Political: Jour - Nal

OF
TH E JOUR.N
POLIT
ECON OMY
Volume LVII DCM x b6
I-liatem Curte Mm4-&n ma 463
~ :?ure TheVr of Gove t Fitugge

A~~~proad~ ~ ~ ~ ~ Ja 3~ :s Buchna?' 496j
and a Nad SAhtBt e Rnd, -. K 506
M..
E.onomks before Couuot .&M.'
.~Sea~~
..d
to Wok
the RRight J7 to. 4$?
' to Sat In,FaXiy Fwaiin Is. CA66Ovutoild
EoenJ.Lang 543
Jlo Reviews (See in de front cow) 547
Bys hedved t;55
THE UNIVERSITY OF:CHICAGOPRESS

CHICAGOkLLINOI-S. U.S.A.
This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
POLITICAL 0
w& O t*8 uex~~*O
C .na ; .. ... @ >

,..i~~~ ~~ ~~~~~ ~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.4 W . . .. t
ftyO* fl00twy W
t r
l Si:
!
'A
53
BCBE= R; ii71
.,, _E r = ' $.~~~~~~~~~~~

13 -iE-;
;3;Ei =i 5.52? . -7-7R~fi P i . ,-
~~~~bW
.n..,etn -8>l 5*0>.;1^tztfps
Ana . . ..n , ~ ~r k ,
';=77 <s-7
s= =j a;K ||. i ;
LF'%tf'.w
.~~~~~~~~~~~~~~~~~~~~4
>'
IT~~~~~~~~~~~~~~~~~ bef
^P~d,# i
Nai'sss~~ak

THE JOURNAL OF
POLITICALECONOMY
Volume LVII DECEMBER 1949 Number 6
THIE MARSHALLIAN DEMAND CURVE

MILTON FRIEDMAN'
ALFRED MARSHALL'S theory of de- lowed to vary, is nowhere given explicit-

mand strikingly exemplifies his ly. The reader is left to infer the contents
"impatience with rigid definition of ceteris paribus from general and vague
and an excessive tendency to let the con- statements, parenthetical remarks, ex-
text explain his meaning."2 The concept amples that do not purport to be exhaus-
of the demand curve as a functional relative, and concise mathematical notes in
tion between the quantity and the price the Appendix.
of a particular commodity is explained In view of the importance of the de-
repeatedly and explicitly in the Princi- mand curve in Marshallian analysis, it is
ples of Economics: in words in the text, in natural that other economists should
plane curves in the footnotes, and in sym- have constructed a rigorous definition to
bolic form in the Mathematical Appen- fill the gap that Marshall left. This oc-
dix. A complete definition of the de- curred at an early date, apparently with-
mand curve, including, in particular, a out controversy about the interpretation
statement of the variables that are to be to be placed on Marshall's comments.
considered the same for all points on the The resulting definition of the demand
curve and the variables that are to be al- curve is now so much an intrinsic part of
I I am deeply indebted for helpful criticism and
current economic theory and is so widely
suggestions to A. F. Burns, Aaron Director, C. WV. accepted as Marshall's own that the as-
Guillebaud, H. Gregg Lewis, A. R. Prest, D. H. sertion that Marshall himself gave no ex-
Robertson, G. J. Stigler, and, especially, Jacob plicit rigorous definition may shock most
Viner, to whose penetrating discussion of the de-
mand curve in his course in economic theory I readers.
can trace some of the central ideas and even de- Yet why this particular interpretation
tails of this article. The standard comment that evolved and why it gained such unques-
none is to be held responsible for the views expressed
herein has particular relevance, since most dis- tioned acceptance are a mystery that re-
agreed with my interpretation of Marshall as pre- quires explanation. The currently ac-
sented in an earlier and much briefer draft of this cepted interpretation can be read into
article.
Marshall only by a liberal-and, I think,
2
C. W. Guillebaud, "The Evolution of Marshall's
Principles of Economics," Economic Journal, LII strained-reading of his remarks, and its
(December, 1942), 333. acceptance implicitly convicts him of log-
463

464 MILTON FRIEDMAN
ical inconsistency and mathematical er- that "same" in this phrase does not
ror at the very foundation of his theory mean ''same over time." The points on a
of demand. More important, the alterna- demand curve are alternative possibili-
tive interpretation of the demand curve ties, not temporally ordered combina-
that is yielded by a literal reading of his tions of quantity and price. "Same"
remarks not only leaves his original work means "same for all points on the de-
on the theory of demand free from both mand curve"; the different points are
logical inconsistency and mathematical to differ in quantity and price and are not
error but also is more useful for the anal- to differ with respect to "other things."3
ysis of most economic problems. In the second place, "all" other things
Section I presents the two interpreta- cannot be supposed to be the same with-
tions of the demand curve and compares out completely emasculating the con-
them in some detail; Section II argues cept. For example, if (a) total money ex-
that a demand curve constructed on my penditure on all commodities, (b) the
interpretation is the more useful for the price of every commodity other than the
analysis of practical problems, whatever one in question, and (c) the quantity pur-
may be the verdict about its validity as chased of every other commodity were
an interpretation of Marshall; Section supposed to be the same, the amount of
III demonstrates that my interpretation money spent on the commodity in ques-
is consistent with Marshall's monetary tion would necessarily be the same at all
theory and with his work on consumer's prices, simply as a matter of arithmetic,
surplus; and Section IV presents the tex- and the demand curve would have unit
tual evidence on the validity of my inter- elasticity everywhere.4 Different specifi-
pretation. Finally, Section V argues that cations of the "other things" will yield
the change that has occurred in the in- different demand curves. For example,
terpretation of the demand curve reflects 3 Of course, when correlations among statistical
a corresponding change in the role as- time series are regarded as estimates of demand
signed to economic theory. curves, the hypothesis is that "other things" have
been approximately constant over time or that
appropriate allowance has been made for changes in
I. ALTERNATIVE INTERPRETATIONS OF
them. Similarly, when correlations among cross-
MARSHALL'S DEMAND CURVE section data are regarded as estimates of demand
curves, the hypothesis is that "other things" are
The demand curve of a particular approximately the same for the units distinguished
group (which may, as a special case, con- or that appropriate allowance has been made for
sist of a single individual) for a particular differences among them. In both cases the problem
of estimation should be clearly distinguished from
commodity shows the quantity (strictly the theoretical construct to be estimated.
speaking, the maximum quantity) of the 4 Yet Sidney Weintraub not only suggests that
commodity that will be purchased by the Marshall intended to keep a, b, and c simultaneously
group per unit of time at each price. So the same but goes on to say: "Clearly Marshall's
assumption means a unit elasticity of demand in the
far, no question arises; this part of the market reviewed and no ramifications elsewhere;
definition is explicit in Marshall and is that was why he adopted it" ("The Foundations of
common to both alternatives to be dis- the Demand Curve," American Economic Review,
XXXII [September,I942], 538-52, quotation from
cussed. The problem of interpretation re- n. I2, p. 541). Weintraub even adds the condition
lates to the phrase, "other things the of constant tastes and preferences to a, b, and c,
same," ordinarily attached to this defi- speaking of a change in tastes as shifting the de-
mand curve. Obviously, a, b, and c together leave
nition. no room for tastes and preferences or, indeed, for
In the first place, it should be noted anything except simple arithmetic.

THE MARSHALLIAN DEMAND CURVE 465
one demand curve will be obtained by b) AN ALTERNATIVE INTERPRETATION

excluding b from the list of "other It seems to me more faithful to both
things"; another, quite different one, by the letter and the spirit of Marshall's
excluding c. writings to include in the list of "other
things" (I) tastes and preferences of the
a) THE CURRENT INTERPRETATION group of purchasers considered, (2) their
The current interpretation of Mar- real income, and (3) the price of every
shall's demand curve explicitly includes closely related commodity.
in the list of "other things" (i) tastes and Two variants of this interpretation can
preferences of the group of purchasers be distinguished, according to the device
adopted for keeping real income the same
considered; (2) their money incom-te,and
at different points on the demand curve.
(3) the price of every other commodity.
One variant, which Marshall employed
The quantities of other commodities are
in the text of the Principles, is obtained
explicitly considered as different at differ-
by replacing "(2) their real income" by
ent points on the demrlandcurve, and still (2a) their money income and (2b) the
other variables are ignored.r "purchasing power of money." Con-
On this interpretation, it is clear that, stancy of the "purchasing power of
while money income is the same for money" for different prices of the com-
different points on the demand curve, nrodity in question implies compensating
real income is not. At the lower of two variations in the prices of some or all
prices for the commodity in question, other commodities. These variations will,
more of some commodities can be pur- indeed, be negligible if the commodity
chased without reducing the amounts in question accounts for a negligible frac-
purchased of other commodities. The tion of total expenditures; but they
lower the price, therefore, the higher the should not be disregarded, both because
real income. empirical considerations must be sharply
separated from logical considerations and
5 Explicit definition of the demand curve in this because the demand curve need not be
way by followers of Marshall dates back at least
to i894 (see F. Y. Edgeworth, article on "Demand
limited in applicability to such commod-
Curves" in Palgrave's Dictionary of Political Econo- ities. On this variant, all commodities
my, edited by Henry Higgs [rev. ed.; London: Mac- are, in effect, divided into three groups:
millan & Co., Ltd., I926]). Edgeworth's article
apparently dates from the first edition, which was
(a) the commodity in question, (b) closely
published in i894. While Edgeworth does not ex- related commodities, and (c) all other
plicitly attribute this interpretation to Marshall, commodities. The absolute price of each
it is clear from the context that he is talking about a
Marshallian demand curve and that he does not
commodity in group b is supposed to be
regard his statements as inconsistent in any way the same for different points on the de-
with Marshall's Principles. Though no explicit mand curve; only the "average" price, or
listing of "other things" is given by J. R. Hicks,
Value and Capital (Oxford, I939), the list given an index number of prices, is considered
above is implicit throughout chaps. i and ii, for group c; and it is to be supposed to
which are explicitly devoted to elaborating and ex- rise or fall with a fall or rise in the price
tending Marshall's analysis of demand. For state-
ments in modern textbooks on advanced economic of group a, so as to keep the "purchasing
theory see G. J. Stigler, The Theory of Price (New power of money" the same.
York: Macmillan Co., I946), pp. 86-90, and Ken-
neth E. Boulding, Economic Analysis (rev. ed., The other variant, which Marshall em-
New York: Harper & Bros., I948), pp. I34-35. ployed in the Mathematical Appendix of

466 MILTON FRIEDMAN
the Principles, is obtained by retaining sumption of compensating variations in
" (2) their real income" and adding (4) other prices is easier to explain verbally
the average price of all other commod- and can be justified as empirically rel-
ities. Constancyof real income for differ- evant by considerations of monetary the-
ent prices of the commodity in question ory, which is presumably why Marshall
then implies compensatingvariations in used this variant in his text. On the other
money income. As the price of the com- hand, the assumption of compensating
modity in question rises or falls, money variations in income is somewhat more
income is to be supposedto rise or fall so convenient mathematically, which is pre-
as to keep real incomethe same. sumably why Marshall used this variant
These two variants are essentially in his Mathematical Appendix.
equivalent mathematically,' but the as- On my interpretation, Marshall's de-
6
mand curve is identical with one of the
Let x and y be the quantity and price, respec-
tively, of the commodity in question; x' and y' constructions introduced by Slutsky in
the quantity and price of a composite commodity his famous paper on the theory of
representing all other commodities; and in, money choice, namely, the reaction of quantity
income. Let
demanded to a "compensated variation
x = g(y, y', in,) ()
of price," i.e., to a variation in price ac-
be the demand curve for the commodity in ques-
companied by a compensating change in
tion, given a utility function,
money income.7 Slutsky expressed the
ET= U(x, xu ), (2)
compensating change in money income
where u is a parameter to allow for changes in in terms of observable phenomena, tak-
taste, and subject to the condition ing it as equal to the change in price
Xy + Xty' = m . (3) times the quantity demanded at the ini-
From eq. (3) and the usual utility analysis, it fol- tial price. Mosak has shown that, in the
lows that eq. (i), like eq. (3), is a homogeneous
function of degree zero in y, y', and m; i.e., that The homogeneity of eqs. (5) and (6) in y, y
and m means that x is a function only of ratios
g(Xy, Xy', Xm, u) = g(y, y' mu)
M, . (4) among them. Thus eqs. (5) and (6) can be written:
On the current interpretation, a two-dimensional
demand curve is obtained from eq. (i) directly by x = g(y, y', m, u,) = g() - I,
giving y' (other prices), m (income), and u (tastes)
fixed values. A given value of y then implies a given
value of x from eq. (i), a given value of x' from
eq. (3), and hence a given value of U (i.e., real in-
come) from eq. (2). The value of U will vary with
y, being higher, the lower y is.
On my alternative interpretation, u and U are / I - X- \
given fixed values and x' is eliminated from eqs. (2) =

UO x g?)I,to ,
and (3). This gives a pair of equations,

x = g(y, y', m, uo), (5)
u0(Xx m, = uUO)
U0 = U0(X, M XY,U), (6)
The choice of price-compensating variations is
equivalent to selecting the forms of these two equa-
where the subscript o designates fixed values. The
tions in the next to the last terms of eqs. (5) and
two-dimensional variant involving compensating
(6'); of income-compensating variations, to selecting
variations in other prices is obtained by eliminating
y' from eqs. (5) and (6) and giving m a fixed value; the forms in the last terms.
the variant involving compensating variations in 7 Eugenio Slutsky, "Sulla teoria del bilancio del
income, by eliminating rn from eqs. (5) and (6) consumatore," Giornale degli economisti, LI (i9I5),S
and giving y' a fixed value. 1-26, esp. Sec.F8.

THE MARSHALLIANDEMAND CURVE 467
limit, the change in income so computed since other prices are supposed to be the
is identical with the change required to same for all points on it.
keep the individual on the same level of From the definition of the demand
utility (on the same indifference curve).8 curve Cc, OC is obviously the maximum
It follows that a similar statement is valid price per unit that an individual would
for compensating changes in other prices. be willing to pay for an infinitesimal ini-
In the limit, the change in other prices re- tial increment of X when his money in-
quired to keep the individual on the come and the prices of other commod-
same indifference curve when his money ities have the values assumed in drawing
income is unchanged but the price of one Cc. Let us suppose him to purchase this
commodity varies is identical with the amount at a price of OC, determine the
change in other prices required to keep
Price of X
unchanged the total cost of the basket of Priceof othercommodfies
commodities purchased at the initial A
prices, i.e., to keep unchanged the usual
type of cost-of-living index number.
C) COMPARISON OF THE INTERPRETATIONS
The relation between demand curves

constructed under the two interpreta-
C < ~~~~~~~~~~~
tions is depicted in Figure i. Curve Cc rep- H
resents a demand curve of an individual

consumer for a commodity X drawn on
the current interpretation. Money in- -V
come and the prices of other commod-
ities are supposed the same for all points o N M
on it; in consequence, real income is Quaritityof X
lower at C than at P, since, if the individ- FIG. I.-Comparison of demand curves con-
ual sought to buy OM of X at a price of structed under the two interpretations.
OC, he would be forced to curtail his pur-
chases of something else. As the curve is maximum price per unit he would be
drawn, of course, he buys none of X at a willing to pay for an additional incre-
price of OC, spending the sum of OHPM ment, and continue in this fashion, ex-
on other commodities that his action at a acting the maximum possible amount
price of OH shows him to value less for each additional increment. Let these
highly than he does OM units of X. The successive maximum prices per unit de-
ordinate is described as the ratio of the fine the curve Cv. The consumer obvi-
price of X to the price of other commod- ously has the same real income at each
ities. For the demand curve Cc this is a point on Cv as at C, since the maximum
question only of the unit of measure, price has been extracted from him for
8 Jacob L. Mlosak, "On the Interpretation of the
each successive unit, so that he has
Fundamental Equation of Value Theory," in 0. gained no utility in the process.
Lange, F. McIntyre, and T. 0. Yntema, Studies in Cv is now a demand curve constructed
MathematicalEconomics and Econometrics (Chicago: according to my interpretation of Mar-
University of Chicago Press, 1942), pp. 69-74, esp.
n. 5, pp. 73-74, which contains a rigorous proof of shall. If other prices are supposed to be
this statement by A. Wald. the same, the necessary compensating

468 MILTON FRIEDMAN
variations in money income as the price prices rather than in money income are
of X falls are given by triangular areas used to keep real income the same, the
exemplified by HCD for a price of OH: absolute price of neither X nor other
OH is the maximum price per unit that commodities can be read directly from
the individual will give for an additional Figure I. For each ratio of the price of X
infinitesimal increment of X when he has to the price of other commodities, the
spent OCDN for ON of X out of his ini- quantity of X purchased will be that
tial income of, say, m; but his situation shown on Cv. But the prices of other
is exactly the same if, when the price of goods will vary along Cv, rising as the rel-
X is OH, his income is (m - HCD) and ative price of X falls, so the absolute
he spends OHDN on X; he has the same price of X can no longer be obtained by
amount left to spend on all other com- multiplying the ordinate by a single
modities, their prices are the same, and scale factor.
he has the same amount of X; accord- Figure I is drawn on the assumption
ingly, his demand price will be the same, that X is a "normal" commodity, that is,
and he will buy ON of X at a price of OH a commodity the consumption of which
and an income of (m - HCD) .9 is higher, the higher the income. This is
If compensating variations in other the reason Cv is drawn to the left of Cc-
at every point on Cv other than C, real
9 In the notation of n. 6, except that u is omitted
for simplicity, the quantities of X and X' that will income is less than at the corresponding
be purchased for any given values of y and y' and point on Cc; hence less X would be con-
any given real income, U0, are obtained by solving sumed.
simultaneously:
Us Y
Curve Aa represents a demand curve
UZI Y
on my interpretation of Marshall for a
~~~~~~~(I)
and real income the same as at point P on
U(x, X') = Uo , (2) Cc; it is like Cv but for a higher real in-
where U. and U,' stand for the partial derivatives
come. Real income is higher on Aa than
of U with respect to x and x', respectively, i.e., for on Cc for prices above OH, lower for
the marginal utility of X and X'. The solution of prices below OH, which is the reason A a
these equations gives the demand curve on my in-
terpretation of Marshall, using compensating vari-
is to the right of Cc for prices above OH
ations in money income. and to the left of Cc for prices below OH.
Uo(o, m/y') is the utility at C in the diagram.
For any given amount of X and given value of y', d) WHY TWO INTERPRETATIONS ARE
the amount of X' purchased is obtained by solving POSSIBLE
The possibility of interpreting Mar-

U(x, x') = UO(O. m) (3)
shall in these two quite different ways
which is identical with eq. (2). The amount paid for arises in part from the vagueness of Mar-
X (the area under Cv) is shall's exposition, from his failure to give
m-xyf . (4) precise and rigorous definitions. A more
fundamental reason, however, is the exist-
The maximum price that will be paid per unit of X
is the derivative of eq. (4), or ence of inconsistency in the third and
later editions of the Principles. In that
dx' Uz ,
Y dxY ~e7'Y 5 edition Marshall introduced the cele-
brated passage bearing on the Giffen
which is identical with eq. (i). It follows that Cv
is a demand curve constructed on my interpreta-
phenomenon. This passage and a related
tion of Marshall. sentence added at the same time to the

Mathematical Appendix fit the current him concepts for organizing materials,
interpretation better than they fit my in- labels in an "analytical filing box." The
terpretation. Although these are the only "commodity" for which a demand curve
two items that I have been able to find in is drawn is another label, not a word for
any edition of the Principles of which a physical or technical entity to be de-
this is true, they provide some basis for fined once and for all independently of
the current interpretation. A hypothesis the problem at hand. Marshall writes:
to explain the introduction of this incon- The question where the lines of division
sistency into the Principles is offered in between different commodities should be
Section IVe below. drawn must be settled by convenience of the
particular discussion. For some purposes it
may be best to regard Chinese and Indian
II. THE RELATIVE USEFULNESS OF THE
teas, or even Souchong and Pekoe teas, as
TWO INTERPRETATIONS different commodities; and to have a separate
The relative usefulness of the two in- demand schedule for each of them. While for
other purposes it may be best to group to-
terpretations of the demand curve can be gether commodities as distinct as beef and
evaluated only in terms of some general mutton, or even as tea and coffee, and to have
conception of the role of economic the- a single list to represent the demand for the
ory. I shall use the conception that un- two combined.12
derlies Marshall's work, in which the
a) THE DISTINCTION BETWEEN CLOSELY
primary emphasis is on positive econom- RELATED AND ALL OTHER
ic analysis, on the forging of tools that COMMODITiES
can be used fairly directly in analyzing A demand function containing as sep-
practical problems. Economic theory arate variables the prices of a rigidly de-
was to him an "engine for the discovery fined and exhaustive list of commodities,
of concrete truth." "Man's powers are all on the same footing, seems largely
limited: almost every one of nature's foreign to this approach. It may be a use-
riddles is complex. He breaks it up, stud- ful expository device to bring home the
ies one bit at a time, and at last combines mutual interdependence of economic
his partial solutions with a supreme phenomena; it cannot form part of Mar-
effort of his whole small strength into shall's "engine for the discovery of con-
some sort of an attempt at a solution of crete truth." The analyst who attacks a
the whole riddle."" The underlying justi- concrete problem can take explicit ac-
fication for the central role of the concepts count of only a limited number of factors;
of demand and supply in Marshall's en- he will inevitably separate commodities
tire structure of analysis is the empirical that are closely related to the one im-
generalization that an enumeration of mediately under study from commodities
the forces affecting demand in any prob- that are more distantly related. He can
lem and of the forces affecting supply will pay some attention to each closely re-
yield two lists that contain few items in lated commodity. He cannot handle the
common. Demand and supply are to more distantly related commodities in
Io Alfred Marshall, "The Present Position of this way; he will tend either to ignore
Economics" (i885), reprinted in Memorials of Alfred 12
Marshall, Principles of Economics (8th ed.;
Marshall, ed. A. C. Pigou (London: Macmillan London: Macmillan & Co., Ltd., 1920), p. ioo1 n.
& Co.,Ltd., I925), p. I59. All subsequent page references to the Principles,
"lAlfred Marshall, "Mechanical and Biological unless otherwise stated, are to the eighth and final
Analogiesin Economics"( I898), ibid.,p. 3 I4. edition.

470 MILTON FRIEDMAN
them or to considerthem as a group. The ing primarilythe commodity in question.

formally more general demand curve On the other hand, any changes in them
will, in actual use, becomethe kind of de- would have a significant effect on that
mand curve that is yielded by my inter- commodity.They are put into the pound
pretationof Marshall. in order to separate problems, to segre-
The part of the Marshallianfiling box gate the particularreactionsunderstudy.
covered by ceteris paribus typically in- They are put in individually and ex-
cludes three quite differentkinds of vari- plicitly because they are so important
ables, distinguished by their relation to that account will have to be taken of
the variable whose adaptation to some them in any applicationof the analysis.
change is directly under investigation Price changes within the group of "all
(e.g., the price of a commodity): (a) vari- other commodities" and an indefinitely
ables that are expected both to be ma- long list of other variables are contained
terially affected by the variable under in group c. These variables are to be ig-
study and, in turn, to affect it; (b) vari- nored. They are too numerousand each
ables that are expected to be little, if at too unimportant to permit separate ac-
all, affected by the variable under study count to be taken of them.
but to materiallyaffect it; (c) the remain- In keeping with the spirit of Marshal-
ing variables, expected neither to affect lian analysis this classification of var-
significantly the variable under study iables is to be taken as illustrative, not
nor to be significantly affected by it. definitive. What particularvariables are
In demand analysis the prices of appropriatefor each group is to be de-
closely related commodities are the var- terrnined by the problem in hand, the
iables in group a. They are put individ- amount of informationavailable, the de-
ually into the pound of ceterisparibusto tail requiredin results, and the patience
pave the way for further analysis. Hold- and resourcesof the analyst.
ing their prices constant is a provisional
step. They must inevitably be affected b) CONSTANCY OF REAL INCOME
by anything that affects the commodity It has just been arguedthat any actual
in question; and this indirect effect can analysis of a concrete economic problem
be analyzed most conveniently by first with the aid of demand curves will in-
isolating the direct effect, systematically evitably adopt one feature of my inter-
tracing the repercussions of the direct pretation of Marshall-consideration of
effect on each closely related commodity, a residual list of commoditiesas a single
and then tracing the subsequentreflexin- group. For somewhatsubtler reasonsthis
fluences on the commodity in question. is likely to be true also of the second fea-
Indeed, in many ways, the role of the de- ture of my interpretationof Marshall-
mand curve itself is as much to provide holding real income constant along a de-
an orderly means of analyzing these in- mand curve. If an analysis, begun with a
direct effects as to isolate the direct demandcurve constructedon the current
effect on the commodity in question. interpretation, is carried through and
The average price of "all other com- made internally consistent, it will be
modities,"incomeand wealth, and tastes found that the demand curve has been
and preferences are the variables in subjected to shifts that, in effect, result
group b. These variables are likely to be from failure to keep real income con-
affectedonly negligiblyby factors affect- stant along the demandcurve.

THE MARSHALLIAN DEMAND CURVE 47I
An examplewill show how this occurs. seen most easily by supposing DD to

Let us suppose that the government have unit elasticity, so that the same
grants to producers of commodity X a amount is spent on X at P' as at P. The
subsidy of a fixed amount per unit of same amount is then available to spend
output, financed by a general income on all other commodities,and, since their
tax, so that money income available for prices are supposedto be the same for all
expenditure(i.e., net of tax and gross of points on DD under the currentinterpre-
subsidy) is unchanged. For simplicity, tation, the same quantity of each of them
suppose, first, that no commodities are will be demanded.But then wheredo the
closely related to X either as rivals or as resources come from to produce the ex-
complements, so that interrelations in tra MN units of X? Obviously, our as-
consumption between X and particular Priceof X
other commoditiescan be neglected; sec- A
ond, that the tax is paid by individualsin D
S
about the same income class and with
about the same consumptionpattern as
those who benefit from the subsidy, so
that complicationsarising from changes
in the distributionof income can be neg-
lected; and, third, that there are no idle
resources.Let DD in Figure 2 be a de-
mand curve for commodityX, and SS be
the initial supply curve for X, and let the D
initial position at their intersection,
point P, be a position of full equilibrium. \A
The effect of the subsidy is to lower the 0 M N
supply curve to S'S'. Since we have ruled Quantityof X
out repercussionsthrough consumption
FIG. 2.-Illustrative analysis of effect of subsidy
relationswith other marketsand through
changes in the level or distribution of
sumptions are not internally consistent.
money income, it is reasonableto expect The additional units of X can be pro-
that the intersection of this new supply duced only by bidding resources away
curve and the initial demand curve, from the production of other commod-
point P', will itself be a position of full ities, in the process raising their prices
equilibrium,involving a lower price and and reducing the amount of them pro-
largerquantity of X. Yet, if the demand duced. The final equilibrium position
curve is constructedon the currentinter- will therefore involve higher prices and
pretation and if the supply curve is not lower quantities of other commodities.
perfectly inelastic,'3point P' is not a po- But, on the current interpretation, this
sition of full equilibrium. This can be means a shift in the demand curve for
I3 If it is perfectly inelastic, neither the price nor X-say, to D'D'-and a final equilibrium
the quantity of X is changed, so the new position of position of, say,P."t4
equilibrium coincides with the old; but the demand
curve will pass through the initial position of equilib- 14 D'D' will not necessarily be to the left of DD
rium whether constructed on the current interpre- even for a "normal" commodity. The reason is that
tation or on mine; hence the two coincide at the the ordinate of Fig. 2 measures the absolute price
one point on them that is relevant. of X, so that ordinates of the same height on DD

472 MILTON FRIEDMAN
The assumption that the elasticity of The fundamental principle illustrated
DD is unity is not, of course, essential for by this example can be put more gen-
this argument. If the elasticity of DD is erally. The reason why a demand curve
less than unity, a larger amount than for- constructed under the current interpre-
merly is available to spend on other com- tation fails to give the correct solution
msodities;at unchanged prices this means even when all disturbing influences can
a larger quantity demanded. In conse- be neglected is that each point on it im.-
quence, while the additional amount of plicitly refers to a different productive
resources required to produce the in- capacity of the community. A reduction
creased amount of X demanded is smaller in the price of the commodity in question
when DD is inelastic than when it has is to be regarded as enabling the com-
unit elasticity, this is counterbalanced by munity, if it so wishes, to consume more
increased pressure for resources to proof some commodities-this commodity
duce other commodities. Similarly, when or others-without consuming less of any
DD is elastic, the additional amount of commodity. But the particular change
resources required to produce the in- in supply whose consequences we sought
creased quantity of X demanded is larger to analyze-that arising from a subsidy
than when DD has unit elasticity, but -does not make available any additional
some resources are released in the first resources to the community; any increase
instance from the production of other in the consumption of the commodity in
commodities. question must be at the expense of other
No such internal inconsistency as that commodities. The conditions for which
just outlined arises if the demand curve the demand curve is drawn are therefore
is constructed by keeping real income the
As was pointed out above (Sec. Ib), in the limit,
same. Curve AA is such a demand curve. holding real income constant is equivalent to hold-
At prices of X less than PM, prices of ing constant the cost of a fixed basket of commodi-
ties. Thus, if P" is considered close to P,
other commodities are supposed to be
sufficiently higher than at P to keep real X1Y1 XI-YI = XIY2 + XIY2. (2)
income the same, which involves the re- In the neighborhood of P, yI can be regarded as the
lease of just enough resources so that the cost per unit of producing X; y', as the cost per unit
of producing X'. The condition that sufficient re-
position of final equilibrium, P", lies on sources be released to permit the production of the
the demand curve so constructed-at requisite additional amount of X is therefore
least for small changes in the price of (X2 - X,)YI (X2 - XI')YIX (3)
X.I5
which is equivalent to
and D'D' represent different ratios of the price of XIYI + XIYI = X2YI + X21Y11 (4)
X to the price of other commodities. If the ordinate
measured the ratio of the price of X to the price of But, in the limit, eqs. (i) and (2) imply eq. (4), as
other commodities, D'D' would always be to the can be seen by subtracting eq. (2) from eq. (i)
left of DD for "normal" commodities, always to the and replacing y2and y' in the result by (Y2 - YI + Y)
right for "inferior" commodities. and (y' - y' + Y), respectively.
More generally, constant real income involves
'5 Let X' be a single composite commodity
keeping a price index unchanged; constant use of
representing all commodities other than X; x and resources involves keeping a quantity index un-
x', the quantities of X and X'; and y and y', their changed; and, in the limit, a constant price index
prices. Let the subscript i refer to values at the and constant total expenditures imply a constant
initial position of equilibrium, P; the subscript 2, quantity index.
to values at the final position, P". The condition Note that AA need not be steeper than DD in a
of constant total expenditures means that graph like Fig. 2. The point in question is that com-
XIyI + Xtyl = X2y2 + X2y2. (I) mented on in n. 14.

inconsistent with the conditions postu- wants depend principally on two factors
lated on the side of supply. On the other -the total resources at his disposal and
hand, if the demand curve is constructed the terms on which he can exchange one
by keeping "real income" the same, no commodity for another, that is, on his
such inconsistency need arise. True, con- real income and on relative prices. The
stant "real income" in the sense of "util- form of analysis that is now fashionable
ity" and constant "real income" in the distinguishes three effects of changes in
sense of outputs attainable from a fixed his opportunities-the income effect aris-
total of resources are different concepts, ing from changes in his money income;
but they converge and can be treated as the income effect arising from changes in
the same in the neighborhood of a position the price of a commodity, with un-
of equilibrium. changed money income and prices of
Of course, not all shifts in supply that other commodities; and the substitution
it is desired to analyze arise in ways that effect arising from a change in the rel-
leave the productive capacity of the com- ative price of a commodity, with un-
munity unaltered. Many involve a changed real income.
change in productive capacity-for ex- The distinction between the so-called
ample, changes in supply arising from "substitution" and "income" effects of a
improvements in technology or the dis- change in price is a direct consequence of
covery of previously unknown resources. defining the demand curve according to
Even in these cases, however, a demand the current interpretation of Marshall.
curve constructed on the current inter- Its basis is the arithmetic truism that at
pretation will not serve. There is no rea- given prices for all commodities but one,
son to expect the differences in produc- a given money income corresponds to a
tive capacity implicit in constant money higher real income, the lower the price
income and constant prices of other of the remaining commodity-at a lower
goods to bear any consistent relation to price for it, more of some commodities
the change in productive capacity arising can be purchased without purchasing
on the side of supply."6 The better plan, less of others. In consequence, a decline
in these cases, is to allow separately and in the price of a commodity, all other
directly for the increase in productive prices constant, has, it is argued, two
capacity by redrawing the demand curves effects: first, with an unchanged real in-
to correspond to an appropriately higher come, it would stimulate the substitution
real income and then to use a demand of that commodity for others-this is the
curve on which all points refer to that substitution effect; second, if the money
higher real income. income of the consumers is supposed to
The main point under discussion can be unchanged, the increase in their real
be put still more generally. The oppor- income as a result of the decline in price
tunities open to a consumer to satisfy his causes a further change in the consump-
tion of that commodity as well as of
I6 Note the difference from the previous case
others-this is the income effect.'7
of constant productive capacity. As stated above,
there is reason to expect constant real income along 17 See Slutsky, op. cit.; Henry Schultz, The
a demand curve to bear a consistent relation to Theory and Measurement of Demand (Chicago:
constant productive capacity in the neighborhood University of Chicago Press, I938), pp. 40-46;
of equilibrium. The reason, in effect, is provided by J. R. Hicks and R. G. D. Allen, "A Reconsideration
one of the conditions of equilibrium: the tangency of the Theory of Value," Economica, XIV (I934),
of consumption and production indifference curves, 52-76 and i96-219; Hicks, op. cit., Part I.

474 MILTON FRIEDMAN
The two different kinds of income III. THE CONSISTENCY OF THE ALTERNA-
effects distinguished in this analysis- TIVE INTERPRETATION WITH OTHER
one arising from a change in money in- PARTS OF MARSHALL'S WORK
come, the other from a change in the Marshall's demand curve is part of a
price of one commodity-are really the coherent body of thought; it is designed
same thing, the effect of a change in real to fit into the rest of his structure of anal-
incomewith given relative prices, arising ysis; and it is used extensively in devel-
in different ways. It is hard to see any oping and applying this structure. It
gain from combining the second income would take us too far afield to demon-
effect with the substitution effect; it strate in detail that my interpretation
seems preferableto combine the two in- of his demand curve is consistent with
come effects and thereby gain a sharp the rest of his work. However, two
contrastwith the substitutioneffect. special topics call for some explicit con-
It has often been stated that Marshall sideration: (I) the relation between the
"neglected the income effect."'8 On my demand curve and Marshall's theory of
interpretationof his demand curve, this money, because, in my view, this ex-
statement is invalid. One must then say plains the particular device that he
that Marshallrecognizedthe desirability adopted for holding real income con-
of separating two quite different effects stant; and (2) the concept of consumer's
and constructed his demand curve so surplus, because this is one of the most
that it encompassedsolely the effect that important applications of the demand
he wished to isolate for study, namely, curve and certainly the most controver-
the substitution effect. Instead of neg- sial and because the passages in the later
lecting the income effect, he "elim- editions of the Principles that are incon-
inated" it. sistent with my interpretation were in-
The conclusionto which the argument troduced into the discussion of con-
of this section leads is identical with that sumer's surplus.
reachedby Frank H. Knight in a recent
article, in which he says: a) THE THEORY OF RELATIVE PRICES AND
THE THEORY OF MONEY
We have to choose in analysis between
holding the prices of all other goods constant Granted that real income is to be held
and maintaining constant the "real income" constant along the demand curve, why
of the hypothetical consumer.... The treat- do so by holding money income and the
ment of the Slutzky school adopts the assump- purchasing power of money constant ra-
tion that . .. the prices of all other goods (and ther than, for example, by holding prices
the consumer's money income) are constant.
Hence, real income must change. Of the two al- of other goods constant and permitting
ternatives, this seems to be definitely the wrong compensating variations in money in-
choice.... The simple and obvious alternative come? What reason is there to treat the
is to draw the demand curves in terms of a prices of all other commodities as moving
change in relativeprices, i.e., to assume that the inversely to the price of the commodity
value of money is held constant, through com-
pensating changes in the prices of other goods, in question?
and not that these other prices are held con- The answer to these questions is given,
stant.19 I think, by one of Marshall's basic or-
'8 Hicks, op. cit., p. 32.
(December, I944), esp. Sec. III, "The
289-3P8,
"Realism and Relevance in the Theory of
It Meaning of a Demand Curve," pp. 298-301.
Demand," Journal of Political Economy, LII Quotation from p. 299.

ganizing principles, namely, the separa- fixed (nominal)stock of money, leave the
tion of the theory of relative prices from community with a larger "stock of com-
monetary theory, the theory of the level mand over commodities"than previous-
of prices.The Principlesis devoted to the ly, and establish an incentive (reflecting
theory of relative prices under given "monetary" forces) to increase expend-
monetary conditions; Money, Credit, and itures and thereby raise prices until the
Commerceto the analysis of monetary fixed stock of money again represented
conditions and their effect on the "pur- the same "stock of command over com-
chasing power of money." With given modities," i.e., until the "purchasing
monetary conditions, is it possible for the power of money" reached its former
prices of all commoditiesother than the level.21 This argument suggests that not
one in question to remain the same, on only was constant purchasing power of
the average, while the price of this one money a device for separatingthe theory
rises or falls? Will not a rise or fall in the of relative prices from monetary theory,
price of the commodityin question set in it was also a bridge between the two.
motion monetaryforces affecting other Marshall separated the two theories in
prices? A complete answer requires ex- his attempt to reduce problems to man-
plicit specification of the content of ageable proportions, but he constructed
"given monetary conditions" and per- them in such a way as to make them
haps also of the sourceof the initial price mutually consistent and thus facilitate
change. ultimate combination.22
Marshall'sselection of a constant pur- C. W. Guillebaud has pointed out to me that
2!
chasing power of money as a means of Marshall typically supposed the desired "stock of
impoundingmonetary forces is presum- command over commodities" to be a given fraction
ably the end-result of a chain of reason- of real income (see ibid.) and that the argument in
the text might not apply if this fraction were taken
ing about the influence of monetary as the fundamental given. The monetary effects of
forces, not the direct content that he a change in one price, other prices given, would
depend on the source of the initial price change.
gave to "given monetary conditions." then If this involved no change in aggregate real in-
The beginning of the chain of reasoning come (e.g., arose from a shift in demand), the argu-
xnaywell be his own version of the quan- ment in the text would remain unchanged. If it did
a change in aggregate real income (e.g., arose
tity theory of money. According to this involvefrom an invention reducing the cost of producing the
version, "the value of money is a func- commodity in question), no inconsistency need arise,
tion of its supply on the one hand, and since the desired "stock of command over commodi-
ties" would change in proportion to the change in
the demand for it, on the other, as mea- real income. These considerations account for the
sured by 'the average stock of command phrase "perhaps also of the source of the initial
over commodities which each person price change" at the end of the preceding paragraph
cares to keep in a ready form.' Given of the text.
"120
" This interpretation would, of course, be con-
monetary conditionswould then imply a tradicted if Marshall had devised his theory of
given stock of money and a given de- money after he had substantially completed his
sired "average stock of command over theory of relative prices, as might be inferred from
commodities." A decline in one price published the fact that Money, Credit, and Commercewas not
until 1923, thirty-three years after the
alone, all other prices remaining the first edition of the Principles. But in Marshall's
same,is inconsistentwith these "givens." case, the order of publication is a poor guide to the
It would increase the real value of a essence order of construction. Keynes tells us that the
of his quantity theory of money is contained
20
J. M. Keynes, "Alfred Marshall, i842-1924," in a manuscript "written about I871"; that "by
Memorials, p. 29. i87i his progress along" the lines of the material

476 MILTON FRIEDMAN
Marshall was, of course, very much arated, they are not basically independ-
aware of the interaction between real ent. From this point of view it is entirely
and monetary factors. The i879 Eco- natural that the recent development of
nomics of Industry contains an extremely alternative monetary theories should
interesting discussion of the trade cycle, have stimulated re-examinationof price
part of which Marshall thought suffi- theory.
ciently important to quote at length in b) CONSUMER'S SURPLUS
i886 in answering questions circulated
Marshall's discussion of consumer's
by the celebratedRoyal Commissionon surplus constitutes one of the most ex-
the Depression of Trade and Industry.23
tensive applications that he made of his
Marshall's decision to keep the pur-
demand curve and has probably given
chasing power of money the same for
rise to more controversy and discussion
differentpoints on a demand curve may
than any other part of his theory. Re-
not be the device best suited to abstract
cently, consumer's surplus has come in
from monetary factors. It serves, how-
for renewedattention, primarilyas a re-
ever, to emphasize the necessity of con- sult of J. R. Hick's attempt to rehabil-
sideringexplicitly the monetary arrange-
itate and reinterpret the concept.24The
ments under which the forces affecting
reasonfor commentingon it here is not to
relative prices are supposed to operate.
contribute to the discussion or to eval-
The best apparatusfor tacklingproblems
uate the merits or demerits of the con-
of relative prices cannot be determined
cept but rather to show the relation be-
independentlyof these arrangementsand
tween Marshall's treatment of con-
of their mode of operation.Though price
sumer's surplus and my interpretation of
theory and monetary theory can be sep-
his demand curve.
contained in the Pure Theory "was considerably Marshall's treatment of consumer's
advanced"; that the Pure Theory itself was "sub- surplus might, offhand, seem inconsistent
stantially complete about i873," though not printed
even for private circulation until i879; that "in with my interpretation of his demand
i877 he turned aside to write the Economics of curve for either of two different, and al-
Industry with Mrs. Marshall"; and that Mrs. most opposed, reasons. In the first
Marshall said "Book III on Demand was largely
thought out and written on the roof at Palermo, place, consumer's surplus refers to a
Nov. i88i-Feb. i882" (Memorials, pp. 28, 21, 23, difference in real income under different
39 n.). These dates are extremely suggestive, par- situations. But, on my interpretation, all
ticularly since the constancy of the purchasing
power of money is not explicitly mentioned in the 24See Hicks, op. cit., pp. 38-41; "The Rehabilita-
Pure Theory, which Marshall was presumably tion of Consumers' Surplus," Review of Economic
working on at about the same time that he was Studies, VIII (February, 194I), io8-I6; "Con-
developing his monetary theory, while it is explicitly sumers' Surplus and Index Numbers," ibid. (sum-
mentioned in the i879 Economics of Industry, mer, I942), I26-37; "The Four Consumer's Sur-
begun some years later. See also nn. 37 and 38 below. pluses,"ibid., XI (winter, I943), 3I-4I. See also A.
23 See Alfred Marshall and Mary Paley Marshall, Henderson, "Consumer's Surplus and the Compen-
Economics of Industry (London: Macmillan & Co., sating Variation," Review of Economic Studies,
ist ed., i879; 2d ed., i88i), Book III, chap. i, VIII (February, I94I), II7-2I; Knight, op. cit.;
pp. I50-57. This and all later references are to the Kenneth E. Boulding, "The Concept of Economic
first edition. "Answers to Questions on the Subject Surplus," American Economic Review, XXXV
of Currency and Prices Circulated by the Royal (December, I945), 85i-69, reprinted in American
Commission on the Depression of Trade and In- Economic Association, Readings in the Theory of
dustry (i886)," Official Papers by Alfred Marshall Income Distribution (Philadelphia: Blakiston Co.,
(London: Macmillan & Co., Ltd., 1926), pp. 7-9. I946), pp. 638-59; E. J. Mishan, "Realism and
See also "Remedies for Fluctuations of General Relevance in Consumer's Surplus," Review of
Prices" (i887), Memorials, pp. i89-92. Economic Studies, XV (I947-48), 27-33.

points on the demand curve are to be re- Marshall's definition of consumer's sur-
garded as corresponding to the same real plus, his suggested estimate of its rrag-
income. A movement along such a de- nitude, and the relation of this estimate
mand curve cannot, therefore, involve a to the correct value under the two al-
change in consumer's surplus. Does this ternative interpretations of the demand
not eliminate the entire notion of con- curve.
sumer's surplus and make Marshall's en- Marshall is more explicit and complete
tire discussion of it pointless? The an- in defining consumer's surplus than was
swer is clearly "No," the reason being his wont, and his definition adm its of
that the two situations compared need little ambiguity: "The excess of the price
not correspond to two points on the same which he would be willing to pay rather
demand curve, even though a single de- than go without the thing, over that
mand curve is used to estimate the differ- which he actually does pay, is the eco-
ence in real income between the two sit- nomic measure of this surplus satisfac-
uations. tion. It may be called consumer's sur-
In the second place, Marshall re- plus. n25
garded his analysis of consumer's surplus Marshall then proceeds to argue that
as valid only for commodities that ac- consumer's surplus can be estimated by
count for a small part of total expend- the famous triangle under the demand
iture. He makes this restriction in order curve. As Hicks remarks, this "associa-
to justify neglecting changes in the mar- tion of Consumer's Surplus with the
ginal utility of money. But, if all points curvilinear triangle under the demand
on the demand curve correspond to the curve . . . is not a definition; it is a the-
same real income, does it not then follow orem, true under certain restrictive as-
that the marginal utility of money is the sumptions, but only true if these assump-
same everywhere on the demand curve? tions are granted."126 The confusion of the
And does it not also follow that his esti- suggested estimate with the definition is
mate of consumer's surplus is exact, so perhaps the chief source of misunder-
that the assumption that a negligible standing on this exceedingly complex
proportion of expenditures is devoted to subject.
the commodity in question becomes un- Figure i, introduced in Section Ic
necessary? Again the answer is "No," above to illustrate the relation between
and for much the same reason. If the two demand curves drawn on the current and
situations compared differ in real income, on my interpretation, can also be used to
the fact that real income is the same show the relation between consumer's
along the demand curve becomes somne- surplus as defined and estimates of it ob-
thing of a vice in using it to measure contained from demand curves constructed
sumer's surplus. The assumption that a according to the two interpretations.
negligible proportion of expenditures is Curve Cc, it will be recalled, is a demand
devoted to the commodity in question curve for the commodity X constructed
cannot be dispensed with on my inter- according to the current interpretation.
pretation; indeed, if anything, it is even Money income and all other prices are
more necessary than on the current in- the same for all points on it. Aa and Cv
terpretation. 25 Principles, p. I24.
To explain and justify these cryptic 26
"Rehabilitation of Consumers' Surplus," p.
answers, it will be necessary to examine I09.

478 MILTON FRIEDMAN
are demand curves constructed accord- der Cv between 0 and M, or OCDGM.
ing to my interpretation-A a for a real The triangulararea CDH minus the tri-
income the same as at P on Cc; Cv for a angular area DPG therefore gives the
real income the same as at C on Cc. At consumer's surplus. This area is neces-
point P on Acaand at point C on Cv, sarily positive; we know he is willing to
money incomeand all other prices are the pay at least OHPM for OM of X; hence
same as on Cc. At other points other OCDGMmust be greaterthan OHPM.
prices are sufficientlydifferent,or money Marshall's estimate of the maximum
income is, to compensate for the differ- sum is the area under the demand curve:
ence in the price of X and thereby keep OCPM if we use the current interpreta-
real incomethe same. tion, OAPM if we use the alternative in-
Now consider the consumer's surplus terpretation. For a "normal" commod-
obtained from this commodity when the ity, the case for which the figure is
consumeris at p.27 This is definedas "the drawn, both are clearly too large. How
excess of the price which he would be large the error is depends on the differ-
willing to pay rather than go without the ence between Aca and Cc, on the one
thing, over that which he actually does hand, and Cv,on the other. Now we have
pay." "Price"is here to be interpretedas seen (in Sec. Ic) that these differences
"total amount" rather than "price per arise entirely from differencesin the real
unit.' 28 Further, it is clear that the sum income associated with the different
he wouldpay ratherthan go without is to curves; if real incomes differ little, so
be determined for circumstances other- will the curves. Here is where Marshall's
wise the same as at P; in particular,his assumptionabout the fractionof expend-
money income and the other prices are to itures devoted to the commodity enters
be the same as at p.29 Now the amount the picture. If this fraction is small, the
that he actually does pay for OM of X is differencesin real income will tend to be
given by the rectangle OHPM in the small, and both estimates will approach
figure. By the argument of Section Ic, the correct value.3o Since the error is
the maximum amount that he would be 30 This statement is not rigorous. As the fraction
willing to pay for OM of X ratherthan go of expenditures devoted to the commodity dimin-
without any of it is given by the area un- ishes, so will aggregate consumer's surplus. It is
not enough that the error become small in absolute
27 For simplicity, the discussion is restricted
terms; its ratio to the correct value must become
to the consumer's surplus obtained from the entire
small. This, in general, will occur, as is well known.
amount of X consumed; and to facilitate this, the The chief qualification has to do with the behavior
demand curves have been drawn to cut the price
of a demand curve constructed under the current
axis. interpretation (e.g., Cc) for small quantities of X.
28 See Mathematical Note II of the Principles
The crucial question is the difference in real income
(p. 838), in which Marshall defines p as "the price between P and C. Expenditure on the commodity
which he is just willing to pay for an amount x of the might be a small fraction of total expenditure at P;
commodity" and then differentiates p with respect yet, if the demand curve constructed under the
to x to get the price per unit. current interpretation were extremely inelastic,
29 None of the reasons cited earlier for keeping not near C. In this case the difference in real income
real income the same along the demand curve apply might be large.
here. The question being asked is purely hypotheti- This qualification is emphasized by Marshall.
cal; no other reactions need be allowed for. Further, For example: "If however an amount b of the com-
to keep his real income the same when he has none modity is necessary for existence, f(z) [sic] [ordinate
of X as when he has OM of X would make the entire of the demand curve] will be infinite, or at least in-
discussion of consumer's surplus pointless. The definitely great, for values of x less than b. We must
whole point of the discussion is to measure the differ- therefore take life for granted, and estimate sepa-
ence in real income between the two situations. rately the total utility of that part of the supply of

larger for Aa than for Cc, it is clear that nition of the demand curve-the prob-
Marshall's assumption is, if anything, lemiof form; (2) the shape of the demand
even more necessary on my interpreta- curve-the problem of content; and (3)
tion of the demand curve than on the the use of the demand curve-the prob-
current one.3' lem of application. In his usual manner
Marshall gives precedence to the prob-
IV. TEXTUAL EVIDENCE ON WHAT
lem of content and does not explicitly
MARSHALL REALLY MEANT
separate his discussion of content from
Marshall's writings on demand bear his discussionof form. His definitionsare
on three different problems: (i) the defi- characteristically given parenthetically
the commodity which is in excess of absolute neces- and implicitly. He went to extreme
saries" (p. 84I). See also pp. I33 and 842.f(z) clearly lengths to present his tools in the context
should bef(x), as it is in the first four editions of the of concrete problems, so that definitions
Principles. See appendix to this paper.
This discussion of the role of the assumption that grewout of the uses to be made of them.32
the commodity absorbs only a small fraction of His discussion of utility and diminishing
income throws some light on an issue about which utility in the chapter of the Principles
there has been considerable discussion, namely,
whether Marshall assumed the marginal utility of which introduces the concept of a de-
money to be roughly constant with respect to a mand curve (Book III, chap. iii, "Grada-
change in price or a change in income. The above tions of Consumers'Demand") is part of
analysis suggests that he assumed constancy with
respect to a change in income. This is also Hicks's
the discussionof content, even though it
conclusion (Value and Capital, p. 40; "Rehabilita- precedes his definition. It is the means
tion," p. io9). Samuelson denies this and asserts whereby he rationalizeshis "one general
that he assumed constancy with respect to a change
in price (see Paul A. Samuelson, "Constancy of the law of demand:-The greater the amount
Marginal Utility of Income," in Studies in Mathe- to be sold, the smaller must be the price
matical Economics and Econometrics, p. 8o). at which it is offeredin orderthat it may
3I The argument can be easily extended to "in-
find purchasers."33 It is not part of his
ferior" goods. The order of the three curves in Fig. i
is then reversed; the estimates then become too definitionof the demandcurve.
small, instead of too large; but the error under Similarly, one of the major applica-
the alternative interpretation remains larger in tions that Marshallmade of the demand
absolute value than under the current interpretation.
In the terminology used by Hicks in "The Four curve was his analysis of consumer's
Consumers' Surpluses," what I have called the con- surplus. This analysis, too, must be dis-
sumer's surplus is what Hicks calls the "quantity- tinguished from his definition of the de-
compensating variation." The estimate of con-
sumer's surplus derived from the demand curve mand curve. Assumptions made in his
constructed under my interpretation (the area 32 Cf. J. M. Keynes, Memorials, esp. pp. 33-38;
APH) Hicks calls the "quantity-equivalent varia-
see also Guillebaud, op. cit.
tion." The area CDH in Fig. i, Hicks calls the
"price-compensating variation." Hicks's fourth 33 Principles, p. 99. Note that on my interpreta-
concept, "price-equivalent variation," is not shown tion this is truly a general law, not subject to the
directly in Fig. i. It is obtained by drawing a exceptions that have been made in recent literature.
horizontal line through C. Designate by E the point It depends for its validity only on (a) the postulate
at which this line cuts Aa. The "price-equivalent that consumers can be treated as if they behaved
variation" is then equal to the area APH minus consistently and attempted to maximize some
AEC. These relations can be checked by noting function of the quantity of commodities consumed;
that curve mep in Hicks's Fig. 3 is A a in our Fig. I; (b) the observed fact that consumers choose a higher
his curve PCM is Cv in our Fig. i. Further, in com- income in preference to a lower, other things the
paring the two figures, the part of Hicks's diagram same; and (c) the observed fact that consumers do
for quantities less than hN should be neglected. That not spend all their income on one commodity. For
is, his point P is equivalent to our point C, his p proof that a demand curve constructed on my inter-
to our P. Our Fig. i is also equivalent to Fig. 3B in pretation must slope negatively see Slutsky, op. cit.,
Boulding, "The Concept of Economic Surplus." Sec. 8.

480 MILTON FRIEDMAN
discussion of consumer's surplus cannot, I87934 but, according to Keynes, "sub-

without additional evidence, be supposed stantially complete about i873" ;35 and
to apply equally to other applications of the constancy of the purchasing power of
the "demand curve." money is in his and Mrs. Marshall's The
Economics of Industry, published in
a) THE CENTRAL PASSAGES IN THE TEXT i879.36 The actual wording of the first
OF THE "PRINCIPLES") and third quotations can be traced back
The central passages in the text of the to the first edition of the Principles
eighth and final edition of the Principles (i890), of the second quotation, to the
bearing on the other things to be kept the second edition (i8gi).37
same seem to me to be three: one govern-
34 Reprinted, together with the companion
ing the entire volume, and two essentially paper, The Pure Theory of Foreign Trade, by the
parenthetical comments in his discussion London School of Economics and Political Science
of the demand curve: (1930).
35 Memorials,p. 23.
We may throughout this volume neglect
36 This work should not be confused with the con-
possible changes in the general purchasing
power of money. Thus the price of anything densation of the Principles, published, under the
will be taken as representative of its exchange same title but with Alfred Marshall as sole author,
in i892.
value relatively to things in general [p. 621.
37 In all editions of the Principles the statement
The larger the amount of a thing that a per-
son has the less, other things being equal (i.e. corresponding to the first quotation is in a sub-
the purchasing pozverof money, and the amount section dealing with the meaning of the word
"value." In the first (i890), second (i89i), and
of money at his command being equal), will be third (i895) editions, the subsection on "value" is
the price which he will pay for a little more of at the end of Book I, "Preliminary Survey," chap. i,
it: or in other words his marginal demand price "Introduction," and contains the statement:
for it diminishes [p. 95; italics added]. "Throughout the earlier stages of our work it will
The demand prices in our list are those at be best to speak of the exchange value of a thing at
which various quantities of a thing can be sold any place and time as measured by its price, that is,
in a market during a given time and under given the amount of money for which it will exchange
conditions. If the conditions vary in any respect then and there, and to assume that there is no
the prices will probably require to be changed; change in the general purchasing power of money"
(p. 9, all three editions). In the first edition this
and this has constantly to be done when the assumption is repeated at the beginning of the
desire for anything is materially altered by a chapter on "The Law of Demand" (Book III,
variation of custom, or by a cheapening of the chap. ii): "The purchasing power of this money
supply of a rival commodity,or by the invention may vary from time to time; but in these early
of a new one [p. ioo; second set of italics added]. stages of our work we assume it to be constant"
(ist ed., p. IVi). This repetition was eliminated in
For our purposes the critical part of later editions, apparently in the process of intro-
ducing into the second edition the chapter on
the second quotation is the italicized pa- "Wants in Relation to Activities." In the fourth
renthesis and, of the third, the second set edition (i898), the subsection on "value" was split,
of italicized phrases. part remaining at the end of Book I, chap. i, the
remainder, including the material on the purchasing
Though these quotations are taken power of money, being transferred to end of Book 11,
from the eighth edition of the Principles, "Some Fundamental Notions," chap. ii, "Wealth."
their substantive content is contained in The wording was changed to essentially its final
form; the only difference is that the first sentence
Marshall's earliest published work on the is in the passive voice, reading: "Throughout this
theory of demand. All except the con- volume possible changes in the general purchasing
stancy of the purchasing power of money power of money will be neglected" (4th ed., p. 130).
In the fifth edition (1907), the rest of the subsection
is in The Pure Theory of (Domestic) Val- on "value" was transferred to the end of Book II,
ues, printed for private circulation in chap. ii, and the quotation revised to its present

THE MARSHALLIAN DEMAND CURVE 48i
b) THE BEARING OF THESE PASSAGES ON 3. "Custom"

THE TWO INTERPRETATIONS 4. Price of "a rival commodity" (to
The "other things" listed in the three avoid "cheapeningof the supply of a
passages cited above are as follows: rival commodity")
I. "Purchasing power of money" marginal utility of money to him, and diminishes
2. "Amount of money at his command" the price that he is willing to pay for any pleasure"
(p. I56). The only change in this statement in
later editions was the substitution of "benefit"
form; even the page number is the same in the for "pleasure" (8th ed., p. 96).
fifth and eighth editions (p. 62). The Economics of Industry also contains a state-
In both editions of The Economics of Industry, ment anticipating the second part of the italicized
subsection 4 in Book II, "Normal Value," chap. i, parenthesis: "The price which he is willing to pay
"Definitions. Law of Demand," contains essentially for a thing depends not only on its utility to him
the same material as the subsection on "value" but also on his means; that is, the amount of
in the Principles referred to in the preceding para- money or general purchasing power at his disposal"
graph, including the following statement: "But
(P- 70).
while examining the theory of Normal value we In all editions of the Principles the statement cor-
shall, for convenience, assume that the purchasing responding to the third quotation is in the final
power of money remains unchanged. So that a rise subsection of the chapter first introducing the de-
or fall in the price of a thing will always mean a mand curve (ist ed., Book III, chap. ii; in later
rise or fall in its general purchasing power or ex- editions, Book III, chap. iii). In the first edition
change value" (pp. 68-69). No corresponding it reads: "It must be remembered that the demand
statement appears in The Pure Theory. schedule gives the prices at which various quantities
The italicized parenthesis in the second quota- of a thing can be sold in a market during a given
tion is identical in the second and all later editions time and under given conditions. If the conditions
of the Principles. The remainder of the quotation vary in any respect the figures of the schedule will
was worded as follows in the second edition: "An probably require to be changed. One condition
increase in the amount of a thing that a person has which it is especially important to watch is the
will, other things being equal ... diminish his price of rival commodities, that is, of commodities
Marginal Demand-price for it" (p. I52). In the which can be used as substitutes for it" (p. i6o). A
third edition, the words "marginal" and "demand" footnote is attached to the word "rival," the first
were not capitalized, and the hyphen was eliminated sentence of which reads: "Or to use Jevons' phrase
after "Demand" (p. I70). In the fourth edition (Theory of Political Economy, Ch. IV), commodities
the end of the statement was expanded to read, that are nearly 'equivalent' "(ist ed., p. i6o, n. 2).
"diminish the price which he will pay for a little The part of the second sentence of the third
more of it: or in other words diminishes his magrinal quotation following the semicolon assumed its
demand price for it" (pp. 169-70). In the fifth final form in the second edition (p. I57), the foot-
edition the quotation takes its present form, except note reference to Jevons being dropped. The rest
for the addition of a comma, even the page number of the quotation is the same in the second and
being the same as in the eighth edition (p. 95). In third editions as in the first and assumes its final form
all editions from the second on, the indicated quo- in the fourth (p. I 74). The change made in the second
tations are in Book III, chap. iii, the chapter first sentence from the first to the second edition argues
introducing the demand curve. This chapter is that the list was not intended to be exhaustive, but
entitled "The Law of Demand" in the second and illustrative. No change in substance is involved (see
third editions, "Gradations of Demand" in the ist ed., p. I55). In all editions the quoted state-
fourth, and "Gradations of Consumers' Demand" ment is followed by the example of tea and coffee
in the fifth and later editions. to illustrate the necessity of assuming the prices
The absence of the statement from the first of rival commodities to be known; in the second
edition reflects a difference in exposition, not in edition the example of gas and electricity was
substance. As noted above, an explicit statement added, and in the third edition the example of
that the purchasing power of money is assumed different varieties of tea. The passage itself, the
constant appears in the chapter on "The Law of changes in it, and the examples all indicate that
Demand" in the first edition. In all editions this Marshall considered the price of "rival" commodi-
chapter contains a statement covering the second ties particularly important. The examples, to-
part of the italicized parenthesis, which is worded gether with the footnote in the first edition, make
as follows in the first edition: "Every increase in it clear that he meant "close" rivals.
his resources increases the price which he is willing For a statement in the Pure Theory covering the
to pay for any given pleasure. And in the same way substance of these quotations, except the constancy
every diminution of his resources increases the of the purchasing power of money, see n. 38, below.

482 MILTON FRIEDMAN
5. Range of rival commodities available 3 to their tastes and preferences, and
(to avoid "invention of a new one")31 item 4 to the price of every other com-
modity rather than of rival commodities
i. The currentinterpretation.-The cur-
alone. It ignores entirely items i and 5.
rent interpretation of Marshall's demand
curve treats item 2 as referring to the Item 2 is not entirely unambiguous. It
money income of the group of purchasers might be interpreted as referring to the
cash balances of the purchasers or to
to whom the demand curve relates, item
their wealth instead of, or in addition to,
38 The adequacy of this list as a summary of their income. On the whole, the most
Marshall's views may be checked by comparing it reasonable course seems to be to inter-
with two others in Marshall's writings. In The Pure
Theoryof (Domestic) Values, he writes: "The periods pret it as referring to both income and
with which we are concerned ... are sufficiently wealth,39 particularly since wealth qual-
long to eliminate .. . casual disturbances.... But ifies for the list of "other things" by vir-
they are sufficiently short to exclude fundamental
changes in the circumstances of demand and in those tue of its possible importance as a factor
of supply. On the side of demand for the ware in affecting consumption. This expansion
question it is requisite that the periods should not of the current interpretation does not
include (i) any very great change in the prosperity
and purchasing power of the community; (ii) any alter it materially; it merely transfers
important changes in the fashions which affect the "wealth" from the category of "other
use of the ware; (iii) the invention or the great things" implicitly supposed to be the
cheapening of any other ware which comes to be
used largely as a substitute for it; (iv) the de- same to the list of things mentioned ex-
ficiency of the supply of any ware for which the ware plicitly.
in question may be used as a substitute, whether
this deficiency be occasioned by bad harvests, by
war, or by the imposition of customs or excise letters in brackets added). This statement dates
taxes; (v) a sudden large requirement for the com- from the first edition (pp. I7-7 I); only trivial
modity, as e.g. for ropes in the breaking out of a editorial changes were made in later editions.
maritime war; (vi) the discovery of new means of Item a in this list corresponds with i in my list;
utilizing the ware, or the opening up of important b with 2; d and presumably e with 3; andf with 4 and
markets in which it can be sold" (p. I5). 5. Item c is presumably in part covered by restric-
tion of the discussion to a demand curve for a
Item i in this list presumably corresponds with
particular market; in part it contains an item that
2 in my list; ii corresponds with 3, and iii and iv
may deserve to be added to the list, namely,
with 4 and 5, iii excluding a fall in the price of a
"wealth." The wording of f is ambiguous, since it
rival commodity and iv a rise. Items v and the first
could refer to substitutes for the good in question,
part of vi would seem to be contained in 3 and largely
to complements, or to both. The subsequent text
redundant with ii. The rest of vi is presumably
and the examples cited make it clear that it refers to
covered by the restriction of the discussion to a
substitutes; one example, of petroleum and pe-
demand curve for a particular market.
troleum lamps, itself ambiguously worded, suggests
The other list is in Marshall's discussion in the
that it may refer to complements as well.
Principles of the difficulties of the statistical study
of demand (Book III, chap. iv), where he writes: 39In the quotations from Book III, chap. iv,
"Other things seldom are equal in fact over periods in the preceding footnote, "wealth" is mentioned
of time sufficiently long for the collection of full explicitly, though separately from "general pros-
and trustworthy statistics.... To begin with, [a] perity" and "total purchasing power." See also the
the purchasing power of money is continually quotations in the fourth and fifth paragraphs of
changing.... Next come the changes in [b] the n. 37. Marshall repeatedly refers to "rich" and
general prosperity and in the total purchasing power "poor" rather than to high- and low-income people
at the disposal of the community at large.... (e.g., pp. I9, 95, 98). However, in an illustrative
Next come the changes due to [c] the gradual growth case, a rich man and a poor man are identified by
of population and wealth.... Next, allowance their annual incomes (p. i9). And in Book III,
must be made for changes in [d] fashion, and taste chap. vi, he remarks: "We have throughout this
and habit, for [e] the opening out of new uses of a and preceding chapters spoken of the rich, the mid-
commodity, for [f] the discovery or improvement or dle classes and the poor as having respectively large,
cheapening of other things that can be applied to medium and small incomes-not possessions"
the same uses with it" (Principles, pp. io9-Io; (p. 134).

Item 3 requires no discussion, since the words in that quotation are the only
only reasonable interpretation of it is that words on the page even remotely sup-
it refers to tastes and preferences.40 porting the substitution of "any other"
The important defect of the current for rival. The specific examples that fol-
interpretation is its treatment of item 4, low the quotation-tea and coffee, gas
which is, in turn, responsible for the neg- and electric lighting, differentvarieties of
lect of items i and 5. "Rival commodity" tea, beef and mutton-make it clear that
is replaced by, or read to mean, "any Marshallwas using the word "rival"in a
other commodity," and hence item 4 is narrowsense and not in that broad sense
taken to mean that the price of every in which it may be said that all commod-
other commodity is to be supposed the ities are "rivals" for the consumer'sin-
same. For example, Henry Schultz says, come.43Whatever the merits of the cur-
as if it were obvious and without citing rent interpretation, it cannot be found
any statements of Marshall: "Marshall explicitly in Marshall.
also assumes, in giving definite form to The interpretation of item 4 as refer-
the law of demand for any one commod- ring to all other commodities makes item
ity, that the prices of all other commod- 5 unnecessary and contradicts item i.
ities remain constant."4' Numerous other Item 5 is unnecessary because the intro-
statements to the same effect could be duction of a new commodity is equiv-
cited. It is an amusing commentary on alent to a decline in its price fromtinfin-
our capacity for self-delusion that the ity to a finite amount; hence is ruled out
only references to Marshall for support if the price of every other commodity is
that I have seen are to the page contain- to be unchanged. Item i is contradicted
ing the third quotation in Section IVa because, if all other prices are unchanged,
above-the source of the words quoted the purchasing power of money will be
in item 4.42 The first set of italicized lower, the higher the price of the com-
40See n. 38, above. In discussing the law of modity in question. The purchasing
diminishing marginal utility, Marshall says: "We power of money cannot, therefore, be the
do not suppose time to be allowed for any alteration same for all points on the demand curve.
in the character or tastes of the man himself"
(p. 94). The redundancy of item 5 on this in-
4' Op. cit., p. 53. Immediately after making this terpretation of item 4 is unimportant;
statement he quotes from Edgeworth's article on this item is in a list that is illustrative
"Demand Curves" cited in n. 5, above, not as rather than exhaustive, and there is no
evidence for the validity of his interpretation of
Marshall but rather as an indication of the diffi- reason why Marshall should have scru-
culties that it raises.
curve] was not intended to be . . . interpreted" as
42 Joan Robinson states without citation: "Mar-
"showing the effect of compensated variations in
shall instructs us to draw up a demand schedule on price." Similar statements, all citing p. ioo of
the assumption that the prices of all other things the Principles as authority, are made by Rob-
are fixed" (The Economics of Imperfect Competition
ert Triffin, Monopolistic Competition and General
[London: Macmillan & Co., Ltd., I9341, p. 20).
Equilibrium Theory (Cambridge: Harvard Uni-
Paul Samuelson says, also without citation: "All versity Press, I940), p. 44; Ruby Turner Norris,
other prices and income are held constant by ceteris The Theory of Consumer's Demand (New Haven:
paribus assumptions" in the "Marshallian partial Yale University Press, I94I), p. 82; and Weintraub,
equilibrium demand functions" (Foundations of op. cit., p. 539.
Economic Analysis [Cambridge: Harvard University
Press, I947], p. 97). In an unpublished exposition of 43 If any doubt remains, it is removed by the
income and substitution effects prepared for class use footnote in the first edition attached to the word
about I939, I stated, also without citation: "There "rival" referring to Jevons' phrase "commodities
is no question but that it [the Marshallian demand that are nearly 'equivalent"' (see n. 37, above).

484 MILTON FRIEDMAN
pulously avoided overlapping. The log- rationalization, item i becomes redun-

ical inconsistencybetween items i and 4 dant, but, in the limit, not logically in-
cannot, however,be dismissedso lightly. consistent with an item 4 taken to refer
Retention of the current interpretation to all other commodities.
requireseither that item i be eliminated, I do not believe that Marshall in-
on the grounds that the quotations on tended to restrict the use of the demand
which it is based are exceptional and curve to commodities accounting for
peripheral,or that Marshallbe convicted only a small fraction of total expenditure.
of logicalinconsistencyon a fundamental He speaks of a demand curve for wheat
point in his theory of demand.44Item I (p. io6), for houseroom (p. I07), and for
cannot, I think, be eliminated. The con- other commodities that he cannot have
stancy of the purchasingpower of money regarded as unimportant. He first explic-
is clearly fundamental in Marshall's itly introduces the restriction to unim-
thought, probably more fundamental portant commodities in connection with
than any other item on our list.45 his discussion of consumer's surplus,
One excuse for retaining the current which comes well after the initial d is-
interpretation of Marshall, despite the cussion of the demand curve in the
logical inconsistency that it introduces, eighth edition, three chapters later; and
is to suppose that Marshall intended to the restriction is repeated at most points
restrict the use of his demand curve to at which the argument depends on it. At
commodities that account for only a one point the restriction is said to be
small fraction of total expenditures. A "generally," not universally, justifiable.
change in the price of such a commodity This evidence may not be conclusive
would have only a small effect on the but it certainly establishes a strong pre-
purchasing power of money, and it sumption that Marshall did not intend
could be argued that Marshall neglected the restriction to carry over to all uses of
it as a "second-ordereffect." On this the demand curve.46
44 The extent to which the current interpretation
46 In connection with the discussion of consumer's
dominates economic thought could not be more
surplus and the assumption of a constant marginal
strikingly illustrated than by the fact that so
utility of money implicit in that discussion, Marshall
acute an economic theorist as J. R. Hicks can write:
says: "The assumption . . . underlies our whole
"No doubt it [the constancy of the marginal utility
reasoning, that the expenditure on any one thing
of money] was . . . associated in his [Marshall's]
. . . is only a small part of his whole expenditure"
mind with the assumption of a constant value of
(p. 842). The first sentence of the paragraph from
money (constant prices of other consumers' goods
which this quotation is taken explicitly limits it to
than the one, or sometimes ones, in question)"
"the discussion of consumers' surplus" (p. 842).
("The Rehabilitation of Consumers' Surplus,"
The quotation is followed by a cross-reference to the
p. io9). Hicks here treats constancy of all other
part of Marshall's famous analysis of the process by
prices as an alternative statement of item I, when,
which equilibrium is reached in a corn-market in
in fact, it is logically inconsistent with item i.
which he discusses "the latent assumption, that
45 See nn. 37 and 38, above. Note also that con- the dealers' willingness to spend money is nearly
stancy of the purchasing power of money was a constant throughout" (p. 334). "This assumption,"
standard assumption of economic theory long he says, "is justifiable with regard to most of the
before Marshall's day. It was made by Ricardo market dealings with which we are practically con-
in his price theory, and Marshall refers to Cournot's cerned. When a person buys anything for his own
discussion of the reasons for making this assump- consumption, he generally spends on it a small
tion (see Marshall, Principles, pp. ix, 62; Augustin part of his total resources" (p. 335).
Cournot, Researchesinto the Mathematical Principles Nowhere in Book III, chap. iii, does Marshall
of the Theory of Wealth [1838], Nathaniel Bacon explicitly restrict his discussion to unimportant
translation [New XYork: Macmillan Co., i897], commodities. The one statement in that chapter
p. 26). that might be regarded as so restricting the dis-

It should be noted that Marshall's ex- in Section IVa, which reads, in part:
plicit introduction of the restriction to "The larger the amount of a thing that a
unimportant commodities has no bearing person has the less . . . will be the price
on the relative validity of the two inter- which he will pay for a little more of it."
pretations of his demand curve. The re- This is a curious form of phrasing on the
striction is necessary on either of the two current interpretation. Why emphasize
interpretations at each point at which the amount of a thing that a person has
Marshall explicitly makes it. So the re- and the marginal expenditure that he
striction cannot be regarded as called for can be induced to make rather than the
by the inconsistency of items i and 4 on amount he purchases and the average
the current interpretation of 4. price he pays? On my interpretation, this
2. The alternative interpretation.-My phrasing follows directly from the argu-
interpretation of the Marshallian de- ment of Section Ic above (and Note II of
mand curve resolves almost all the diffi- Marshall's Mathematical Appendix), ac-
culties that plague the current interpre- cording to which a demand curve con-
tation, since it accepts at face value the structed on my interpretation can be
five "other things" listed at the begin- viewed as showing the maximum price
ning of Section IVb. Marshall's words per unit that a person can be induced to
can be taken to mean what they say pay for successive increments of the com-
without uncomfortable stretching, and modity.
there is no logical inconsistency in the One minor puzzle remains on my inter-
constancy of both item i, the purchasing pretation. Why does Marshall restrict
power of money, and item 4, the prices of his attention to "rival" commodities?
rival commodities. Item 5, the range of Why not to "closely related" commod-
rival commodities available, is still re- ities, whether rivals or complements? His
dundant, since, if "rival" has the same use of the word "rivals" in discussing the
meaning in 4 and 5, the invention of a demand curve is apparently not a mere
new rival commodity means a change in verbal accident. He uses the word re-
its price from infinity to a finite value. peatedly; almost all his examples deal
My interpretation explains also the with the effect of, or through, substi-
precise wording of the second quotation tutes. I have no very good answer to this
puzzle; the only one that seems at all per-
cussion is the statement on p. 95 that "the marginal suasive is that he thought the concept of
utility of money to him is a fixed quantity." But "joint demand" and the associated an-
the context argues and Note II in the Mathe-
matical Appendix demonstrates that this is merely a
alytical apparatus better suited to prob-
verbal statement of an identity (if income is un- lems involving complementary goods.47
changed, so is marginal utility of money), and thus is My interpretation follows so directly
not really relevant to the issue. In the eighth edi-
tion, Note II is referred to only at the end of the 47 In Note VII of the Mathematical Appendix,
subsection following the paragraph containing the Marshall qualifies a suggested formula for combin-
passage quoted. However, in the first edition, the ing consumer's surplus from different commodities
corresponding note (Note III) is referred to at the by saying: "if we could find a plan for grouping
end of the paragraph containing the passage quoted, together in one common demand curve all those
and hence clearly covers it (pp. I55-56, 737-38). things which satisfy the same wants, and are rivals;
The above quotations are essentially unchanged and also for every group of things of which the
from the first edition on. The restriction to unim- services are complementary (see Book V, chap. vi)
portant commodities is, however, mentioned neither . . ." (p. 842). Book V, chap. vi, contains the dis-
in Marshall and Marshall, Economics of Industry, cussion of joint demand. The qualification quoted
nor in the Pure Theory. appears first in the third edition.

486 MILTON FRIEDMAN
from Marshall's words that further de- affected by the rise in the price of a par-
fense of it would be unnecessarywere it ticular commodity.48
not for the unquestioned dominance of
the current interpretation in the eco- d) THE EVIDENCE OF THE MATHEMATICAL
APPENDIX
nomic thinking and writing of the past
half-century.This circumstanceexplains The Mathematical Appendix to the
the presentation of additional textual Principles confirmsand extends the evi-
evidence bearing on the validity of the dence already presented from the text of
alternative interpretation. the Principles and from Marshall'sother
writings. Note II (III in the first edition)
C) COUNTEREVIDENCE FROM THE TEXT OF explicitly derives a relation between
THE "PRINCIPLES" price and quantity demanded that is
I have been able to find only one pas- identical with a demand curve on my in-
sage in the text of the eighth edition of terpretation,in which real income is kept
the Principles that is in any way incon- constant by compensating variations in
sistent with my interpretation of Mar- money income. Indeed, my derivation of
shall. This is the celebratedpassage, ad- such a demand curve in Section Ic above
verted to above, which deals with the so- is a verbal paraphrase of Marshall's
called "Giffen phenomenon"and which mathematics. Marshall does not explic-
was first introducedin the third edition: itly say that the relation he derives is a
demandcurve, but Note II is attached to
For instance, as Sir R. Giffen has pointed his initial discussionof the demand curve
out, a rise in the price of bread makes so large a
drain on the resources of the poorer laboring
(Book III, chap. iii, in the eighth edition)
families and raises so much the marginal utility and is given as the authority for state-
of money to them, that they are forced to cur- ments made about the demand curve;
tail their consumption of meat and the more hence there can be no doubt that it pre-
expensive farinaceous foods: and bread being sents the pure theory of his demand
still the cheapest food which they can get and
will take, they consume more, and not less of
curve.
it [p. I32; italics added]. In all editions of the Principles Note
VI, attached to Marshall's discussion of
This passage clearly offsets an income consumer'ssurplus, contains a sentence
effect against a substitution effect, that is definitely wrong on the current
whereas, on my interpretation of Mar- interpretation of his demand curve but
shall, real income is the same at all correcton my interpretation.
points on the demand curve, so there is Finally, a sentence added to Note VI
no "incomeeffect" (see Sec. JIb, above). in the third edition, referred to in the
The passage is thus in the spirit of the text of the Principles in connection with
current interpretation. Yet the words I the material added on the Giffen phe-
have italicized indicate that it does not nomenon, contains an implicit math-
necessarily contradict my interpretation ematical proposition that is correct on
of Marshall. The purchasing power of the current interpretation but incorrect
money and the real income of the com- on my interpretation.The mathematical
munity at large may remain constant; point in question is considerably more
yet the real income of a particulargroup
Marshall's explicit discussion of, and em-
in the community that has a special con- phasisSee
48
on, this possibility in "Remedies for Fluctua-

suullption pattern may be adversely tions of GeneralPrices" (i887), Memorials,p. 207.

subtle than those referredto in the two tation not only is consistent with both
preceding paragraphs, so it cannot be the letter and the spirit of the entire text
given the sameweight. of the first edition of the Principlesbut is
These two notes are examinedin some almost conclusivelyestablishedby the ev-
detail in the appendix to this paper, to idence cited above from two notes in the
which the reader is referredfor proof of Mathematical Appendix of the first edi-
the above statements. tion. In his determinedeffort to be per-
suasive and to make his work accessible
e) A SYNTHESIS OF THE EVIDENCE to educated laymen, Marshallmight well
There are two differencesbetween the have been vague in his verbal presenta-
current interpretation of Marshall's de- tion, though even there it seems unlikely
mand curve and my interpretation: (i) that he would have been logically incon-
On the currentinterpretation,account is sistent. It is hardly credible that he
taken of the price of each other commod- would have been not merely vague but
ity individually; on my interpretation, downright wrong on simple mathemnat-
only of the average price of all commod- ical points stated in mathematical lan-
ities other than the one in question and guage, especially since the mathematical
its close rivals. (2) On the current inter- points in question could hardly even
pretation, real income varies along the have arisenif he had been explicitly using
demand curve with the price of the good the currentinterpretationof the demand
in question; on my interpretation, real curve.
income is constant along the demand I am inclined to believe, however, that
curve. by the time Marshallmade the revisions
On the first, and less important,point, incorporatedin the third edition of the
it is mathematically convenient to con- Principles-presumably between i89i,
sider each other price separately, and when the second edition appeared, and
this procedure might well have recom- i895, when the third edition appeared-
mendeditself to the writer of mathemat- he had himself been influenced by the
ical Notes XIV and XXI. On the other currentinterpretation,probably without
hand, it is impossible to consider each realizing that it was different from his
priceseparatelyin a practicalanalysis;so own. This conjecture is based primarily
the use of an averageprice would clearly on the two passages cited above as in-
have recommendeditself to the writer of consistent with my interpretation: the
the text of the Principles and is entirely passage dealingwith the Giffen phenom-
in the spirit of Marshall'sexplicit meth- enon and the last sentence of Note VI of
odological statements (see Sec. Ha, the Mathematical Appendix. Both were
above). Marshall does not discuss this added in the third edition--and these
point explicitly; hence the textual ev- are the only passages I have been able
idenceis all indirect. to find in any edition of the Prin-
On the secondand basic point of differ- ciples that fit the current interpretation
ence the evidence leaves little room for better than they fit my interpretation.
doubt: Marshall's theory of demand, in Further,both show someevidence of con-
the form in which it is presented in the fusion about the fine points of his theory
first edition of the Principles, is explic- of demand (see last paragraphof appen-
itly based on constancy of real income dix to this paper).
along the demand curve. This interpre- The hypothesis that Marshall did not

488 MILTON FRIEDMAN
recognizethe contradiction between the Marshall himself writes: "My main
current interpretation and his earlier position as to the theory of value and dis-
workwould hardly be tenable if the lapse tribution was practically completed in
of time between the work incorporated the years i867 to i870.... By this time
in the first and the third editions of the [from the context, i8741 I had practically
Principleswere as short as between their completed the whole of the substance of
publication. But, as already noted, this my Mathematical Appendix.s52 Thus
is not the case. The essence of both his Marshall appears to have completed his
theory of demand and his analysis of fundamental work on the theory of de-
consumer's surplus is contained in the mand in the early i870's and to have
Pure Theory of Domestic Values, which, made no important substantive changes
though not printed until I879, "must thereafter. The third edition appeared
have been substantially complete about some twenty or more years later-an
I873."f49 The one important point in the ample lapse of time for the precise details
theory of demand that is not in the Pure of an essentially mathematical analysis
Theory-explicit mention of constancy to have become vague and their differ-
of the purchasingpower of money-is in ence from a superficiallysimilarset of de-
the 1879 Economics of Industry. The only tails to pass unnoticed. This seems es-
important addition in the Principles is pecially plausible in view of the accept-
the concept of "elasticity of demand"; ance of the current interpretation by
and even this concept, which is not rel- others and the absence of controversy
evant to the present problem, was about it.
worked out in I88I-82.50 No important
substantive changes were made in the b, b2,b3 . . . are necessary for existence, if y =f(x)
theory of demand in successive editions y = f2(x), y = f3(x) . . . be the equations to their
demand curves. . . , then the total utility of his
of the Principles, though the exposition wealth, subsistence being taken for granted, is
was amplifiedand rearranged,the word- represented by
ing changedin detail, and some examples E t f(x)dx"
modified.The only important change of
substance introducedinto the discussion (ist ed., p. 741).
of consumer'ssurplus (in the third edi- The eighth edition does not contain the first
tion) was in connectionwith a point that statement. Instead, the text contains an explicit
warning against adding consumer's surpluses for
has no bearingon the present issues' different commodities, and a footnote says: "Some
49Keynes, Memorials, p. 23. ambiguous phrases in earlier editions appear to have
suggested to some readers the opposite opinion"
SOIbid., P.39, n. 3.
(p. IAI). Note VII in the Mathematical Appendix
5I This change does not reflect favorably on Mar- was modified by replacing "his wealth" by "in-
shall's willingness to admit error. The first edition come" and, of more importance, "is represented"
states: "Subject to these corrections then we may by "might be represented" and by adding after the
regard the aggregate of the money measures of the formula the significant qualification, "if we could
total utility of wealth as a fair measure of that find a plan for grouping together in one common
part of the happiness which is dependent on wealth" demand curve all those things which satisfy the
(pp. I79-80), the corrections referred to being for same wants, and are rivals; and also for every group
"differences in the wealth of different purchasers" of things of which the services are complementary.
(p. I78) and "elements of collective wealth which . . . But we cannot do this; and therefore the formula
are apt to be overlooked" (p. I79). A footnote to remains a mere general expression, having no prac-
the first quotation refers to mathematical Note tical application" (p. 842). As noted, these changes
VII, in which he says, subject to the same two date from the third edition.
qualifications: "if a,, a2, a3 . . . be the amounts
consumed of the several commodities of which 52 Letter to J. B. Clark, Memorials, p. 4i6.

Further circumstantial evidence that V. ALTERNATIVE CONCEPTIONS OF

Marshall did not recognize the contra- ECONOMIC THEORY
diction between the current interpreta- There remains the mystery how the
tion and his earlier work is provided by current interpretation of Marshall's de-
the apparent absence of any explicit dis- mand curve gained such unquestioned
cussion of the question in the writings of dominance at so early a date and re-
either Marshall or the more prominent of tained it so long, not only as an interpre-
his students or even of any comments tation of Marshall, but also as "the" defi-
that could reasonably be interpreted as nition of "the" demand curve.
implying recognition of the existence of One obvious explanation is that math-
alternative interpretations of the de- ematical economists were more likely
mand curve. Yet, as noted earlier (n. 5), than others to state explicitly and pre-
the current interpretation is explicitly cisely their assumptions about the be-
given by Edgeworth as early as i894 in havior of other prices; that mathemat-
an article on "Demand Curves" in Pal- ical economists were likely to be familiar
grave's Dictionary of Political Economy with Walras' independent definition and
that Marshall must be presumed to have to take it as a point of departure; and
read. Though the assumption of constant that, in any event, the current interpre-
prices of commodities other than the one tation is mathematically more conven-
in question is not explicitly attributed to ient. Other economists, it could be
Marshall, most of the article is based on argued, followed the lead of the math-
Marshall; and there is no suggestion that ematical economists, and thus the cur-
this assumption does not apply to Mar- rent interpretation was taken for granted
shall's demand curve. Further, Walras' and accepted without question.
definition of the demand curve, which This explanation seems to me a signifi-
presumably influenced Edgeworth, is cant part of the answer; however, I do
not believe that it is the entire answer.
identical with the current interpretation
If, as I have argued above, my interpre-
of Marshall's demand curve, and Mar-
tation of Marshall is more useful for
shall refers to Walras several times in the
most practical problems, why has its use
first edition of the Principles, though it been so rarely proposed; why has there
seems clear that Marshall developed his been no general feeling of dissatisfaction
theory of demand independently of Wal- with the current interpretation? There
ras.53 So Marshall must have been ex- must, it would seem, be something about
posed to a definition of the demand the role that has been assigned to eco-
curve corresponding to the current inter- nomic theory that has made the current
pretation at a time when he was still interpretation acceptable.
making substantial revisions in the Prin- I amninclined to believe that this is, in
ciples. If he had recognized that this in- fact, the case; that, by slow and gradual
terpretation was incorrect, would he not steps, the role assigned to economic the-
have taken the opportunity to clarify his ory has altered in the course of tine until
statements in later editions? today we assign a substantially different
role to theory than Marshall did. We
53Principles (ist ed.), pp. xi, xii, 425; Keynes,
Memorials, pp. 19-24; Marshall's letter to J. B. curtsy to Marshall, but we walk with
Clark, ibid., pp. 4i6-i8. Walras.

490 MILTON FRIEDMAN
The distinction commonly drawn be- Facts by themselves are silent.... The most
tween Marshall and Walras is that Mar- reckless and treacherous of all theorists is he
who professes to let facts and figures speak for
shall dealt with "partial equilibrium," themselves, who keeps in the background the
Walras with "general equilibrium." This part he has played, perhaps unconsciously, in
distinction is, I believe, false and unim- selecting and grouping them, and in suggesting
portant. Marshall and Walras alike dealt the argument post hoc ergo propter hoc....
with general equilibrium; partial equilib- The economist . .. must be suspicious of any
direct light that the past is said to throw on
rium analysis as usually conceived is but
problems of the present. He must stand fast
a special kind of general equilibrium by the more laborious plan of interrogating
analysis-unless, indeed, partial equilib- facts in order to learn the manner of action of
rium analysis is taken to mean erroneous causes singly and in combination, applying this
general equilibrium analysis. Marshall knowledge to build up the organon of economic
wrote to J. B. Clark in i908: "My whole theory, and then making use of the aid of the
organon in dealing with the economic side of
life has been and will be given to present- social problems.s6
ing in realistic form as much as I can of
my Note XXI."54 Note XXI, essentially Economic theory, in this view, has two
unchanged from the first edition of the intermingled roles: to provide "system-
Principles to the last, presents a system atic and organized methods of reasoning"
of equations of general equilibrium. It about economic problems; to provide a
ends with the sentence: "Thus, however body of substantive hypotheses, based
complex the problem may become, we on factual evidence, about the "manner
can see that it is theoretically determi- of action of causes." In both roles the
nate, because the number of unknowns is test of the theory is its value in explain-
always exactly equal to the number of ing facts, in predicting the consequences
equations which we obtain."55 The ex- of changes in the economic environment.
planation given above why Marshall Abstractness, generality, mathematical
might have decided to hold the purchas- elegance-these are all secondary, them-
ing power of money constant was entire- selves to be judged by the test of applica-
ly in terms of constructing the demand tion. The counting of equations and un-
curve so that it would be consistent with knowns is a check on the completeness of
general equilibrium in those parts of the reasoning, the beginning of analysis, not
system not under direct study. an end in itself.
The important distinction between the Doubtless, most modern economic the-
conceptions of economic theory implicit orists would accept these general state-
in Marshall and Walras lies in the pur- ments of the objectives of economic the-
pose for which the theory is constructed ory. But our work belies our professions.
and used. To Marshall-to repeat an ex- Abstractness, generality, and mathemat-
pression quoted earlier-economic the- ical elegance have in some measure be-
ory is "an engine for the discovery of come ends in themselves, criteria by
concrete truth." The "economic orga- which to judge economic theory. Facts
non" introduces "systematic and organ- are to be described, not explained. The-
ized methods of reasoning." Marshall ory is to be tested by the accuracy of its
wrote: "assumptions" as photographic descrip-
54 Memorials,p. 417.
56The quotations are all taken from Marshall,
55Principles, p. 856. This note was numbered "The Present Position of Economics" (i885),
XX in the first edition. memorials, pp. 159, i6i, i64., 66, 068, 171.

THE MARSHALLIANDEMAND CURVE 49I
tions of reality, not by the correctnessof Keynes's theory of employment is Wal-

the predictionsthat can be derived from rasian.58
it. From this viewpoint the current in- VI. CONCLUSION
terpretation of the- demand curve is
Modern economic theory typically de-
clearly the better. It is more general and
fines the demandcurve as showingthe re-
elegant to includethe price of every com-
lation between the quantity of a com-
nodity in the universe in the demand
modity demandedand its price for given
function rather than the average price of
tastes, money income, and prices of other
a residual group. Any price may affect
commodities. This definition has also
any other, so a demand equation in-
been uniformly accepted as a correct in-
cluding every price is a more accurate
terpretation of the demand curve de-
photographic description. Of course, it
fined and used by Alfred Marshall in his
cannot be used in discovering "concrete
Principles of Economics.Rarely has the
truth"; it contains no empiricalgeneral-
ization that is capable of being contra- view been expressedthat a differentdefi-
nition would be preferable.
dicted-but these are Marshallianobjec-
Despite its unquestioned acceptance
tions. From the "Walrasian"viewpoint,
for over half a century, this interpreta-
to take one other example from recent
tion of Marshall is, in my view, wrong.
developments in economic theory, it is
Marshall'searly writings, the text of the
a gain to eliminate the concept of an "in-
Principles, and, even more definitely, the
dustry," to take the individual firm as
Mathematical Appendix provide almost
the unit of analysis, to treat each firm as
a monopoly, to confine all analysis to conclusiveproof that Marshall'sdemand
either the economics of the individual curve differsin two respectsfrom the one
firm or to a general equilibriumanalysis commonly used and attributed to him:
of the economy as a whole.37From the first, commodities other than the one in
question and its close rivals are treated
Marshallianviewpoint this logical termi-
nus of monopolisticcompetition analysis as a group rather than individually, and
is a blind alley. Its categories are rigid, only their average price is explicitly
taken into account; second, and far
determinednot by the problem at hand
more important, real income is consid-
but by mathematical considerations. It
ered the same at all points on the de-
yields no predictions,summarizesno em-
mand curve, whereas constant money
pirical generalizations,provides no use-
income and other prices imply a higher
ful frameworkof analysis.
real income, the lower the price of the
Of course, it would be an overstate-
commodity in question. Two variants of
ment to characterize all modern eco-
Marshall's demand curve can be distin-
nomic theory as "Walrasian" in this
guished:one, employed in the text of the
sense. For example, Keynes's theory of
Principles, uses variations in the prices
employment, whatever its merits or de-
of other commodities to compensate for
merits on other grounds, is Marshallian
variations in the price of the commod-
in method. It is a general equilibrium
ity in question and thereby keeps the
theory containing important empirical
purchasingpowerof money constant; the
content and constructed to facilitate
other, employed in the Mathematical
meaningful prediction. On the other
58 0. Lange, Price Flexibility and Employment
hand, much recent work based on (Bloomington, Ind.: Principia Press, I944), is
57See Triffin, op. cit., pp. i88-89. perhapsas good an exampleas any.

492 MILTON FRIEDMAN
Appendix, uses variations in money in- "Every increase in his means diminishes the
come to compensate for variations in the marginal utility of money to him; ....
"Therefore, du/dx, the marginal utility to
price of the commodity in question. him of an amount x of a commodity remaining
The only textual evidence that con- unchanged, an increase in his means . . . in-
flicts with this interpretation is a passage creases dp/dx, that is, the rate at which he is
in the text and a related sentence in the willing to pay for further supplies of it. Treating
Mathematical Appendix that were added u as a variable, that is to say, allowing for pos-
sible variations in the person's liking for the
to the third edition of the Principles. The
commodity in question, we may regard dp/dx
inconsistency of these with the rest of the as a function of m, u, and x. ..."
Principles can be explained by the hy- The wording in the eighth edition is identical
pothesis that Marshall himself was after except that "marginal utility of money" is
a point influenced by the current inter- replaced by "marginal degree of utility of
money" and that "du/dx" and the words
pretation of the demand curve without
"Treating u . . . in question" are omitted from
recognizing its inconsistency with his the last paragraph quoted (pp. 838-39). The
earlier work. Some circumstantial evi- changes were first made in the third edition.
dence also supports this hypothesis. In the second sentence of this note, the word
The alternative interpretation of the "price" is to be interpreted as "total amount"
not as "price per unit." This is clear from the
demand curve not only is faithful to both
context and is demonstrated by the equation
the letter and the spirit of Marshall's that follows and the designation of dp/dx
work but also is more useful for the anal- as "the rate at which he is willing to pay for
ysis of concrete problems than is the de- further supplies of it." The words "just will-
mand curve commonly employed. The ing" in the second sentence and the equa-
tions that follow demonstrate that p is the
acceptance of a less useful definition
maximum amount he can pay for an amount
seems to me to be a consequence of a x and have the same utility as if he had none
changed conception of the role of theory of the commodity. Thus Marshall is describ-
in economic analysis. The current inter- ing a process like that outlined in Section Ic
pretation of the demand curve is Wal- of this paper, whereby the maximum possible
rasian; and so is current economic theory amount is extracted from the individual for
each successive increment of the commodity,
in general. the individual retaining the same "real in-
come," that is, remaining on the same in-
APPENDIX ON TWO NOTES IN THE
difference curve, throughout the process.
MATHEMATICAL APPENDIX TO The last sentence quoted shows that u is to
THE PRINCIPLES be regarded as a parameter to allow for changes
I. NOTE II OF THE EIGHTH EDITION in tastes. The rest of that sentence simply
describes a function like that obtained by elimi-
This note is numbered III in the first edition nating y' from equations (5) and (6) of note 6
of the Principles, II in the rest. In the first of this paper. The parameter m in Marshall's
edition the relevant parts are worded as follows function takes the place of U0 in our footnote,
(pp. 737-38): since dp/dx is still to be regarded as the price
"If m is the amount of money or general per unit paid for an additional increment of the
purchasing power at a person's disposal at any commodity rather than as the price per unit at
time, and ,u represents its total utility to him, which any amount can be purchased. In conse-
then dyu/dm represents the marginal utility quence, no explicit statement is needed as yet
of money to him. about the compensating variations in income
"If p is the price which he is just willing to that are implicit in Marshall's analysis.
pay for an amount x of the commodity which The word "demand" does not appear in this
gives him a total pleasure u, then note. But the note is attached to the chapter
in the Principles in which Marshall first intro-
dLp and dm d =-du
dm =Au;
' dim dx orx duces the demand curve (Book III, chap. ii, in

the first edition; Book III, chap. iii, in later In its final form Note VI seems internally
editions) and is cited as proof of statements inconsistent: the second sentence is wrong on
about the demand curve; hence there can be the current interpretation of Marshall's de-
no doubt that the "function" mentioned in the mand curve, correct on my interpretation; the
last sentence quoted is the counterpart of final sentence, added in the third edition, seems
Marshall's demand curve. correct on the current interpretation, wrong on
I have been able to construct no interpre- my interpretation.
tation of this note that would render it con-
sistent with the current interpretation of a) THE SECOND SENTENCE
Marshall's demand curve.
The second sentence is wrong on the current
2. NOTE VI interpretation, which holds money income and
other prices constant along the demand curve,
This note has the same number in all edi- since the ordinate of the demand curve for any
tions. In the first edition the relevant parts are quantity x cannot then exceed money income
worded as follows (p. 740): divided by x, and this is not "indefinitely
"If y be the price at which an amount x of a great" for a fixed value of x-say, x0-whether
commodity can find purchasers in a given xs is greater or less than b. True, f(x) might ap-
market, and y = f(x) be the equation to the proach infinity as x approaches zero, but this
demand-curve, then the total utility of the
commodity is measured by f
ra
f(x) dx, where a
is not what Marshall says; he says it is "indefi-
nitely great, for values of x less than b," i.e.,
is the amount consumed. for any particular value of x less than b
"If however an amount b of the commodity say, xN = o.99b.
is necessary for existence, f(x) will be infinite, On the variant of my interpretation involv-
or at least indefinitely great, for values of x ing compensating variations in money income-
less than b. We must therefore take life for the variant that the note numbered II in the
granted, and estimate separately the total eighth edition leads me to believe Marshall
utility of that part of the supply of the com- used in the Mathematical Appendix-this
modity which is in excess of absolute neces- sentence is entirely valid. As x declines from a
value larger than b, the compensating varia-
saries: it is of course f
1(x)dx ... tion in money income required to keep the
"It should be noted that, in the discussion of individual's real income the same becomes
Consumers'Rent, we assume that the marginal larger and larger, approaching infinity as x
utility of money to the individual purchaser is approaches b, the minimum amount necessary
the same throughout...." for existence. This permits the ordinate of the
Only trivial changes were made in these demand curve likewise to approach infinity as
sentences in subsequent editions: a typographi- x approaches b. On the variant involving com-
cal error in the fifth edition, which remained pensating variations in other prices-the one
uncorrected thereafter, substituted f(z) for Marshall used in the text-the definition of the
f(x) in the second sentence; and "consumers' demand curve breaks down for values of x
surplus" replaced "Consumers' Rent." In the less than b: for a finite price of the commodity
third edition the following sentence was added in question, sufficiently high so that the given
at the end of the note: money income could purchase only a quantity
"If, for any reason it be desirable to take x less than b, there will exist no set of non-
account of the influence which his expenditure negative prices for the remaining commodities
on tea exerts on the value of money to him, it is that will keep the purchasing power of money
only necessary to multiply f(x), within the constant in the sense of enabling the same
integral given above by that function of xf(x) money income to provide the same level of
(i.e. of the amount which he has already spent utility; money income and real income cannot
on tea) which represents the marginal utility both be held constant and at the same time all
to him of money when his stock of it has been prices be kept nonnegative. This sentence can
diminished by that amount" (3d ed., p. 795). therefore be defended as valid on either variant
The only subsequent changes were the addition of my interpretation.
of a comma after "reason" and the deletion One possible ground for dismissing this
of the comma before "within" (8th ed., p. 842). sentence as evidence against the current inter-

494 MILTON FRIEDMAN
pretation is that the so-called "error" on that requires; or that, if at one stage it had been,
interpretation is of my own making, arising Marshall would have failed to see the simple
from a too subtle and too literal reading of the mathematical error implicit in a literal reading
note. Marshall, it could be argued, was using of his words on the current interpretation of the
"demand curve" to mean "utility curve" and demand curve? It seems to me far more credible
f(x) to mean "marginal utility," and therefore that he meant what he said and that the cor-
he did not consider whether the sentence would rectness of what he said on my interpretation
be valid if f(x) were to be interpreted literally of his demand curve is strong evidence for that
as the ordinate of the demand curve. A note interpretation.
that Marshall published in i893 on "Consumer's
Surplus" could be cited as evidence for this con- b) THE FINAL SENTENCE
tention. In this note he quotes part of Note VI The explanation that follows of the final
as follows: " 'If, however, an amount b of the sentence added to Note VI in the third edition,
commodity is necessary for existence, [the though not completely satisfactory, is reason-
utility of the first element] a will be infinite.' "59 ably so, and I have been able to construct no
The bracketed expression that Marshall sub- other even remotely satisfactory explanation.
stituted forf(x) would support the notion that Let U be the utility function of the "indi-
he was using "demand curve" and "utility vidual purchaser" and U_ the marginal utility
curve" interchangeably. of x units of tea to him, i.e., the partial deriva-
I do not myself accept this argument; it tive of U with respect to x. Now the increase in
seems to me to do much less than justice to utility attributable to having a rather than b
Marshall. In the first place, I am inclined to give units of tea-consumer's surplus in utility
little weight to an incidental, explanatory, units-is given by
phrase inserted by Marshall as late as i892 ra
or i893, some twenty years after the funda- 1 Udx (1)
mental analysis incorporated in Note VI had
been completed. I have noted above and shall where the integral is computed for constant
presently cite evidence that Marshall may have quantities of other commodities equal to the
been somewhat confused about the fine points amounts consumed when a units of tea are
of his own theory of demand by the early consumed and other conditions are those corre-
i890's. In the second place, and more important, sponding to the demand curve y = f(x).
Marshall dearly distinguishes in the earlier At every point along the demand curve,
notes in the Mathematical Appendix between
a utility curve and a demand curve, repeatedly U= ny = n (x) f (x) (2)
using the word "utility," and in the first sen-
tence of Note VI says that "the total utility of where n is the marginal utility of money-
itself, of course, a function of x along the de-
the commodity is measuredby Jf(x) dx" (ist mand curve. Integrating both sides of equa-
ed., p. 740; italics added). If he had been using tion (2) gives
f(x) to stand for marginal utility, the words I ra ra
have italicized could have been omitted. f d= a n(x) f (x) dx. (3)
Finally, Note VI, like most of the rest of the
Mathematical Appendix, summarizes a subtle, The left-hand side of equation (3) is sym-
closely reasoned, and by means obvious, mathe- bolically identical with equation (I); yet there
matical argument, in which, so far as I know, is an important difference between them. In
few errors have ever been found. Is it credible equation (I), U, is computed, holding the quanti-
that it would have been worded as loosely and ties of other commodities constant as x varies;
carelessly as the argument being criticized in equation (3), Us is computed, holding con-
stant whatever is held constant along the de-
59 "Consumer's Surplus," Annals of the American
mand curve (money income and other prices
Academy of Political and Social Science, III (March,
i893), 6I8-2I (brackets in original). This note is a
on the current interpretation; real income on my
reply to some comments by Simon Patten. The letter interpretation). In general, quantities of other
a after the brackets which appears in the Annals commodities vary along the demand curve
note does not appear in the Principles, and I can (on either interpretation), and Ux may depend
explain it only as a typographical error. on the quantities of other commodities, so the

U. in equation (3) may be numerically differ- On my interpretation, either money income

ent from the U, in equation (i) for a value of varies along the demand curve, so as to keep
x other than a. This difficulty disappears if Us real income constant, or other prices do; hence
is supposed to be independent of the quantities the preceding argument is no longer valid.
of other commodities-an assumption that That the two forms of statement are no longer
Marshall pretty clearly makes as a general always equivalent can be shown by a counter-
rule (e.g., see Nn. I and II of the Mathematical example. If other prices are held constant and
Appendix). On this assumption, then, the right- compensating variations of income are used to
hand side of equation (3) measures consumer's keep real income constant, U = V+ V'x is a
surplus in utility units. utility function that gives different values of n
It is at this point that difficulties of inter- for different values of x yielding the same
pretation arise; for the right-hand side of value of xf(x). If money income is held constant
equation (3) is obtained by multiplying "f(x) and compensating variations of other prices
within the integral given above by that function are used to keep real income constant, U =
of" x "which representsthe marginal utility ...
3+ x - X2+ x/x' is such a utility func-
of money." Why does Marshall say "that function. Hence Marshall's use of xf(x) instead of
tion of xf(x)" rather than of x alone? And is x is invalid on either variant of the alternative
it valid to make this substitution? One can interpretation.
argue that to each value of x there corre- This explanation leaves a number of Mar-
sponds a value of f(x) and hence of xf(x), so shall's verbal statements wrong or ambiguous,
that the two forms of statement are equivalent: whichever interpretation of the demand curve
Marshall has simply made the transformation is accepted. (I) The parenthetical explanation
z = xf(x) and converted n(x) into n(z). This of the meaning of xf(x) seems wrong-why
argument is not, however, rigorous. In general, the word "already"? If one is thinking of going
x will not be a single-valued function of z;
through the process of extracting as much as pos-
hence to any given value of z there may corre-
sible from the consumer for each successive unit
spond more than one value of x and hence more of tea and is supposing the maximum price that
than one value of n. The two forms of statement
he will pay for successive units to be given by
are equivalent if and only if n is a single-valued
function of z, i.e., if n(x) is the same for all the demand curve, then f f(x)dx and not xf(x)
values of x for which xf(x) is the same. is the amount he has "already spent on tea."
Given independence between the marginal If one is thinking of the amount spent on tea
utility of tea and the quantity of other com- at a given price for tea, then xf (x) is the amount
modities, this condition is always satisfied on spent when the price is f(x), not the amount
the current interpretation of the demand curve "already spent." The explanation offered above
but not on the alternative interpretation. Let x' accepts the latter rendering of the parenthesis,
stand for the quantity of a composite com-
i.e., supposes the word "already" omitted. (2)
modity representing all commodities other
The last clause-"when his stock of money has
than tea, y' for its price, and Us', for its marginal
utility. At each point on the demand curve, been diminished by that amount"-is am-
biguous. To make it consistent with the ex-
Uz Uz
-= ,= n. planation offered above, one must add "and
Y tea is unavailable, so that the balance is spent
On the current interpretation of the demand solely on other commodities at the prices as-
curve, money income and the prices of other sumed in drawing the demand curve for tea."
commodities are the same for all points on the The referenceto "stock of money" suggests that
demand curve. It follows that, for all values of Marshall was supposing money income con-
x that yield the same value of xf(x), the same stant and so, independently of the rest of the
amount will be spent on other commodities; quotation, would tend to rule out compensating
so x' is the same (since y' is by definition); so variations in money income. It should be
Ux' is (since, on the assumption of inde- noted that there are no such ambiguities in the
pendence, Uxe depends only on x'); and so n is. original version of Note VI, either in the parts
Marshall's use of xf (x) instead of x is thus valid quoted above or in the parts not quoted.
on the current interpretation of the demand
curve. OFCHICAGO
UNIVERSITY


Political: Jour - Nal

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Political: Jour - Nal

Uploaded by

Copyright:

Available Formats

OF

I-liatem Curte Mm4-&n ma 463

~ :?ure TheVr of Gove t Fitugge

and a Nad SAhtBt e Rnd, -. K 506

Bys hedved t;55

THE UNIVERSITY OF:CHICAGOPRESS

C .na ; .. ... @ >

.,, _E r = ' $.~~~~~~~~~~~

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

THIE MARSHALLIAN DEMAND CURVE

ALFRED MARSHALL'S theory of de- lowed to vary, is nowhere given explicit-

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

one demand curve will be obtained by b) AN ALTERNATIVE INTERPRETATION

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

given fixed values and x' is eliminated from eqs. (2) =

and (3). This gives a pair of equations,

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

C) COMPARISON OF THE INTERPRETATIONS

The relation between demand curves

resents a demand curve of an individual

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

The possibility of interpreting Mar-

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

them or to considerthem as a group. The ing primarilythe commodity in question.

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

An examplewill show how this occurs. seen most easily by supposing DD to

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

discussion of consumer's surplus cannot, I87934 but, according to Keynes, "sub-

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

b) THE BEARING OF THESE PASSAGES ON 3. "Custom"

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

pulously avoided overlapping. The log- rationalization, item i becomes redun-

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

on, this possibility in "Remedies for Fluctua-

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

Further circumstantial evidence that V. ALTERNATIVE CONCEPTIONS OF

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

tions of reality, not by the correctnessof Keynes's theory of employment is Wal-

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

U. in equation (3) may be numerically differ- On my interpretation, either money income

This content downloaded from 141.218.001.105 on August 02, 2016 10:28:37 AM

You might also like