. Protective and preferential import duties.
London: Macmillan.
Pigou, A., ed. 1925. Memorials of Alfred Marshall 1842
1924. London: Macmillan.
Pigou, A., and D. Robertson. 1931. Economic essays and
addresses. London: King.
Price, L. 1946. Memoirs and notes on British
economists !,S,SI 1946. MSS. Brotherton Library,
University of Leeds.
Sargan, J. 1976. Econometric estimators and the Edge- worth
expansion. Econometrica 44: 421-448.
Schumpeter, J. 1954. History of economic analysis. London:
Allen and Unwin.
Stigler, S. 1978. Francis Ysidro Edgeworth, statistician.
Journal of the Royal Statistical Society, A 141: 287 322.
Sully, J. 1918. My life and friends. London: Fisher Unwin.
Vickrey, W. 1960. Utility, strategy and social decision rules.
Quarterly Journal of Economies 74: 507 535.
Walras, I.. 1874. Elements of pure economics. Trans. W.
Jaffe. London: Allen and Unwin. 1954.
Whitaker, J., ed. 1975. The early economic writings of Alfred
Marshall 1867-1890. London: Macmillan.
Edgeworth, Maria (1767-1849)
J. P. Croshaw
Bom in England of an Irish land-owning family,
Maria Edgeworth began her career as amanuensis
and co-author to her father Richard Lovell
Edgeworth, the educator and amateur inventor. Her
first publications were a series of moral tales for
children (The Parents' Assistant, 1796, and Early
Lessons, 1802) which aimed to instil the virtues she
saw as essential to a 'good' individual and so a
'good' society: honesty, frugality and hard work.
These characteristics match rather precisely those of
Adam Smith's ‘prudent man'
Edgeworth, Maria (1767-1849)
in the Wealth of Nations. Her tales teach the value
of a work ethic, sharply contrasting the evils of
sloth and idleness with the pleasures of diligence
and achievement. Indeed, her attimde towards this
aspect of labour did not exclude her own privileged
class of landowners, who, as she witnessed in her
own country, frequently abused the landlord-tenant
contract.
In 1800 she published the work which is, per-
haps, of most interest to economists, Castle
Rackrent. Through the character of Thady Quirk, an
ancient retainer of the Rackrent family, she recounts
the history of three generations of absentee
landlords, of their tenants and of the depths to
which the Rackrent fortunes had fallen through
successive generations of dissolute lifestyle. The
book not only influenced prominent literary figures
of the time (for example, Turgenev and Walter
Scott) but also established a literary precedent for
the development of fictional characters within the
context of a realistic historical, social and economic
setting - an approach which, in England, could be
said to reach its peak with George Eliot's
Middlemarch. In the 19th century the name
Rackrent came to stand for the embodiment of the
vices of the landed aristocracy and was freely used
as such by writers like Carlyle and, later, her
nephew F. Y. Edgeworth.
Maria Edgeworth continued her critical exam-
ination of the landlord-tenant relationship in novels
like The Absentee (1812) and Ennui (1825) where
she addressed issues such as leases, population and
economic progress and the impact of manufacture
on a traditional agricultural economy. Her letters to
David Ricardo confirm her interest in the poverty
and distress among the Irish agricultural peasantry.
She initiated and engaged in a vigorous
correspondence with Ricardo over the potato
question and the effects of famines in the 1820s. On
this subject she differed with both Ricardo and
Malthus arguing that the essential cause of the
difficulty lay in mismanagement. She rather
amusingly suggested that instead of theorizing from
afar, Ricardo should travel to Ireland and see for
himself
Education in Developing Countries
Education in Developing Countries
Paul Glewwe
Abstract
In many developing countries, children com-
plete few years of schooling and learn little
during their time in school. There are many
estimation problems that confound attempts to
understand the impact of education policies on
years of schooling and learning while in school.
Recent research has focused on implementing
randomized trials to get around estimation
problems based on retrospective data. While
some useful results have been found, many
additional studies are still under way. As these
results accumulate it is likely that general
conclusions can be drawn, but the evidence to
date is too limited to draw general policy
recommendations.
Keywords
Attenuation bias; Credit; Education in devel-
oping countries; Education production func-
tions; Returns to schooling; Value of time
JEL Classifications
01
Most economists who study economic growth agree
that an educated citizenry is necessary for sustained
economic growth, and virtually all international
development organizations concur (UNDP ; World
Bank ), and so those organizations provide
substantial financial resources and policy advice to
promote education in developing countries. Yet in
many developing countries, especially the poorest,
many children leave school at a young age and
learn little during the time they spend in school.
These problems have led many economists and
other social
3489
scientists to turn their attention to education in
developing countries.
This article summarizes recent research on the
factors that affect the amount of time that children
spend in school and the factors that determine how
much they leam during their time in school. Thus, it
focuses on the factors that shape education out-
comes as opposed to the impact of education on
income, economic growth and other phenomena
(for a recent assessment of the impact of education
on other socioeconomic outcomes, see Glewwe ).
This article also omits, due to space constraints, a
discussion of estimation issues (see Glewwe , and
Glewwe and Kremer for thorough discussions of
estimation problems and possible solutions).
Factors that Determine Years
of Schooling
In developing countries, parents usually decide how
many years their children will attend school. Each
year, parents consider the costs and expected
benefits of an additional year of schooling and then
enrol their children for another year if the expected
benefits outweigh the estimated costs. The main
costs are school fees and other payments required
by schools, transportation and (occasionally) meals
and housing, and the opportunity cost of the
children’s time. There may also be an additional,
‘psychic’ cost; some parents may dislike particular
values that schools attempt to instil in students. For
many parents, the largest of these costs is the value
of their children’s time; in developing countries,
especially in rural areas, children’s time is valuable
because they can help in household farming
activities.
The main benefits of schooling are the skills
learned (which usually reap substantial monetary
returns in the labour market), increased employment
opportunities that come with educational
credentials, and the direct satisfaction and social
approval that parents receive from having educated
children. While the decision rule to continue
schooling when the benefits outweigh the costs
3490
would seem to hold as a tautology, there are cir-
cumstances in which children are not enrolled in
school even when the economic benefits outweigh
the costs. This could occur because the costs are
incurred today while the benefits accrue over many
years in the future. In particular, parents who have
low incomes and cannot obtain credit may not send
their children to school even though the present
discounted value at prevailing interest rates is
positive.
Given this type of decision making by parents,
policies to increase school enrolment must focus on
reducing the costs of schooling, increasing the
benefits of education, or providing access to credit.
Reductions in fees are easy to implement, and in
some countries (such as Mexico) parents with low
incomes receive monthly payments if their children
are enrolled in school. Of course, this entails
potentially large budgetary costs, so some
governments try to limit fee exemptions and
outright subsidies to households or communities
that are particularly needy. Evidence from many
developing countries indicates that reducing fees or
providing payments conditional on school
enrolment can lead to large increases in enrolment;
studies in Honduras, Kenya, Mexico and Nicaragua
document these impacts (see Glewwe and Kremer ,
for further details and references).
The main alternative policy for increasing
school enrolment is to increase the expected returns.
These returns will increase if the relative price of
skilled labour increases, and if schools become
more effective at providing academic skills. While
some economists have shown that increased returns
to education does raise school enrollment (Foster
and Rosenzweig ), most policy research has focused
on what makes schools more efficient at raising
students’ skills. This research is discussed in the
next section.
Three additional points regarding policies to
increase years of schooling deserve attention. First,
improvements in the health and nutritional status of
both very young and school-age children are
another potentially important route to increase the
time that children spend in school (see Glewwe and
Miguel , for a review of this literature). Second,
many policy discussions
Education in Developing Countries
presume that the main reason children are not in
school is that no school is available, yet in most
countries schools are available but parents opt not to
enrol their children because they judge that the costs
outweigh the benefits (see Glewwe and Zhao ).
Third, the role of credit constraints in determining
years in school is an underresearched topic, in terms
of both the impact of credit constraints and policies
that could loosen those constraints.
Factors that Determine Student
Learning
In principle, student learning can be depicted as a
production process in which student, household,
teacher and school characteristics combine to pro-
duce students’ academic skills. While the existence
of an academic skills production function is true
almost by definition, there are serious problems that
confound attempts to estimate this process. The
main problem is omitted variables bias: students,
households, teachers and schools can vary in
hundreds of ways, and no data-set contains all
variables that are potentially important. Indeed,
important factors such as student innate ability,
teacher effort and parental encouragement are
almost impossible to measure and likely to be
correlated with the observed variables. This prob-
lem applies to virtually all studies based on retro-
spective (non-experimental) data; indeed, it is
probably the main reason that different studies find
very different results (the main alternative
explanation is that educational production functions
are very different in different countries). A second
serious estimation problem is attenuation bias.
Much of the data on students, households, teachers
and schools has a substantial amount of
measurement error. This typically leads to
underestimation of the true impacts of variables,
which may explain, at least in part, why many
variables in estimates of the determinants of student
learning are statistically insignificant.
In recent years economists and other social
scientists have turned to natural experiments and
randomized trials to estimate the impacts of
Education in Developing Countries
particular school characteristics, policies and pro-
grammes on student academic achievement. Natural
experiments result from institutions and policies
that cause random variation in school or student
characteristics, which can be used to analyse the
impact of those characteristics on student learning
(and on time spent in school). Randomized trials are
controlled experiments designed by researchers and
school officials that generate random variation in a
school characteristic or policy, which again allows
one to estimate the impact of the characteristic or
policy on learning. Natural experiments are
relatively rare, but in recent years randomized trials
have been implemented in many countries in Africa,
Asia and Latin America.
One of the first randomized trials was conducted
in Nicaragua in the late 1970s. The results indicated
that workbooks and radio instruction had significant
impacts on pupils’ math scores. In the Philippines in
the early 1980s, provision of textbooks raised
students’ performance on academic tests, but in
Kenya in the late 1990s the only effect of textbooks
was among the better students, perhaps because the
textbooks provided were too difficult for most
students. Other randomized trials conducted in
Kenya suggest little impact on test scores from
reductions in class size, provision of flip charts, and
provision of deworming medicine. On a more
positive note, school meals in Kenya raised test
scores in schools that had well-trained teachers, but
not in schools with poorly trained teachers. In
public schools in an urban area of India, a remedial
education programme increased test scores at a
relatively low cost Finally, a computer-assisted
learning programme in India also appears to have
increased test scores. The positive impacts of radio
education in Nicaragua and computer instruction in
India suggest that using modem technologies may
be particularly helpful in schools with weak
teachers. (For citations and more detailed
discussion, see Glewwe and Kremer .)
While natural experiments and especially ran-
domized trials may seem to avoid the estimation
problems that plague retrospective studies, more
randomized studies are needed before general
conclusions can be drawn that can guide policy
3491
in countries that have not yet had such studies.
Moreover, randomized trials can also suffer from
estimation problems. One problem is that parents of
students in the control schools (or schools excluded
from the evaluation) may try to enrol their children
in the treatment schools. This may affect the results
by increasing class size (if class size affects
learning). This would not occur if the policy were
implemented nationwide. In addition, children who
transfer into treatment schools may not be a random
sample of the general student population. A related
problem is that marginal students in the treatment
schools are less likely to drop out (if the
intervention raises student achievement), which
leads to underestimation of the impact of the policy
on learning if comparisons are made based on all
students currently enrolled in school. A final
problem with randomized trials is that the
evaluation itself may lead the treatment group to
change its behaviour, or the control group to change
its behaviour, because both groups know that their
results are being used in an evaluation.
In summary, recent research on education in
developing countries has provided fairly convincing
evidence of the impact on time in school and on
learning for particular policies in particular coun-
tries. Many additional studies are currently under
way, and as these results accumulate it is likely that
general conclusions can be drawn. This should lead
to better education policies, which will contribute to
higher economic growth and, ultimately, a higher
quality of life in developing countries.
See Also
► education Prc
► iuman Capiti
► Letums to
Scl
Bibliography
Foster, A., and M. Rosenzweig. 1996. Technical change
and human capital returns and investments: Evidence
from the Green Revolution. American Economic
Review 86: 931-953.
3492
Glewwe, P. 2002. Schools and skills in developing countries:
Education policies and socioeconomic outcomes. Journal
of Economic Literature 40: 436-482.
Glewwe, P., and M. Kremer. 2006. Schools, teachers and
education outcomes in developing countries. In Hand-
book on the economics of education, ed. E. Hanushek and
F. Welch. Amsterdam: North-Holland.
Glewwe, P., and E. Miguel. 2006. The impact of child health
and nutrition on education in less developed countries. In
Handbook of agricultural economics, vol. 4, ed. R.
Evenson and T. Schultz. Amsterdam: North-Holland.
Glewwe, P., and M. Zhao. 2005. Attaining universal primary
completion by 2015: How much will it cost?. Department
of Applied Economics/University of Minnesota.
UNDP (United Nations Development Programme). 1990.
Human development report. New York: UNDP.
World Bank. 2001. World development report 2000/2001:
Attacking poverty. Washington, DC: World Bank.
Education Production Functions
Eric A. Hanushek
Abstract
The accumulated economic analysis of educa-
tion suggests that current provision of schooling
is very inefficient. Commonly purchased inputs
to schools - class size, teacher experience, and
teacher education - bear little systematic rela-
tionship to student outcomes, implying that con-
ventional input policies are unlikely to improve
achievement. At the same time, differences in
teacher quality have been shown to be very
important. Unfortunately, teacher quality,
defined in terms of effects on student perfor-
mance, is not closely related to salaries or read-
ily identified attributes of teachers.
Keywords
Education production functions; Random
assignment; School attainment; School
resources; Student outcomes; Teacher quality
JEL Classifications
12
Education Production Functions
A simple production model lies behind much of the
analysis in the economics of education. The
common inputs are things like school resources,
teacher quality, and family attributes; and the out-
come is student achievement. Knowledge of the
production function for schools can be used to
assess policy alternatives and to judge the effec-
tiveness and efficiency of public provided services.
This area is, however, distinguished from many
because the results of analyses enter quite directly
into the policy process.
Historically, the most frequently employed
measure of schooling has been attainment, or sim-
ply years of schooling completed. The value of
school attainment as a rough measure of individual
skill has been verified by a wide variety of studies
of labour market outcomes (for example, Mincer ;
Psacharopoulos and Patrinos ). However, the
difficulty with this common measure of outcomes is
that it assumes a year of schooling produces the
same amount of student achievement, or skills, over
time and in every country. This measure simply
counts the time spent in schools without judging
what happens in schools - thus, it does not provide a
complete or accurate picture of outcomes.
Recent direct investigations of cognitive
achievement find significant labour market returns
to individual differences in cognitive achievement
(for example, Lazear ; Mulligan ; Mumane et al. ).
Similarly, society appears to gain in terms of
productivity; Hanushek and Kimko ( ) demonstrate
that quality differ
ences in schools have a dramatic impact on pro-
ductivity and national growth rates. (A parallel line
of research has employed school inputs to measure
quality but has not been as successful. Specifically,
school input measures have not proved to be good
predictors of wages or growth.)
Because outcomes cannot be changed by fiat,
much attention has been directed at inputs -
particularly those perceived to be relevant for policy
such as school resources or aspects of teachers.
Analysis of the role of school resources in
determining achievement begins with the Coleman
Report, the US government’s monumental study on
educational opportunity released
Education Production Functions
in 1966 (Coleman et al. ). That study’s greatest
contribution was directing attention to the
distribution of student performance - the outputs as
opposed to the inputs.
The underlying model that has evolved as a
result of this research is very straightforward. The
output of the educational process - the achievement
of individual students - is directly related to inputs
that both are directly controlled by policymakers
(for example, the characteristics of schools,
teachers, and curricula ) and are not so controlled
(such as families and friends and the innate
endowments or learning capacities of the students).
Further, while achievement may be measured at
discrete points in time, the educational process is
cumulative; inputs applied sometime in the past
affect students’ current levels of achievement.
Family background is usually characterized by
such socio-demographic characteristics as parental
education, income, and family size. Peer inputs,
when included, are typically aggregates of student
socio-demographic characteristics or achievement
for a school or classroom. School inputs typically
include teacher background (education level,
experience, sex, race, and so forth), school
organization (class sizes, facilities, administrative
expenditures, and so forth), and district or
community factors (for example, average
expenditure levels). Except for the original
Coleman Report, most empirical work has relied on
data constructed for other purposes, such as a
school’s standard administrative records. Based
upon this, statistical analysis (typically some form
of regression analysis) is employed to infer what
specifically determines achievement and what is the
importance of the various inputs into student
performance.
Measured School Inputs
The state of knowledge about the impacts of
resources is best summarized by reviewing avail-
able empirical studies. Most analyses of education
production functions have directed their attention at
a relatively small set of resource measures, and this
makes it easy to summarize the results
3493
(Hanushek ). The 90 individual publications that
appeared before 1995 contain 377 separate
production function estimates. For classroom
resources, only nine per cent of estimates for
teacher education and 14% for teacher-pupil ratios
yielded a positive and statistically significant
relationship between these factors and student
performance. Moreover, these studies were offset
by another set of studies that found a similarly
negative correlation between those inputs and
student achievement. Twenty-nine per cent of the
studies found a positive correlation between teacher
experience and student performance; however, 71%
still provided no support for increasing teacher
experience (being either negative or statistically
insignificant). Studies on the effect of financial
resources provide a similar picture. These indicate
that there is very weak support for the notion that
simply providing higher teacher salaries or greater
overall spending will lead to improved student
performance. Per pupil expenditure has received the
most attention, but only 27% of studies showed a
positive and significant effect. In fact, seven per
cent even suggested that adding resources would
harm student achievement. It is also important to
note that studies involving pupil spending have
tended to be the lowest-quality studies as defined
below, and thus there is substantial reason to
believe that even the 27% figure overstates the true
effect of added expenditure.
These studies make a clear case that resource
usage in schools is subject to considerable ineffi-
ciency, because schools systematically pay for
inputs that are not consistently related to outputs.
Study Quality
The previous discussions do not distinguish among
studies on the basis of any quality differences. The
available estimates can be categorized by a few
objective components of quality. First, while
education is cumulative, frequently only current
input measures are available, which results in
analytical errors. Second, schools operate within a
policy environment set almost always at higher
levels of government. In the United
3494
States, state governments establish curricula, pro-
vide sources of funding, govern labour laws,
determine mles for the certification and hiring of
teachers, and the like. In other parts of the world,
similar policy setting, frequently at the national
level, affects the operations of schools. If these
attributes are important - as much policy debate
would suggest - they must be incorporated into any
analysis of performance. The adequacy of dealing
with these problems is a simple index of study
quality.
The details of these quality issues and
approaches for dealing with them are discussed in
detail elsewhere (Hanushek ) and only summarized
here. The first problem is ameliorated if one uses
the ‘value added’ versus ‘level’ form in estimation.
That is, if the achievement relationship holds at
different points in time, it is possible to concentrate
on the growth in achievement and on exactly what
happens educationally between those points when
outcomes are measured. This approach ameliorates
problems of omitting prior inputs of schools and
families, because they will be incorporated in the
initial achievement levels that are measured
(Hanushek ). The latter problem of imprecise
measurement of the policy environment can
frequently be ameliorated by studying performance
of schools operating within a consistent set of
policies - for example, within individual states in
the USA or similar decisionmaking spheres
elsewhere. Because all schools within a state
operate within the same basic policy environment,
comparisons of their performance are not strongly
affected by unmeasured policies (Hanushek et al. ).
If the available studies are classified by whether
or not they deal with these major quality issues, the
prior conclusions about research usage are
unchanged (Hanushek ). The best quality studies
indicate no consistent relationship between
resources and student outcomes.
An additional issue, which is particularly
important for policy purposes, concerns whether
this analytical approach accurately assesses the
causal relationship between resources and perfor-
mance. If, for example, school decision-makers
provide more resources to those they judge as most
needy, higher resources could simply signal
Education Production Functions
students known for having lower achievement.
Ways of dealing with this include various regres-
sion discontinuity or panel data approaches. When
done in the case of class sizes, the evidence has
been mixed (Angrist and Lavy ; Rivkin et al. ).
An alternative involves the use of random
assignment experimentation rather than statistical
analysis to break the influence of sample selection
and other possible omitted factors. With one major
exception, this approach nonetheless has not been
applied to understand the impact of schools on
student performance. The exception is Project
STAR, an experimental reduction in class sizes that
was conducted in the US state of Tennessee in the
mid-1980s (Word et al. ). To
date, it has not had much impact on research or our
state of knowledge. While Project STAR has
entered into a number of policy debates, the inter-
pretation of the results remains controversial
(Krueger ; Hanushek ).
Magnitude of Effects
Throughout most consideration of the impact of
school resources, attention has focused almost
exclusively on whether a factor has an effect on
outcomes that is statistically different from zero. Of
course, any policy consideration would also
consider the magnitude of the impacts and where
policies are most effective. Here, even the most
refined estimates of, say, class size impacts does not
give very clear guidance. The experimental effects
from Project STAR indicate that average
achievement from a reduction of eight students in a
classroom would increase by about 0.2 standard
deviations, but only in the first grade of attendance
in smaller classes (kindergarten or first grade) (see
Word et al. ; Krueger ). Angrist and Lavy ( ),
with their regression discontinuity
estimation, find slightly smaller effects in grade five
and approximately half the effect size in grade four.
Rivkin et al. ( ), with their fixed effects
estimation, find effects half of Project STAR in
grade four and declining to insignificance by grade
seven. Thus, from a policy perspective the
alternative estimates are both small in economic
Education Production Functions
terms when contrasted with the costs of such large See Also
class size reductions and inconsistent across
studies. ► -f
Do Teachers and Schools Matter?
Because of the Coleman Report and subsequent
studies discussed above, many have argued that
schools do not matter and that only families and
peers affect performance. Unfortunately, these
interpretations have confused measurability with
hue effects.
Extensive research since the Coleman Report
has made it clear that teachers do indeed matter
when assessed in terms of student performance
instead of the more typical input measures based on
characteristics of the teacher and school. When
fixed effect estimators that compare student gains
across teachers are used, dramatic differences in
teacher quality are seen.
These results can also be reconciled with the
prior ones. These differences among teachers are
simply not closely correlated with commonly
measured teacher characteristics (Hanushek ;
Rivkin et al. ). Moreover, teacher credentials and
teacher training do not make a consistent difference
when assessed against student achievement gains
(Boyd et al. ; Kane et al. ). Finally, teacher quality
does not appear to be closely related to salaries or
to market decisions. In particular, teachers exiting
for other schools or for jobs outside of teaching do
not appear to be of higher quality than those who
stay (Hanushek et al. ).
Some Conclusions and Implications
The existing research suggests inefficiency in the
provision of schooling. It does not indicate that
schools do not matter. Nor does it indicate that
money and resources never impact achievement.
The accumulated research surrounding estimation
of education production functions simply says there
currently is no clear, systematic relationship
between resources and student outcomes.
3495
► _,oc
► Leturns
Bibliography
Angrist, J.D., and V. Lavy. 1999. Using Maunondides’ rule to
estimate the effect of class size on scholastic achieve-
ment. Quarterly Journal of Economics 114: 533-575.
Boyd, D., P. Grossman, H. Lankford, S. Loeb, and J.
Wyckoff. 2006. How changes in entry requirements alter
the teacher workforce and affect student achievement.
Education Finance and Policy 1: 176-216.
Coleman, J.S., E.Q. Campbell, C.J. Hobson, J. McPartland,
A.M. Mood, F.D. Weinfeld, and R.L. York. 1966.
Equality of educational opportunity. Washington, DC:
US Government Printing Office.
Hanushek, E.A. 1979. Conceptual and empirical issues in the
estimation of educational production functions. Journal
of Human Resources 14: 351-388.
Hanushek, E.A. 1992. The trade-off between child quantity
and quality. Journal of Political Economy 100:84-117.
Hanushek, E.A. 1999. Some findings from an independent
investigation of the Tennessee STAR experiment and
from other investigations of class size effects. Educa-
tional Evaluation and Policy Analysis 21: 143-163.
Hanushek, E.A. 2003. The failure of input-based schooling
policies. Economic Journal 113: F64-F98.
Hanushek, E.A., and D.D. Kimko. 2000. Schooling, labor
force quality, and the growth of nations. American
Economic Review 90: 1184—1208.
Hanushek, E.A., S.G. Rivkin, and L.L. Taylor. 1996.
Aggregation and the estimated effects of school
resources. Review of Economics and Statistics 78: 611-
627.
Hanushek, E.A., J.F. Kain, D.M. O’Brien, and S.G. Rivkin.
2005. The market for teacher quality, Working paper no.
11154. Cambridge, MA: NBER.
Kane, T.J., J.E. Rockoff, and D.O. Staiger. 2006. What does
certification tell us about teacher effectiveness? Evidence
from New York City, Working paper no. 12155.
Cambridge, MA: NBER.
Krueger, A.B.. 1999. Experimental estimates of education
production functions. Quarterly Journal ot Economics
114: 497-532.
Lazear, E.P. 2003. Teacher incentives. Swedish Economic
Policy Review 10(3): 179-214.
Mincer, J. 1970. The distribution of labor incomes: A survey
with special reference to the human capital approach.
Journal of Economic Literature 8: 1-26.
Mulligan, C.B. 1999. Gabon versus the human capital
approach to inheritance. Journal of Political Economy
107(pt 2): S184-S224.
3496
Murnane, R.J., J.B. Willett, Y. Duhaldeborde, and J.H. Tyler.
2000. How important are the cognitive skills of teenagers
in predicting subsequent earnings? Journal of Policy
Analysis and Management 19: 547-568. '
Psacharopoulos, G., and H.A. Patrinos. 2004. Returns to
investment in education: A further update. Education
Economics 12: 111-134.
Rivkin, S.G., E.A. Hanushek, and J.F. Rain. 2005. Teachers,
schools, and academic achievement. Econometrica 73:
417^458.
Word, E., I Johnston, H.P. Bain, B. DeWayne Fulton, J.B.
Zaharies, M.N. Lintz, C.M. Achilles, J. Folger. and C.
Breda. 1990. Student/Teacher Achievement Ratio (STAR),
Tennessee's K-3 class size study: Final summary report,
1985-1990. Nashville: Tennessee State Department of
Education.
Educational Finance
William A. Fischel
Abstract
The American system of government-financed
education is decentralized among 50 states and
more than 15,000 local school districts. Local
funds are derived from local property taxes, and
this system tends to make local spending
unequal. State- government efforts to equalize
education spending involve manipulating the
local ‘tax price’ with matching grants. School
districts with low tax prices are not, however,
necessarily populated by rich people, so the
distribution of state funds may penalize many
low-income districts with large amounts of non-
re sidential property.
Keywords
Educational finance; Local government; Median
voter; Property taxation; School districts (USA);
School vouchers; Spatial competition; Tax price
of school spending; Tiebout hypothesis
JEL Classification
12
Educational Finance
This article deals with the government-financed
system of education in the United States, which is
referred to as ‘public’ education. Educational
finance in the United States is different from that of
other nations, which typically fluid education from
national taxes. Within each American state, a
substantial portion of education is financed by local
governments, although the proportion financed
locally has declined from 83.2% in 1920 to 43.2%
in 2000.
The state-local system of finance stems from the
history and geography ofthe United States and the
federal nature of its government. The 50 states are,
in the eyes of the national government, primarily
responsible for education. In most states,
implementation of this responsibility is delegated to
local municipal corporations called ‘school dis-
tricts’. ’fhc school district is more than a local
administrative agency of the state. It is a distinct
political entity that usually has some correspon-
dence with the geographic area of a municipality.
The district, however, has a separate board of
directors, which is locally elected. The board then
selects a superintendent of schools to manage the
district's education. Boards have the authority to
levy taxes, which are almost always on property
within then district, and spend the revenue they
derive from them. The state government may pre-
scribe curricular standards for public schools, but
the method of achieving these standards is the
responsibility of the local district.
School districts and school boards were once the
most common form of local government in the
United States, numbering about 200,000 in 1900.
The number of school districts declined steadily
throughout the twentieth century, which can largely
be accounted for by the consolidation of rural one-
room school districts into larger units. By 1970,
one-room schools were essentially extinct, and
since 1970 the total number of school districts has
declined only slightly, numbering about 16,000 at
the beginning of the twenty-first century.
Despite their numerical decline in rural areas,
there are many school districts in most metropolitan
areas. Urban households that are already on the
move for job- related reasons have the luxury of
choosing a home within one of several school
Educational Finance
Educational Finance,
Fig. 1 School spending in
rich and poor districts
districts in most regions of the nation. Choosing
among school districts and the resulting competi-
tion among districts to obtain residents is consistent
with the model proposed by Tiebout ( ).
Numerous tests of the Tiebout model indicate that
the quality of schooling is important to most home
buyers (Oates ; Bradbury et al. ). There is also
evidence that spatial competition makes school
districts more efficient in delivering education
services (Hoxby ).
One-room schools of the nineteenth century
were usually ‘ungraded’. Students were instead
divided into skill-specific recitation groups, formed
without regard for chronological age. In this
system, uniformity of education was not critical.
New pupils could be placed according to what they
knew in particular subjects rather than by age. But
when almost all schools were age-graded, it paid for
each district to offer an age-specific curriculum that
allowed both teachers and pupils to be
interchangeable among schools and districts
(Fischel ).
Standardization of age-graded curricula became
widespread by about 1940 and was brought about
by two forces, one local and the other statewide.
Property-owning voters in a given district would
find that potential homebuyers would shun them if
they did not offer a standard, public-school
education. Voters would thus support taxes
necessary to fund standardized schools. However,
differences in the
3497
economic make-up and tax-bases of local districts
sometimes made this difficult to do.
Figure illustrates the problem for attempts to
fund schools from local sources. It depicts a trade-
off between local school spending and other goods
for the median voter (the voter with the median
income, assumed always to be in the majority in
local elections) in two separate communities, a rich
district and a poor district. The decisive voter
chooses the mix of school spending and private
goods that achieves the highest indifference curve
that his private-public budget line allows
(Bergstrom and Goodman ). Because at the local
level education is essentially a private good, the
slope of the budget lines is the ‘tax price’ of school
spending for the median voter in each community.
The tax price is not a tax rate. A school district
composed exclusively of mansions will have, for a
given level of spending, a much lower property tax
rate than a district composed of modest-sized
homes. But if the second moment of the distribution
of wealth is the same in both communities, the tax
price faced by the median voter in each will be the
same. A 1000 dollar increase in per-pupil spending
will cost the median voter the same amount of
money in both cases, if one assumes that the
number of public-school children per household is
the same in both.
The other generalization that Fig. illustrates is
that average income of a district accounts for
3498
Educational Finance, Y = Private
Fig. 2 S u b s i d i e s t o
poor districts
much of the differences in spending per pupil. Even
though the tax prices are the same, the positive
income elasticity of demand for education
(estimated at somewhere between 0.5 and 1.0)
causes the richer community to choose a higher
level of school inputs (Bergstrom et al. ). While
much of the criticism of these differences is based
on equity concerns, there are efficiency reasons to
promote a relatively uniform system of education
(Benabou ).
The way most states have attempted to equalize
education opportunities is to reduce the tax price of
spending in poorer districts. State funds (from
statewide taxes) are offered to the poorer
community in proportion to the district’s own tax
effort. The poorer median voter thus perceives, as
indicated by the dotted budget line in Fig. , that for
every dollar raised locally, the state will send it
another dollar. The tax price has been cut in half in
the graphical example, so that the poorer commu-
nity will choose to spend an amount closer to that
of the richer district.
By manipulating the local tax price, state gov-
ernments can in principle induce a substantial
equality of school spending in nominally inde-
pendent districts, though state officials still seem
surprised that there is an income effect as well as a
substitution effect from lowering the tax price.
They seem to expect that the arrow in Fig. should
point horizontally to the right. Instead,
Educational Finance
local voters use the subsidy (the reduced tax price)
to both increase local spending on schools, which is
the desired substitution effect, and to reduce their
own local taxes (nudging the arrow’s direction
upwards), which is the income effect.
Another factor can also account for differences
in local tax prices. The poorer district may have a
substantial amount of non-residential property to
tax. Commercial and industrial uses do not come
with children attached (at least in metropolitan
areas, where workers can live in other communi-
ties), and so their tax revenues amount to a subsidy
to their school district. The effect of this is the same
as a matching-grant subsidy by the state. And the
effect is not trivial. Nationally, almost one-half of
all property taxes are paid by non-residential
property owners, which puts them on the same
order of magnitude as state funds for public
education.
Although both state subsidies and a large non-
residential tax base reduce the tax price, they have
been treated differently in recent years. The school
finance litigation movement began with Serrano v.
Priest in California in 1971 (Brunner and
Sonstelie ). Its objective was to use state
constitutional directives (equal protection and
school funding clauses) to improve schools in poor
districts. For strategic reasons, the movement
focused its remedial efforts on
Educational Finance
differences in tax base per pupil rather than dif-
ferences in spending per pupil or on educational
outcomes. Many state courts thus ruled that unequal
tax bases, not unequal spending, were
constitutionally suspect and ordered legislatures to
transfer hinds from the ‘property rich’ to the
‘property poor’.
What this remedy overlooked is that low-income
communities are as likely to be ‘property rich’ (on
the widely used ‘tax base per-pupil’ standard) as
high-income communities. This is because many
urban districts have a large non-residential property
tax base that offsets the lower valued residential tax
base. (The poor may have migrated there for jobs or
rezoned land to attract industry, something most
affluent suburbs are reluctant to do.) Besides this,
poorer cities often have relatively few children in
public schools because of an aged population or
because low-quality public schools encourage the
use of private schools. In any case, many of the
court- induced ‘equalization’ remedies have
actually caused state hinds to be removed horn low-
income (but ‘property rich’) districts to higher-
income districts that are ‘property poor’ because of
their modest nonresidential tax base and large
school-age population.
An alternative response to the difficulties of
distributing state funds to school districts is simply
to have the state government run the schools
without the intermediation of local school boards
and districts. Another is a voucher system, in which
the state gives public funds to parents and allows
them to select whatever school they want. Both are
certainly viable means of school finance, and it is
worth asking why they have not been embraced.
Full state funding forgoes the local monitoring
of school performance by voters. Capitalization of
school quality in local home values creates a feed-
back mechanism for local governance. The median
voter in most jurisdictions is a homeowner, and
voters therefore care about the consequences of
school governance. School superintendents who
waste local taxpayers’ money will find that their
tenure is short as voters become dissatisfied. Even if
they keep their jobs, the declines in taxable property
value due to
3499
inefficient policies will leave them with less rev-
enue to spend in the future (Hoxby ). Neither of
these desirable feedback effects is likely to occur
under a state-managed system.
The drawback of school vouchers appears to be
that voters are reluctant to embrace them as a
general practice. American voters appear to per-
ceive benefits from local public schools that go
beyond educational qualities. One benefit I have
advanced is that public schools create location-
specific social capital among adults (Fischel ).
Adults with children are more likely to know the
parents of their children’s schoolmates. This creates
a network of adult social capital that lowers the
transaction costs of public participation in
municipal affairs. A voucher system disperses
children to various schools and thus does not create
the same location- specific social capital that public
schools do. In any case, America’s continuing
embrace of locally run and locally financed public
education reflects the school’s central role in
facilitating local self-governance.
See Also
► Lc cal Pub
► • ..
►
►
► Tiet
Bibliography
Benabou, R. 1996. Heterogeneity, stratification, and growth:
Macroeconomic implications of community structure and
school finance. American Economic Review 86: 584-609.
Bergstrom, T.C., and R.R Goodman. 1973. Private demand
for public goods. American Economic Review 63: 280-
296.
Bergstrom, T.C., D.L. Rubinfeld, and P. Shapiro. 1982.
Micro-based estimates of demand functions for local
school expenditures. Econometrica 50: 1183 1205.
Bradbury, K.L., K.E. Case, and C. Mayer. 2001. Property tax
limits, local fiscal behavior, and property values:
Evidence from Massachusetts under proposition 2 1/2.
Journal of Public Economics 80: 287-311.