Coalitions
Demange, G., and R. Guesnerie. 2001. On coalitional stability
of anonymous interim mechanisms. E c o n o m i c
T h e o r y 18: 367-389.
Diamantoudi, E., and L. Xue. 2003. Farsighted stability in
hedonic games. S o c i a l C h o i c e a n d W e l f a r e 21:
39-61.
Dreze, J., M. Le Breton, and S. Weber. 2006. Rawlsian pricing
of access to public facilities: A unidimensional illustration.
J o u r n a l o f E c o n o m i c T h e o r y 136: 759-766.
Dubey, P., and A. Neyman. 1984. Payoffs in nonatomic
economies: An axiomatic approach. E c o n o m e t r i c a
52: 1129-1150.
Durlauf, S. 2002. The memberships theoiy of poverty: The
role of group affiliations in determining socioeconomic
outcomes. In U n d e r s t a n d i n g Poverty in
A m e r i c a , ed. S. Danziger and R. Haveman.
Cambridge, MA: Harvard University Press.
Dutta, B., A. van den Nouweland, and S. Tijs. 1998. Link
formation in cooperative situations. I n t e r n a t i o n a l
J o u r n a l o f G a m e T h e o r y 27: 245-256.
Dutta, B., and R. Vohra. 2005. Incomplete information,
credibility and the core. M a t h e m a t i c a l S o c i a l
S c i e n c e s 50: 148-165.
Echenique, F., and J. Oviedo. 2004. Core many-to-one
matchings by fixed-point methods. J o u r n a l o f E c o -
n o m i c T h e o r y 115: 358-376.
Economides, N.S. 1986. Non-cooperative
e q u i l i b r i u m c o a l i t i o n s t r u c t u r e s , University
Discussion Paper No. 273, Columbia University.
Einy, E., D. Moreno, and D. Monderer. 1999. On the least core
and the Mas-Colell bargaining set G a m e s a n d
E c o n o m i c B e h a v i o r 28: 181-188.
Einy, E., D. Moreno, and B. Shitovitz. 2000. On the core of an
economy with differential information. Journal o f
E c o n o m i c T h e o r y 94: 262-270.
Ellickson, B., B. Grodal, S. Scotchmer, and W. Zame. 2001.
Clubs and the market: Large finite economies. J o u r n a l
o f E c o n o m i c T h e o r y 101: 40-77.
Engelbrecht-Wiggans, R., andD. Granot. 1985. On market
prices in linear production games. Mathematical Pro-
gramming 32: 366-370.
Epstein, L., and M. Marinacci. 2001. The core of large
differentiable TU games. J o u r n a l o f E c o n o m i c
T h e o i y 100: 235-273.
Farrell, J., and S. Scotchmer. 1988. Partnerships. Quarterly
Journal of Economics 103: 279-297.
Forges, F., J.-F. Mertens, and R. Vohra. 2002a. The ex ante
incentive compatible core in the absence of wealth effects
.Econometrica 70: 1865-1892.
Forges, F., E. Minelli, and R. Vohra. 2002b. Incentives and the
core of an exchange economy: A survey. Journal of
Mathematical Economics 38: 1-41.
Gale, D., and L.S. Shapley. 1962. College admissions and the
stability of marriage. A m e r i c a n Mathematical
Monthly 69: 9-15.
Garratt, R., and C.-Z. Qin. 1997. On a market for coalitions
with indivisible agents and lotteries. Journal of Economic
Theory 77: 81-101.
Gellekom, J.R.G., J.A.M. Potters, J.H. Reijnierse, M.C. Engel,
and S.H. Tijs. 2000. Characterization of
1721
the Owen set of linear production processes. G a m e s
a n d E c o n o m i c B e h a v i o u r 32: 139-156.
Gillies, D.B. 1959. Solutions to general non-zero-sum games.
In C o n t r i b u t i o n s t o t h e t h e o r y o f g a m e s ,
vol. 4, ed. A.W. Tucker and R.D. Luce. Princeton:
Princeton University Press.
Granot, D. 1986. A generalized linear production model: A
unifying model. M a t h e m a t i c a l P r o g r a m m i n g
34: 212-222.
Gravel, N., and S. Thoron. 2007. Does endogenous formation
of jurisdictions lead to wealth-stratification? J o u r n a l
o f E c o n o m i c T h e o r y 132: 569-583.
Greenberg, J. 1995. Coalition structures. In H a n d b o o k o f
game theoiy with economic applications,
ed. R.- J. Aumann and S. Hart. Amsterdam: North-
Holland.
Greenberg, J., and S. Weber. 1986. Strong Tiebout equilib-
rium under restricted preferences domain. J o u r n a l o f
E c o n o m i c T h e o r y 38: 101-117.
Greenberg, J., and S. Weber. 1993. Stable coalition structures
with a unidimensional set of alternatives. J o u r n a l o f
E c o n o m i c T h e o r y 60: 62-82.
Hahn, G., and N. Yannelis. 1997. Efficiency and incentive
compatibility in differential information economies.
E c o n o m i c T h e o r y 10: 383-411.
Haimanko, O., M. Le Breton, and S. Weber. 2004. Voluntary
formation of communities for the provision of public
projects. J o u r n a l o f E c o n o m i c T h e o r y 115: 1-
34.
Hammond, P., M. Kaneko, and M. Wooders. 1989. Con-
tinuum economies with finite coalitions: Core, equilibria,
and widespread externalities. Journal of
E c o n o m i c T h e o r y 49: 113-134.
Harsanyi, J. 1974. An equilibrium point interpretation of
stable sets and a proposed alternative definition. M a n -
a g e m e n t S c i e n c e 20: 1472-1495.
Hart, S., and M. Kurz. 1983. Endogenous formation of
coalitions. Econometrica 51: 1047-1064.
Hart, S., and A. Mas-Colell. 1996. Bargaining and value.
Econometrica 64: 357-380.
Herings, J-J., A. Mauleon, and V. Vannetelbosch. 2006.
Farsightedly stable networks. RM/06/041, University of
Maastricht.
Herings, J.-J., G. van der Laan, and D. Talman. 2007. The
socially stable core in structured transferable utility
games. Games and Economic Behavior 59: 85-104.
Holmstrom, B., and R. Myerson. 1983. Efficient and durable
decision rules with incomplete information. Econometrica
51: 1799-1819.
Ichiishi, T., and A. Idzik. 1996. Bayesian cooperative choice
strategies. International Journal of Game Theory 2E. 455-
473.
Iehle, V. 2006. The core partition of a hedonic game, Working
paper, Maison des Sciences Economiques, Universite
Pantheon-Sorbonne.
Izquierdo, J.M., and C. Rafels. 2001. Average monotonic
cooperative games. Games and Economic B e h a v i o r 36:
174-192.
Jackson, M., and A. van den Nouweland. 2005. Strongly stable
networks. Games and E c o n o m i c B e h a v i o r 51: 420-
1722
Jackson, A., and A. Watts. 2002. The evolution of social and
economic networks. Journal of Economic Theory 106:265
295.
Jackson, A., and A. Wolinsky. 1996. A strategic model of
social and economic networks. Journal of Economic
Theory 7 44 74.
Jordan, J.S. 2006. Pillage and property. Journal of Economic
Theory 131: 26—14.
Kajii, A. 1992. A generalization of Scarf’s theorem: An '/-core
existence theorem without transitivity or completeness.
Journal of Economic Theory 56: 194 205.
Kalai, E., A. Postlewaite, and .1. Roberts. 1979. A group
incentive compatible mechanism yielding core allocations.
Journal of Economic Theory 20: 13 22.
Kalai, E., and D. Schmeidler. 1977. An admissible set
occurring in various bargaining situations. Journal of
Economic Theory 14: 402-411.
Kalai, E., and E. Zemel. 1982. Generalized network problems
yielding totally balanced games. Operations Research 30:
998 1008.
Kaneko, .VI., and VI. Wooders. 1982. Cores of partitioning
games. Mathematical Social Sciences 3: 313-327.
Kaneko, VI., and VI. Wooders. 1986. The core of a game with
a continuum of players and finite coalitions: The model
and some results. Mathematical Social Sciences 12: 105
137.
Kaneko, M., and VI. Wooders. 1989. The core of a continuum
economy with widespread externalities and finite
coalitions: from finite to continuum economics. Journal of
Economic Theory 49: 135 168.
Kaneko, M., and M. Wooders. 1994. Widespread externalities
and perfectly competitive markets. In Imperfections and
behavior in economic organizations, ed. R. Gilles and P.
Ruys. Boston: Kluwer Academic Publishers.
Kaneko, M., and M. Wooders. 1996. The nonemptiness of
the /core of a game without side payments. International
Journal of Game Theory 25: 245 258.
Keiding, H. 1976. Cores and equilibria in an infinite economy.
In (’(imputing equilibrium: How and
why, ed. J. I.os and M.W. I.os. Amsterdam: North-
Holland.
Koczy, A., and I.. Lauwers. 2004. The coalition structure core
is accessible. Games and Economic Behavior 48: 86-93.
Konishi, H., M. I.e Breton, and S. Weber. 1997. Equilibrium
in a model with partial rivalry. Journal of Economic
Theory 72: 225 237.
Konishi, H., M. Le Breton, and S. Weber. 1998. Equilibrium
in a finite local public goods economy. Journal of
Economic Theory 79: 224 244.
Konishi, H., and D. Ray. 2003. Coalition formation as a
dynamic process. Journal of Economic Theory 110:
1- 41.
Kovalenkov, A., and M. Wooders. 2001a. Epsilon cores of
games with limited side payments; nonemptiness and
equal treatment. Games and Economic Behavior 36: 193-
218.
Coalitions
Kovalenkov, A., and VI. Wooders. 2001b. An exact bound on
epsilon for non-emptiness of the epsilon-cores of games.
Mathematics of Operations Research 26: 654 678.
Kovalenkov, A., and M. Wooders. 2003. Approximate cores of
games and economies with clubs. Journal of Economic
theory 110: 87-120.
Kovalenkov, A., and M. Wooders. 2005a. Many sided
matchings, clubs and market games. In Group formation
in economics: networks, clubs and
coalitions, ed. G. Demange and M. Wooders. Cambridge:
Cambridge University Press.
Kovalenkov, A., and VI. Wooders. 2005b. Laws of scarcity for
a finite game - Exact bounds on estimations. Economic
Theory 26: 383-396.
Kovalenkov, A., and M. Wooders. 2006. Comparative statics
and laws of scarcity for games. In Rationality and
equilibrium; a symposium in honor of Marcel K. Richter,
ed. C.D. Aliprantis et al. Berlin: Springer- Verlag.
Kurz, .VI. 1989. Game theory and public economics. In
Handbook of game theory, vol. 2. ed. R.J. Aumann and S.
Hart. Amsterdam: North-Holland.
Le Breton, .VI., I. Ortuno-Ortin, and S. Weber. 2006. Gamson
s Law and hedonic games, Working Paper No. 420, IDEI,
Toulouse.
Le Breton, .VI., G. Owen, and S. Weber. 1992. Strongly
balanced cooperative games. International Journal of
Game Theory 20: 419 427.
Le Breton, .VI., and S. Weber 2001. The art of making
everybody happy: How to prevent a secession. Working
Paper No. 01 176. International Monetary Frmd.
Lee, D., and O. Volij. 2000. The core of economies with
asymmetric information: An axiomatic approach. Journal
of Mathematical Economics 38: 43-63.
Legros, P., and A.F. Newman. 2002. Monotone matching in
perfect and imperfect worlds. Review of Economic
Studies 69: 925-942.
Legros, P., and A.F. Newman. 1996. Wealth effects, distri-
bution, and the theory of organization. Journal of Eco-
nomic Theory 70: 312-341.
Logistics Research Project. 1957. Reports of three informal
conferences on the theory of games. Department of
Mathematics, Princeton University. Accession No.
(OCoLC)ocm39726894. Conferences held 20-1 March,
1953; 31 January-1 February 1955, and
11- 12 March, 1957, Princeton.
Lucas, W., andR. Thrall. 1963. n-person games in partition
function form. Naval Research Logistics Quarterly 10:
281-298.
Maschler, M., and G. Owen. 1992. The consistent Shapley
value for games without side payments. In Rational
interaction, ed. R. Selten. Berlin: Springer-Verlag.
Maschler, M., and B. Peleg. 1966. A characterization, exis-
tence proof and dimensions bounds for the kernel of a
game. Pacific Journal of Mathematics 18: 289-328.
Maschler, M., and B. Peleg. 1967. The structure of the- kernel
of a cooperative game. SLAM Journal of Applied
Mathematics 15: 569-604.
Coalitions
Maschler, M., B. Peleg, andL.S. Shapley. 1971. The kernel and
bargaining set for convex games. International Journal of
Game Theory 1: 73-93.
Mas-Colell, A. 1989. An equivalence theorem for a bargaining
set. Journal of Mathematical Economics 18: 129-139.
Mauleon, A., and V.J. Vannetelbosch. 2004. Farsightedness
and cautiousness in coalition fomration games with
positive spillovers. Theory and Decision 56: 291-324.
McLean, R., and A. Postlewaite. 2005. Core convergence with
asymmetric information. Games and Economic Behavior
50: 58-78.
Milchtaich, I. 1996. Congestion games with player-specific
payoff functions. Games and Economic Behavior 13:
111- 124.
Milchtaich, I., and E. Winter. 2002. Stability and segregation
in group formation. G a m e s a n d E c o n o m i c
B e l l a v - i o r 38: 318-346.
Milleron, J.C. 1972. Theory of value with public goods: A
survey article. Journal of Economic Theory 5: 419-477.
Monderer, D., and L.S. Shapley. 1996. Potential games.
Games and Economic Behavior 14: 124-143.
Muench, T. 1972. The core and the Lindahl equlibrium of an
economy with public goods; an example. Journal of
Economic Theory 4: 241-255.
Myerson, R.B. 1977. Graphs and cooperation in games.
Mathematics of Operations Research 2: 225-229.
Myerson, R.B. 1995. Sustainable matching plans with adverse
selection. G a m e s a n d E c o n o m i c B e h a v i o r 9:
35-65.
Myerson, R.B. 2007. Virtual utility and the core for games
with incomplete information. Journal of
E c o n o m i c Theory 136: 260-285.
Nash, .1. 1953. Two-person cooperative games.
E c o n o m e t r i c a 21: 128-140.
Owen, G. 1975. On the core of linear production games.
M a t h e m a t i c a l P r o g r a m m i n g 9: 358-370.
Owen, G. 1977. Values of games with a priori unions. In
Essays i n m a t h e m a t i c a l e c o n o m i c s a n d
g a m e theory, ed. R. Henn and O. Moeschlin.
Berlin/Heidel- berg/New York: Springer.
Page Jr., F.H. 1997. Market games with differential infor-
mation and infinite dimensional commodity spaces: The
core. E c o n o m i c T h e o r y 9: 151-159.
Page Jr., F.H., and M. Wooders. 1996. The partnered core and
the partnered competitive equilibrium. E c o n o m i c s
Letters 52: 143-152.
Page, F.H. Jr., and M. Wooders. 2006. S t r a t e g i c b a s i n s
of attraction, the path dominance core,
a n d n e t w o r k f o r m a t i o n g a m e s , Working
Paper No. 06-W14 (revised), Department of Economics,
Vanderbilt University.
Page, F.H. Jr., and M. Wooders. 2007. Networks and clubs.
Journal of Economic Behavior & O r g a n i z a t i o n
(forthcoming).
Page Jr., F.H., M. Wooders, and S. Karnat. 2005. Networks
and farsighted stability. Journal of Economic Theory 120:
257-269.
1723
Papai, S. 2004. Unique stability in simple coalition formation
games. G a m e s a n d E c o n o m i c B e h a v i o r 48:
337-354.
Pauly, M. 1970. Cores and clubs. P u b l i c C h o i c e 9: 53-65.
Peleg, B. 1985. The axiomatization of the core of cooperative
games without side payments. J o u r n a l o f M a t h e -
m a t i c a l E c o n o m i c s 14: 203-214.
Perry, M., and P.J. Reny. 1994. A noncooperative view of
coalition formation and the core. E c o n o m e t r i c a 6 2 :
795-817.
Predtetchinski, A., and J.-J. Herings. 2004. A necessary and
sufficient condition for non-emptiness of the core of a non-
transferable utility game. J o u r n a l o f E c o n o m i c
T h e o r y 116: 84-92.
Qin, C.-Z. 1994. The inner core of an n-person game. G a m e s
a n d E c o n o m i c B e h a v i o r 6: 431-444.
Ray, D., and R. Vohra. 1997. Equilibrium binding agree-
ments. J o u r n a l o f E c o n o m i c T h e o r y 73: 30-78.
Ray, D., and R. Vohra. 1999. A theory of endogenous coalition
structures. G a m e s a n d E c o n o m i c B e h a v i o r
26: 286-336.
Reny, P.J., andM. Wooders. 1996a. The partnered core of a
game without side payments. J o u r n a l of
E c o n o m i c Theory 70: 298-311.
Reny, P.J., and M. Wooders. 1996b. Credible threats of
secession, partnership, and commonwealths. In U n d e r -
standing strategic interaction: Essays in
h o n o r o f R e i n h a r d S e l t e n , ed. W. Albers et al.
Berlin: Springer- Verlag.
Roth, A.E. 1984. Stable coalition formation: Aspects of a
dynamic theory. In C o a l i t i o n s a n d c o l l e c t i v e
a c t i o n , ed. M. Holler. Wiirzberg: Physica-Verlag.
Samet, D., and E. Zemel. 1984. On the core and dual set of
linear programming games. M a t h e m a t i c s of
O p e r a t i o n s R e s e a r c h 9: 309-316.
Scotchmer, S., and M. Wooders. 1988. M o n o t o n i c i t y i n
games that exhaust gains to scale,
Technical Report No. 525, IMSSS, Stanford University.
Seidmann, D.J., and E. Winter. 1998. Gradual coalition
formation. R e v i e w o f E c o n o m i c S t u d i e s 65:
793-815.
Serrano, R., and R. Vohra. 2006. Information transmission in
coalitional voting games. J o u r n a l o f E c o n o m i c
T h e o r y (forthcoming).
Selten, R. 1981. A non-cooperative model of characteristic
function bargaining. In E s s a y s i n g a m e t h e o r y
and mathematical economics in honor of
O . M o r g e n s t e r n , ed. V. Bohm and H. Nachtkamp.
Mannheim: Bibliographisches Institut.
Shapley, L.S. 1953. A value for n-person games. In
Contributions to the theory of games II,
ed. A.W. Tucker and R.D. Luce. Princeton: Princeton
University Press.
Shapley, L.S. 1967. On balanced sets and cores. N a v e l
R e s e a r c h L o g i s t i c s Q u a r t e r l y 9: 45-48.
Shapley, L.S. 1971. Cores of convex games. I n t e r n a t i o n a l
J o u r n a l o f G a m e T h e o r y 1: 11-26.
Shapley, L.S., and M. Shubik. 1966. Quasi-cores in a
monetary economy with nonconvex preferences.
E c o n o m e t r i c a 34: 805-827.
1724
Shapley, L.S., and M. Shubik. 1969a. On the core of an
economic system with externalities. A m e r i c a n E c o -
n o m i c R e v i e w 59: 678-684.
Shapley, L.S., and M. Shubik. 1969b. On market games.
J o u r n a l o f E c o n o m i c T h e o r y 1: 9 25.
Shapley, L.S., and .VI. Shubik. 1972. The assignment game 1:
The core. I n t e r n a t i o n a l J o u r n a l o f G a m e
T h e o r y 1: 11-30.
Shapley, L.S., and VI. Shubik. 1975. Competitive outcomes in
the cores of market games. I n t e r n a t i o n a l J o u r n a l
o f G a m e T h e o r y 4: 229-237.
Sharkey, WAV'. 1990. Cores of games with fixed costs and
shared facilities. I n t e r n a t i o n a l Economic
R e v i e w 31: 245-262.
Shenoy, P.P. 1979. On coalition formation: A game- theoretic
approach. I n t e r n a t i o n a l J o u r n a l o f G a m e
T h e o r y 8: 133 164.
Shenoy, P.P. 1980. A dynamic solution concept for abstract
games. Journal of Optimization Theory and Applications
32: 151 169.
Shubik, M. 1959. Edgeworth market games. In C o n t r i b u -
tions to the theory of games IV, annals
o f m a t h e m a t i c a l s t u d i e s 4 0 , oil. F.R. Luce and
A. Tucker. Princeton: Princeton University Press.
Shubik, M. 1671. The bridge game economy: An example of
indivisibilities. J o u r n a l o f P o l i t i c a l E c o n o m y
79: 909-912.
Shubik, M., and M. Wooders. 1982. Near markets and market
games. Cowles Foundation Discussion Paper No. 657, also
published as ‘Clubs, near markets and market games’. In
Topics in mathematical economies and
game theory: Essays in honor of Robert
J . A u m a n n , ed. M. Wooders. Fields Institute
Communication 23. Providence: American Mathematical
Society, 2000.
Shubik, M., and M. Wooders. 1983a. Approximate cores of
replica games and economies: Part I. Replica games,
externalities, and approximate cores. M a t h e m a t i c a l
S o c i a l S c i e n c e s 6: 27-48.
Shubik, M., and M. Wooders. 1983b. Approximate cores of
replica games and economies: Part II. Set-up costs and
firm formation in coalition production economies.
M a t h e m a t i c a l S o c i a l S c i e n c e s 6 : 285 306.
Tanman, Y., A. Urbano, and J. Watanabe. 1997. A model of
multiproduct price competition. J o u r n a l of
E c o n o m i c T h e o r y 77: 377 401.
Ticbout, C. 1956. A pure theory of local expenditures.
J o u r n a l o f P o l i t i c a l E c o n o m y 64: 416 424.
Tijs, S. 1681. Bounds for the core and the tau-value. In G a m e
t h e o r y a n d m a t h e m a t i c a l e c o n o m i e s , ed. O.
Moeschlin and D. Pallaschke. Amsterdam: North-
Holland.
van den Nouweland, A. 2005. Models of network formation in
cooperative games. In G r o u p f o r m a t i o n i n
economies: Networks, clubs and
coalitions,
ed. G. Demange and M. Wooders. Cambridge: Cam-
bridge University Press.
von Neumann, .1., and O. Morgenstem. 1953. T h e o r y o f
g a m e s a n d e c o n o m i c b e h a v i o r . Princeton:
Coase Theorem
Yohra. R. 1999. Incomplete information, incentive com-
patibility and the core. J o u r n a l o f E c o n o m i c
T h e o r y 86: 123-147.
Yolij. O. 2000. Communication, credible improvements and
the core of an economy with asymmetric information.
International Journal of Game Theory
29: 63-79.
Watts. A. 2001. A dynamic model of network formation.
G a m e s a n d E c o n o m i c B e h a v i o r 34: 331-341.
Weber. S. 1979. On e-cores of balanced games. I n t e r n a -
t i o n a l J o u r n a l o f G a m e T h e o r y 8: 241-250.
Weber. S. 1981. Some results on the weak core of a non-
sidepayment game with infinitely many players.
J o u r n a l o f M a t h e m a t i c a l E c o n o m i c s 8:
101-111.
Weber. S.. and H. Wiesmeth. 1991. The equivalence of core
and cost share equilibria in an economy with a pubhc
good. J o u r n a l o f E c o n o m i c T h e o r y 54: 180-
197.
Wilson. R. 1978. Information, efficiency and the core of an
economy. E c o n o m e t r i c a 46: 807-816.
Winter. E.. and M. Wooders. 1994. An axiomatization of the
core for finite and continuum games. S o c i a l C h o i c e
a n d W e l f a r e 11: 165-175.
Wooders, VI. 1978. Equilibria, the core, and jurisdiction
structures in economies with a local public good. J o u r -
n a l o f E c o n o m i c T h e o r y 18: 328-348.
Wooders, VI. 1980. The Ticbout hypothesis: Near optimality
in local pubhc good economies. E c o n o m e t r i c a 48:
1467-1486.
Wooders, VI. 1983. The epsilon core of a large replica game.
J o u r n a l o f M a t h e m a t i c a l E c o n o m i c s 11:
277-300.
Wooders, M. 1994a. Equivalence of games and markets.
E c o n o m e t r i c a 62: 1141-1160.
Wooders, M. 1994b. Large games and economies with
effective small groups. In G a m e - t h e o r e t i c
methods in general equilibrium analysis,
ed. J.-F. Mertens and S. Sorin. Dordrecht: Kluwer
Academic Publishers.
Wooders, M., and W.R. Zame. 1987. Large games; fair and
stable outcomes. J o u r n a l o f E c o n o m i c T h e o r y
42: 59 93.
Xue, L. 1998. Coalitional stability under perfect foresight.
Coase Theorem
Francesco Parisi
Abstract
The Coase Theorem holds that, regardless of the
initial allocation of property rights and choice of
remedial protection, the market will
Coase Theorem
determine ultimate allocations of legal entitle-
ments, based on their relative value to different
parties. Coase’s assertion has occasioned
intense debate. This article provides an intel-
lectual history of Coase’s fundamental theorem
and surveys the legal and economic literature
that has developed around it. It appraises the
most notable attacks to the Coase Theorem, and
examines its methodological implications and
normative and practical significance in legal
and policy settings.
Keywords
Adverse selection; American Law and Eco-
nomics Association; Asymmetric information;
Bargaining; Coase theorem; Contract enforce-
ment; Efficient allocation; Entropy; Externali-
ties; Free rider problem; Hold-up; Inalienability;
Incentives; Income effect; Law, economic
analysis of; Liability rules; Pigou, A. C.;
Pigouvian taxes; Plant, A.; Private information;
Property fragmentation; Property rights; Public
goods; Scarcity; Social cost; Stigler, G.;
Strategic behaviour; Tort; Transaction costs;
Voluntary transfers
JEL Classifications
D62
Mutuality of advantage from voluntary exchange is
one of the most fundamental concepts in eco-
nomics. The well-known proposition of Ronald H.
Coase ( ) - generally known as the Coase
theorem - builds on this simple and yet fundamental
insight. The law creates many rights and legal
entitlements, establishing the initial allocation of
rights and liabilities. Whenever there are no legal or
factual impediments to exchange, the dynamic of
the market will determine the final allocation of
such rights.
In this context, Coase suggests that the trans-
ferability of rights in a free economy leads toward
their best use and an efficient final allocation.
Whenever the initial allocation is not optimal, the
owners of the rights will have an incentive to
transfer them to other individuals who value them
more. Such an exchange will continue until there
1725
is no further potential for reciprocal profit, which
will not be exhausted until each right is in the hands
of the highest-valuing individual. The Coase
theorem predicts that, in a competitive market
environment without legal or factual impediments
to exchange, the final allocation of rights will be
efficient.
This article discusses the pervasive methodo-
logical implications of Ronald Coase’s idea to the
field of law.
A Brief Intellectual History
Coase’s assertion that an initial assignment of
property rights is often irrelevant to overall welfare
has occasioned one of the most intense and
fascinating debates in the history of legal and
economic thought. Private property is often
explained as the unavoidable by-product of scarcity
in a world where common pool losses outweigh the
sum of contracting costs and enforcement of
exclusive property rights. At the turn of the 20th
century, the underlying assumption in the economic
literature was that private property emerged out of a
spontaneous evolutionary process because of the
desirable features of private property regimes in the
creation of incentives for constrained optimization.
This understanding of the relationship between
scarcity and emergence of legal entitlements char-
acterized mainstream property right theory when
Coase entered the academic world. Coase began his
undergraduate studies at the London School of
Economics in 1929, as a candidate for a Bachelor
Degree in Commerce. In those years, one of
Coase’s teachers, Sir Arnold Plant, was re-
examining the theme of property rights from a novel
perspective. According to Plant, the traditional
justification for private property - scarcity - was
incapable of serving as the sole intellectual
foundation for this institution. Plant showed that
incentives, rather than scarcity, lay at the core of the
property right problem (Plant ).
Coase’s use of legal rules as an object of eco-
nomic research in his analysis of incentive structure
and alternative final resource allocations reveals a
remarkable technical affinity with the
1726
work of his undergraduate teacher. In his Nobel
memorial lecture, Coase acknowledges the impor-
tance of his encounter with Plant as a ‘great stroke
of luck’ that cultivated his interest in property rights
theory (Coase , p. 715). For Coase, Plant’s teaching
that ‘[t]he normal economic system works itself
(Salter , pp. 16-17) and that prices in a competitive
market lead resources to their highest valuing uses
was a revelation into the dynamic of the economic
system: ‘I was then 21 years of age, and the sun
never ceased to shine. I could never have imagined
that these ideas would become some 60 years later a
major justification for the award of a Nobel Prize.
And it is a strange experience to be praised in my
eighties for work / did in my twenties’ (Coase , p.
716).
The experience of the following years at the
London School of Economics laid the methodo-
logical foundations of what would later become
Coase’s theorem on the problem of social costs. All
the ingredients of his revolutionary analysis on the
debated theme of social cost had been profiled
during his LSE years (see Williamson and Winter ,
pp. 34-5). But it is not until the late 1950s that
Coase verbalized such a simple and yet ingenious
idea. He had first expounded the core of his later
theorem in an article published in 1959. In those
pages, one grasps what would later become the
central theme of Coase’s celebrated argument:
Whether a newly discovered cave belongs to the man
who discovered it, the man on whose land the
entrance to the cave is located, or the man who owns
the surface under which the cave is situated is no
doubt dependent on the law of property. But the law
merely detemrines the person with whom it is nec-
essary to make a contract to obtain the use of the
cave. Whether the cave is used for storing bank
records, as a natural gas reservoir, or for growing
mushrooms depends, not on the law of property, but
on whether the bank, the natural gas corporation, or
the mushroom concern will pay the most in order to
be able to use the cave. ( , p. 25)
The discussion of the rationale of property
rights under Coase’s highest bidder framework
obviously contained an attack on the Pigouvian
approach (Pigou ) to the problem. The point was
rather self-evident to Coase, but not so for
Coase Theorem
some of the Chicago economists. George Stigler
was among Coase’s early critics:
Ronald Coase criticized Pigou’s theory rather casu-
ally, in the course of a masterly analysis of the
regulatory philosophy underlying the Federal Com-
munication Commission’s [FCC] work. Chicago
economists could not understand how so fine an
economist as Coase would make so obvious a mis-
take. Since he persisted, we invited Coase (he was
then at the University of Virginia) to come and give a
talk on it. Some twenty economists from Chicago and
Ronald Coase assembled one evening at the home of
Aaron Director. ... In the course of two hours of
argument the vote went from twenty against and one
for Coase to twenty-one for Coase. What an
exhilarating event! (Stigler , pp. 75-6)
According to Coase, the objections to his FCC
paper are at the origin of his later 1960 article on
the problem of social costs. Coase recalls that he
was urged to omit that section of his FCC article,
something he refused to do. In retrospect, Coase
believes that had it not been for the Chicago
economists’ attacks his full-fledged idea would
have never been formulated ( , p. 250).
The Positive Coase Theorem
The arguments that were refined in the course of
such debate were later put together in the form of an
article for the Journal of Law and Economics in
1960, titled ‘The Problem of Social Cost’. This
article - later known as the Coase theorem - soon
became a milestone in legal and economic litera-
ture. In the course of his austere discussion, Coase
does not reveal any sign of anticipated realization of
the revolutionary power of his insight. Indeed,
Coase insists that he never intended to convey his
thoughts in the precise and analytical form of a
theorem ( , p. 157).
A few years after the publication of ‘The Prob-
lem of Social Cost’, a sizeable number of com-
mentaries and theoretical elaborations were
developed on Coase’s newly presented theme. The
unpretentious style of Coase’s article had thus been
crowned by a notoriety rarely attained by legal
writings of any sort (Shapiro , p. 1540). Part of the
uproar is explained by the fact that the article
challenged an established
Coase Theorem 1727

principle of public finance (see Manne pp. 123-6).novelty of his approach inspired an entire gener-
Before ‘The Problem of Social Cost’, very little ation of scholars - pioneers in this new branch of
attention had been given to the possibility that the
applied economics. Only a few months prior to
problem of externalities could be resolved throughreceiving the Nobel Prize for economics, in occa-
free market exchanges. sion of the First Annual Meeting of the American
Coase boldly attacked the conclusions reachedLaw and Economics Association, Ronald H. Coase
by the Pigouvian tradition by suggesting its influ-
was recognized, together with Guido Calabresi,
ence was in part due to the lack of clarity in itsHenry G. Manne and Richard A. Posner, as a
exposition ( , p. 39). Coase departs from the
founding father of Law and Economics. This
Pigouvian approach by demonstrating that, in the recognition follows many years of challenging
absence of transaction costs, generators and victims
debate. Many of the writings that developed around
of externalities will negotiate an efficient allocation
‘The Problem of Social Cost’ tested the premises of
of resources, independent of the initial assignment
Coase’s model, seeking to undermine the conditions
of rights among them. In confuting the conclusionsof his model and stressing the lack of practical
of the Pigouvian tradition, Coase gave life to a reach of his analysis.
model with the potential for the evaluation of an Further criticisms pertained to three fundamen-
unlimited number of legal and social issues. tal points. One group of critics observed that the
George Stigler was the first scholar to restate
Coase Theorem disregarded the inter-industrial
Coase’s model in the form of a theorem: ‘[U]nder long-term effects of the system (Calabresi ;
perfect competition private and social costs will be
Wellisz ). These critics argued that Coase ignored
equal’ ( , p. 113). Demsetz ( , p. 349)
the possible disequilibria which may occur after the
defined the theorem in the following terms: ‘Therenegotiation and the likely dynamic changes in the
are two striking implications of this process that are
initial equilibrium. In the context of Coase’s well-
true in a world of zero transaction costs. The output
known example, if the right has been assigned to the
mix that results when the exchange of property ranchers, the farmer will have to pay local ranchers
rights is allowed is efficient and the mix is until they all relinquish their right of pasture. The
independent of who is assigned ownership (except entire cost will, thus, burden the farming industry.
that different wealth distributions may result in Farmers will either have to bear the burden of the
different demands)’. Soon thereafter, Guido injury caused by the livestock or agree to pay the
Calabresi stated the same principle more price demanded by the ranchers, whichever is less,
descriptively: ‘Thus, if one assumes rationality, no
on the assumption that negotiation is costless.
transaction costs, and no legal impediments to Under this liability rale, the cost of ranching will
bargaining, all misallocations of resources would be
not reflect the cost imposed on the farmers. The
frilly ciued in the market by bargains’ (Calabresi ,
transfer of rights and liability from one group to
p. 68). another will, therefore, result in a shift in the
The implicit premise of Coase’s analysis draws
relative wealth and costs associated with the two
upon a fundamental postulate of microeconomic industries. The criticism claims that, in the long ran,
theory: the free exchange of goods in the market every shift of wealth will lead to an inter-industrial
moves goods towards their optimal allocation. The disequilibrium.
voluntary transfer of individual rights in the In 1968, Calabresi, one of the initial proponents
marketplace, thus, will ciue a non-optimal alloca-of this criticism, reconsidered it, noting that in the
tion of legal entitlements. presence of determined conditions the conclusions
of Coase remain as true in the long ran as in the
The Coasean Methodological Revolution short term ( , p. 67). Calabresi’s
Coase’s article constitutes, according to many later analysis re-established the authority of the
commentators, the first example of an economic Coase Theorem, at least on this point. It became
analysis of law in North American literature. The clear that Coase did not ignore the long-term
1728
effects of his model. Perhaps not explicitly, he had
considered them to their logical extreme. Calabresi
observes: ‘The reason is simply that (on the given
assumptions) the same type of transactions which
cured the short mn misallocation would also occur
to cure the long mn ones. ... This process would
continue until no bargain could improve the
allocation of resources’ (1967, pp. 67-8).
In 1972, Harold Demsetz joined this debate,
demonstrating with a more systematic analysis that
the conclusions reached by Coase are not corroded
by the long-term effects of a change in the
assignment of property rights. Demsetz’s reasoning
finds its basis in the principle that the process of
allocation of scarce resources among alternative
uses is analogous to the process of constrained
optimization of the single owner of two conflicting
activities.
An additional critique, formulated by Calabresi (
) and Wellisz ( ), suggests
that strategic behaviour in the bargaining process
risks compromising Coase’s results. These authors
observe that the change in the rule of law creates the
conditions for possible extortion on the part of the
right holders against the other individuals who are
bound by the rule. The argument is that individuals
are likely to threaten the use of their own rights in a
measure which exceeds the optimal level, in order
to maximize the gain from the release of their own
legal entitlements. By introducing the possibility of
strategic behaviour in the negotiation, the result
may differ from the optimal equilibrium. Demsetz (
, p. 21) supplied a convincing answer to
this criticism. According to Demsetz, the possibility
of strategic behaviour in the negotiations does not
alter the efficiency in the final allocation of
resources between the two activities. Strategies will
be capable of altering the internal distribution of the
contractual surplus between the parties, but not the
final outcome of the negotiation.
It should be noted, however, that the entire
analysis presupposes that the so-called income
effect can be ignored. In general, a different allo-
cation of property rights implies a different distri-
bution of wealth between the individuals involved.
Different initial endowments generate
Coase Theorem
different final allocations, notwithstanding an equal
level of efficiency. In order for the final allocations
to be identical, it is necessary that the utility
functions of the individuals involved are almost
linear. The absence of the income effects implies, in
this sense, that the demand functions for the good
are independent of the income level.
It should be further observed that the credibility
of the threat made in the course of strategic
bargaining finds its limits in the market structure in
which the Coasean negotiation takes place. In
general, the competitive structure of the market
eliminates much of the advantage that can be
obtained through strategic behavior in the negoti-
ation process. Inasmuch as the market of resources
is competitive, strategic bargaining is not capable of
bringing about any abnormal return.
The criticism, however, appears to be on the
mark when it argues that, in some marginal situa-
tions, the curing role of the free exchange may still
be impeded. For example, consider reversing the
assignment of property rights between the rancher
and the farmer. In such a situation, the farmer is
likely not to have an equally large number of
alternatives. The transfer of a farm from one place
to another is costly, and farming unavoidably
requires the undertaking of location-specific
investments. Since some capital investment is
irreversibly locked in that specific location, the
farmer has less opportunity to relocate than the
rancher. The rancher, consequently, finds himself in
a position of local monopoly in the sale of his
property right. Demsetz considers the monopoly
that affects this feature of the Coasean exchange
identical to the standard monopoly of microeco-
nomic analysis ( , p. 24). According to
Demsetz, the concerns for possible monopolistic
structures in the market of rights considered by
Coase must not, however, be used to raise again the
already resolved problem of the initial allocation of
rights, since reversing the rale of liability would
simply result in the farmer now having monopoly
power ( , pp. 24-5).
A second group of critics concentrated on the
distributive effects of the model (Regan ; Nutter ).
They argued that a final efficient allocation of
resources requires transfers of wealth
Coase Theorem
induced by the changed legal rule. Further, these
critics observed that, even if one disregards the
distributive effects of the rule, a different assign-
ment of the right could in some cases create the
conditions for strategic behaviour in negotiation
capable of disturbing the efficiency of the final
allocation.
A third group of authors focused on the scarce
realism of the no-transaction-cost assumption (see
Cooter , p. 457). According to this criticism, the
true Achilles’ heel of Coase’s analysis was in the
unrealistic assumption of absence of costs in the
process of negotiation and transfer of the right.
These authors observed that the idea of a transac-
tion without cost is a logical fiction cloaking a mere
tautology.
The Normative Coase Theorem
The utility of models predicting behaviour in a zero
transaction-cost world is that they guide the law -
whose object is to develop rules which approximate
the zero transaction-cost world as closely as
possible - in responding to legal problems arising in
a positive transaction-cost environment (Epstein ).
The vast literature that developed around Coase’s
theorem formulated important normative corollaries
of it, based on the evaluation of the relative costs of
alternative assignments of rights.
According to the positive Coase theorem, absent
transaction costs, the final allocation of scarce
resources would coincide with the use that an
individual who is the single owner of different
activities would make of his endowments,
regardless of the initial assignment of rights and
choice of remedial protection. When transaction
costs are present, however, an exchange will be
pursued only to the point at which its marginal
benefit equals the marginal cost of the transaction.
If transaction costs exceed the benefits of a contract,
no exchange will take place in the market. For a
right to be exchanged it is necessary that transaction
costs be less than the difference between the
demand and supply prices. If this condition is not
met, the Coasean bargaining will not be carried out,
and both initial
1729
assignment of rights and choice of remedies will
affect final allocations.
The Relevance of Transaction Costs and the
Simple Normative Coase Theorem
The notion of transaction costs has acquired par-
ticular importance in law and economics as the
absence of transaction costs represents a funda-
mental condition for the applicability of the positive
Coase theorem. Although at first impression
transaction costs play a role analogous to trans-
portation costs in international trade or, more gen-
erally, to the contracting costs in the economics of
exchange (Demsetz , p. 20), in Coase’s world the
role of transaction costs has much greater normative
implications.
For purposes of the theorem, the notion of
transactions costs should include not only
bargaining costs associated with the negotiation and
conclusion of the contract but also all costs
associated with the strategic behaviour of the parties
and the execution and enforcement of the
transaction. The notion of transaction costs should
thus include ex ante costs due to asymmetric
information, adverse selection, free riding, and
hold-up strategies, as well as ex post costs
associated with monitoring and enforcing the
contracts.
Strategic behaviour may be an important source
of transaction costs in a Coasean setting. In Coase’s
various examples, the property rights which are
exchanged are private goods, characterized by their
excludability. Difficulties arise when the object of
the Coasean bargaining is an entitlement which has
the nature of a public good (see Cheung , pp. 49-
70). Due to the well-known problems associated
with the supply of public goods, the Coasean
bargaining solution may fail to cure a non-optimal
allocation of rights that falls within this category.
Consider a scenario in which the object of the
Coasean negotiation consists of a non-excludable
right (for example, the right to enjoy pollution-free
air in a residential environment). As well known,
individuals will not reveal their own preferences for
public goods through the price system, placing
public goods among those cases that are most
resistant to the Coasean antidote.
1730
A first simple normative reformulation of the
Coase theorem focuses on transaction costs and the
role that legal systems may play in reducing these
impediments to voluntary bargaining. Legal rules
can lower obstacles to private bargaining, such as
by reducing transaction costs and minimizing other
costs associated with transfer (strategic, legal, and
so on). For this reason, transactional cost
considerations should be fundamental to any
analysis of legal regimes and the design of
contracting processes, governance mechanisms and
institutions.
The Complex Normative Coase Theorem
The first original formulation of Coase’s proposi-
tion can be restated as a normative theorem: in the
presence of positive transaction costs, the efficiency
of the final allocation is not independent of the
choice of the legal rale, and that the preferable
initial assignment of rights is that which minimizes
the effects of such transaction costs. The various
normative restatements of the Coase Theorem aim
at identifying legal rules and remedies that replicate
the outcomes of a hypothetical Coasean bargaining
or to mimic the solution that would be chosen by
the single owner of interfering resources.
Important normative reformulations of the Coase
Theorem focus on two important elements:
relevance of initial assignment of rights and rele-
vance of remedial protection. Demsetz ( ) and
Calabresi and Melamed ( ) were among the
first to discuss systematically the problems resulting
from lifting the assumption of zero transaction
costs. Articulating the normative core of the Coase
theorem, Demsetz observes that the introduction of
significant transaction costs into the choice of
liability rule analysis does affect resource
allocation. One liability rule may be superior to
another because the difficulty of avoiding costly
interactions is usually different for the interacting
parties. Accordingly, the normative predicament
indicates that the rale of liability should be based on
which party can avoid the costly interaction at the
lowest cost.
When two or more parties have conflicting
interests in the same resource, the law must decide
which party shall prevail, that is, which party shall
Coase Theorem
receive the entitlement. Once the entitlement deci-
sion is made, the law must decide how the entitle-
ment is to be protected and whether it may be
transferred. Articulating a concept of entitlements
protected by property, liability or inalienability
rales, Calabresi and Melamed ( ) develop a
framework that integrates the approaches of prop-
erty and tort. Entitlements can be protected by
property rules (transfer of the entitlement involves a
voluntary sale by its holder), liability rales (the
entitlement may be destroyed by another party if he
is willing to pay an objectively determined value for
it), or rales of inalienability (transfer of the
entitlement is not permitted, even between a willing
seller and a willing buyer). Calabresi and Melamed
allow for a wide range of concerns to be balanced
through the assignment of a particular entitlement.
Calabresi and Melamed outline how, given the
reality of transaction costs, an economic efficiency
approach selects one allocation of entitlements over
another. Entitlements cannot be enforced solely
through property rules because, even if the transfer
would benefit all parties, high transaction costs
(especially the hold-up problem) may prevent an
efficient reallocation. Calabresi and Melamed
demonstrate how liability rales often achieve a
combination of efficiency and distributive results
that would be difficult to achieve under a property
rale. Calabresi points out that Coase’s analysis
offers invaluable instruments for the identification
of the areas in which public intervention becomes
desirable (Calabresi , pp. 72-3). In its normative
version, the theorem indicates that legal rules that
minimize the effects of such costs are to be
preferred for being relatively more efficient
(Polinksy , p. 14). In its more complex formulation,
the Coase theorem provides, indeed, a guide for
such a choice.
The following is a classic illustration (Polinsky ,
pp. 11-14). The smoke of a factory soils laundry
which is line drying on five neighbouring
properties. The losses amount to $150 for each
neighbour, for a total of $750. The damage could be
eliminated through the installation of a purifying
filter on the industrial smokestack or through the
acquisition of electric dryers on the part of each one
of the neighbouring owners. The cost
Coase Theorem
of the filter would amount to $300, while the dryers
would impose a cost of $100 per household, for a
total of $500. The first solution is obviously more
efficient, since the acquisition of five dryers would
require a greater expenditure than the single filter.
The Coase theorem predicts that in the absence of
transaction costs the efficient solution will be
chosen independently of the initial assignment of
property rights. Even if we assume an initial
allocation of polluting right to the industry (that is,
fully legalizing industrial emissions), the
landowners would jointly offer to buy the industrial
filter at their expense. Sharing the cost of the filter
in equal parts, each owner would face a cost of only
$60, with a relative saving of $40 compared with
the otherwise necessary acquisition of a personal
dryer.
If we relax the initial assumption of no trans-
action costs, the initial allocation of property rights
no longer is immaterial. Imagine that each owner
has to face a cost of $120 in order to negotiate the
contract with his neighbours and with the owner of
the industrial plant. If the right is assigned to the
industry, each landowner will have to choose
whether to bear the loss of his soiled laundry for
$150, to acquire the electric dryer for $100, or,
finally, to undertake the negotiation process for a
total pro-rata cost of $180. Considering these
alternatives, each rational land- owner would
choose to acquire his own dryer, generating a
socially non-optimal outcome. However, the
assignment of property rights to the neighbouring
residents rather than to the polluting industry would
minimize the effect of positive transaction costs,
since the industry would have incentives to install
the filter, without any need for Coasean bargaining
with the neighbours.
Two impediments to bargaining (that is, sources
of transaction costs) take the form of externalities
and hold-up, which Epstein ( ) rales, is the work by Ayres and Talley (
shows stand in inverse relationship to each other.
He defines the optimal legal rule as that which
minimizes the sum of these externality and holdout
costs in any particular institutional setting. Epstein
demonstrates, through examples in property,
restitution and tort, how Coase’s transaction costs
model plays the central organizing role in
developing legal responses to many private law
1731
problems. Notwithstanding the obvious measure-
ment and information problems, Epstein ( )
stresses the importance of the ‘single owner test’:
where resources are under the command of two or
more persons, the legal arrangement should attempt
to induce all the parties to behave in the same way
that a single owner would. Epstein concludes that,
where the single owner test yields a unique result,
that result should be adopted as the legal rule.
Where the single owner test does not yield clear
results, however, no corollary principle will provide
a decisive answer to the particular problem.
Further exploring the choice between property
and liability rules suggested by Calabresi and
Melamed, Kaplow and Shavell ( ) address
several factors casting doubt on the equivalence of
these alternatives in low transaction-cost envi-
ronments. Their analysis considers several objec-
tions to Coasean costless bargaining, including the
inability of a party to ascertain what the other is
willing to pay or accept, victims’ ability to mitigate
harm, the problem posed by one party being
judgment proof, and administrative costs. Kaplow
and Shavell find a presumption in favour of liability
rales over property rules in the context of harmful
externalities, but that this may be overcome as a
result of one of more of the factors they describe.
After considering some of the proffered
justifications for the use of property rales to protect
possessory interests, the authors find a strong
theoretical case for the protection of these interests
using property rales. The normative Coase theorem
thus underlies the choice of the optimal system to
ensure the protection of various types of property
rights.
Also bridging the gap between Coase, where
liability rales and property rales are equally effi-
cient, and Calabresi and Melamed, where high
transaction costs lead to a preference for liability
) on
private information as a transaction cost. The
inefficiency occurs when parties misrepresent their
own valuations to gain strategic advantage in the
bargaining process. Focusing on the effect of
splitting an entitlement between two rivalrous users
rather than among buyers or among sellers, these
authors find that, when two parties have
1732
private information about how much they value an
entitlement, endowing each party with a partial
claim to the entitlement can reduce the incentive to
behave strategically during bargaining by inducing
greater disclosure. A bargainer has two Coasean
alternatives: buy the other party’s claim or sell
one’s own claim. The normative formulation of
Ayres and Talley is that a liability rule regime is
preferable because it allows a party’s decision to
pursue one of these alternative transactions to
function as a credible signal of a low or high
valuation, thereby encouraging more efficient trade.
Building upon the literature on property frag-
mentation (Heller ; Buchanan and Yoon ), Parisi ( )
and Schulz et al. ( ) suggest that property is
subject to a fundamental law of entropy. In the
property context, entropy induces a one-directional
bias. This bias is driven by asymmetric transaction
costs - it is often harder to reunite separated
property bundles than to break them apart. Parisi
hypothesizes that courts and legislators account for
the presence of asymmetric transaction costs and
correct for problem through the selective use of
remedies and by selecting default rules designed to
minimize the total deadweight losses of property
fragmentation. Parisi ( ) that area. If the Pigouvian tax is imposed on the
offers a
reformulation of the normative Coase theorem in
situations characterized by asymmetric transaction
and strategic costs, such as when complementary
fragments of property are attributed to different
owners. The asymmetry arises from the fact that it
is often harder to reunite separated property bundles
than to break them apart. This variant of the Coase
theorem turns on (a) an initial allocation of
entitlements that minimizes the effects of the
positive transaction costs, and (b) the selection of
legal rules that reduce social welfare losses by
facilitating optimal levels of reunification.
The Coase Theorem and Its Legacy in
Law and Economics
In 1960 Coase entrusted legal and economic
scholars with the challenging task of deriving
Coase Theorem
the implications of his theorem in their areas of
research. Coase’s invitation was taken up by a
number of economists and lawyers who
experimented with the unparalleled analytical
potential of Coase’s theorem in their research.
According to Coase, economists in the Pigouvian
tradition fail to consider the possible reciprocity of
the effects of individual choices. By labelling one
agent as injurer and the other as victim, the
Pigouvian tradition presumes an initial allocation of
rights (Comes and Sandler , p. 59). In such a
manner this approach falls into a serious
methodological error, notwithstanding empirical
psychological studies suggesting otherwise (see
Kahneman et al.
, pp. 1325^18). By taxing the generator of the
externality in a measure corresponding to the
difference between the private cost and the social
cost of his own activity, the followers of Pigou fail
to consider the effects of potential victims’
behaviour. If the social cost of the industrial
emissions is calculated by aggregating the economic
disadvantages of the residents who are negatively
affected by the smoke, the figure will vary with the
number of individuals who fix their residence in
industrial activity only, there will be less incentive
for each resident to consider moving into a different
neighbourhood. New individuals may actually
locate their residence in that area, without
considering the potential increase in the costs
imposed on the industrial activity.
Through these arguments, Coase’s analysis
demonstrates the incapacity of the Pigouvian
approach to consider the interdependence of the
harmful effects generated by individual choices.
Coase’s analysis occasioned a paradigmatic shift in
legal and economic analysis, and, as Henry Manne
once observed, ‘it is hard to imagine law ever again
being free of the influence of the techniques and
findings of objective economic analysis’ ( , p.
4). His theorem, short of
providing a simplistic formula for the social cost
problem, suggests an alternative approach based on
the evaluation of the relative costs of alternative
assignments of rights and legal protection.
Coase, Ronald Harry (Born 1910)
See Also
► lold Problem
► 'rc-peity Rights
Bibliography
Ayres, L, and E. Talley. 1995. Solomonic bargaining: Dividing
a legal entitlement to facilitate Coasean trade. T h e
Y a l e L a w J o u r n a l 104: 1027-1117.
Buchanan, J., and Y.J. Yoon. 2000. Symmetric tragedies:
Commons and anticommons property. J o u r n a l o f
L a w a n d E c o n o m i c s 29: 1-13.
Calabresi, G. 1965. The decision for accidents: An approach
to non-fault allocation of costs. H a r v a r d L a w
R e v i e w 78: 713-745.
Calabresi, G. 1968. Transaction costs, resource allocation and
liability rules: A comment J o u r n a l o f L a w a n d
E c o n o m i c s 11: 67-68.
Calabresi, G., and A.D. Melamed. 1972. Property rules,
liability rules, and inalienability: One view of the cathe-
dral. H a r v a r d L a w R e v i e w 85: 1089-1128.
Cheung, S. 1970. The structure of a contract and the theory of
a non-exclusive resource. J o u r n a l o f L a w a n d
E c o n o m i c s 13: 49-70.
Coase, R.H. 1959. The Federal Communications Commission.
J o u r n a l o f L a w a n d E c o n o m i c s 2: 1-40.
Coase, R.H. 1960. The problem of social cost. J o u r n a l o f
L a w a n d E c o n o m i c s 3: 1-44.
Coase, R.H. 1988. T h e f i r m , t h e m a r k e t , a n d t h e
l a w . Chicago: University of Chicago Press.
Coase, R.H. 1992. The institutional structure of production.
A m e r i c a n E c o n o m i c R e v i e w 82: 713-719.
Coase, R.H. 1993. Law and economics at Chicago. J o u r n a l
o f L a w a n d E c o n o m i c s 36: 239-254.
Cooter, R. 1987. Coase theorem. In T h e n e w P a l g r a v e :
A d i c t i o n a r y o f e c o n o m i c s , ed. J. Eatwell, M.
Milgate, and P. Newman, vol. 1. London: Macmillan.
Comes, R., and T. Sandler. 1986. T h e t h e o i y o f
externalities, public goods, and club
g o o d s . Cambridge: Cambridge University Press.
Demsetz, H.M. 1967. Toward a theory of property rights.
A m e r i c a n E c o n o m i c R e v i e w 57: 347-359.
Demsetz, H.M. 1972. When does the mle of liability matter?
T h e J o u r n a l o f L e g a l S t u d i e s 1: 13-28.
Epstein, R.A. 1993. Holdouts, externalities, and the single
owner: One more salute to Ronald Coase. J o u r n a l o f
L a w a n d E c o n o m i c s 36: 553-586.
Heller, M.A. 1998. The tragedy of the anticommons: Property
in the transition from Marx to markets. H a i v a r d L a w
R e v i e w 111: 621-688.
Kahneman, D., J.L. Knetsch, and R.H. Thaler. 1990. Exper-
imental tests of the endowment effect and the Coase
theorem. J o u r n a l o f P o l i t i c a l E c o n o m y 98:
1325-1348.
Kaplow, L., and S. Shavell. 1996. Property mles versus
liability mles: An economic analysis. H a i v a r d L a w
R e v i e w 109: 723-754.
1733
Kuhn, T.S. 1970. T h e s t r u c t u r e o f s c i e n t i f i c
r e v o l u t i o n s . 2nd ed. Chicago: University of Chicago
Press.
Manne, H.G. 1975. T h e e c o n o m i c s o f l e g a l
r e l a t i o n s h i p s . St. Paul: West Publishing Co.
Manne, H.G. 1993. T h e i n t e l l e c t u a l h i s t o r y o f
George Mason University School of Law.
Arlington: School of Law, George Mason University.
Nutter, G.W. 1968. The Coase theorem on social cost: A
footnote. J o u r n a l o f L a w a n d E c o n o m i c s 16:
503-507.
Parisi, F. 1995. Private property and social costs. E u r o p e a n
J o u r n a l o f L a w a n d E c o n o m i c s 2 : 149-173.
Parisi, F. 2002. Entropy in property. A m e r i c a n J o u r n a l
o f C o m p a r a t i v e L a w 50: 595-632.
Parisi, F. 2006. Entropy and the asymmetric Coase theorem.
In P r o p e r t y r i g h t s d y n a m i c s : A l a w a n d
e c o n o m i c s p e r s p e c t i v e , ed. D. Porrini and G.
Ramello. London: Routledge.
Pigou, A.C. 1920. T h e e c o n o m i c s o f w e l f a r e .
London: Macmillan.
Plant, A. 1974. S e l e c t e d e c o n o m i c e s s a y s a n d
a d d r e s s e s . London: Routledge and Kegan Paul.
Polinsky, A.M. 1989. A n i n t r o d u c t i o n t o l a w a n d
e c o n o m i c s . Boston: Little, Brown and Company.
Regan, D.H. 1972. The problem of social cost revisited.
J o u r n a l o f L a w a n d E c o n o m i c s 15: 427-433.
Salter, J.A. 1921. A l l i e d s h i p p i n g c o n t r o l . Oxford:
Clarendon Press.
Schulz, N., F. Parisi, and B. Depoorter. 2002. Fragmentation
in property: Towards a general model. J o u r n a l o f
Institutional and Theoretical Economics
158: 594613.
Shapiro, F.R. 1985. The most cited law review articles.
C a l i f o r n i a L a w R e v i e w 73: 1540-1554.
Stigler, G.J. 1966. T h e t h e o r y o f p r i c e . 3rd ed. New
York: Macmillan.
Stigler, G.J. 1988. M e m o i r s o f a n u n r e g u l a t e d
e c o n o m i s t . New York: Basic Books.
Wellisz, S. 1964. On external diseconomies and the govern-
ment assisted invisible hand. E c o n o m i c a 31: 345-362.
Williamson, W., and S.G. Winter. 1991. T h e n a t u r e o f
Coase, Ronald Harry (Born 1910)
Steven G. Medema
Abstract
Ronald Coase made seminal contributions to
law and economics and to the theory of the firm,
for which he received the 1991 Nobel Prize. The
importance of understanding the
1734
role of transaction costs in economic activity
and the influence of alternative institutional
structures on economic performance are hall-
marks of Coase’s scholarship, and both the
economic analysis of law and the new institu-
tional economics are outgrowths of Ins work.
Coase occupies a significant although somewhat
controversial place in the history of the Chicago
School of economics.
Keywords
Accounting; Average cost pricing; Bargaining;
Chicago School; Coase conjecture; Coase the-
orem; Coase, R. H.; Cobweb theorem; Con-
sumer theory; Externalities; Firm, organization
of; Government failure; Imperfect competition;
Knight, F.; Law and economics; Lemer, A.;
Marginal cost pricing; Market failure;
Mathematics and economics; Monopoly; Multi-
part pricing; New institutional economies;
Opportunity cost; Posner, R.; Public goods;
Public utilities; Public utility pricing; Rational
expectations; Social cost theory; Special While most economists identify Coase with his two
interests; Transaction costs; Viner, J.
JEL Classifications
B31
Ronald Harry Coase was born on 29 December
1910 in the London suburb of Willesden. He
received the BSc in Commerce from the London
School of Economics in 1932 and while there was
greatly influenced by Arnold Plant, who, as Coase
has said, taught him many of the lessons that later
came to be associated with the Chicago School.
Interestingly, Coase did not take a single economics
course while he was at the LSE, which he suggests
gave him 'a freedom in thinking about economic
problems which [he] might not otherwise have had'
(1990, p. 3).
Upon completing his sUidies at the LSE, Coase
took up a position at the Dundee School of Eco-
nomics and Commerce, where he taught with his
friend and public choice pioneer Duncan Black
from 1932 to 34. Coase moved on to the University'
of Liverpool in 1934-35 before returning to the
LSE, where he remained from 1935 until
Coase, Ronald Harry (Born 1910)
1951. His time at the LSE was interrupted by the
Second World War, during which he served as a
statistician at the Forestry Commission (1940-41)
and in the Central Statistical Office, Offices of the
War Cabinet (1941—46). Coase left the LSI' for the
US and the University of Buffalo in 1951,
remaining there until 1958. Following a year spent
at the Center for Advanced Study in the Behavioral
Sciences at Stanford, he accepted an appointment at
the University of Virginia in 1959.
Although Coase is most closely associated with
the Chicago School, his two most influential works
- ‘The Nature of the Firm’ (1937) and ‘The Problem
of Social Cost’ (I960) - were written before he
arrived at Chicago, in 1964, to teach at the Law
School and to join Aaron Director in editing the
Journal of Law and Economics. Coase retired from
the University of Chicago in 1981 and was awarded
the Nobel Prize in Economics in 1991.
Scholarly Work
classic articles on the film and social costs, his
published output is very extensive and ranges across
topics such as accounting, advertising, public
goods, consumer surplus, public utility pricing,
monopoly theory, blackmail, the economic role of
government, and the history of economic thought.
Several themes appear throughout Coase’s work:
the important role played by institutions - in
particular the film, the market and the law - in
determining economic structure and performance,
the role of transaction costs in economic activity,
the need for a comparative institutional approach to
economic policy, and a distaste for abstract
theorizing. These themes come through
unmistakably in The Finn, the Market and the Law
(1988) and Essays on Economics and Economists
(1994), which, together, collect many of Coase’s
most significant writings.
The lion’s share of Coase's work during the first
part of his career dealt, in one way or another, with
film behaviour and organization. His earliest
contributions analysed the formation of
Coase, Ronald Harry (Born 1910)
producers’ expectations (for example, Coase and
Fowler 1935), using the pig cycle as the case study.
The conventional cobweb theorem explanation for
these cycles was that producers expected current
prices and costs to continue into the future. The
adjustments in supply that resulted then gave rise to
disequilibrium cycles. Coase and Fowler found that
this explanation was incorrect - that producers did in
fact adjust their expectations of prices and costs
very quickly, and that the prediction errors arose
from the difficulty of predicting variations in
demand and in foreign supply. This work was later
cited by J.F. Muth ( , p. 21) in one of his
classic papers on rational
expectations. Coase also collaborated with Fowler
and Ronald Edwards on a series of pieces dealing
with the interrelations between accounting and
economics (for example, Coase 1938; Coase et al.
1938). These writings, which were very much in the
LSE cost tradition, demonstrated that traditional
accounting practices do not adequately capture the
true (opportunity) nature of costs and also pointed to
the problematic nature of designing workable
accounting methods to do so.
Coase also wrote a number of articles dealing
with monopoly and imperfect competition, a few of
which bear mention of here. Two of his theoretical
pieces are of particular import. ‘Durability and
Monopoly’ (1972) demonstrated that a monopoly
firm which produces a good that is infinitely durable
will be forced to sell the good at the competitive
price, unless it can decrease the durability of the
good or make contractual arrangements through
which it promises to limit its production - a result
which has come to be known as ‘the Coase con-
jecture’. ‘The Marginal Cost Controversy’ (1946) is
Coase’s most significant work on monopoly and
deals with public utility pricing and regulation.
Abba Lemer and others had claimed that marginal
cost pricing accompanied by a government subsidy
is the efficient pricing policy for public utilities.
Against this, Coase argued that marginal cost
pricing is inferior to a system of multi-part pricing
and may in fact be inferior to average cost pricing.
This paper, and three related papers that followed it,
are illustrative of one of the central themes in
Coase’s work - that, in assessing the efficiency of
economic outcomes, one must
1735
focus broadly, rather than narrowly, on benefits,
costs, and incentives.
Coase’s work on public utilities also has an
historical strand. Articles on the British Post Office
discuss the rise of the penny postage in Great
Britain under Rowland Hill and the attempts by the
Post Office to enforce its monopoly against
inclusions by private entrepreneurs, including the
messenger companies (for example, 1955). His
study of British broadcasting analyses the devel-
opment of wireless and wire radio broadcasting, as
well as of television broadcasting and the rise of the
BBC as the monopoly supplier of all of the above
(1950, 1954). His interest in the government’s role
in broadcasting carried over to the United States and
an analysis of the role of the Federal
Communications Commission (1959, 1966) in the
allocation of broadcast frequencies. In fact, it was
from this study that ‘The Problem of Social Cost’
came to be written.
While the foregoing gives a sense for the
breadth of Coase’s contributions, it is unquestion-
able that his most influential work is contained in
two papers - ‘The Nature of the Firm’ (1937) and
‘The Problem of Social Cost’ (1960), the two works
cited by the Royal Swedish Academy in awarding
Coase the Nobel Prize. In the former, Coase set out
to explain why firms exist and what determines the
extent of a firm’s activities. He found the answer in
a concept to which most economists had until
recently paid scant attention - transaction costs.
Coase suggested that we tend to see firms emerge
when the cost of internal organization is lower than
the cost of transacting in the market, and that the
limit of a firm’s activities (or, the extent of internal
organization) comes at the point where the cost of
organizing another transaction internally exceeds
the cost of transacting through the market. Although
published in 1937, ‘The Nature of the Firm’
attracted little attention until the early 1970s, when
Oliver Williamson, Armen Alchian, Harold
Demsetz and others began to build on or take off
from Coase’s contribution to bring transaction
costs, the contracting process, and firm organization
to the fore in economic analysis.
‘The Problem of Social Cost’ took the
transaction-cost paradigm in a different direction -
1736
the legal-economic arena and situations of conflicts
over rights. Although ‘The Problem of Social Cost’
is one of the most cited articles in all of the
economics and legal literatures, it has also been
widely misunderstood. From this paper comes the
now-famous Coase theorem - actually codified as
such by George Stigler ( ) - which
says that when transaction costs are zero and rights
are folly specified, parties to a dispute will bargain
to an efficient outcome, regardless of the initial
assignment of rights. But Coase recognized that the
transaction costs are pervasive and will generally
preclude the working of this bargaining mechanism.
Coase thus concludes that legal decision-makers
should assign rights so as to maximize the value of
output in society - a concept that lies at the heart of
the modem law and economics movement (Medema
; Medema and Zerbe ).
The crux of ‘The Problem of Social Cost,’
however, is Coase’s attempt to demolish the
Pigovian tradition of social cost theory (Pigou ).
The analysis that came to be known as the Coase
theorem was used to demonstrate that, under
standard neoclassical assumptions, Pigovian
remedies for externalities are unnecessary:
costlessly functioning markets, like the costlessly
functioning governments of Pigovian welfare
theory, will generate efficient outcomes. The
problem, as Coase pointed out, is that neither
markets nor governments function costlessly, and
thus neither will generate optimal solutions. This
leaves policymakers with a choice among imperfect
alternatives, and Coase advocates a close
examination of the benefits and costs associated
with the alternative policy options, in order to
facilitate the adoption of policies (including doing
nothing at all) which maximize the value of output.
That government failure is at least as pervasive
as market failure, and that economists are too quick
to advocate tax, subsidy, and regulatory solutions
without a careful examination of the situation, are
recurring themes in Coase’s work. His analyses of
social cost issues, public utility pricing, and his
classic article on role of the lighthouse in public
goods theory as against the actual history of private
lighthouse provision in Great
Coase, Ronald Harry (Born 1910)
Britain (1974) are excellent examples of Coase’s
position here. When Coase looks at government, he
sees agencies captured by special interests, making
policies that usually make matters worse rather than
better, and operating in virtual ignorance of the
virtues of the market. Yet a careful reading of Coase
suggests that he is not ‘anti- government’ but,
rather, an advocate for economic theorizing and
policymaking which recognizes that policy choices
are always among imperfect alternatives.
These criticisms are part of Coase’s more gen-
eral concern about the way that economists practice
their trade (1994). He is suspicious of consumer
theory as a whole and of the way in which
mathematical and quantitative techniques have been
used in modem economics. His own writings
evidence some graphs and some technical intuitive
analysis, but, reflecting Coase’s lifelong distaste for
using mathematics in his work, there is not an
equation to be found. Coase believes that
economists are obsessed with what he calls ‘
blackboard economics’, an economics where curves
are shifted and equations are manipulated on the
blackboard, with little attention to the
correspondence (or lack thereof) between these
models and the real-world economic system. This,
he says, has manifested itself in economists’
ignorance of the role played by transaction costs
and economic institutions generally, and in an
approach to public policy that fails to examine in
any kind of depth the consequences of alternative
policy actions.
Coase and Chicago
Coase’s critical attitude toward the practice of
economics does not stop at the doors of the Uni-
versity of Chicago. Indeed, his close association
with the Chicago School belies a degree of tension
in the relationship and highlights the risks involved
in thinking in terms of a homogeneous Chicago
school. In spite of his position as a founding father
of law and economics and, by extension, the
expansion of the boundaries of economics so
closely associated with Chicago, Coase has been
critical of economic imperialism
Coase, Ronald Harry (Born 1910) 1737

generally and of the economic analysis of law in 1946. The marginal cost controversy. Economica,
particular (Coase 1977, 1993). Coase’s interest is New Series 13: 169-182.
not the economic analysis of law, but rather the 1 95 0. British broadcasting: A study in monopoly.
study of how the legal system impacts the economic L on do n: L on gm an s , G r e e n an d C o.
system - old-style Chicago law and economics of 1954. The development of the British television
the sort being published in the Journal of Law and service. Land Economics 30: 207-222.
Economics in the 1960s and 1970s. As such, his 1955. The postal monopoly in Great Britain: An
interest and intellectual commonalities lie much historical survey. In Economic essays in com-
more with the older Chicago school of Frank memoration of the Dundee School of Economics
Knight and Jacob Viner than with the Becker- 1931-1955, ed. J.K. Eastham. London: William
Stigler-Posner generation, and he has a much Culcross and Sons.
greater interest in the new institutional economics1959. The Federal Communications Commission.
(of which he is also regarded as a founding father) Journal of Law and Economics 2: 1-40.
than in the modem economic analysis of law 1960. The problem of social cost. Journal of Law
movement a la Richard Posner. Coase has been and Economics 3: 1-44.
chastised by Posner ( ) on 1966. The economics of broadcasting and govern-
this and other counts, but he remains unapologetic. ment policy. American Economic Review 56:
That Coase has a place within the Chicago tradition 440-447.
goes without saying, but he has also remained his 1972. Durability and monopoly. Journal of Law and
own man - dissenting from the received doctrine Economics 15: 143-149.
when it did not fit with his views. 1974. The lighthouse in economics. Journal of Law
and Economics 17: 357—376.
1977. Economics and contiguous disciplines. In
The organization and retrieval of economic
knowledge, ed. M. Perlman. Boulder: Westview
See Also Press.
1988. The firm, the market, and the law. Chicago:
► Ciicagc
University of Chicago Press.
► Chicago ■
1990. Accounting and the theory of the firm.
► Coas° Th eorem
Journal of Accounting and Economics 12: 3-13.
1992. The institutional structure of production.
American Economic Review 8 2: 7 13 - 71 9.
1993. Law and economics at Chicago. Journal of
Selected Works Law and Economics 36: 239-254.
1994. Essays on economics and economists.
1935. (With R.F. Fowler.) Bacon production and Chicago: University of Chicago Press.
the pig-cycle in Great Britain. Economica, New
Series 2: 142-167. Bibliography
1937. The nature of the firm. Economica, New
Series 4: 386^105. Medema, S.G. 1994. R o n a l d H . C o a s e . London/New
1938. Business organisation and the accountant. In York: Macmillan/St. Martin’s Press.
Medema, S.G. 1999. Legal fiction: The place of the Coase
Studies in costing, ed. D. Solomons. London: theorem in law and economics. E c o n o m i c s a n d
Sweet and Maxwell, 1952. P h i l o s o p h y 15: 209-233.
1938. (With R.S. Edwards and R.F. Fowler.) Medema, S.G., and R.O. Zerbe Jr. 2000. The Coase theorem.
Published balance sheets as an aid to economic In The encyclopedia of law and
e c o n o m i c s , ed. B. Bouckaert and G. De Geest
investigation'. Some difficulties. London: Aldershot: Edward Elgar Publishing.
Accounting Research Association Publication
No. 3.
1738
Muth, J.F. 1961. Rational expectations and the theory of price
movements. E c o n o m e t r i c a 29: 315-335.
Pigou, A.C. 1932. T h e e c o n o m i c s o f w e l f a r e . 4th
ed. London: Macmillan.
Posner, R.A. 1993. Ronald Coase and methodology. J o u r -
n a l o f E c o n o m i c P e r s p e c t i v e s 7 (4): 195-210.
Stigler, G.J. 1966. T h e t h e o r y o f p r i c e . 3rd ed. New
York: Macmillan.
Williamson, O.E. 1975. Markets and hierarchies: Analysis and
antitrust implications. New York: Free Press.
Cobb-Douglas Functions
Murray Brown
Abstract
Perhaps the most common form of production
function in economics, the Cobb-Douglas
function has a range of attractive properties. The
input demand and supply of output functions
have the property of continuous differentiability
everywhere on their respective domains; and the
form has a function coefficient that is identical
to its degree of homogeneity, calculated by
summing the factor production elasticities. Its
restrictions have made it an object of disdain for
some. But the Cobb-Douglas form is remarkably
robust in a vast variety of applications and is
therefore very likely to endure.
Keywords
Aggregation (production); CES production
function; Cobb, C.; Cobb-Douglas functions;
Douglas, P. H.; Elasticity of substitution; Factor
substitution; Frontier production functions;
Production functions; Technical change; Walras,
L.; Wicksell, J. G. K.; Wicksteed, P. H.
JEL Classifications
E23
The Cobb-Douglas function is perhaps the most
ubiquitous form in economics, owing its popularity
to the exceptional ease with which it can be
Cobb-Douglas Functions
manipulated and to the fact that it possesses the
minimal properties that economists consider
desirable. It appeared early (at least by 1916; see
Wicksell , p. 133), notably in the theory of
distribution where it was used to prove the adding-
up theorem of factor shares when the production
elasticities sum to unity. It is the first form that
many embryonic mathematical economists squeeze
and buffet to obtain nice expressions for marginal
products and utilities. It has been applied
econometrically countless times, still surprising
people that it can explain the data so well
(Mairesse ). It forces itself into relatively new areas
such as frontier production functions (see Forsund
et al. ). And it has been used both as a utility and
production function in analyses of growth,
development, macroeconomics, public finance,
labour and just about any other applied area in
economics. Yet it possesses restrictive properties
and perhaps for that reason it has become for some
an object of disdain, often regarded as a child’s toy
in the world of real economics. But for others, the
Cobb-Douglas is at least a venerable form and,
effectively, it and its putative inventor are regarded
fondly.
In its unrestricted form, the Cobb-Douglas can
be written as f(x) = i=lxf, where A is an efficiency
parameter, a, is the elasticity of/(x) with respect to
x,, and x is confined to /?"++. Defining the x, as goods
consumed, it has been used as a utility function;
defining them as inputs in the production process, it
is a production function; as normalized prices, it is
an indirect utility function; and so on. We focus
here on its use as a production function for a single
output.
A large part of the appeal of the form stems
basically from the fact that if 0 < a, < l,./(x) is
strongly pseudo-concave on its domain. That entails
that if the firm is a profit maximizer and factor
supply and product demand functions are
continuously differentiable on their domains, then
the input demand and supply of output functions
have the immensely useful property of continuous
differentiability everywhere on their respective
domains. Also, if Y.,a, < 1 and if factor supply and
product demand functions are well-behaved, the
input demand functions are downward sloping with
respect to own price and the output supply
Cobb-Douglas Functions
function does not slope downward with regard to
product price. What could be better and, moreover,
it is all so simple to demonstrate.
Another attractive property of the form is that it
has a function coefficient that is identical to its
degree of homogeneity, calculated by summing the
factor production elasticities. Thus, 52,-a,- 1
for all i easily and succinctly characterizes
decreasing, constant and increasing returns to scale,
respectively. This characteristic also has important
implications for the cost, profit and revenue duals of
the production function. For example, the cost
function of a price-taking firm which has a Cobb-
Douglas technology decomposes into two parts, one
a linear homogeneous function of factor prices and
the other a function of output cj, that is C(q, w) = B\
{ i=iwfqc0, where if is a positive constant, w is a
(positive) price vector of the inputs, ct = cq/yZ^i and
CO = 1 fEjClj.
The list of attractive properties extends to the
aggregation problem since the Cobb-Douglas is
homogeneous and weakly separable. First consider
the question of aggregation across inputs. Suppose
one can write a generalized Cobb-Douglas function
as follows:
«= n( nv ? ) r '
where by = ciy/ZZ/i.y, Ys = 'iZjasj , Js is the number
of factors in the sth group, S is the number of
groups, 5=1, 2,..., S, and j = 1, 2,..., Js. Notice that
Yjjbsj = 1. Since each expression in the parentheses
is homogeneous of degree one for each s, the profit
maximization procedure can be decomposed into
two stages and there exist quantity and price
indexes (call them xs and Ws, respectively) such that
the expenditure on the sth group is WsXs for s =
1,2,..., S.
With respect to aggregation across firms, sup-
pose the rth firm’s production function were
q, = Arx\l;xcz;,
where = 1 and i = 1, 2, ..., R. It is evident that the
expansion paths for all firms are straight lines
through their respective origins. Then under
1739
the extremely restrictive conditions that the
expansion paths for each firm are parallel (i.e. if cit.
= cit = Cj for each i and for all r and /), and that the
first order conditions are satisfied, the R functions
consistently aggregate to
a nicely behaved aggregate production function.
There is another way to look at the aggregation-
across-firms problem that involves the Cobb-
Douglas function. Suppose that factors in each firm
are used in fixed proportions with the Leontief
coefficients being distributed across all firms
according to a Pareto distribution. Then a surprising
result by Houthakker (see Sato ) i s that the
aggregate production function of the industry is a
Cobb-Douglas form.
Of course, there is a price for these desirable
implications and most of it is owing to the fact that
the Cobb-Douglas technology entails that the elas-
ticity of substitution takes on the knife-edge value
of unity. If there is no technological change, a unit
substitution elasticity implies that the income shares
of all factors of production remain constant in the
face of changes in things that are deemed germane
such as saving, the rate of growth of the economy
and relative factor supplies. Only the state of the
technology matters in this instance, a highly
disputable outcome. When technological change is
allowed to proceed in a Cobb-Douglas world, it is a
fact that Hicks-, Solow- and Harrod- neulral
technological change are equivalent, thus blurring
these distinctions. Another implication of the unit
substitution elasticity of the (linear homogeneous)
Cobb-Douglas form is that, used in growth models,
it guarantees the existence and stability of
equilibrium growth, again obscuring an important
problem in economics.
Furthermore, it is a fact that the Cobb-Douglas
form requires that each factor of production be
essential in the sense that no factor may be
completely substituted for another. Hence the
domain of the function must be confined to the set
of strictly positive real numbers. This is not
particularly disturbing for situations in which the
factors can be taken to be large aggregates but it
does limit the analysis in other contexts.
1740
Technological change is represented in the
Cobb-Douglas by changes in the efficiency
parameter A which are Hicks neutral, by changes in
the scale of the factor inputs which are factor
augmenting and also Hicks neutral, and by changes
in the elasticities of production which may be Hicks
non-neutral. However, the unit elasticity of
substitution is restrictive in still another way: it
cannot represent a technological advance that
results in a change in the ease of substitution among
factors of production.
What is the form’s provenance? It is generally
attributed to Paul Douglas and although he grace-
fully acknowledged (Douglas ) that
Wicksteed and Walras were cognizant of it, he
neglected to add Wicksell’s name to the list. Be that
as it may, Douglas relates in his gentle comments
that in 1927 he asked a professor of mathematics,
Charles Cobb, to devise a formula that could be
used to measure the comparative effect of each of
two factors of production upon the total product to
satisfy a linear log-log relationship in his input and
output data. His work encountered a host of
theoretical concerns (see Brown for a discussion)
aside from the capital, output and labour measures
for which he was faulted. But the production form
remained in spite (or perhaps because) of its
restrictive properties.
Subsequent work has demonstrated that the
Cobb-Douglas is a special case of a variety of forms
and approaches. The constant elasticity of
substitution (CES) production function is perhaps
the most well known of the forms that yield the
Cobb-Douglas as a special case, either by using
L’Hopital’s rule when the elasticity of substitution
goes to unity or it can be derived from certain
expressions used in deriving the CES function (see
Brown and De Cani ). Parenthetically, the CES,
itself, is known to mathematicians as a
mean of order t [i.e. (XI \ I 'I) 'for TF=- 0] so that, if
A X
one takes the limit as > 0, of course, the Cobb-
Douglas emerges. Also, it can be derived from the
translog production form (Christensen et al. ) and
many others, besides, by judiciously restricting
certain parameters. A different approach to the
derivation of the Cobb-Douglas form has been
taken by P. Zarembka ( ), who
Cobb-Douglas Functions
specifies each variable as z(A) = (zA - 1 )/A for A / 0
and z(A) In z for A 0. Then, applying this
transformation to the production function, we would
have z0 = q and z;- = x, for all i. Thus, if the z/£ (k =
0, 1, ..., n) are related linearly, the transformation
turns out to be a useful procedure in econometrics to
treat the general problem of functional form, an
important special case of which is the Cobb-
Douglas.
In sum, though it is restrictive and sometimes
regarded as an economic toy, the Cobb-Douglas
form is remarkably robust in a vast variety of
applications and that it will endure is hardly in
question.
See Also
Bibliography
Brown, M. 1966. On the theory and measurement of tech-
nological change. Cambridge: Cambridge University Press.
Brown, M., and J. De Cani. 1963. Technological change and
the distribution of income. International
E c o n o m i c R e v i e w 4 : 289-309.
Christensen, L.R., D.W. Jorgenson, and L.J. Lau. 1973.
Transcendental logarithmic production frontiers.
R e v i e w o f E c o n o m i c s a n d S t a t i s t i c s 55: 28-
45.
Douglas, P.H. 1948. Are there laws of production?
A m e r i c a n E c o n o m i c R e v i e w 38: 1-41.
Douglas, P.H. 1967. Comments on the Cobb-Douglas
production function. In T h e theoiy and
e m p i r i c a l a n a l y s i s o f p r o d u c t i o n , edited by.
M. Brown, National Bureau of Economic Research,
Studies in Income and Wealth No. 31. New York:
Columbia University Press.
Forsund, F.R., C.A.K. Lovell, and P. Schmidt. 1980. A survey
of frontier production functions and of their relationship
to efficiency measurement. Journal of
E c o n o m e t r i c s 13: 5-25.
Mairesse, J. 1974. C o m p a r i s o n o f p r o d u c t i o n
f u n c t i o n e s t i m a t e s . Paris: Institut National de la
Statistique et des Etudes Economiques.
Sato, K. 1975. P r o d u c t i o n functions and
a g g r e g a t i o n . Amsterdam: North-Holland.
Wicksell, K. 1958. S e l e c t e d p a p e r s o n e c o n o m i c
t h e o r y , ed. Erik Lindahl. Cambridge, MA: Harvard.
Cobbett, William (1763-1835)
Zarembka, P. 1987. Transformation of variables in econo-
metrics. In T h e n e w P a l g r a v e : A d i c t i o n a r y ’
o f e c o n o m i c s , ed. J. Eatwell, M. Milgate, and P.
Newman, vol. 4. London: Macmillan.
Cobbett, William (1763-1835)
N. W. Thompson
William Cobbett was bom, appropriately, at the
‘Jolly Farmer’ in Famham, Surrey, in 1763. He was
by turns soldier, clerk, teacher, joumahst and
pohtical agitator but whether in his early literary
career as an anti-Jacobin pamphleteer or later in his
role as combative radical his voice was that of the
small farmer threatened by the forces of economic
and social change which characterized the early
phases of Britain’s industrial revolution.
Cobbett’s acerbic castigation of its major the-
orists shows he had little time for political economy
and he would certainly have greeted with a wry
guffaw his inclusion in a dictionary of economic
thought. Nevertheless it was the case that much of
his writing in the Twopenny Register (1816-20) and
works such as Paper against Gold (1815) was given
over to a discussion of economic questions and in
particular the economic difficulties that confronted
Britain in the post-Napoleonic war period, and even
in his Rural Rides (1830) few opportunities were
lost of discoursing on matters economic.
For Cobbett material impoverishment was a
consequence of the pohtical decisions affecting
taxation, the Debt, the convertibility of the currency
etc., which emanated from a Parliament corrupted
by the influence of tax-gatherers, ‘Change Alley
men, sinecurists, placemen, Jews and Borough-
tyrants. Crucial here was the passage of the Bank
Restriction Act of 1797, which in suspending cash
payments by the Bank of England had created a
‘Paper money system’ that robbed the ‘industrious’
of their ‘earnings’ producing ‘that monster in civil
society, starvation in the midst of abundance’. This
paper money system was seen in
1741
itself as a means of appropriating the product of
labour but most importantly, with the return to cash
payments in 1819, it created a situation where the
nation was compelled to pay in an appreciated
medium of exchange debts contracted in a depre-
ciated paper currency. The industrious were there-
fore forced to pay in gold what had been contracted
in paper. Thus both the inflation of the Revolution-
ary and Napoleonic wars and the deflation that
followed were seen by Cobbett as redistributing
wealth in favour of the idle generally and the
fundholders and bankers in particular. Justice
demanded, therefore, that there should be an equi-
table adjustment in the level of taxation to take
account of the appreciation in the value of money.
For this, parliamentary reform and an end to polit-
ical corruption were necessary prerequisites.
Under the undifferentiated heading of ‘The
Thing’, Cobbett sought to attack all those who
jeopardized his vision of a stable, hierarchically
structured, rural economy and society, dominated by
the independent yeoman farmer and free from ‘the
all-devouring Jew and tax-eater’. Specifically, he
condemned ‘the Funding and Manufacturing and
Commercial and Taxing System’ for their tendency
to draw ‘wealth into great masses ... [and] man also
into great masses’, like ‘the Great Wen’ [London],
which both corrupted its inhabitants and
impoverished the country as a whole.
Cobbett’s was an anti-industrial, anti-
commercial, anti-urban and anti-City pohtical
economy with its ideological roots in the pohtical
radicalism of the eighteenth century and with
Paine’s Decline and Fall of the English System of
Finance (1797) as its basic text. Yet if, with the
growth of industrial capitalism, this ‘old corruption’
critique became increasingly irrelevant and
inadequate; if by the 1830s it was being supplanted
in the working-class press by anticapitalist and
socialist analysis of contemporary ills which
deployed to critical effect the tools and concepts of
pohtical economy, this should not obscure the
remarkable longevity of Cobbettian ideas. Thus
even a superficial survey of Chartist literature will
show just how indelible was the mark left by the
Twopenny Register upon a generation of pohtical
radicals. In a sense too Cobbett’s vision of a pastoral
England peopled
1742
by free-born, hearty independent cultivators will
remain indestructible, drawing sustenance as it does
from an ineradicable human craving for the
permanence, the certainty, the order and the sta-
bility which rural self-sufficiency seems to offer.
Selected Works
1802-36. Cobbett spolitical register. The paper was
published throughout this period but the cheap,
popular post-Napoleonic war edition of the
paper referred to as Cobbett’s Twopenny
Register ran from 12 October 1816 until 20 July
1820.
18 15 . Paper against Gold and Glory against
prosperity. L on do n.
18 30 . Rural rides in the counties of Surrey, Kent,
Sussex___Lo nd on .
References
Beer, M. 1953. .1 h i s t o t y o f B r i t i s h s o c i a l i s m , vol.
2. London: Allen & Unwin.
Cole, G.D.H. 1947. T h e l i f e o f W i l l i a m C o h h e t t ,
3rd ed. London: Home & van Thai.
Cole, G.D.H. 1977. .1 history of socialist thought. 5 vols. Vol. 1,
Socialist thought: the Forerunners, 1789-1850. London:
Macmillan.
Green, D. 1983. Great ('obbett: The Noblest agitator.
Oxford: Oxford University Press.
Sambrook, J. 1973. W i l l i a m C o b b e t t . London:
Routledge & Regan Paul.
Spater, G. 1982. William Cobbett: The poor man's friend.
vol. 2. Cambridge: Cambridge University Press.
Thompson, N.W. 1984. T h e p e o p l e ' s s c i e n c e : T h e
popular political economy of exploitation
a n d c r i s i s . Cambridge: Cambridge University Press.
Cobden, Richard (1804-1865)
William D. Grampp
Keywords
Anti-Corn Law League (UK); Hank of England;
Bright, J.; Cobden, R.; Colonialism; Corns
Laws; Factory Acts; Free trade;
Cobclen, Richard (1804-1865)
Manchester School; McCulloch, J. R.; New
Poor Law; Pacifism; Tooke, T.
JEL Classifications
B31
Cobden led the campaign that repealed the Com
Laws in 1846, after which there was free trade in
grain. The son of a Middlesex farmer, he sought his
fortune in Manchester, became an owner of a mill
that employed 2,000 workers and was noted for
excellence of its calicoes. At 35, he was a rich man.
His calling, however, was politics. After taking
part in the successful effort to incorporate Man-
chester, he entered the movement against the Com
Laws in 1838. Until then it had been conducted by
middle class radicals and various business interests,
among them the Manchester Chamber of
Commerce. Cobden, John Bright, and others like
them wanted to enlarge the movement, make it bold
and uncompromising. They were exasperated by the
businessmen who so wanted to look respectable that
they could not see where their interest lay. Thomas
Tooke had said the same about the London
merchants, when on their behalf he drafted the
celebrated petition of 1820 for free trade and they
were reluctant to sign it.
The militants of Manchester formed the
National Anti-Corn Law League and agitated for
free trade up and down the country. They become
known as the Manchester School of Economics and
were celebrated as arch advocates of laissez- faire.
AcUially they were a coalition of diverse interests
that agreed on only one issue - repeal of the Com
Laws - and each did so for its particular reasons.
Cobden's reason was peace. He believed free
trade would break down national barriers and give
everyone a material interest in avoiding war. This
was not an argument gotten up for the occasion but
the expression of a view he had long held. When
young he wrote two long tracts on foreign policy
which denounced alliances among nations and
political engagements of all kinds, decried the idea
of a balance of power, was especially disapproving
of colonics, then went on to extol
Cobweb Theorem
free trade as the way to peace and its guarantor.
Years later, after he and Bright had brought down
the Com Laws, he told him, ‘I have always had an
instinctive monomania against the system of foreign
interference, protocolling, diplomatising, etc.’
That scarcely expressed the horror he had of
violent action, even the suggestion of it. When the
southern states of America seceded, he thought
Lincoln was wrong in bringing the issue to battle
although he had no sympathy with them (except
their fondness for free trade). He was shocked by
the massacres in India and was opposed to wars of
independence and to revolution. He thought duel-
ling was barbarous, was against capital punishment,
objected to boxing, couldn’t stand brass bands, and
asked the Pope to prohibit bull fighting in Spain. He
favoured free trade so long as its effect was
peaceful, as he believed it usually was, but when he
believed it was not he quickly put it aside. He
opposed the sale of foreign bonds in the London
market if the proceeds were to be used to buy arms.
‘No free trade in cutting throats’, he said.
Pacifism, not laissez-faire, was Cobden’s
guiding principle; and he applied laissez-faire less
to domestic than to foreign markets. He did not care
for the Factory Acts but only spoke, never voted,
against them. He approved of increasing the
monopoly powers of the Bank of England and of
regulating aspects of railway construction. He had
no use for the New Poor Law, of which most
economists of the day approved, and spoke
derisively of McCulloch’s ‘usual dogmatism’. But
he carefully read the latter’s edition of the Wealth of
Nations and wrote in the margins of the chapters
that moved him. His notes are especially lively
where Smith condemns the colonial policy of Great
Britain. However, where he describes the operation
of the invisible hand, the margins are quite
untouched.
See Also
► Nv. T : ;>ni
► J.GI v
1743
Selected Works
1835. England, Ireland and America, by a Man-
chester Manufacturer. L on do n.
1836. Russia, by a Manchester Manufacturer.
E d i nb u rg h.
1 84 9. Speeches on Peace, Financial Reform and
Other Subjects. L on do n.
1867. Political Writings. London: Ridgeway.
1868. Speeches on Questions of Public Policy, 2
vols, ed. J. Bright and J.E. Thorold Rogers.
Oxford: Oxford University Press.
Bibliography
Ashworth, H. 1877. Recollections of Richard Cobden, M.P., and
the Anti-Corn Law League. London: Cassell. Morley, J. 1881.
The Lije of Richard Cobden. London: Fisher Unwin.
Ritchie, J.E. 1865. T h e L i j ' e o f R i c h a r d C o b d e n .
London: Ward and Lock.
Cobweb Theorem
B. Peter Pashigian
Abstract
The cobweb theorem purports to explain
persistent fluctuations of prices in selected
agricultural markets. It was first developed in
the 1930s under static price expectations where
the predicted price equalled actual price in the
last period. Muth’s rational expectations
hypothesis posited that forecast errors will not
be serially correlated and the pattern of past
forecast errors cannot be used to improve the
accuracy of the forecasts. The fundamental
question of whether observed price cycles are
better explained by systematic errors in price
forecasts or by the cumulative impact of
unpredictable shocks has not as yet been
definitively addressed.
1744
Keywords
Adaptive expectations; Cobweb theorem;
Expectations; Kaldor, N.; Price cycles; Price
expectations; Rational expectations
JEL Classifications
Qll
The persistent fluctuations of prices in selected
agricultural markets have attracted the attention of
economists from time to time, and the theory of the
cobweb was developed to explain them. The theory
is applicable to those markets where production
takes time, where the quantity produced depends on
the price anticipated at the time of sale, and where
supply at time of sale determines the actual market
price.
One strand of the cobweb literature (the term
was coined by Kaldor ) concentrates on how
expectations are formed and the effect of the price
expectations mechanism on the stability of equi-
librium. Cobweb theory was first developed under
static price expectations where the predicted price
equalled actual price in the last period. The cobweb
theorem proved that the market price would (not)
converge to (long-run) equilibrium price if the
absolute value of the price elasticity of demand was
greater (smaller) than the price elasticity of supply.
This stability condition was modified later as more
sophisticated expectations models were adopted.
Early articles by Tinbergen, Ricci and Schultz
appeared in 1930 in German (see Waugh , for a
review of this literature). Ezekiel’s important article
( ) spells out in
greater detail the conditions for convergence,
divergence or perpetual oscillation and shows how
cycles of different lengths could be generated under
static expectations.
Why the theory was developed in the 1930s and
not earlier is a bit of mystery, for recurring price
cycles for some agricultural products had been
reported by agricultural economists for some time.
Economists may have been attracted to the cobweb
theory in the 1930s because of the events of the
Depression. A theory that explained both oscillation
and long departures from stationary equilibrium
was more attractive after the
Cobweb Theorem
events of the Depression. The fact that Ezekiel’s
paper was reprinted in the 1944 American Eco-
nomic Association volume on business cycles lends
credence to this view.
The impression left by Ezekiel and subsequent
contributors is that the cobweb theory is a valuable
tool for explaining price cycles. Ezekiel was aware
of the simplicity of static expectations and not
unmindful of the importance of shocks on the
demand and the supply sides of the market in
causing aberrant price fluctuations (for example,
weather and the randomness of yields). Even so,
agriculture economists, who were presumably more
familiar with price fluctuations in agricultural
markets, have been more prone to accept the theory,
while other theorists have given the theory more of
a mixed reception.
The price expectations mechanism has under-
gone many refinements over the years. In 1958
Nerlove proposed the use of adaptive expectations.
This suggestion is motivated by the findings of
econometric studies which showed the price
elasticity of demand to be less than the price
elasticity of supply for many agricultural goods.
Under these conditions the static expectations
version of the cobweb model predicts a price cycle
of increasing amplitude. However, the observed
price cycles in agricultural markets showed no sign
of being explosive. Nerlove attempted to reconcile
theory with evidence and to show that convergence
is possible under a broader set of conditions
provided expectations are adaptive. During the
1930s the attractiveness of the cobweb model
seemed to be in its ability to explain persistent or
even explosive price cycles. By the late 1950s these
were no longer attractive features, and Nerlove felt
compelled to offer an explanation of why price
cycles of increasing amplitude are not observed
even when demand elasticities are smaller than
supply elasticities. Waugh ( ) took a different tack
and attempted
to reconcile the theory with the evidence of stable
price cycles by suggesting that the price elasticity of
supply becomes smaller (larger) than the price
elasticity of demand at prices well above (below)
the long-run equilibrium price. Under this
assumption, a stable price cycle will eventually be
reached.
Cobweb Theorem
The length of the cobweb price cycle is deter-
mined by the length of the production process. If it
takes one year to bring a fattened hug to the market,
then the complete price cycle should take two years.
At first, little attention and superficial explanations
were given to explain why the predicted length is
often shorter than the actual length of the price
cycle. It was left to the critics to point out these
discrepancies.
The critics are responsible for the other strand of
the literature. They appeared early but were not
very influential at first although their criticisms
were ultimately given more weight. The critics
questioned the rationality of using an arbitrary
expectations mechanism by otherwise profit-
maximizing agents, and pointed out that the theory
implies that producers would expect to lose wealth
if they entered and remained in an industry with a
cobweb price cycle. In a perceptive article on the
pig cycle in England, Coase and Fowler ( ) Nerlove’s suggestion did rectify one limitation of
questioned the realism of static expecta
tions. They showed that the price of a bacon
(mature) pig less the cost of feeding for the next
five months and less the cost of a feeder (young)
pig, which would be stable in a competitive market
if farmers had static expectations, fluctuated over
time. Hence the empirical evidence contradicted the
assumption of static expectations. They presented
evidence that pig breeders reacted quickly to a
change in expected profits, and this implied that the
pig price cycle should be only two years instead of
the observed four-year period. The fluctuation in
the profits per pig was attributed to the difficulty of
predicting both demand and foreign imports. The
Coase-Fowler paper advanced, if only in faint
outline, the essence of the rational expectations
hypothesis which was to blossom some 35 years
later. They hinted that anticipated prices would not
be formed in a mechanistic way because profits
would be higher the more accurate are the forecasts.
Prediction errors were due to the difficulty of
predicting shifts in demand and in foreign supply.
Buchanan’s paper ( ) criticized the cobweb
model because it implied that producers suffer
aggregate losses over the price cycle when output is
determined by the long-run supply curve. He
pointed out that the theory was based on the
1745
dubious assumption of a continued supply of
entrepreneurs standing ready to dissipate their
capital. The critics were also disturbed by the
ambiguity of whether the supply curve is of the
short-or long-run variety, and the failure to clarify
how the adjustment from the short-run to the long-
run supply curve is made. These early criticisms
and ambiguities aside, references to the cobweb
theory continued to appear in textbooks.
Nerlove’s paper ( ) briefly rekindled the
controversy. His purpose was to resurrect the theory
and show that it could explain price behaviour if
adaptive expectations were employed. Mills ( )
criticized the use of adaptive and other
autoregressive expectations mechanisms in the
deterministic model because they implied a simple
pattern of forecast errors that producers could
detect, incorporate into their forecasts and thereby
improve the accuracy of their price forecasts. While
the cobweb theory, it did not address the critical
issue of why producers relied on any particular
forecasting mechanism. Muth ( )
developed the implications of rational expectations
for cobweb theory in his now famous paper. Muth
postulated that expectations were the predictions of
the economic structure of the market and
incorporated all available information. Under
certain conditions the predicted price equals the
conditional expectation of price, given currently
available information. Adaptive expectations can be
rational only under special conditions, and the
coefficient of adaptation is determined by the values
of the slopes of the demand and supply curves.
The rational expectations formulation has
powerful implications for cobweb theory. If the
price forecasts incorporate all available information
and are on average correct, then forecast errors will
not be serially correlated and the pattern of past
forecast errors cannot be used to improve the
accuracy of the forecasts. Moreover, what is then
left of the supposed ability of the cobweb theory to
explain the cyclical behaviour of prices? Price
fluctuations would have to be explained either by
the cyclical pattern of exogenous variables or by the
summation of random shocks (Slutsky ). Muth’s
paper represents a
1746
frontal attack on the traditional cobweb model. He
notes that the traditional model tends to predict a
shorter price cycle than is observed and indicates
that the rational expectations version predicts a
longer price cycle.
Interest in the cobweb model has ebbed in
recent years and few articles on it have appeared in
the major journals. Economists have found it more
rewarding to apply the rational expectations
hypothesis to areas like monetary or business- cycle
theory than to the study of particular markets, even
though the analysis of markets with inventories
raises issues that are just as difficult and subtle. The
question of whether the cobweb does or does not
explain price cycles has not really been resolved.
Freeman ( ) has suggested that
the traditional cobweb model explains cycles in the
markets for lawyers, physicists and engineers. Tests
of the rational expectations hypothesis have been
suggested by Pashigian ( ) when expec
tations data are available and by Hoffman and
Schmidt ( ) when expectations data are
unavailable. So the methodology exists for
distinguishing between the competing hypotheses.
Few econometric tests have been made of the
rational expectations hypothesis in markets where
the assumptions of the cobweb model apply. The
fundamental question of whether observed price
cycles are better explained by systematic errors in
price forecasts or by the cumulative impact of
unpredictable shocks has not as yet been
definitively addressed.
See Also
► Rational Expectations
Bibliography
Buchanan, N. 1939. A reconsideration of the cobweb theorem.
J o u r n a l o f P o l i t i c a l E c o n o m y 47: 67-81.
Coase, R.H., and R.F. Fowler. 1935. Bacon production and the
pig-cycle in Great Britain. E c o n o m i c a 2: 142-167.
Coddington, Alan (1941-1982)
Ezekiel, M. 1938. The cobweb theorem. Q u a r t e r l y
J o u r n a l o f E c o n o m i c s 52: 225-280.
Freeman, R.H. 1971. T h e m a r k e t f o r C o l l e g e -
t r a i n e d m a n p o w e r . Cambridge, MA: Harvard
University Press.
Hoffman, D.L., and P. Schmidt 1981. Testing the restrictions
implied by the rational expectations hypothesis.
J o u r n a l o f E c o n o m e t r i c s 15: 265-287.
Kaldor, N. 1934. A classificatory note on the determinateness
of equilibrium. R e v i e w o f E c o n o m i c S t u d i e s 1:
122-136.
Mills, E.S. 1961. The use of adaptive expectations in stability
analysis: Comment Q u a r t e r l y J o u r n a l o f E c o -
n o m i c s 75: 330-335.
Muth, J.F. 1961. Rational expectations and the theory of price
movements. E c o n o m e t r i c s 29: 315-335.
Nerlove, M. 1958. Adaptive expectations and cobweb
phenomena. Quarterly Journal of
E c o n o m i c s 72: 227-240.
Pashigian, B.P. 1970. Rational expectations and the cobweb
theory. J o u r n a l o f P o l i t i c a l E c o n o m y 78:338-
352.
Slutzky, E.E. 1937. The summation of random causes as the
source of cyclical processes. E c o n o m e t r i c a 5: 105-
146.
Waugh, F.V. 1964. Cobweb models. J o u r n a l o f F a r m
Coddington, Alan (1941-1982)
Victoria Chick
Bom in Doncaster, Yorkshire, Coddington began
his academic career with a degree in physics at
Leeds University. After a year teaching mathemat-
ics in a school in York, he returned to university in
that city for his D.Phil. in Economics. On taking his
degree in 1966 he was appointed Assistant Lecturer
in Economics at Queen Mary College, London,
where he rose steadily to become Professor in 1980.
His theoretical work has three main strands: an
early interest in the theory of bargaining, resulting
in several articles and a much-respected book
(1968); various aspects of environmental eco-
nomics; and, continuing throughout his career,
methodology and the history of 20th-century eco-
nomic thought.
Codetermination and Profit-Sharing
It is this last area which one immediately asso-
ciates with Coddington’s name. From his work on
interpretations of Keynes, his characterizations of
the ‘neoclassical synthesis’ of the textbooks as
‘hydraulic Keynesianism’ and of the ‘fundamen-
talists’ as ‘Chapter 12 Keynesians’ (referring to the
chapter of the General Theory which discusses the
precariousness of long-term expectations) have
entered economists’ everyday language.
His coverage of modem economic thought
included not only Keynes, but also Hicks, Shackle,
Friedman, Hahn, Malinvaud, Glower and
Leijonhufvud. His article on Hicks (1979) and much
of his work of Keynes, with an invaluable new
chapter (‘The Keynesian Dichotomy’) were
reworked into his posthumously published book,
Keynesian Economics: The Search for First Prin-
ciples, to which every reviewer responded first with
enthusiasm, then with regret that they had been
deprived by Coddington’s suicide of the intellectual
stimulus of debating with him.
Selected Works
1968. Theories of the bargaining process. London:
George Allen & Unwin.
1973. Bargaining as a decision process. Swedish
Journal of Economics 75(4): 397-405.
1974a. The economics of conservation. In
Conservation in practice, ed. A. Warren and
F.B. Goldsmith. New York: Wiley.
1974b. Creaking semaphore and beyond: A
consideration of Shackle’s epistemics and
economics. British Journal for the Philosophy
of Science 26(2): 151-63.
1975. The rationale of general equilibrium theory.
Economic Inquiry 13(4): 539-58.
1978. Review of the theory of unemployment
reconsidered (Malinvaud). Journal of Economic
Literature 16(3): 1012-18.
1979. Hicks’ contribution to Keynesian economics.
Journal of Economic Literature 17(3): 970-88.
19 83 . Keynesian economics: The search for first
principles. Lo nd on : G e o rg e A l l e n &
Un w i n .
1747
Codetermination and Profit-Sharing
D. M. Nuti
The contract regulating labour employment by
capitalist firms usually embodies three basic ele-
ments: a fixed money wage rate per unit of time, the
subjection of workers to the employer’s authority in
the workplace and the short-term nature of the
hiring commitment. Explicit or implicit departures
from this standard can be observed; they are the
result of individual or collective negotiations in the
labour market, which balance out their advantages
and disadvantages for each party, either directly or
through accompanying changes in other parameters
of the labour contract. Government legislation and
economic policy set limits or fix actual values for
some of these parameters and stipulations; within
these bounds the market determines the rest.
Long tenure, i.e. the employee’s option on
continued employment, like all options has a value
(for the employee) and a cost (for the employer),
which is matched by correspondingly lower pay
than that associated with shorter-term contracts. The
partial and delayed indexation of money wages to a
consumer price index for the period between
successive rounds of wage negotiations favours
employees when inflation decelerates and
employers when it accelerates. Piece- rates, i.e.
wages related to individual performance, give
employees a short term reward (penalty) for effort
supply higher (lower) than that which otherwise
could be contractually fixed, as well as automatic
participation in productivity gains due to learning
by doing, subject to a ratchet effect on the
determination of subsequent rates; employers save
on the costs of recruitment, supervision, and
contractual enforcement, lose short term produc-
tivity gains but can use more fully their contractual
power in exacting effort and speeding up progress
when rates are reviewed. Government policy
influences directly or indirectly market
1748
choice, in the pursuit of policy targets such as
distributive fairness, employment, price stability,
efficiency and growth.
The same combination of private interest and
government policy determines the degree of
workers’ participation in decision-making processes
(codetermination) and in the performance (profit-
sharing) of enterprises (for a bibliographical survey,
see Bartlett and Uvalic ).
Codetermination
Employee participation in enterprise decision-
making in cooperatives amounts to full entrepre-
neurship through participation in assemblies, the
election of representative organs and involvement in
the appointment of managers. In other enterprises it (iii) The greater correspondence between workers’
takes the form of access to information and right to
consultation, participation in decisions on
conditions and organization of work and on internal
social questions, through a workers’ council or
similar organ; right up to the minority (or even
parity) participation and vote in the board of
directors of a joint-stock company (as in German
Mitbestimmung; see Nutzinger ) with a possibility
of influencing decisions about employment, the
level and structure of investment and other crucial
factors were the other board members to be
sufficiently divided.
The effects of codetermination are three-fold: Profit-Sharing
(i) The reduction in labour disutility obtainable
when workers have a say in the division of
labour and work organization, since enter-
prises may neglect workers’ preferences about
the specific uses to which their labour is put or
at any rate respond to the needs of a
hypothetical average worker: if the number of
enterprises is not large enough, workers’
control is necessary to reduce disutility and
alienation. The effect of workers’ control on
productivity has an indeterminate sign
(Pagano ).
(ii) The reduction of the number and intensity of
conflicts in the workplace in general and, in
particular, the more likely acceptance by
workers of unpopular decisions by
Codetermination and Profit-Sharing
management, when workers receive detailed
and credible information and participate in
decision making, identifying themselves
partly with the enterprise and above all
lengthening their time horizon in view of
continued participation in decision-making
(Aoki ; Cable ; Fitzroy and Mueller ). Of
course conflicts within the firm are made
more tractable by the introduction of
codetermination but afterwards are bound to
reappear over time (Furobotn ); also there
remains a basic conflict between employed
and unemployed workers which may even be
exacerbated by the employment protection
policies conceivably encouraged by those
already employed in their exercise of
codetermination.
powers and responsibilities, code- termination
being the counterpart of workers’ exposure to
enterprise risks. The very fact that workers,
unlike capitalists, cannot diversify between
different enterprises when selling their
services exposes them to an employment and
income risk which induces them to make a
claim to control; a claim which up to a point
the employer may prefer to accept instead of
granting higher wages or longer tenure.
In pre-capitalist systems workers’ participation in
the results of their enterprise took the forms - now
little used - of sharecropping in agriculture and of
sliding scales (indexing wage rates to the price of
the product), for instance, in English coalmines. In
modem capitalism such participation - for which
‘profit-sharing’ is a shorthand label - takes the form
of cooperatives’ net revenue sharing, production
prizes based on group or overall performance,
participation in gross/net revenue/profit, share
options, participation in investment funds and pay
increases graded according to productivity growth.
The effects of an element of profit-sharing in
labour earnings are three-fold:
Codetermination and Profit-Sharing
(i) An expected increase in labour productivity.
This is not due to workers gaining from the
product of individual extra-effort (as in the
case of piece-rates) since each of n workers
employed will only get at most 1 hi of the
product of his own extra-effort (Samuelson
) and on the contrary may reduce effort
if he can, being exposed to at most only 1/;? of
the output loss from his own lower effort. The
productivity gain can be expected from
workers, costlessly to themselves, gaining
from intelligent and effective use of any given
individual level of effort, from cooperating
with other workers and management and from
monitoring and supervising each other’s
effort, efficiency and cooperation (Reich and
Devine ; Fitzroy and Kraft ).
(ii) Cyclical flexibility of labour earnings and
therefore greater stability of profit levels and
rates. Employment will not be stabilized
during the cycle by labour earnings flexibility
obtained through profit-sharing because the
marginal cost of labour to firms - i.e. the fixed
component of pay - does not vary
automatically. Workers, who are normally
risk-averse, will prefer a fixed sum of money
to a profit-sharing formula of equivalent
amount while employers, who are normally
risk-lovers, may or may not prefer greater
stability of profit rates (according to their
actual attitude to risk and the alternative cost
of reducing risk through diversification) to the
point of granting higher average earnings on a
profit-sharing formula than a fixed wage to
mutual advantage. Therefore profit- sharing is
favoured primarily in risky ventures;
otherwise on this ground alone profit- sharing
would be favoured by firms only in a
recession (when workers would only accept it
as an alternative to a permanent wage cut) and
by workers only during a boom (when firms
would only accept it as an alternative to a
permanent wage increase).
(iii) Higher level of labour employment, for a
given level of labour earnings with respect to
a fixed wage regime, due to the lower
marginal cost of labour to profit-sharing
1749
firms. Vanek finds that higher employment
will be associated with higher aggregate
income, lower prices (because of higher out-
put), higher export volume and domestic
import substitution (with undetermined effects
on the balance of payments depending on
price and income elasticities), lower after-tax
and after-labour-share profits and higher
labour-share in national income (Vanek ).
Rediscovering Vanek’s macroeconomic benefits
from profit-sharing (though not its impact on net
profits and relative income shares), Weitzman
claims that these benefits are neglected by indi-
vidual firms, as in other instances of ‘public goods’,
‘externalities’ and ‘market failures’, therefore
necessitating public policy measures (Weitzman ,
). However, there is no
reason why a firm should object to granting a given
increase in earnings under the guise of a profit-
share instead of an equivalent fixed amount unless
that represents forced insurance against profit
variability; and why workers - at least at the level of
nation-wide collective bargaining - should not take
into account the potential employment and price
stability benefits of this formula and offset them
against the greater variability of their earnings in
between negotiations, due to both cyclical factors
and random factors affecting their firm’s
performance.
Contrary to Weitzman’s belief, in fact, profit-
sharing is not absolutely superior to wage contracts.
For workers, profit-sharing transforms the
probability distribution of uncertain employment at
a fixed and certain income into a probability
distribution of employment with a higher mean
(because of lower marginal cost of labour) but no
less variable over the cycle, at a more variable
income (both over the cycle and for other factors
affecting dispersion of enterprise performance) and
at a higher (real) mean. For firms it transforms a
more into a less variable probability distribution of
money profit rates around the same mean (or a
lower mean if workers are protected from actual
losses; the effect on real profit rates depending on
accounting conventions and choice of numeraire).
In the pursuit of greater employment and price
1750
stability of course a government may grant tax relief
to shared profits, just as effectively and with just as
much reason as it may subsidize the marginal cost
of labour to firms under a wage regime. Otherwise
there is no reason why profit- sharing should be
forced upon unwilling workers and firms by well-
meaning reformers, beyond the extent they are
prepared to consider in their market transactions.
These propositions are developed further below (see
also Nuti , ).
Interdependence Between
Codetermination and Profit-Sharing
The respective effects of codetermination and of
profit-sharing are not independent. The productivity
increase expected from profit-sharing can be raised
by workers having collective discretion over the
organization of labour; or the productivity fall
which might derive from workers’ control over
labour organization might be tempered by profit-
sharing. Greater variability of earnings - during the
cycle and across firms - strengthens under profit-
sharing the case for codetermination already present
in workers’ exposure to employment risk in the
wage regime. The income premium required by
risk-averse workers to replace some of their fixed
wage with a variable profit-share can be reduced by
their involvement in the decisions which expose
them to income variability in the first place. The
reduction in conflict frequency and intensity
expected from codetermination is enhanced by
profit- sharing because for each worker it partly
internalizes the conflict between ‘us’ and ‘them’
otherwise manifested and enacted externally; in any
case it is a requirement of any effective incentive
system that power and responsibility should not be
separated.
The quantification of degrees of ‘codetermina-
tion’ and to a lesser extent of ‘profit-sharing’ raises
conceptual and practical difficulties (though see
Cable 1985). By and large we can observe a certain
correlation between the two: both codetermination
and profit-sharing are zero in the pure capitalist
enterprises and unity in cooperatives and other
forms of partnerships of capital
Codetermination and Profit-Sharing
and labour; minor forms of codetermination (or
conversely of profit-sharing) tend to go hand in
hand with minor forms of profit-sharing (or of
codetermination); a high degree of one without the
other is virtually unknown.
The combination of 100 per cent codetermina-
tion (= self-determination) and 100 per cent profit-
sharing ( net revenue sharing) obtained in
cooperative firms, according to conventional
literature, is subject to economic stimuli of a
somewhat ‘perverse’ kind. These are primarily:
restrictive employment ( membership) policies;
destabilizing and Pareto-inefficient reactions (or at
best inelasticities) to price changes and technical
progress; a low propensity towards self-financed
investment (Ward ; Vanek ). In empirical studies of
cooperative firms there is no incontrovertible
evidence of these phenomena, which are probably
partly offset by other economic (job security,
growth-mindedness, etc.) andnon-economic stimuli;
but there is a presumption that - albeit in a weak
form - the same tendencies and, in particular,
employment restrictive policies might be associated
with codetermination. We can also presume that
workers’ eagerness to press and ability to assert
demands for codetermination, as in the case of other
demands, increase as unemployment diminishes.
Hence the employment-generating benefits of
profit-sharing can be at least partly offset by the
restrictive employment policies possibly associated
with codetermination brought about by profit-
sharing and by greater proximity to full
employment. Recent empirical studies suggest
modest but sizeable improvements in economic
performance from codetermination and profit-
sharing (Cable and Fitzroy ; Estrin et al. ) when and
where they occur but there may have been costs that
remained unobserved and, in any case, the
improvements cannot be generalized.
Markets and Policy
Degrees of codetermination and profit-sharing may
well be regarded as desirable on ‘political’ (as
opposed to ‘purely technical’) grounds such as
Codetermination and Profit-Sharing 1751

equity and social peace. They may also be the best be justified in forcing profit-sharing on to a yet
policy instruments in the pursuit of public objectives unconverted and disbelieving public, thus achieving
such as stabihty, employment and growth, in the full employment, price stability and growth at a
sense of having the least cost in terms of public stroke. Unfortunately miracles exist only for the
funds or offering the most attractive trade-offs uninformed and the faithful, but do not bear the
between alternative targets. Otherwise, as Jensen weight of sober scrutiny. First, excess demand for
and Meckling argue for codetermination and one labour at full employment cannot be sustained and
can also argue for profit-sharing, if it is truly can only be a temporary disequilibrium. Second,
beneficial to both stockholders and labour no laws permanent excess demand for labour is inconsistent
would be needed to force firms to undertake with lack of codetermination, and when this is
reorganization ( , p. 474). Yet renewed and introduced restrictive employment policies will alter
insistent calls for public intervention in favour of the picture. Third, and most important, there is no
profit-sharing without codetermination have been guarantee that full employment can necessarily be
put forward by M.L. Weitzman in recent writings ( achieved. Without these benefits the alleged ‘public
, , ,, ). The proposal has good’ merits of the sharing contract disappear.
been enthusiastically received in certain academic
and political circles and hailed as a breakthrough in
the specialist press. Excess Demand for Labour at Full
Weitzman’s novelty, the foundation for this Employment
renewed fascination with profit-sharing, is the rash
assertion of two propositions. First, that long-run Suppose that the share economy reaches a state of
full employment equilibrium under profitsharing is full employment. Weitzman maintains the presence
associated with permanent but non-inflationary and persistence of excess demand for labour in
excess demand for labour, which cushions off the long-run equilibrium on the basis of the following
economy from contractionary shocks and gives new argument:
dignity and status to labour. In adman’s language we
are told, for instance: labour total pay = marginal revenue value of
A share system has the hard-boiled property of
excess demand for labour, which turns into a tena-
labour productivity at full
cious natural enemy of stagnation and inflation. The employment
share economy possesses a built-in, three-pronged
assault on unemployment, stagnant output, and the (1)
tendency of prices to rise. This is a hard combination
to beat. (Weitzman , p. 144. because long-run equilibrium must be full-
Second, that even in the short run the share employment equilibrium and because of the under-
economy can achieve and maintain frill employ- lying homomorphism of profit-sharing and wage
ment. For instance: contracts in long-run equilibrium (Weitzman ). By
The share system, . . . , has a strong built-in mecha- definition of profit-sharing
nism that automatically stabilizes the economy at full
employment, even before the long-run tendencies
have had the chance to assert their dominance. labour total pay = fixed pay
... a share economy has the direct ‘strong force’ of + share of net profits
positive excess demand for labor ... pulling it towards (2)
full employment. ... the strong force of the share
system will maintain full employment. (Weitzman , p.
97)
where fixed pay is greater than or equal to zero, and
the share of net profits is greater than zero. It
Were these claims well founded an enlightened
follows from ( ) and ( ) that:
government possessing these troths would
1752 Codetermination and Profit-Sharing

marginal revenue value of share of net profit is zero: with the sharing com-
labour productivity at full ponent of earnings the ‘share economy’ also van-
ishes and reverts to the fixed wage economy
employment > fixed pay = marginal cost of
without any excess demand for labour.
labour to firms Third, workers perceiving excess demand for
(3 ) labour are likely to reduce their supply of effort
and/or increase turnover - as they do in the only
i.e. firms will wish to employ more workers than
known instances of permanent excess demand for
are available. A permanent state of excess demand
labour, i.e. Soviet-type economies (see Lane ) - if
for labour will exist, which will protect full
not right down to the point where their marginal
employment from contractionary shocks, as long as
product equals fixed pay at least as close to that
shocks do not reduce the marginal revenue value of
level as they are allowed to get by monitoring and
labour productivity at full employment below the
supervising arrangements. This is another
fixed element of pay, in which case the maintenance
mechanism which can reduce and eliminate excess
of over-full employment requires a reduction of the
demand for labour if it occurred.
fixed element without cutting earnings as much as
necessary in the wage regime.
There are three grounds for refuting this syllo-
Codetermination and Employment
gism. First, firms should be well aware that, what-
ever their pay formula, they can only attract The lack of codetermination is an explicit precon-
workers by offering the going rate for labour total dition of Weitzman’s claims (though not of
pay and should regard this, and not the fixed Vanek’s, who does not claim full and over-full
element of pay, as marginal cost of labour. If firms employment of labour and does not need this
behave as they should, excess demand for labour restriction). (In the earlier version of his analysis
disappears. Weitzman takes a sanguine view of the possibility
Second, if firms regard the fixed element of pay of keeping codetermination in check: ‘ . . . the
as the marginal cost of labour they should find its bargaining power of labor unions is not a natural
being lower than the marginal revenue value of right . . . ’ (1984a, p. 109); ‘ . . . the decisions on
labour productivity disquieting enough to experi- output, employment and pricing are essentially
ment with alternative combinations of pay param- made by capitalists’ in his model (p. 132); ‘I can see
eters without raising total pay above labour no compelling reason why a capitalist firm should
productivity. Risk-averse workers preferring fixed be more prone to allow increased worker
pay to potentially variable earnings of identical participation in company decision making under
mean, risk-neutral or risk-loving employers will one contract form than under another’ (p. 133,
reduce their labour cost by raising the fixed element emphasis added). His latest version is more open-
of pay at the expense of workers’ profit share; even minded: workers’ participation in decisionmaking
without taking into account attitude to risk it is becomes not only possible but desirable as ‘a
plausible to expect managers to experiment with question of justice and practical politics’ as long as
alternative pay parameters and not to rest until they it excludes employment decisions ( ).
have equalized their marginal cost and marginal It is extremely hard to imagine any major decision,
value of labour, i.e. in which workers might have a voice, that would
not directly or indirectly also affect employment.
marginal revenue value of labour productivity at Either this limitation or workers’ participation
full employment = fixed pay would have to give way.) We know that it is
(3') possible to exclude workers from code- termination
in the presence of persistent unemployment; such
which can only be reconciled with the definition ( ) exclusion might be difficult at full employment, and
of a profit-sharing contract if the workers’ it would certainly be very
Codetermination and Profit-Sharing
difficult with excess demand for labour, but the
persistent state of excess demand for labour pos-
tulated by Weitzman should make the exclusion of
codetermination, whether or not employment
questions are directly involved, impossible without
an authoritarian or military regime. This is not a
moral, or legal, or legalistic proposition; it is a
question of ‘practical politics’.
Once workers have a say on output, employment
and pricing and related questions (investment,
innovation, etc.) they will try and resist the very
possibility of dilution of their own shares just as
shareholders usually resist the dilution of share
capital; for better or worse they are likely to adopt,
or are tempted to adopt, other things being equal,
restrictive employment policies in the possibly
misguided and self-defeating purpose of raising or
maintaining individual earnings. This is not a case
against profit-sharing, but an argument for not
expecting that overfull employment, if achievable,
can be sustained necessarily, i.e. an argument
against the plausibility of Weitzman’s model (see
Nuti ).
Profit-Sharing and Full Employment
The foundation of Weitzman’s claims on behalf of
profit-sharing is the assertion that, even in the short
run, the share economy ‘delivers’ frill employment
of labour. (‘Resources are always frilly utilised in a
share system’ (Weitzman , p. 949); real world
frictions, inertias and imperfections are mentioned
only to be exorcised, and to reassert the full
employment claim at least as a ‘natural tendency’
(p. 949, p. 952) of the share economy which, we are
told, ‘delivers full employment’ ( ); see also
Weitzman ,
p. 97.)
For a share economy to ‘deliver’ full employ-
ment three necessary conditions must be satisfied
simultaneously:
(i) The physical marginal productivity of labour
at full employment must be positive;
(ii) The marginal revenue obtained by firms from
that physical marginal product of labour must
also be positive;
1753
(iii) The fixed element of pay in share contracts
must be flexible enough to fall down to the
level of the marginal revenue product of
labour at full employment, positive as it may
be.
The first condition rales out the possibility of
classical unemployment, i.e. due to lack of equip-
ment, land or other resources in the quantities
necessary to employ all workers efficiently. Yet,
after over a decade of deep and protracted reces-
sion, deindustrialization and decapitalization, even
advanced industrialized countries such as Britain of
France today cannot be expected to be able to
satisfy this condition as a matter of course, not to
speak of Italy or, say, Spain, or of less developed
countries. In his formal model Weitzman ( )
postulates constant physical
productivity of labour; this is a plausible assump-
tion up to near-full capacity but Weitzman gives no
reason why the capacity should be constrained by
labour instead of other resources.
The second condition rales out the possibility of
Keynesian unemployment, i.e. aggregate demand
constraints making the marginal product of labour
valueless before full employment is reached. Even
if the first condition was satisfied, imperfect
competition - which in all of Weitzman’s work
provides the environment in which the share
contract is to operate - provides an excellent reason
why firms might not give to additional physical
products a positive value. Weitzman can assert that
‘... a “pure” sharing system not having any base
wage would possess an infinite demand for labor’ (
, p. 944),
which implies positive marginal revenue for any
level of output, because of the very special
assumption that the elasticity of demand is greater
than unity (p. 938), which makes demand curves
absurdly and indefinitely elastic even for imper-
fectly competitive firms. The proposition cannot
have any claim to general validity.
Even if demand for labour were to be infinite in
the pure share economy, i.e. with a zero fixed
element of pay, it would not necessarily be infinite,
or even large enough to reach full employment, for
a positive fixed element of pay. Weitzman neglects
the determination of the
1754
relative weight of the fixed and variable compo-
nents of the share contract but recognizes the
impossibility of total dependence of pay on profit;
yet he takes for granted, for no good reason, that the
fixed element of pay can be compressed down to
whatever is the full employment marginal revenue
product of labour, which we do not even know for
sure is positive.
It is a non-controversial feature of the sharing
contract, known from Vanek ( ), that the
replacement of part of the wage by a profit-share of
identical average cost to firms will lead to greater
employment, higher output and lower prices - in the
absence of large enough adverse feedback on
investment (which Weitzman recognizes as a
possible short run effect of the introduction of
sharing) and in the absence of large enough
feedbacks of accompanying codetermination on
firms’ employment policy. But there is a world of
difference between higher employment and full
employment and another world of difference
between full employment and persistent over-full
employment; no serious work can afford to switch
indifferently and cavalierly from one to the other.
Share Contracts and Public Good
If the share economy could really guarantee, as
general and necessary consequences of its estab-
lishment, the achievement and stability of full
employment without adverse drawbacks there
would be a case for public policy treating the share
contract as ‘public good’ to be pressed on an
unenlightened public still largely unaware of
potential benefits, as in the case of safe vaccination
against infectious disease. The case for the share
economy would not be much greater than that for
enforced wage flexibility, which would also
guarantee full employment and stability under the
same circumstances. A downward flexible wage
would not deliver excess demand for labour but this
is a questionable achievement and would not be
necessary to absorb contractionary shocks if wages
were flexible; downward flexible wages would also
require a greater fall of money earnings to achieve
full employment in the short run and may be more
likely to bring about adverse
Codetermination and Profit-Sharing
effects on aggregate demand; otherwise there is
little to choose between the two, except for the
lower degree of public resistance that can be
expected for share contracts with respect to wage
cuts.
In fact if the share contract could really deliver
and maintain full employment, while a wage
economy could not, the greater variability of
workers’ earnings associated with profit-sharing
over the cycle would disappear and, between firms,
could be eliminated by labour freely redeploying
itself at will across labour-hungry firms; the
variability of employment would also disappear;
workers would have de facto free access to a job in
any firm of their choice, as in forgotten utopias
(Hertzka ; Chilosi ). Thus it could be said that ‘... a
move towards profit sharing represents an
unambiguous improvement for the working class’
(Weitzman , p. 954). But we have seen that profit-
sharing cannot guarantee full (let alone over-full)
employment. Without full employment, the higher
variability of earnings associated with profit-sharing
remains and it may or may not be compensated by
the higher mean value of employment probability
and perhaps real earnings. Outside over-full
employment, in fact, the share economy is just as
vulnerable to contractionary shocks as the wage
economy because, in spite of flexibility of labour
earnings in the share regime, the marginal cost of
labour to firms (which is the fixed component of
workers’ pay) remains constant just as does the
wage. Thus the higher stability of employment to be
found in Japan simply cannot be the result of profit-
sharing, as Weitzman firmly believes, seeing that
Japan has never known a state of over-full
employment; higher employment stability would
require workers’ shares in GNP instead of their
enterprise’s profits.
The fact that the adoption of a share contract,
without the guarantee of stable frill employment,
has a cost for workers, eliminates the necessity, but
not the possibility, of the share contract having
‘public good’ features. A vaccine may be somewhat
unsafe, its degree of unsafety acceptable to all if
vaccination is universal and all benefit from
reduced exposure to infection, yet individuals
benefit from free-riding strategies and the
Coghlan, Timothy (1855-1926)
enforcement of universal vaccination as ‘public
good’ can still be beneficial to all. If labour con-
tracts were negotiated exclusively at the level of
individuals or firms the external beneficial effects
of the share contract might be lost from sight; but
these external benefits - unlike the case of genuine
‘public goods’ - are completely internalized in
nationwide negotiations between associations of
employers and employees. Admittedly the benefits,
such as they are, of profit-sharing may be still
unknown to the public at large and deserve wider
publicity. But it is counterproductive to foist a good
medicine on a sceptical public by claiming that it
can guarantee longevity or immortality. At the first
signs that such excessive claims are unfounded it
may be thrown away despite its real lesser benefits.
Bibliography
Aoki, M. 1984. The co-operative game theory of the firm.
Oxford: Oxford University Press.
Bartlett, W., and M. Uvalic. 1985. B i b l i o g r a p h y o n
labour- managed firms and employee
p a r t i c i p a t i o n . European University Institute Working
Paper, No. 85/198, Florence. Cable, J.R 1984. E m p l o y e e
participation and firm performance: A
p r i s o n e r s ' d i l e m m a f r a m e w o r k . European Uni-
versity Institute Working Paper, No. 84/126, Florence. Cable,
J.R., and F.R. Fitzroy. 1980. Productive efficiency, incentives
and employee participation: Some preliminary results for
west Germany. K y k l o s 33(2): 100-121. Chilosi, A. 1986.
The right to employment principle and self-
managed market socialism: A historical
account and an analytical appraisal of
s o m e o l d i d e a s . European University Institute Working
Paper, No. 86/214, Florence.
Estrin, S., D.C. Jones, and J. Svejnar. 1984. The varying
nature, importance and productivity effects of worker
participation: Evidence for contemporary producer cooperatives
in industrialised Western societies. CIRIEC Working Paper, No.
84/04, University of Liege. Fitzroy, F.R. and K. Kraft. 1985.
Profitability and profilsharing. Discussion Papers of the
International Institute of Management, WZB, Berlin, IIM/IP 85
41, December.
Fitzroy, F.R., and D.C. Mueller. 1984. Cooperation and
conflict in contractual organisations. Q u a r t e r l y
R e v i e w o f E c o n o m i c s a n d B u s i n e s s 24(4):
24-49.
Furobotn, E.G. 1985. Codetermination, productivity gains
and the economics of the firm. O x f o r d E c o n o m i c
P a p e r s 37: 22-39.
Hertzka, T. 1980. F r e i l a n d . Ein soziales
Z u k u n f t s b i l d . Dresden: Pierson. English translation,
London: Chatto & Windus, 1891.
1755
Jensen, M.C., and W.H. Meckling. 1979. Rights and pro-
duction functions: An application to labor-managed firms
and codetermination. J o u r n a l o f B u s i n e s s 52:
469-506.
Lane, D. (ed.). 1985. E m p l o y m e n t a n d l a b o u r i n
t h e U S S R . London: Harvester Press.
Nuti, D.M. 1985. The share economy:
Plausibility and viability of Weitzman’s
m o d e l . European University Institute Working Paper,
No. 85/194, Florence; Italian translation in P o l i t i c a
e d E c o n o m i c 1, January 1986.
Nuti, D.M. 1986. A rejoinder to Weitzman. (In Italian.)
P o l i t i c a e d E c o n o m i a 4, April.
Nutzinger, H.G. 1983. Empirical research into German
codetermination: Problems and perspectives. E c o -
nomic Analysis and Workers'
M a n a g e m e n t 17(4): 361-382.
Pagano, U. 1984. W e l f a r e , p r o d u c t i v i t y a n d s e l f -
management. European University Institute
Working Paper, No. 84/128, Florence.
Reich, M., and J. Devine. 1981. The microeconomics of
conflict and hierarchy in capitalist production. R e v i e w
o f R a d i c a l P o l i t i c a l E c o n o m i c s 12(4): 27-45.
Samuelson, P.A. 1977'. Thoughts on profit-sharing.
Zeitschriftfur die Gesamte
S t a a t s w i s s e n s c h a f t (special issue on profit-
sharing).
Vanek, J. 1965. Workers’ profit participation, unemployment
and the Keynesian equilibrium.
W e l t w i r t s c h a f t l i c h e s A r c h i v 94(2): 206-214.
Vanek, J. 1970. The general theory of labor-managed market
economies. Ithaca: Cornell University Press.
Ward, B.M. 1958. The firm in Illyria: Market syndicalism.
A m e r i c a n E c o n o m i c R e v i e w 48(4): 566-589.
Weitzman, M.L. 1983. Some macroeconomic implications of
alternative compensation systems. E c o n o m i c J o u r -
n a l 93(4): 763-783.
Weitzman, M.L. 1984. T h e share economy.
Cambridge, MA: Harvard University Press.
Weitzman, M.L. 1985a. Profit sharing as macroeconomic
policy. A m e r i c a n E c o n o m i c R e v i e w : P a p e r s
a n d P r o c e e d i n g s 75(2): 41-45.
Weitzman, M.L. 1985b. The simple macroeconomics of profit
sharing. A m e r i c a n E c o n o m i c R e v i e w 75(5):
937-953.
Weitzman, M.L. 1986. Reply to Nuti. (In Italian.) P o l i t i c a
Coghlan, Timothy (1855-1926)
Colin G. Clark
Coghlan was the son of a poor family in Sydney,
and obtained his education through scholarships.
Like several other distinguished economists, his
1756
original career was in engineering, in this case civil
engineering, which led to his taking an interest in
statistics. New South Wales had been established as
a penal colony under military rule, and therefore
right from the colony's beginning people and
governments were accustomed to statistical
enumeration much more thorough than in the rest of
the world. Coghlan was appointed statistician by the
colonial government, and produced over a long
period of years a large output of excellent statistical
papers.
Before Federation in 1901, there were six
separate colonial governments in Australia, often
pursuing different objectives. This was particularly
the case in the highly controversial matter of free
trade or protection. The two most populous colonies
were Victoria and New South Wales. Victoria had
attracted a large population through the abundance
of alluvial gold in its riverbeds. This was soon
exhausted; and deep mining fell far short of taking
its place. The Victorian Government considered that
the only way to provide employment to make up for
the decline in mining was to establish
manufacturing industries under high tariff
protection. New South Wales (which admittedly had
not had such a destabilizing gold rush) was
determined to adhere to free trade, with the object
of promoting agricultural and pastoral production.
Some of Coghlan’s writings were polemic, strongly
stating the free trade case. In the Wealth and Pro-
gress of New South Wales, in succeeding editions,
the available statistics of the colony were
assembled, pointing to a superior rate of growth to
that of Victoria.
Coghlan was fifty years ahead of his time in
making national product estimations for New South
Wales. Without any experience in any other country
to guide him, he devised methods which would pass
muster today. His work however was not followed
up. There was no further estimate of Australian
national product until the study by Benham in the
1920s, using less precise methods, and then another
gap until Sir John Crawford and I prepared a study
in 1937.
Coghlan prepared a series of publications on a
more extensive basis. The Seven Colonies of Aus-
tralasia (i.c. including New Zealand; it was
Coghlan, Timothy (1855-1926)
thought at the time New Zealand might enter the
Australian federation).
The free trade case was lost in 1908, when
Victoria, with its allies in some other states, per-
suaded the new federal government to impose
highly protective tariffs.
Coghlan was posted to London, where he
performed most of the duties (but without the title,
which went to the political head) of Agent General
for New South Wales. The duties were varied and
responsible, including the raising of large sums in
loans.
In partial retirement, Coghlan embarked upon a
magnum opus. Labour and Industry in Australia,
which was published in England in 1918, giving a
most frill and detailed account of its subject, but
illuminated by stories and anecdotes which make it
readable.
Selected Works
1 89 3. Sheep and wool in New South Wales, with
history and growth of the pastoral industry of
the colony as regards both these items of pro-
duction. S yd n e y : P o t t e r .
1 89 8a . Notes on the financial aspect of Australian
Federation, n . t. p . S y dn e y : W . A . G u l l i c k .
1 89 8b . Notes on the financial aspect of Australian
Federation. The Position of Tasmania and
Western Australia. Sy dn e y : W . A . G u l l i c k .
1 89 8c . Notes on the financial aspect of Australian
Federation. The incidence of the federal tariff.
S yd n e y : W . A. G u l l i c k .
1 89 8d . Tables to accompany notes on financial
aspects of federation. Sy dn e y : W . A .
Gullick.
1 89 8e . A statistical account of the seven colonies
of Australasia 1861-1897. S yd n e y : W . A .
Gullick.
1902. The progress of Australasia in the nineteenth
century. London, Philadelphia: The Linscott
Publishing Company.
1903a. (With T.T. Ewing.) The progress of Aus-
tralia in the century. London Edinburgh: W. &
R. Chambers Ltd.; Philadelphia,Detroit: The
Bradley-Garretson Co., Ltd.
1 90 3b . The decline in the birth-rate of New South
Wales and other phenomena of childbirth: An
Cognitive Ability
19 18 . Labour and industry> in Australia, from the
first settlement in 1788 to the establishment of
the commonwealth in 1901. Lo nd on / N e w
Yo r k : Ox f o r d Un i v e r s i t y P r e s s .
Cognitive Ability
William T. Dickens
Abstract
Modem psychological theory views cognitive
ability as multidimensional while acknowledging
that the many different abilities are themselves
positively correlated. This positive correlation
across abilities has led most psychometricians to
accept the reality of a general cognitive ability
that is reflected in the full scale score on major
tests of cognitive ability or IQ. This article
provides an introduction to the history of
cognitive testing and some of its major
controversies. Evidence supporting the validity
of measures of cognitive ability is presented and
the natiue and implications of group differences
are discussed along with evidence on its
malleability.
Keywords
Ability tests; Achievement tests; Cognitive
ability; Cultural bias; External validity; Factor
analysis; Heritability; IQ; Intelligence; Stereo-
type threat
JEL Classifications
D1
Some people are obviously and consistently quicker
than others to understand new concepts; they solve
unfamiliar problems faster, see relationships that
others don’t and are more knowledgeable about a
wider range of topics. We call such people smart,
bright, quick, or intelligent. Psychologists have
developed tests to measiue this trait.
1757
Originally called IQ tests (for Intelligence Quotient
because the measiues were constructed as the ratio
of mental age to chronological age multiplied by
100), that name has fallen out of favoiu. Instead,
such tests are now often referred to as tests of
cognitive ability. Although the term IQ is still
sometimes used to refer to what such tests measiue,
none constructs a ratio.
History
Spearman ( ) first popularized the observation
that individuals who do well at one type of mental
task also tend to do well at many others. For
example, people who are good at recognizing
patterns in sequences of abstract drawings are also
good at quickly arranging pictures in order to tell a
story, telling what three-dimensional shapes drawn
in two dimensions will look like when rotated, tend
to have large vocabularies and good reading
comprehension, and are quick at arithmetic. This
pattern of moderate to strong positive correlations
across the whole spectmm of mental abilities led
Spearman to hypothesize the existence of a general
mental ability similar to the common notion of
intelligence. A person’s ability with any particular
type of task would be equal to the sum of that
person’s general ability plus considerations unique
to that particular task. Thus general ability could be
measured by constructing sub-tests of a number of
similar items (individual tasks of the same type such
as arithmetic problems) of differing complexity.
Each sub-test would present items of a different
type, and individual scores across sub-tests could be
aggregated. Task specific factors would average out
leaving the final score as mainly a measure of
general ability or ‘g’. Using an approach like this
Binet ( ) developed the first IQ test as a way of
identifying students’ academic potential. That test
was adapted for use in English by Terman and in
1916 became the Stanford-Binet IQ tests - still one
of the most commonly administered tests of
cognitive ability.
Spearman’s hypothesis of a single general
mental ability and many specific abilities was
challenged by Thurstone ( ), who popularized
1758
the notion that people had a number of independent
primary mental abilities rather than a single general
mental ability. Both Spearman and Thurstone made
contributions to the development of factor analysis
as a way to identify the presence of unobserved
variables (abilities) that affect a number of
observable variables (sub-test or item scores).
Today, the Spearman-Thurstone debate has been
resolved with a compromise. The most common
view among psychometricians who study cognitive
ability is that there are a number of different
abilities. Some people are better at solving problems
verbally while others are good at solving problems
that involve visuahzation. Some people who are
good at both of these things may be only average at
tasks that rely heavily on memory. However, there
is a tendency for people who perform well in any of
these broad areas to perform well in all others as
well (Carroll ). Most modem tests of cognitive
ability provide both a full-scale score that is most
reflective of general intelligence, and a number of
special- ability specific sub-scores as well.
Validity
Binet’s is considered the first successful test of
cognitive ability in that it was able to accurately
predict teachers’ assessments of their students on
the basis of a relatively short verbally administered
test. Scores on tests of cognitive ability correlate
well with common perceptions of how bright or
smart someone is. They are also strongly correlated
with measures of academic achievement such as
achievement test scores, grades and ultimate
educational attainment (typically .5 or better). They
are less highly correlated (.5 or less ) with many
important life outcomes including reported annual
income and job status. Performance on a wide range
of jobs and work tasks is positively related to
cognitive test scores with performance on more
demanding jobs having higher correlations. Some
have claimed that general cognitive ability is
responsible for most of this explanatory power (Ree
and Earles ; Ree et al. ). This was a major theme of
the controversial best-seller The Bell Curve
(Hermstein
Cognitive Ability
and Murray ). Heckman ( ), in a review
of that book, argues that even though g has sig-
nificant explanatory power, many other factors,
both cognitive and non-cognitive, matter as well.
Finally, test scores are correlated with a number
of social behaviours including unwed motherhood,
criminal activity, and welfare receipt (Jensen , ch.
9). While these correlations are substantial, and
cognitive test scores are typically better predictors
of most of these outcomes than any other single
personal attribute, they still explain less than half
the variance.
Individuals’ scores on tests of cognitive ability
also tend to be strongly correlated over time - much
more so for adults than for children. A study of
older adults found their full-scale IQ scores to be
correlated .92 when tested at two points in time
three years apart (Plomin et al. ). In contrast, a
study of children tested at two points in time
roughly two years apart found correlations of only .
46 for those who were less than one year old at first
testing and .76 for those who were one year old at
first testing (Johnson and Bradley-Johnson ).
It is common to draw a distinction between tests
of achievement and tests of ability. Achievement
tests measure how much knowledge the test taker
has accumulated in a particular area while ability
tests endeavour to measure how quickly a person
can solve unfamiliar problems. Typically, scores on
the two types of tests are highly correlated. In fact,
all tests of ability are, to some degree, tests of
achievement as it is impossible to measiue ability
without also measuring the test taker’s reading or
verbal comprehension at least. Further, to the extent
that the task being tested relies on knowledge of
geometry, arithmetic, general knowledge, and so
on, the rolls of the achievement test and ability test
are confounded.
Cultural bias has been a concern with
knowledge-based tests. Some knowledge is more
accessible to some people than others. For example,
we would expect that a child growing up with upper
middle-class parents in New York or Paris to find it
easier to learn the distance between the two cities (a
general knowledge question that was once on one
of the popular IQ tests) than someone from the
slums of St. Louis or a tribesman from
Cognitive Ability 1759

the bush in Africa. For this reason a number of tests abstract or symbolic problem-solving. Differences
have been constructed that require a minimal tend to be smaller on seemingly culturally rich
amount of prior knowledge, such as CatteH’s cul- items such as general knowledge. Hermstein and
ture fair test (Cattell ) or Raven’s progressive Murray ( , Appendix 5) provide a review
matrices (Raven ). of the evidence on bias.
The best evidence that tests can be biased in at
least some circumstances comes from studies of a
Group Differences phenomenon called stereotype threat. It has been
shown that reminding peopleof their group identity
No matter what test is administered, men and
can cause them to perform in ways more consistent
women of the same background tend to have very
with stereotypes of the group’s abilities. For exam-
similar average scores on tests of cognitive ability,
ple, blacks have been found to perform worse on
though they differ slightly in their performance on
some particularly difficult vocabulary items when
some sub-tests (Jensen pp. 531—6). However, there
given a questionnaire that asked them to state their
are large differences across ethnic groups and
race before taking the test or when the test was
geographic areas. The difference that has generated
represented as a test of intelligence as opposed to a
the most controversy is the difference in average
test of vocabulary. Women who were told that the
scores of US blacks and whites, which is typically
difficult math test they were taking generally
reported to be about one white standard deviation,
showed gender differences performed worse than
though this gap has declined some in recent years
those taking the same test who were told the test
(Dickens and Flynn ). Do these represent real
showed no differences. Men showed the opposite
differences in cognitive ability or do they reflect
effect and performed better when told the test
cultural bias in the tests?
showed a gender difference (Steele ). However, it
Defenders of the tests offer several pieces of
has not been demonstrated that stereotype threat
evidence suggesting that they are unbiased. Fore-
produces substantial bias on standard tests in
most is the evidence of ‘external validity’ - that the
standard test-taking circumstances.
same regression equation that predicts outcomes
While most evidence is consistent with the view
such as job performance, grades, or educational
that tests provide a fair measure of the underlying
attainment for one group will typically do a
concept of cognitive ability across ethnic groups, it
similarly good job for any other group. Also,
is not conclusive. For example, since tests rarely
different groups find the same questions more or
explain as much as half the variance in the
less difficult. Members of different groups with
outcomes in studies of external validity, there is
similar scores will have similar patterns of right and
always the possibility that the tests underestimate
wrong answers. If some questions are more
black cognitive ability but that other disadvantages
culturally biased than others, the disadvantaged
pull down black performance. If true, the validity of
group should find those items more difficult than
the tests as predictors of practical outcomes is an
the mainstream group does. But researchers looking
artifact of offsetting biases. This could explain why
for such cultural bias have found no evidence of it
it is that when regressions of white performance on
(an exception occurs when one of the groups being
white test scores fail to predict black performance
compared is made up of non-native speakers of the
they tend to predict better performance than is
language in which the test was administered, in
observed. Further, common- sense notions that
which case scores on questions requiring a better
people from different cultural backgrounds
knowledge of the language will be lower).
probably have less opportunity to acquire certain
Surprisingly, to the extent that there are black-white
types of information or practise certain skills should
differences across test items, blacks do worse on
be given some weight. If studies find that blacks do
what seem to be some of the least culturally
no worse than similarly scoring whites on highly
dependent items - those involving
culturally loaded items, that could indicate that the
poorscoring whites
1760
were similarly disadvantaged. If disadvantage is
more common for blacks than whites due to dis-
crimination, that disadvantage could still explain
some of the score gap. However, the strong cor-
relation of even the culturally reduced tests with
performance, and the similar magnitude of the gap
on those tests between groups, suggest that much of
the measured gap in ability between groups reflects
real differences in average developed ability. This
conclusion naturally leads to the consideration of
the sources of those differences.
The question of whether individual, and par-
ticularly group, differences in cognitive ability are
due more to nature or to mature has been enor-
mously controversial for the last century. Dickens (
) presents a summary of the evidence on the
origin of black-white differences and concludes that
they are most likely not substantially genetic in
origin. Rushton and Jensen ( ) reach the
opposite conclusion. Whatever the right answer,
whether the black-white gap has genetic origins is
probably the wrong question. It seems that people
are concerned with the issue mainly because they
confuse having a genetic cause with immutability.
While genes almost certainly play a large role in
explaining individual differences in cognitive
ability within ethnic groups raised in similar cir-
cumstances, it also seems that developed cognitive
ability is highly malleable.
Malleability
A large amount of evidence has accumulated on the
role of genes in explaining individual differences in
cognitive ability. Several reviews of this literature
conclude that differences in genetic endowment
explain somewhere between 60 per cent and 80 per
cent of the variance in cognitive ability in
representative samples of the adult population in
developed countries. The percentage for children is
lower than for adults, with most estimates placing it
around 40 per cent for six-year- olds (Plomin et al.
; Neisser et al. ). The
figure is also estimated to be lower among disad-
vantaged populations (Turkheimer et al. ) though
not consistently (Asbury et al. ). This figure is
referred to as the heritability of
Cognitive Ability
cognitive ability. It is estimated by contrasting
people with different degrees of relatedness raised
in the same home or people with similar relatedness
raised in different homes. For example, the
correlation of the cognitive ability of identical twins
raised in completely independent environments will
be equal to the heritability of cognitive ability under
the assumptions typically employed to make such
estimates. While this evidence establishes that
genes play a large role in determining individual
differences, little is known about which genes are
involved or how they influence cognitive ability
(Plomin et al. ).
The high heritability of cognitive ability has led
some to conclude that people’s environments play
little role in shaping their ability and that, therefore,
individual differences are largely immutable and
group differences must be largely due to differences
in average genetic endowment. It has been argued
that, if all of the observable differences in
environment between people produce only 40 per
cent or less of the variance in cognitive ability, then
the large differences between blacks and whites
could not result from the relatively small
differences in environment between the average
white and the average black. Thus differences in
genetic endowment must play a substantial role. A
formal version of this argument was first presented
by Jensen ( ,
pp. 135-9). A similar argument was made by
Hermstein and Murray ( , pp. 298-9).
Yet despite the high heritability of cognitive
ability, it does seem to be quite sensitive to envi-
ronmental changes. In a review of the effects of
early education programmes, Lazar and Darlington
( , p. 44) noted that ‘The conclusion that a
well-run cognitively oriented early education pro-
gram will increase the IQ sores of low-income
children by the end of the programs is one of the
least disputed results in educational evaluation’.
The gains they surveyed were often quite large,
though they also tended to decline substantially
after children left the programmes. There is also
evidence that being in a cognitively demanding
environment can increase measured cognitive
ability. Ceci ( ) surveys the evidence on the
effects of school attendance on measured ability and
finds it to be substantial.
Cognitive Ability
Finally, the most profound changes in measured
cognitive ability have taken place over time. James
Flynn has documented huge gains in cognitive
ability - as much as a standard deviation or more a
generation - in more than 14 countries. Numerous
other authors have found gains on other tests and in
other countries (Flynn
, ). This phenomenon of large and per
vasive gains has been dubbed ‘the Flynn Effect’.
How is it that large gains are possible in the face
of high heritability estimates? The chief flaw in the
argument that high heritability implies a limited
role for environment is that it misunderstands what
heritability is measuring. It ignores the possibility
that genetic and environmental influences might be
correlated. In particular, it ignores the possibility
that genetic influences on ability are largely the
work of environmental advantages that come about
due to modest physiological advantages.
Consider a sports analogy. Identical twins raised
apart have a shared genetic endowment that tends to
make them notably taller than their peers. As such
they are both better basketball players. Even though
they are raised apart, both are likely to spend more
time playing basketball than other children their
age. They are good at it and thus enjoy it more than
other activities in which they do not naturally excel.
Consequently they both get more practice at
basketball than their peers, and that makes them
better at the game. Being better players than their
peers, they are more likely to be picked by coaches
for high-school teams and more likely to receive yet
more practice and more intensive coaching. If this
leads to them playing in college they will both be
enormously better players than the average person.
A small physiological difference, which would
make only a very modest difference in their
performance on the court if they were untrained and
inexperienced, has mushroomed into a huge
difference in performance because it has been
reinforced by the environmental influences of
practice and coaching.
It is not hard to imagine the same thing hap-
pening with cognitive ability. Someone who is
slightly quicker or has an emotional disposition
amenable to thought and contemplation will be
more likely to spend more time in intellectual
pursuits. Such a person will likely receive positive
1761
reinforcement from teachers and be more likely to
be tracked into more demanding classes and to
develop friendships with other similarly disposed
children. Such a child will have much more
opportunity to practise intellectual work and receive
more ‘coaching’ in intellectual pursuits. A small
initial physiological difference could mushroom
into a large difference in ability through a process
whereby the advantage leads to a better
environment which improves ability and gives
access to even better environments.
If such reciprocal causation is at work in the
development of cognitive ability, then small persis-
tent exogenous differences in environment could
produce large differences in cognitive ability. Dick-
ens and Flynn ( ) lay out a formal model of
such a process. If in a cross section of people in the
same ethnic group most exogenous environmental
differences are transient, then they will not accu-
mulate through reciprocal causation and will not
explain much variance across individuals. However,
small persistent differences between groups or
generations could cause laige differences if they
drive the engine of reciprocal causation. Similarly,
preschool programmes which enrich children’s
cognitive environment can have large effects, but
once the children are removed from the programme
the process can work in reverse and unravel the
gains. The exogenous decline in the quality of the
environment from the removal of the programme’s
stimulation sets off a downward spiral of poorer
performance leading the child into poorer environ-
ments, yet poorer performance and so on.
Conclusion
Modem psychology views cognitive ability as
having a number of dimensions, all of which seem
to be correlated with one another. Many interpret
this correlation as reflecting an underlying general
cognitive ability, or g, that is measured by the full-
scale scores on the major tests of cognitive ability or
IQ. General cognitive ability is an important
predictor of a wide range of economic and life
outcomes, with similar predictive validity across
groups with different average levels of ability. Still,
cognitive test scores typically explain
1762
far less than half the variance in life outcomes, so
cognitive ability is only one important factor among
many that explain success.
Adult differences in cognitive ability within
representative samples of ethnic groups raised in
similar circumstances are subject to substantial
genetic influence, but this does not mean that group
differences are genetic in origin. Despite the large
role played by genetic differences in explaining
adult variance in cognitive ability, there is
considerable evidence that intelligence is highly
malleable and the life outcomes influenced by
intelligence even more so.
See Also
► Behavio
Bibliography
Asbury, K., T. Wachs, andR. Plomin. 2005. Environmental
moderators of genetic influence on verbal and nonverbal
abilities in early childhood. I n t e l l i g e n c e 33: 643-661.
Binet, A. 1905. New methods for the diagnosis of the
intellectual level of subnomrals. L'Annee
P s y c h o l o g i q u e 12, 191-244. English edition: 1916.
The Development of Intelligence in
C h i l d r e n (trans: Elizabeth S. Kite). Vineland:
Publications of the Training School at Vineland.
Carroll, J. 1993. Human cognitive abilities: A sun’ey of factor-
analytic studies. Cambridge, MA: Cambridge University
Press.
Cattell, R. 1960. M e a s u r i n g i n t e l l i g e n c e w i t h
c u l t u r e f a i r t e s t s . Champaign: Institute for
Personality and Ability Testing.
Ceci, S. 1991. How much does schooling influence general
intelligence and its cognitive components? A reassessment
of the evidence. D e v e l o p m e n t a l P s y c h o l o g y ’ 27:
703-722.
Dickens, W. 2005. Genetic differences and school readiness.
T h e F u t u r e o f C h i l d r e n 15: 55-69.
Dickens, W., and J. Flynn. 2001. Heritability estimates versus
large environmental effects: The IQ paradox resolved.
P s y c h o l o g i c a l R e v i e w 108: 346-369.
Dickens, W., and J. Flynn. 2006. Black Americans reduce the
racial IQ gap: Evidence from standardization samples.
P s y c h o l o g i c a l S c i e n c e 17: 913-920.
Flynn, J. 1987. Massive gains in 14 nations: What IQ tests
really measure. P s y c h o l o g i c a l B u l l e t i n 101: 171-
191.
Flynn, J. 1998. IQ gains over time: toward finding the causes.
In T h e r i s i n g c u r v e l o n g - t e r m g a i n s i n I Q
and
Cognitive Ability
r e l a t e d m e a s u r e s , ed. U. Neisser. Washington, DC:
American Psychological Association.
Flynn, J. 2006. Efeito Flynn: repensando a inteligencia e seus
efeitos [The Flynn effect: rethinking intelligence and what
affects it]. In I n t r o d u q d o a P s i c o l o g i a d a s
Diferenqas Individuals [Introduction to
the psychology of individual
d i f f e r e n c e s ] , ed. C. Flores-Mendoza and R. Colorn.
Porto Alegre: ArtMed.
Heckman, J. 1995. Lessons from the bell curve. J o u r n a l o f
P o l i t i c a l E c o n o m y 103: 1091-1120.
Hermstein, R., andC. Murray. 1994. T h e B e l l C u r v e .
New York: Free Press.
Jensen, A. 1973. Educability and group
d i f f e r e n c e s . New York: Harper and Row.
Jensen, A. 1998. The g factor: The science of Mental Ability.
Westport: Praeger.
Johnson, C., and S. Bradley-Johnson. 2002. Construct
stability of the cognitive abilities scale-second edition for
infants and toddlers. Journal of
Psychoeducational
A s s e s s m e n t 20: 144-151.
Lazar, I., and R. Darlington. 1982. Lasting effects of early
education: A report from the consortium for longitudinal
studies. In M o n o g r a p h s o f t h e S o c i e t y f o r
R e s e a r c h i n C h i l d D e v e l o p m e n t , vol. 47.
Chicago: Chicago University Press.
Neisser, U., G. Boodoo, T. Bouchard Jr., A. Boykin, N. Brody,
S. Ceci, D. Halpem, J. Loehlin, R. Perloff, R. Sternberg,
and S. Urbina 1996. Intelligence: Knowns and unknowns.
A m e r i c a n P s y c h o l o g i s t 51: 77-101.
Plomin, R., J. Defries, G. McCleam, and P. McGuffin. 2001.
B e h a v i o r a l g e n e t i c s . 4th edn. New York: Worth
Publishers.
Plomin, R., N. Pedersen, P. Lichtenstein, and G. McCleam.
1994. Variability and stability in cognitive abilities are
largely genetic later in life. B e h a v i o r G e n e t i c s 24:
207-215.
Raven, J. 1941. Standardization of progressive matrices.
British Journal of Medical Psychology
19: 137-150.
Ree, M., and J. Earles. 1992. Intelligence is the best predictor
of job performance. C u r r e n t D i r e c t i o n s i n
P s y c h o l o g i c a l S c i e n c e 1: 86-89.
Ree, M., J. Earles, and M. Teachout. 1994. Predicting job
performance: Not much more than g. J o u r n a l o f
A p p l i e d P s y c h o l o g y ’ 79: 518-524.
Rushton, J., and A. Jensen. 2005. Thirty years of research on
race differences incognitive ability. P s y c h o l o g y ,
P u b l i c P o l i c y , a n d L a w 11: 235-294.
Speamran, C. 1904. General intelligence, objectively
detemrined and measured. A m e r i c a n J o u r n a l o f
P s y c h o l o g y 15: 201-293.
Steele, C. 1997. A threat in the air: how stereotypes shape
intellectual identity and performance. A m e r i c a n
P s y c h o l o g i s t 52: 613-629.
Thurstone, L. 1935. T h e v e c t o r s o f t h e m i n d .
Chicago: University of Chicago Press.
Turkheimer, E., A. Haley, M. Waldron, B. D’Onofrio, and I.
Gottesman. 2003. Socioeconomic status modifies
Cohen, Ruth Louisa (Born 1906) 1763

See Also
Cohen Stuart, Arnold Jacob (1855-1921)

►
Arnold Heertje
Selected Works

1 88 9. Bijdrage tot de theorie der progressieve

inkomstenbelasting. Th e Ha gu e : N i j h o f f .
Keywords
Cohen Stuart, A. J.; Cournot, A. A.; Marginal
Bibliography
utility of income; Oligopoly; Optimal taxation;
Pierson, N. G.; Progressive and regressive Edgeworth, F.Y. 1897. P a p e r s r e l a t i n g t o p o l i t i c a l
taxation e c o n o m y . Vol. 2. London: Macmillan. 1925.
Mirrlees, J.A. 1971. An exploration in the theory of optimum
income taxation. R e v i e w o f E c o n o m i c S t u d i e s
38: 175-208.
Musgrave, R.A., and A.T. Peacock, eds. 1958. C l a s s i c s i n
JEL Classifications
t h e t h e o i y o f p u b l i c f i n a n c e . London:
B31
Macmillan.

Born in The Hague, Cohen Stuart was an engineer

who took up the challenge put forward by the
famous Dutch economist and politician N.G.
Cohen, Ruth Louisa (Born 1906)
Pierson to study the mathematical foundations of
what we would call nowadays an optimal tax Phyllis Deane
structure. His thesis (Cohen Stuart 1889) has been
reprinted in part (Musgrave and Peacock )•
The international attention to Cohen Stuart’s
exposition is due to the thorough discussion by F.
Ruth Cohen is one of the select band of leading
Y. Edgeworth in his article on the pure theory of
professional economists whose importance is
taxation (Edgeworth ). Following a lead by Pierson,
measured more by her emphatically common- sense
Cohen Stuart studied the impact of the principle that
influence on colleagues and pupils than by the
each taxpayer should sacrifice an equal proportion
length of her publications list. Bom in 1906, she
of the total utility which he derives from material
entered Newnham College, Cambridge, in 1926 to
resources. He proved that it depends on the decrease
read for the Economics Tripos within a Faculty that
of marginal utility of income, whether the income
contained a galaxy of outstanding individuals and
taxed above a certain minimum will be progressive,
included such stars as Keynes, Pigou, Robertson,
regressive or proportional in relation to the level of
Maurice Dobb and Piero Sraffa. She had already
income. Cohen Stuart argues that in most practical
developed a research interest in agricultural
cases a modest progressive tax rate will emerge.
economics when, after spending a couple of years
Although based on old-fashioned concepts of
as Commonwealth Fund Fellow at Stanford and
measurable utility, Cohen Stuart’s contribution to
Cornell, she joined the Oxford University
the analysis of the optimal income tax is part of the
Agricultural Economics Research Institute in 1933.
modem theory of optimal taxation (Mirrlees ) and
For the next six years she pursued that
therefore comparable to Cournot’s role in the
specialization and completed two well-received
development of the theory of oligopoly.
books - one an analytical and statistical history of
milk prices and the other a text on the economics of
agriculture which J.M. Keynes invited her to write
for the
1764
Cambridge Economic Handbooks, of which he was
then editor.
In common with most of the leading British
economists of her generation she became a tempo-
rary civil servant soon after the outbreak of World
War II, serving first at the Ministry of Food and
later at the Board of Trade when questions of
postwar reconstruction were rising to the top of the
economic policy agenda of government. In 1945 she
returned to Cambridge as lecturer in the Faculty of
Economics and Politics and as Director of Studies
in Economics at Newnham College, to which she
had been elected a Fellow in 1939 and of which she
was to become an active Principal in 1954. Soon
after retiring from her university post in 1972 she
was elected to the Cambridge City Council, where
for a decade and a half she has devoted her
formidable energy and talent as an applied
economist to local government problems.
Ruth Cohen’s reputation among contemporary
economists has rested largely on her capacity to
offer forceful, direct and perceptive oral comments
on issues of current economic debate - theoretical as
well as applied. This is the kind of salutary
influence on the discipline that is rarely
acknowledged in print, except in the occasional
footnote. In the event, however, her typically terse
intervention in the torrid capital theory debates that
raged in the learned journals of the 1950s and 1960s
has been duly credited with having triggered off a
spate of articles in the capital-switching and capital-
reversing phase of the debate - a phase which
eventually ended in general agreement that her
observation had revealed what some would describe
as a fatal flaw, and others an awkward anomaly, in
the orthodox neo-classical theory of the production
function. To have stimulated this degree of consen-
sus in a theoretical controversy which has carried
unusually heavy methodological and ideological
undertones is no mean achievement.
Selected Works
1936. A history of milk prices. Oxford: Agricultural
Economics Research Institute.
1940. The economics of agriculture. Cambridge:
Cambridge University Press.
Cohn, Gustav (1840-1919)
References
I larcourt, G.C. 1972. Some Cambridge
c o n t r o v e r s i e s i n t h e theory of c a p i t a l .
Cambridge: Cambridge University Press.
Johnson, I LG. 1978. Ruth Cohen: A neglected contributor to
contemporary capital theory. In T h e s h a d o w o f
Keynes, oil. E.S. Johnson. Oxford: Blackwell.
Cohn, Gustav (1840-1919)
H. C. Reektenwald
Lecturer (Privatdozent) at the University of Hei-
delberg in 1869, Cohn was appointed professor of
economics at the Riga Polytechnic Institute (1869-
72). After spending some years in England he was
appointed to a chair of economics at the Zurich
Polytechnic Institute in 1875 and then in Gottingen
in 1884. There he lived up to his death in 1919.
Cohn is noted for his pioneering contributions to
the theory and policy of transportation and public
finance. In his Untersuchungen (1874-75),
Eisenbahnpolitik (1883) and System (1898, vol. 3),
utilizing biased materials produced by parliamentary
commissions, he strongly recommended railway
centralization and government ownership while
opposing canal construction; yet he failed to test the
efficiency of these policy recommendations. In the
field of public finance he is (like Rau, Roscher, von
Stein and Wagner) the typical German exponent of
Smith's liberal principles (not his moral theory)
coupled with his own historical and ethical ideas,
which were based on relatively poor analysis and
synthesis. Cohn attributed to the state an economic
and moral competence which he unquestioningly
assumed. Advocating the legitimacy of value
judgements (Werturteile) and ethical norms in
economic science, he dealt with equity in taxation,
particularly the controversial subsistence level and
progressive taxation. Seligman rightly classes Cohn
among the founders of the science of public finance.
Cointegration 1765

His writings on general economics (1885—98, likelihood; Trend/cycle decomposition; Unit

1886) are distinguished by a philosophical foun- roots; Variance; Vector autoregressions; Vector
dation and a brilliant essayistic style which earned error correction model; Vector moving average
him a great reputation among his contemporaries. models

JEL classifications
Selected Works C32

18 74 —5 . Untersuchungen iiber die englische

Eisenbahnpolitik. L e i p z i g : D un c k e r &
Hu mb l o t 1 88 3. Die englische Eisenbahnpolitik
der letzten zehn Jahre ( 1873-1883). L e ip z i g :
Du n c k e r & H um b lo t .
1885—98. System der Nationalokonomie. 3 vols,
Stuttgart: Enke.
18 86 . Nationalokonomische Studien. S tu t t g a r t :
En k e .
Bibliography
Recktenwald, H.C. (ed.). 1973. P o l i t i c a l e c o n o m y :
A h i s t o r i c a l p e r s p e c t i v e . London: Collier-
Macmillan.
Cointegration
Mark W. Watson
Cointegration means that two or more time series
share common stochastic trends. Thus, while each
series exhibits smooth or trending behaviour, a
linear combination of the series exhibits no trend.
For example, short-term and long-term interest
rates are highly serially correlated (so they are
smooth and in this sense exhibit a stochastic trend),
but the difference between long rates and short rates
- the ‘term spread’ - is far less persistent and shows
no evidence of a stochastic trend. Long rates and
short rates are cointegrated.
The concept of cointegration was formalized by
Clive WJ. Granger in a series of papers in the 1980s
(Granger ; Granger and Weiss ; Granger ; Engle and
Granger ), and in 2003 Granger received the Nobel
Prize in Economics for this work. A flurry of
research activity followed Granger’s original
contributions in this area and produced a practical
set of econometric procedures for analysing
cointegrated time series.

Abstract
Mathematical Structure of I
This article summarizes the mathematical
(1) Cointegrated Models
structure of cointegrated time series models and
discusses econometric procedures commonly Let X, denote a scalar 1(1) stochastic process, with
used to analyse cointegrated time series. This moving average representation A, c(L)st> where E t
discussion is carried out in the context of
is a scalar white noise process, and •'W = £“ QCiL‘
stochastic trends that follow driftless 1(1) or
is a polynomial in the lag operator L, and where the
‘unit root’ processes. The article concludes with
moving average coefficients, Cj decay sufficiently
a brief discussion of cointegration in the context
rapidly so that < 00■ The Beveridge-Nelson
of more general stochastic trends.
decomposition (see trend/cycle decomposition)
implies that X, can be represented as Xt xt + at,
Keywords
where 7/ is a random walk, so that xt xt. i + et, where
Beveridge-Nelson decomposition; Cointegration;
Dickey-Fuller unit root tests; Error-correction et is white noise and at has a moving average
terms; Granger, C.W.J.; Heteroskedasticity and representation at = d{L)et, where
autocorrelation corrections; Maximum
1766
£>i < oo. Thus, X, can be expressed as the sum of a
stochastic trend, xt, and an 1(0) process, at.
When X, is an n x 1 vector of 1(1) processes, a
similar result implies that A) = Axt + at, where A is a
matrix of constants, xt is a vector of random- walk
stochastic trends, and a, is a vector of I (0)
processes. Because X, contains n elements, the
vector xt will generally contain n stochastic trends.
However, when xt contains only k < n stochastic
trends, A is n x k, so that a!A = 0, for any vector a in
the null space of the column space of A. This means
that XX, = a'a t, so that the linear combination XX,
does not depend on the stochastic trends. In this
case, the time series making up Xt are said to be
cointegrated. Any non-zero vector a that satisfies a!
A 0 will annihilate the stochas
tic trend in XX,, and vectors with this property are
called cointegrating vectors. When A has full col-
umn rank, the number of linearly independent
cointegrating vectors is r = n — k, which is called
the cointegrating rank of the process.
For example, suppose that Xt contains n = 3
series representing interest rates on one-month,
three-month and six-month US treasury bills.
Suppose tliat X„ = xt + a„, for / 1,2, 3, where
xt is a common stochastic trend shared by the three
interest rates. Then Xt Ax, + a,, where k 1 (there is a
single stochastic trend), A = (11 1)' (the trend has an
equal effect on each of the interest rates) and a.\ =
(1 0 1)' and a2 = (1 0 —1)' are
two linearly independent cointegrating vectors, so
that r = 2 and XxX, and X2X, denote the interest rate
term spreads.
Vector moving average models (VMAs) and
vector autoregressions (VARs) are often used to
represent the linear properties of vector stochastic
processes. The Granger representation theorem (see
Engle and Granger ) shows that VMAs and VARs
for cointegrated processes have special structures.
In general, the VMA for an 1(1) vector process is
"X, = /.XL);:,, where is white noise with full rank
covariance matrix. When Xt is not cointegrated, the
n x n matrix D( I), which contains the sum of the
moving average coefficients, has rank n. But, when
X, is cointegrated, D( I) has rank k < n, where k
denotes the number of
Cointegration
stochastic trends. When X, is not cointegrated, the
VAR for X, can be written in terms of AX, and has
the form d>(L)X, = e„ where O(L) is a stable lag
polynomial (so its roots are outside the unit circle)
and s, is white noise. When X, is cointegrated, the
VAR has the form O (L) X, = fjcXX,_i + £,, where a
is an n x r matrix with columns that are the linearly
independent cointegrating vectors. Thus, the
cointegrated VAR expresses the elements of AX, as
functions of its own lags, but also includes the r
regressors XX, | in each of the VAR’s n equations.
The variables aX,_, are called ‘error-correction
terms’ and the cointegrated VAR is called a ‘vector
error correction model’ (VECM).
Watson ( ) provides a summary of the
algebra linking these various representations of the
cointegrated model.
Testing for Cointegration
The time series making up X, are cointegrated if the
linear combinations XX, are 1(0) random variables.
If X, is not cointegrated, then XX, will be I (1) for
any non-zero vector a. Tests of cointegration ask
whether XX, is 1(1) or 1(0).
Consider the simple case in which there is only
one potential cointegrating vector, so that XX, is a
scalar. Cointegration can then be tested using a unit
root test applied to XX,. The straightforward
application of a unit root test requires that a is
known, so that the scalar variable XX, can be
calculated directly from the data. This is possible in
many empirical applications (such as the interest
rate example described above) where the value of a
can be pre-specified.
Thus, suppose that a is known, and consider the
competing hypotheses HI(1): XX, is 1(1) and HI(0l:
XX, is 1(0). The hypothesis Hi(i) means that the
elements ofX, are not cointegrated and the hypoth-
esis HI(0) means that the elements are cointegrated.
Under HI(1j the autoregressive model for XX,
contains a unit root, while under HI(0), the auto-
regressive model for XX, is stable.
The null Hi(1) can be tested against the alterna-
tive HI(0) using an augmented Dickey-Fuller
Cointegration
(ADF) unit root test or the modified ADF test
developed in Elliott, Rothenberg and Stock ( ).
The null HI(0) can be tested against HI(1)
using the best local test proposed by Nyblom ( ),
modified for serial correlation as
described in Kwiatkowski et al. ( ), or a
point-optimal test as discussed in Jansson ( ).
(There are important practical consider
ations associated with the choice of the long-run-
variance estimator (see heteroskedasticity and
autocorrelation corrections) used in tests for the
H^o) null hypothesis because of the high degree of
serial correlation under the alternative. See Muller (
) for discussion. )
When a is not known, the unit root tests
described in the last paragraph use a'Xt in place of
a!Xt, where a is an estimator of a. For example,
Engle and Granger ( ) suggest estimating a by
regressing the first element of Xt onto the other
elements of Xt using OLS, and carrying out an ADF
test using the residuals from this regression.
Estimation of a changes the distribution of the ADF
test statistic from what it is when a is known, so that
critical values for the Engle-Granger test are
different than the standard ADF critical values. As
described in Phillips and Ouliaris ( ) and Hansen (
) the correct
critical values depend on the number of elements in
X and on the properties of the deterministic trends
in the model. Stock and Watson (
tabulate choices of critical values from the Phillips
and Ouliaris ( ) and Hansen ( ) papers
that are appropriate for data that follow 1(1) pro-
cesses that may or may not contain drift, and thus
serve as conservative critical values. Modifications
for tests of the Hi(0) null versus the Hi(i) alternative
are discussed in Shin ( ) and
Jansson ( ).
The tests outlined above are useful for testing
whether a single series a.Xt is I (0) or 1(1), but in
many applications there may be more than one
potential cointegrating relation (r> 1) so that it is
useful to have tests for hypothesis that postulate
different values of r. That is, it is useful to entertain
hypotheses of the form Hj : r = j, for j = 0, 1, ... n.
The hypothesis r 0 means that there is no
cointegration, r= 1 means that there is a single
1767
cointegrating vector, and so forth. As discussed in
Johansen ( ), these tests are easily formulated
and carried out using the VECM model. Recall that
the VECM model has the form 0(L) Xt = [>otXt | +
st. Consider the null and alternative hypotheses
H0 :r=r0 v.v. Ha:r= ra where ra>rm and write the
VECM as <J>(L)A, = fj0 u.'()Xt i + /i a'X, | + s:h
where aQ contains the rQ cointegrating vectors under
the null and a contains the additional cointegrating
vectors under the alternative. Under the null
hypothesis, the variables a'X, \ do not enter the
VECM, while under the alternative these variables
enter the VECM. Thus, the null and alternative can
be written as//„ : /) = 0versus IIa : [1 / 0. As in the
case of r = 1, the tests depend on whether the
cointegrating vectors are known or unknown. When
the cointegrating vectors are known, the regressors
c/^X, , and a.'X, t can be constructed from the data,
and the Wald test for [1 0 can be constructed
using the usual
regression formula. When the cointegrating vectors
are unknown, the testing problem is more difficult,
but Johansen ( ) provides an simple
formula for the likelihood ratio test statistic. In
either case, the critical values for the test are ‘non-
standard’, that is, they are not based on the y2
or /'’distributions. Critical values for the tests
depend on the values of ra — r0, the number of
cointegrating vectors that are known and unknown,
)
and the presence or absence of constants and time
trends in the model. The various critical values are
tabulated in Horvath and Watson ( ).
Estimating Unknown Cointegrating
Coefficients
Unknown coefficients in cointegrating vectors are
typically estimated using least squares and Gaussian
maximum likelihood estimators (MLEs). The
properties of these estimators can be understood by
considering a simple bivariate model
Ai, = 6X21 + i] 1,
X21 = A2,-1 + 1)2/
1768
where i], = [i]ui]2l]'~ iid N (0, E). In this example,
there is one common trend that coincides with X2t,
the cointegrating vector is a = (1 — 9)' where 9 is an
unknown parameter, the error correction term is a!
X, = i]u which is potentially correlated with the
innovation in the common trend, i]2„ and the
assumption of normality is used to motivate the
Gaussian MLE of 9.
The OLS estimator of 9 has several interesting
properties (Stock ). Even though X2t and r]U arc
correlated, the OLS estimator is consistent; indeed
~ OLS
it is ‘super-consistent’ in the sense that 0
/ _______i \ xOLS 1 'xOLS _
~ Op (T )(l , so that 9 converges to 9 taster
than the usual \/Y rate familiar from regressions
involving 1(0) variables. These results follow
because, in the cointegrated model, the regressor
X2t is 1(1) and therefore is much more variable
than an 1(0) regressor r=lx\, ~ °P(r2) in
this 1(1) regression instead of Op(T ') in the
usual 1(0) regression), and the correlation between
X2t and //i, is non-zero, but vanishes as the sample
size becomes large. (The covariance is constant, but
the variance of X2t increases linearly with t, so the
correlation vanishes as t increases.)
Despite these intriguing and powerful features,
the OLS estimator has two properties that make it
unsatisfactory for many uses. First, while OLS is
consistent, the correlation between the regressor and
error term induces a bias in the large sample
distribution of the estimator, and this bias can be
severe in sample sizes typically encountered in
applied work (Stock ). Second, the large- sample
distribution of the OLS estimator is
non-normal, and this complicates statistical infer-
^ OLS
ence. For example, the standard interval 9
1.96Se(9°LS^J does not provide a 95 per cent
confidence set even in large samples. Interestingly,
Gaussian maximum likelihood estimators share the
super consistency properties of OLS, but do not
suffer from these unsatisfactory properties
(Johansen ; Phillips ).
To construct the Gaussian MLE, factor the joint
density of into the density of
|x1;| (X2,)f=1| ^ and the density of {X2t)] ,.
Cointegration
The density of {X2t}Tt=l does not depend on 6 , and
the density of X\t\{X2t)lt , is characterized by the
Gaussian linear regression Xlt = 0X2l + [jX2, + vt,
where [1 is the regression coefficient from the
regression of rju onto fj2t (=AX2t), vt is the error in
this regression, and vt\{X2t)t , ~ iid N(0, a2). Simple
calculations (Phillips ) can then be used to show
that 9 —9 ~ Op(T ')
~MLE T
and that 0 \{X2t)t , ~ Njd, L), where
V depends on (X2t)Tt=l. Thus, 0 is consistent, is
—6 conditionally normally distributed and unbiased,
and ^0 — 9^/V1 ^ 2 ~ N(0, 1), so that
inference about 9 can be carried out using standard
methods associated with the Gaussian linear
regression model. Thus, for example, 9 ±1.96
Sn(() ^ provides a valid 95 per cent confidence
set for 9, where SE (^)MLE^ is computed using the
usual regression formula.
While these results may appear quite special (Xt
is bivariate and rjt is normally distributed and
serially uncorrelated) they carry over to more
general models with minor modifications. For
example, X\t and X2t may each be vectors and the
regressionXlf = 9X2t + fiX2, + vt becomes a
multivariate regression. Under weak assumptions
on the distribution of r/,, there is sufficient
averaging so that V~^2 (j)MLE — d'j > /V(0, /),
meaning that the assumption of normality for rj is
not critical (although 0 still refers to the MLE
computed by maximizing the Gaussian likelihood).
Serial correlation in r/t can be handled in a variety
of ways. For example, Saikkonen ( ) and Stock
and Watson ( ) consider the
± ‘dynamic OLS’ (DOLS) regression Xlt = 9X2, + Ef=
_k pjX2,-j + vh which includes enough leads and lags
of XX2t to insure that vt is (linearly) independent of
(X2t')lt , . Phillips and Hansen ( ) and Park (
) develop adjustments based
on long-run covariance matrix estimators, and
Johansen ( ) derives the exact Gaussian
MLE based on the VECM. Under general assump-
tions, all of the estimators are asymptotically
equivalent.
Cointegration 1769

Alternative Models for the Common constructed using large-sample normal approxi-
Trends mation applied to Gaussian MLEs.
Several papers have sought to address the Elliott
The concept of cointegration involves variables that critique by developing methods with good
share common persistent ‘trend’ components. The performance for a range of autoregressive roots
statistical analysis outlined above utilized a close to, but not exactly equal to 1.0. For example,
particular model of the trend component, namely, Wright ( ) argues that if 90 is the true value of
the driftless unit root process xt = tt \ + et. Analysis a cointegrating coefficient, then X\, — 60 X21 will be
of this model highlights many of the key features of 1(0), but if 90 is not the true value then Xlt — 60 X21
cointegrated processes, but more general models are will be highly persistent. He suggests testing that 0
often needed for empirical analysis. For example, = 90 by testing the HI(0) null for the series Xlt —
constant terms are often added to the model to 9oX2t. Alternative testing procedures in this context
capture non-zero means of error correction terms or are proposed in Stock and Watson ( ) and
drifts in the trend process. These constant terms Jansson and Moreira ( ).
change the distribution of test statistics for
cointegration in ways familiar from the effect of
constants and time trends in Dickey-Fuller unit root See Also
tests (see Flamilton ). Flansen ( ) and
Johansen ( ) con
tain useful discussion of the key issues. Higher-
order integrated processes (for example, 1(2) pro-
cesses) are discussed in Johansen ( ), Granger ►
and Lee ( ), and Stock and Watson ( ).
Hylleberg et al. ( ) discuss cointegration at
seasonal frequencies. Robinson and Hualde ( ) and Bibliography
the references cited therein discuss Elliott, G. 1998. The robustness of cointegration methods
cointegration in fractionally integrated models. when regressors almost have unit roots.
Elliott ( ) discusses cointegrated models in E c o n o m e t r i c a 66: 149-158.
which the trend follows a ‘nearunit-root’ process - Elliott, G., T.J. Rothenberg, and J.H. Stock. 1996. Efficient
tests for an autoregressive unit root. E c o n o m e t r i c a
an AR process with largest autoregressive root very 64: 813-836.
close to 1.0. (Formally, the asymptotics use a local- Engle, R.F., and C.W.J. Granger. 1987. Co-integration and
to-unity nesting with largest root AR root equal to 1 error correction: Representation, estimation, and testing.
E c o n o m e t r i c a 55: 251-276.
c/1] where c is a constant.) Elliott shows that, while
Granger, C.W.J. 1981. Some properties of time series data and
the basic cointegrated model remains unchanged in their use in econometric specification. J o u r n a l o f
this case, the properties of Gaussian maximum E c o n o m e t r i c s 16: 121-130.
likelihood estimators of unknown cointegrating Granger, C.W.J. 1986. Developments in the study of co-
integrated economic variables. O x f o r d B u l l e t i n o f
coefficients change in important ways. In particular,
E c o n o m i c s a n d S t a t i s t i c s 48: 213-228.
the Gaussian MLEs are no longer conditionally Granger, C.W.J., and T.H. Lee. 1990. Multicointegration.
unbiased, and confidence intervals constructed A d v a n c e s i n E c o n o m e t r i c s 8: 71-84.
using Gaussian approximations (for example, 9 ± Granger, C.W.J., and A.A. Weiss. 1983. Time series analysis of
error-correction models. In S t u d i e s i n e c o n o m e t -
1.96SE ( ll j j can be very misleading. Elliott’s rics, time series, and multivariate
s t a t i s t i c s , i n h o n o r o f T . W . A n d e r s o n , ed.
critique is important because small deviations from
S. Karlin, T. Amemiya, and L.A. Goodman. San Diego:
exact unit roots cannot be detected with high Academic.
probability, and yet small deviations may Hamilton, J.D. 1994. T i m e s e r i e s a n a l y s i s . Princeton:
undermine the validity of statistical inferences Princeton University Press.
Hansen, B. 1992. Efficient estimation and testing of
cointegrating vectors in the presence of deterministic
trends. J o u r n a l o f E c o n o m e t r i c s 53: 86-121.
1770
Horvath, M.T.K., and M.W. Watson. 1995. Testing for
cointegration when some of the cointegrating vectors are
prespecified. E c o n o m e t r i c T h e o r y 11: 952 984.
Ilylleberg, S., R.F. Engle, C.W.J. Granger, and B.S. Yoo.
1990. Seasonal integration and cointegration. J o u r n a l
o f E c o n o m e t r i c s 44: 215 238.
Jansson, M. 2004. Stationary testing with covariates.
E c o n o m e t r i c T h e o r y 20: 56 94.
Jansson, M. 2005. Point optimal tests ofthe null of hypothesis
of cointegration. J o u r n a l o f E c o n o m e t r i c s 124:
187-201.
Jansson, M., and M. Moreira. 2006. Optimal inference in
regression models with integrated regressors.
E c o n o m e t r i c a 74: 681 714.
Johansen, S. 1988. Statistical analysis of cointegrating vectors.
Journal of Economic Dynamics and
C o n t r o l 12:231 254.
Johansen, S. 1994. The role ofthe constant and linear terms in
cointegration analysis of non-stationary variables.
E c o n o m e t r i c R e v i e w s 13: 205 229.
Johansen, S. 1995. A statistical analysis of cointegration for
1(2) variables. E c o n o m e t r i c T h e o r y 11: 25-59.
Kwiatkowski, D., P.C.B. Phillips, P. Schmidt, and Y. Shin.
1992. Testing the null hypothesis of stationarity against
the alternative of a unit root. J o u r n a l o f
E c o n o m e t r i c s 54: 159-178.
Muller, U.K. 2005. Size and power of tests for stationarity in
highly autocorrelated time series. J o u r n a l o f
E c o n o m e t r i c s 128: 195-213.
Nyblom, J. 1989. Testing for the constancy of parameters over
time. J o u r n a l o f t h e A m e r i c a n S t a t i s t i c a l
A s s o c i a t i o n 84: 223 230.
Park, J.Y. 1992. Canonical cointergrating regressions.
E c o n o m e t r i c a ( A Y . 119 143.
Phillips, P.C.B. 1991. Optimal inference in cointegrated
systems. E c o n o m e t r i c a 59: 283-306.
Phillips, P.C.B., and B.li. Hansen. 1990. Statistical inference
on instrumental variables regression with 1(1) processes.
R e v i e w o f E c o n o m i c S t u d i e s 57: 99-124.
Phillips, P.C.B., and S. Ouliaris. 1990. Asymptotic properties
of residual based test for cointegration.
E c o n o m e t r i c a 58: 165-193.
Robinson, P.M., and J. Hualde. 2003. Cointegration in
fractional systems of unknown orders. E c o n o m e t r i c a
71: 1727-1766.
Saikkonen, P. 1991. Asymptotically efficient estimation of
cointegrating regressions. E c o n o m e t r i c T h e o r y 7:
1-21."
Shin, Y. 1994. A residual-based losl of the null of
cointegration against the alternative of no cointegration.
E c o n o m e t r i c T h e o r y 10: 91 115.
Stock, J.H. 1987. Asymptotic properties of least squares
estimates of cointegrating vectors. E c o n o m e t r i c a 55:
1035-1056.
Stock, J.H., and M.W. Watson. 1993. A simple estimator of
cointegrated vectors in higher-order integrated systems.
E c o n o m e t r i c a 61: 783-820.
Stock, J.H., and M.W. Watson. 1996. Confidence sets in
regression with highly serially correlated regressors.
Colbert, Jean-Baptiste (1619-1683)
Manuscript, Department of Economics, Princeton
University.
Stock, J.H., and M.W. Watson. 2007. I n t r o d u c t i o n t o
e c o n o m e t r i c s . 2nd ed. Boston: Pearson-Addison
Wesley.
Watson, M.W. 1994. Vector autoregression and cointegration.
b i H a n d b o o k o f e c o n o m i c s , e d . R.F. Engle and
D.L. McFadden, vol. 4. Amsterdam: North- Holland.
Wright, J.H. 2000. Confidence sols for cointegrating coef-
ficients based on stationarity tests. J o u r n a l o f
B u s i n e s s a n d E c o n o m i c S t a t i s t i c s 18:211
222.
Colbert, Jean-Baptiste (1619-1683)
D. C. Coleman
Keywords
Colbert, J.-B.; Colbertism; Mercantile system;
Mercantilism
JEL Classifications
B31
Colbert was born at Reims on 29 August 1619 and
died on 6 September 1683. In no way at all could he
be called an economist. He was, however, one of
the most powerful administrators, known to history,
of measures affecting the economic life of a nation,
to such an extent and with such lasting influence
that his name is preserved in the notion of
Colbertism.
He came of a mercantile family which had
acquired some public offices. He learned his job as
economic administrator by entering the service, in
1651, of a man he was effectively to succeed,
Cardinal Mazarin. Once successfully installed in
the service of Louis XIV, after Mazarin’s death in
1661 his climb to power was rapid. He soon came
to hold numerous offices of state: finance, com-
merce, buildings, the navy, and more besides. His
achievements rested in part upon his exercising
virtually undisputed power for 22 years as the
dominant minister of the grandest of absolute
Colbert, Jean-Baptiste (1619-1683)
monarchs, and in part upon his own qualities of
character which he brought to bear upon the eco-
nomic problems of France as he perceived them.
Those qualities included energy, tenacity, shrewd-
ness, honesty, a notable ability to deploy the tech-
niques of the courtier, and a wholly remarkable
capacity for hard work. His hand was felt in every
aspect of French economic life; and everywhere he
exercised that passion for order which is so often
the hallmark of the bureaucrat. Adam Smith sniffed
at him as a ‘laborious and plodding man of business
... accustomed to regulate the different departments
of public offices’ (Smith , p. 627). But he was a lot
more than that. Cold, humourless, and devoted, he
was the superservant of a super-king.
Those qualities did not, on the other hand,
include any original economic ideas whatever. He
had absorbed, with characteristic thoroughness, all
the assumptions, maxims, dogmas, and assorted
notions about economic matters which circulated in
16th- and 17th-century Europe, and to which the
label of mercantilism has become attached.
Consequently, by dint of his position and activities,
and because a very large volume of his papers have
survived for the historian, he has come down to
posterity as the embodiment of conventional
mercantilism in practice. Nonexistent as a
theoretical entity, mercantilism has acquired the
appearance of a coherent economic policy probably
more from Colbert’s activities than from any other
single historical source. And because it appeared,
and was continued after his death, in the grandeur
which was France, it was copied or adapted in other
aspiring monarchies. French mercantilism or
Colbertism thus became a recognizable reality in a
way that the English ‘mercantile system’ did not.
The nature of his economic ideas can often be
gathered from the explanatory memoranda which he
addressed to Louis XIV (who was not always as
interested in such matters as Colbert thought he
should be). They have a familiar ring. He wanted
money circulating in the kingdom, not because he
identified money with wealth, but because it
facilitated the payment of taxes and helped to
stimulate economic activity; those branches of
overseas trade which brought in precious metals
1771
were therefore to be especially favoured.
Manufacturing industry deserved encouragement
because it lessened French dependence on imports,
because it was the basis of an export trade which
brought in wealth, and because it employed the idle
(the Catholic Colbert had the zeal for work and the
disapproval of idleness normally thought of as
peculiar to Puritanism). In the interest of the
economic unification of France, internal trade and
transport needed improvement by the removal of
tolls and the repair of roads and bridges. Royal
support was needed, and was secured, for the
construction of canals - of which the most
spectacular achievement was the opening in 1681 of
the Canal des Deux Mers, providing a waterway
between the Atlantic and the Mediterranean.
Colbert shared the pervasive belief in a fixed
cake of trade, so that, as he patiently explained to
Louis in March 1669, the whole trade of Europe
was carried in a fixed number of vessels and
therefore ‘le commerce cause un combat perpetuel
en paix et en guerre entre les nations de l’Europe, a
qui on emportera la meilleure partie’. The Dutch,
the English and the French were the ‘acteurs de ce
combat’ (Lettres VI, p. 266). France’s gain was to
be secured by Holland’s and/or England’s loss. It
followed that shipbuilding should be encouraged
and the French navy and mercantile marine greatly
enlarged. France should move in on trades hitherto
dominated by her rivals. Hence his setting up in the
1660s of privileged trading companies: a French
East India Company, a French West India Company
to improve and exploit French colonies, and the
Company of the North to tap the Baltic trade. Such
views also provided an economic justification for
the war which Louis launched against Holland in
1672. Colbert had to find the revenue for these and
others of his master’s military activities.
Consequently, he devoted much time to trying to
reform the royal finances. Many of his measures -
for example, to improve the collection of taxes or to
unify the customs system - were thus again part of a
policy designed to improve the performance of the
economy so that it could in turn yield more wealth
to the greater glory of le roi soleil.
1772
How much success attended Colbert’s policies
has been a matter of debate. Laissez-faire econo-
mists and economic historians of similar views have
inevitably disparaged them and stressed the
rigidities which were built into the French economy
in the 18th century. His efforts to unify the chaotic
diversity of French fiscal and customs
administration were only very partially successful;
his overseas trading companies were inadequately
financed and generally unprofitable; his compara-
tive neglect of agriculture left the basis of the
economy in a poor state. But his work did greatly
improve the size and efficiency of the French navy
and mercantile marine; stimulate - albeit at a high
cost - certain areas of French manufacturing indus-
try; and encourage French merchant enterprise in
branches of trade hitherto the preserve of others.
Not all of this was evident in his own lifetime. But
one thing was: Colbert died a very rich man, enno-
bled as Marquis de Seignelay, his brothers and
sisters and cousins amply provided with lucrative
sinecures, his sons as ministers or army officers, and
his three daughters married off to dukes. Such were
the 17th-century rewards of administering an
economy.
Bibliography
Clement, P. (ed.). 1861-2. Lettres, Instructions et Memoires de
Colbert, 8 vols. Paris.
Cole, C.W. 1939. C o l b e r t a n d a c e n t u i y o f Crown
F r e n c h m e r c a n t i l i s m , 2 vols. New York: mercantilism was muddled into existence. For this
Columbia University Press.
Smith, A. 1776. In A n i n q u i r y i n t o t h e n a t u r e a n d
c a u s e s o f t h e w e a l t h o f n a t i o n s , ed. E.
Carman. New York: Modem Library. 1937.
Colbertism
D. C. Coleman
Colbertism is a term used to describe the economic manpower, he set about the country’s industrial
policies associated with the French statesman, Jean-
Baptiste Colbert; and sometimes,
Colbertism
confusingly, as a synonym for mercantilist policies
in general.
In the course of his account, and denunciation,
of the mercantile system, Adam Smith presented it
as something foisted upon governments by con-
spiring businessmen. Extending this view from
England to France, he said of Colbert that he had
been ‘imposed upon by the sophistry of merchants
and manufacturers’ (Smith , p. 434). Whatever
degree of truth there may be in his account so far as
it related to England - and there is some - it wholly
misrepresents the mind of Colbert and the nature of
Colbertism. Distrusting the self-interest of
businessmen as a power for the greater good of
society, Colbert believed profoundly that, although
their pursuit of profits should be encouraged, the
way to ensure that such activities redounded to the
greater wealth, and hence power and glory, of
France was by regulation and order. So Colbertism
was essentially a systematic treatment of economic
activities imposed from above by the King through
his servant. It could be described as a version of the
mercantile system appropriate to an absolutist state.
It owed little or nothing to mercantile or
manufacturing pressures brought to bear on
governments. Although there were some similarities
between Colbertism and English mercantilism, both
in the ideas which lay behind it and in its outward
forms as it affected overseas trade, the creation of
Colbertian policies did not in the least resemble the
process of bargaining and compromise between
and Parliament by which English
reason alone the term ‘Parliamentary Colbertism’,
coined by Cunningham and used by him to describe
English economic policy, 1689-1776
(Cunningham , II, pp. 403-68), was singularly
inappropriate. It was also inapt for the different
reason that Colbertism was distinguished by a
concern for the direct control of production which
was wholly absent from the English version of
mercantilist policies.
The quintessence of Colbertism is strikingly
illustrated in Colbert’s approach to manufactures.
Observing that France had great industrial potential,
with many and scattered crafts and substantial
Colbertism
rehabilitation. He used a variety of weapons: sub-
sidies, special tax reductions or exemptions, pro-
tection against foreign imports, the encouragement
of early marriage and large families, grants of
special privileges, and the establishment of
manufactures royales. Disapproving, for example,
of the way in which his countrymen imported and
wore the woollen cloth or serges of Holland and
England, he set up manufactures royales to
stimulate their production in France; and in 1667
very sharply increased import duties against the
offending English and Dutch imports. Similar
techniques were used to promote the making of
lace, silk stockings, tapestries, carpets, glassware,
tinplate, soap, naval supplies, and cannon. Luxury
items and textiles received particular attention. It
has been said that ‘the greatest industry in France
was supplying the wants of the King and his court’
(Cole , II, p. 303). In quantitative terms this was
probably untrue but its significance was very real;
and such a statement could not possibly be made
about English industry. Stimulation demanded
regulation. So Colbert established a Code of
Commerce, promulgated for textiles elaborate
controls covering precise lengths, widths and other
details of all types of textiles; established an
apparatus of industrial inspection; and insisted upon
all labour being organized within the guild
structure.
Three points need to be stressed about these
measures. First, Colbertism was here a continuation
and codification, a new ordering of old practices; it
was part of an etatisme with medieval roots.
Second, at the time that Colbert was imposing these
measures on the French economy, their English
counterparts were withering away; the last legisla-
tive attempt at general regulation of the English
cloth industry failed in 1678. Third, Colbert’s reg-
ulative achievements were continued after his
death: Colbertism brought many more detailed
regulations in the seventy years after 1683.
Colbert’s founding of privileged monopolistic
trading companies shows a certain resemblance to
the prior establishment of their counterparts in
Holland and England. Again, however, the special
nature of Colbertian mercantilism is evident both in
the preponderance of royal and government finance
in the early years of these companies
1773
because of inadequate mercantile enthusiasm for
them; and in the degree of personal control which
Colbert himself exercised, especially over the
French East India Company. So far from being a
product of mercantile pressures Colbertism ran foul
of merchants on more than one occasion. Colbert
made himself very unpopular with those of
Marseilles, for example, when, obsessed by the
need to keep money circulating so that taxes could
be paid, he tried to prevent them from exporting
coin in order to conduct their trade with the Levant.
And the highly protective anti-Dutch tariff of 1667
attracted internal opposition because it so obviously
invited retaliation.
The vast regulative apparatus built up by Col-
bert and his successors showed more contempt than
understanding of the role of businessmen. French
commercial and industrial advance during the 18th
century, though owing something to Colbert’s
initiating stimuli, continued despite, rather than
because of, the perpetuation of Colbertism. Indeed,
one of the reasons for the final reaction against it
was the extent to which the bureaucratic machine
had become both corrupt in its operation and
irrelevant to the needs of the French economy. It
helped the proliferation in 18th-century France of a
congerie of fiscal officeholders and a concomitant
trade in offices and privileges functioning in and
around an overblown court. Such practices certainly
existed before Colbert’s day; but just as Colbert
brought a new administrative zeal to old economic
ideas, so Colbertism came to provide a still more
fertile soil for the growth of ancient corruptions.
Meanwhile, however, it appealed to other states -
Pmssia and the German principalities, Russia,
Austria, Spain - intent on building up or repairing
economic bases for the support of absolutist courts,
territorial ambitions, or the urge for military glory.
The sorts of mercantilism which they adopted all
varied a good deal, despite the common name and
some common economic ideas. But those of central,
eastern and southern Europe were often much
nearer in spirit to Colbertism than to the mercantile
system which Smith discerned in England or to the
particular variety which the Dutch had erected in
Holland. Colbertism was in this sense sui generis.
1774
See Also
► Mercant
Bibliography
Cole, C.W. 1939. C o l b e r t a n d a c e n t u r y o f
F r e n c h m e r c a n t i l i s m , 2 vols. New York:
Columbia University Press.
—. 1943. F r e n c h m e r c a n t i l i s m , 1 6 8 3 - 1 7 0 0 .
New York: Columbia University Press.
Cunningham, W. 1907. T h e g r o w t h o f E n g l i s h
i n d u s t r y a n d c o m m e r c e , 3 vols. Cambridge:
Cambridge University Press.
Smith, A. 1776. A n i n q u i r y i n t o t h e n a t u r e a n d
c a u s e s o f t h e w e a l t h o f n a t i o n s . Ed. E.
Cannan. New York: Modem Library edn, 1937.
Cole, George Douglas Howard (1889-
1959)
Anthony Wright
A British socialist intellectual, G.D.H. Cole was
bom in Cambridge in 1889. He grew up in London
and was educated at Balliol College, Oxford. As a
young Oxford don, Cole came to prominence during
the second decade of the century as a leading
advocate of guild socialism (a doctrine of workers’
control in industry) and adviser to the labour and
trade union movements. After the collapse of guild
socialism, Cole continued to be the outstanding
socialist theorist and Labour Party intellectual in
Britain during the interwar and immediate postwar
periods, always combining academic work with
political commitment. An encyclopaedist and
polymath, Cole’s published output was prodigious
in both volume and range. He produced over a
hundred books, and at different periods held
academic posts in three disciplines (philosophy,
economics, political theory) and could easily have
held posts in at least two others.
Cole’s central and lifelong preoccupation was with
the advocacy of a decentralized, self
Cole, George Douglas Howard (1889-1959)
managing and participatory form of socialism. It is
against this background that his work in economics
has to be seen. Although he immersed himself in
economic matters during the interwar period (when
he was Reader in Economics at Oxford), he
regarded this as a labour of necessity. In a basic
sense, he did not like economics, and railed against
the ‘algebraic sterilities’ of those economic theorists
who divorced the subject both from social values
and from the solution of pressing problems in the
real world. He was, anyway, not equipped to enter
the higher reaches of theoretical economics, and his
own economic theory therefore remained essentially
derivative. His early guild socialism had been
remarkably innocent of any serious economic
theory at all.
Yet, instead of confirming Cole as of only minor
importance, what this really serves to emphasize is
the remarkable nature of his contribution to
practical economics between the wars. If his
economic theory was derivative, he derived it from
sources that enabled him to construct radical policy
proposals to combat slump and unemployment.
Drawing particularly upon the ‘underconsumption’
(or ‘over-saving’) analysis of capitalism developed
by J.A. Hobson, Cole mounted a sustained critique
of economic orthodoxy in relation to unemployment
throughout the 1920s and argued the need for
demand stimulation and a bold programme of
public works and investment. The great merit of
Hobsonian economic theory for a socialist like Cole
was that it provided the materials from which
capitalism could be both indicted and reformed.
Cole’s recovery programme remained substan-
tially the same in the 1930s, but from the early
years of that decade he displayed a clearer under-
standing of how such a programme was to be
financed. His policy proposals were already proto-
Keynesian, but from the early 1930s (when he
worked with Keynes on the Economic Advisory
Council) he analysed the economic situation from a
recognizably Keynesian perspective. Reviewing
Keynes’s General Theory in the New Statesman
(the house magazine of the British Left), Cole
described it as ‘the most important theoretical
economic writing since Marx’s Capital, or, if only
classical economics is to considered
Collective Action 1775

as comparable, since Ricardo’s Principles'. Above
all, it provided the theoretical credentials for his
own dissenting economics.
However, if Keynes had to be absorbed by the
Left, and mobilized for a recovery programme,
Cole also took the view that it was necessary to
look beyond the conditions of short-term stabili-
zation and towards the development of a ‘new’
economics of socialism. He therefore emerged as a
leading advocate of socialist economic planning in
the 1930s, but for the rest of his life (and after the
war from the vantage point of the Chichele Chair of
Social Political Theory at Oxford) he continued to
search for a form of socialist economy consistent
with his prior commitment to a form of non-
bureaucratic socialist democracy.
Selected Works
19 29 . The next ten years in British social and
economic policy. L on do n: Ma c m i l l a n .
19 32 . Economic tracts for the times. Lo nd on :
Ma c m i l l a n .
19 35 . Principles of economic planning. Lo nd on :
Ma c m i l l a n .
1950. Socialist economics. London: Gollancz.
of collective action are a public good: benefi-
ciaries will gain whether or not they participate
in promoting them, while their individual efforts
cannot secure them. Small groups can use
selective incentives to ensure that their members
contribute to promoting their common interests.
This typically results in the paradoxical
‘exploitation of the great by the small’. The
logic of collective action helps explain many
notable examples of economic growth and
stagnation since the Middle Ages.
Keywords
Anarchy; Bargaining; Cartels; Class conflict;
Collective action; Collective bargaining; Col-
lusion; Common interests; Countervailing
power; Encompassing organizations; Excess
burden; Exploitation; Galbraith, K.; Group the-
ory; Industrial revolution; Invisible hand; Latent
groups; Lobbying; Mercantilism; Non-
excludability; Olson, M.; Patronage dividends;
Public choice; Public goods; Revelation of
preferences; Samuelson, P.; Selective incentives;
Smith, A.; Strategic behaviour; Technical
progress; Trade unions; Wicksell, J.

JEL Classifications
References D71
Cole, M. 1971. T h e l i f e o f G . D . H . C o l e . London:
For a long while, economists, like specialists in
Macmillan.
other fields, often took it for granted that groups of
Wright, A. W. 1979. G . D . H . C o l e a n d s o c i a l i s t
d e m o c r a c y . Oxford: Clarendon. individuals with common interests tended to act to
further those common interests, much as individuals
might be expected to further their own interests. If a
group of rational and self-interested individuals
realized that they would gain from political action
of a particular kind, they could be expected to
Collective Action engage in such action; if a group of workers would
gain from collective bargaining, they could be
Mancur Olson
expected to organize a trade union; if a group of
firms in an industry would profit by colluding to
achieve a monopoly price, they would tend to do so;
if the middle class or any other class in a country
Abstract had the power to dominate, that class would strive
The logic of collective action undermines the to control the government and run the country in its
assumption that common interests are always own interest. The idea that there was some tendency
promoted by their beneficiaries. Where the for groups to
number of beneficiaries is large, the benefits
1776
act in their common interests was often merely
taken for granted, but in some cases it played a
central conceptual role, as in some early American
theories of labour unions, in the ‘group theory’ of
the ‘pluralists’ in political science, in J.K.
Galbraith’s concept of ‘countervailing power’, and
in the Marxian theory of class conflict.
More recently, the explicit analysis of the logic
of individual optimization in groups with common
interests has led to a dramatically different view of
collective action. If the individuals in some group
really do share a common interest, the furtherance
of that common interest will automatically benefit
each individual in the group, whether or not he has
borne any of the costs of collective action to further
the common interest. Thus the existence of a
common interest need not provide any incentive for
individual action in the group interest. If the farmers
who grow a given crop have a common interest in a
tariff that limits the imports and raises the price of
that commodity, it does not follow that it is rational
for an individual farmer to pay dues to a farm
organization working for such a tariff, for the
farmer would get the benefit of such a tariff whether
he had paid dues to the farm organization or not,
and his dues alone would be most unlikely to
determine whether or not the tariff passed. The
higher price or wage that results from collective
action to restrict the supply in a market is similarly
available to any firm or worker that remains in that
market, whether or not that firm or worker partic-
ipated in the output restriction or other sacrifices
that obtained the higher price or wage. Similarly,
any gains to the capitalist class or to the working
class from a government that runs a country in the
interests of that class, will accrue to an individual in
the class in question whether or not that individual
has borne the costs of any collective action. This, in
combination with the extreme improbability that a
given individual’s actions will determine whether
his group or class wins or loses, entails that a
typical individual, if rational and self-interested,
would not engage in collective action in the interest
of any large group or class.
Analytically speaking, the benefits of collective
action in the interest of a group with a
Collective Action
common interest are a public or collective good to
that group; they are like the public goods of law and
order, defence, and pollution abatement in that
voluntary and spontaneous market mechanisms will
not provide them. The fundamental reality that
unifies the theory of public goods with the more
general logic of collective action is that ordinary
market or voluntary action fails to obtain the
objective in question. It fails because the benefits of
collective or public goods, whether provided by
governments or non-governmental associations, are
not subject to exclusion; if they are received by one
individual in some group, they automatically also
go to the others in that group (Olson ).
Since many groups with common interests
obviously do not have the power to tax or any
comparable resource, the foregoing logic leads to
the prediction that many groups that would gain
from collective action will not in fact be organized
to act in their common interests. This prediction is
widely supported. Consumers have a common
interest in opposing the legislation that gives var-
ious producer groups supra-competitive prices, and
they would sometimes also have a common interest
in buyers’ coalitions that would countervail
producer monopolies, but there is no major country
where most consumers are members of any
organization that works predominantly in the
interest of consumers. The unemployed similarly
share a common interest, but they are nowhere
organized for collective action. Neither do most
taxpayers, nor most of the poor, belong to organi-
zations that act in their common interest (Austen-
Smith ; Brock and Magee ; Chubb ; Hardin ; Moe ;
Olson ).
Though some groups can never act collectively
in their common interest, certain other groups can,
if they have ingenious leadership, overcome the
difficulties of collective action, though this usually
takes quite some time. There are two conditions
either of which is ultimately sufficient to make
collective action possible. One condition is that the
number of individuals or firms that would need to
act collectively to further the common interest is
sufficiently small; the other is that the groups
should have access to ‘selective incentives’.
Collective Action
The way that small numbers can make collective
action possible at times is most easily evident on
the assumption that the individuals in a group with
a common interest are identical. Suppose there are
only two large firms in an industry and that each of
these firms will gain equally from any government
subsidy or tax loophole for the industry, or from
any supra-competitive price for its output. Clearly
each firm will tend to get the benefit of any
lobbying it does on behalf of the industry, and this
can provide an incentive for some unilateral action
on behalf of the industry. Since each firm’s action
will have an obvious impact on the profits of the
other, the firms will have an incentive to interact
strategically with and bargain with one another.
There would be an incentive to continue this
strategic interaction or bargaining until a joint
maximization or ‘group optimal’ outcome had been
achieved. This same logic obviously also applies to
collective action in the form of collusion to obtain a
supra- competitive price, and thus we obtain the
well- known incentive for oligopolistic collusion in
concentrated industries whenever there are signif-
icant obstacles to or costs of entry. As the number
in a group increases, however, the incentive to act
collectively diminishes; if there are ten identical
members of a group with a common interest, each
gets a tenth of the benefit of unilateral action in the
common interest of the group, and if there are a
million, each gets one millionth. In this last case,
even if there were some incentive to act in the
common interest, that incentive would cease long
before a group-optimal amount of collective action
had taken place. Strategic interaction or voluntary
bargaining will not occur since no two individuals
have an incentive to interact strategically or to
bargain with one another. This is because the failure
of one individual to support collective action will
not them have any perceptible effect on the
incentive any other individual faces so there is no
incentive for strategic interaction or rational
bargaining. Thus we obtain the result that, in time,
sufficiently small groups can act collectively, but
that this incentive for collective action decreases
monotonically as the group gets larger and
disappears entirely in sufficiently large or ‘latent’
groups.
1777
When the parties that would profit from collec-
tive action have very different demand curves, the
party with the highest absolute demand for collec-
tive action will have an incentive to engage in some
amount of collective action when no other member
of the group has such an interest. This leads to a
paradoxical ‘exploitation of the great by the small’.
This is true to a greater degree and is evident much
more simply if income effects are ignored, as in the
demand curves for a collective good depicted in the
figure below. When the party with the highest
demand curve for the collective good, /J h, has
obtained the amount of the collective good, Qi, that
is in its interest unilaterally to provide, any and all
parties with a lower demand curve, such as /J s, will
automatically receive this same amount, and thus
have no incentive to provide any amount at all!
(Olson ). When income effects and certain ‘private
good’ aspects of some collective goods are taken
into account the results are less extreme, but a
distribution of burdens disproportionality
unfavourable to the parties with the absolutely
larger demands tends to remain. This disproportion
has been evident, for example, in various military
affiances and international organizations, in cartels,
and in metropolitan areas in which metropolis-wide
collective goods are provided by independent
municipalities of greatly different size (Olson and
Zeckhauser ; Sandler ) Fig. .
The other condition, besides small numbers, that
can make collective action possible, is ‘selective
incentives’. Those large groups that have been
organized for collective action for any substantial
period of time are regularly found to have worked
out special devices, or selective incentives, that are
functionally equivalent to the taxes that enable
governments to provide public goods (Olson ;
Hardin ). These selective incentives either punish or
reward individuals depending on whether or not
they have borne a share of the costs of collective
action, and thus give the individual an incentive to
contribute to collective action that no good that is or
would be available to all could provide. The most
obvious devices of this kind are the ‘closed shop’
and picket line arrangements of labour unions,
which often make union membership a condition of
1778
Collective Action, Fig. 1
employment and control the supply of labour during
strikes (see, for example, McDonald ; Gamson ).
Upon investigation it becomes clear that labour
unions are not in this respect fundamentally
different from other large organizations for
collective action, which regularly have selective
incentives that, though usually less conspicuous
than the closed shop or the picket line, serve the
same function.
Farm organizations in several countries, and
quite notably in the United States, obtain most of
their membership by deducting the dues in farm
organizations from the ‘patronage dividends’ or
rebates of farm cooperatives and insurance com-
panies that are associated with the farm organiza-
tions. The professional associations representing
such groups as physicians and lawyers character-
istically have either relatively discreet forms of
compulsion (such as the ‘closed bar’) or subtle
individual rewards to association members, such as
access to professional publications, certification,
referrals, and insurance. In small groups, and
sometimes in large ‘federal’ groups that are
composed of many small groups, social pressure
and social rewards are also important sources of
selective incentives.
The selective incentives that are needed if large
groups are to organize for collective action are less
often available to potential entrants or those at the
lower levels of the social order than to established
and well-placed groups. The unemployed, for
example, obviously do not have the option of
making membership of an organization working in
their interest a condition of employment, nor do
they naturally congregate as the employed do at
Collective Action
workplaces where picket lines may be established.
Those who would profit from entering a cartelized
industry or profession are similarly almost always
without selective incentives. Experience in a variety
of countries also confirms that those with higher
levels of education and skill have better access to
selective incentives than lower income workers;
highly trained professionals such as physicians and
attorneys usually come to be well organized before
labour unions emerge, and the unions of skilled
workers normally emerge before unions
representing less skilled workers. The correlation
between income and established status and access
to selective incentives works in the same direction
as the lesser difficulty of collective action of small
groups of large firms in relatively concentrated
industries explained above. Together these two
factors generate a tendency for collective action to
have, in the aggregate though not in all cases, a
strong anti-egalitarian and pro-establishment impact
(Olson ).
The study of collective action goes back to the
beginnings of economics, but then came to be
strangely neglected during most of the rest of the
history of the subject. Though this is not generally
realized, the study of collective action, admittedly
only in an inductive and intuitive way, was a crucial
part of Adam Smith’s analysis of the inefficiencies
and inequities in the economies he observed
(Smith ). Smith even noted that the main
beneficiaries of collective action in his time were
by no means the poor or those of average means.
He also emphasized the tendency for urban interests
to profit from collective action at the expense of
rural people, because the
Collective Action
geographical dispersion of agricultural interests
areas made it more difficult for them to combine to
exert political influence or to fix prices; this
emphasis presumably owed something to the poor
transportation and communication systems in his
day, which presumably obstructed the organization
of rural interests more in his time than it does in
developed countries now.
The label that Adam Smith gave to the set of
public policies, monopolistic combinations, and
ideas that he attacked was, after all, ‘mercantilism’,
because the single most important source of the
evils was the collective action of merchants, or
merchants and ‘masters’, especially those organized
into guilds or ‘corporations’. In his discussions of
the ‘Inequalities Occasioned by the Policy of
Europe’ and of ‘The Rent of Land’ (Bk. I, ch. 10,
pt. iiandch. IT), Smith emphasized that ‘whenever
the legislature attempts to regulate the differences
between masters and their workmen, its counsellors
are always the Masters’. Similarly,
it is everywhere much easier for a rich merchant to
obtain the privilege of trading in a town corporate,
than for a poor artificer to obtain that of working in
i t . . . Though the interest of the labourer is strictly
connected with that of the society ... his voice is little
heard and less regarded.
The rural interests are similarly at a disadvan-
tage, according to Smith, especially as compared
with those in ‘trade and manufacturers’:
The inhabitants of a town, being collected into one
place, can easily combine together. The most insig-
nificant trades carried on in towns have accordingly,
in some place or another, been incorporated ...
voluntary associations and agreements prevent that
free competition which they cannot prohibit... .The
trades which employ but a small number of hands
run most easily into such combinations. ... People of
the same trade seldom meet together, even for
merriment and diversion, but the conversation ends
in a conspiracy against the public, or in some con-
trivance to raise prices.
By contrast, ‘the inhabitants of the country,
dispersed in distant places, cannot easily combine
together’.
These passages, though not in the order they
appear in Smith, nonetheless correctly convey his
alertness to collective action. Though the
1779
handicap that nual interests face in organizing for
collective action is far less in developed countries
today than it was in Smith’s time, even this part of
his argument still generally holds true in the
developing countries, where transportation and
communication in the nual areas are poor, peasants
are generally unrepresented, and agricultural
commodities normally underpriced (Anderson and
Hayami ; Schultz ; Olson ).
Adam Smith’s insights into collective action and
its consequences were ignored until recent times.
Presumably one reason is that most economists in
the 19th and early 20th centimes were mainly
interested in the logic of the case for competitive
markets. The logic of collective action, by contrast,
is really a general statement of the logic of market
failure; it embodies the central insight of the
theories of public goods and externalities, that
markets and voluntary market-type arrangements do
not generally work in those cases where the
beneficiaries of any collective good or benefit
cannot be excluded because they have not paid any
purchase price or dues (Baumol ). It was not until
Knut Wicksell’s New Principle of Just Taxation’
was published in German in 1896 (Musgrave and
Peacock ) that any economist revealed a clear
understanding of the nature of public goods, and
only with the publication of Samuelson’s articles in
the 1950s (Samuelson ) that this idea came to be
generally understood in the English-speaking world.
A second obstacle to the development of the
logic of collective action was that collective action
by governments was normally taken for granted.
Notwithstanding the difficulties of collective action,
anarchy is relatively rare because a government that
provides some sort of law and order quickly takes
over. This in turn is due to conquerors and the gains
they obtain in increased tax revenues from
establishing some system of law and order and
property rights. In the absence of the provision of
these most elemental collective goods, there is not
much for a conqueror to take, so the historic first
movement of the invisible hand is evident in the
incentive conquerors have to establish law and
order. Those who lead the governments that succeed
conquerors obviously must
1780
maintain a system of law and order if they are to
continue collecting significant tax revenues. Since
governments providing basic collective goods have
been ubiquitous, the classic writers on public goods
like Wicksell and Samuelson did not even ask how
collective goods emerged in the first place. They
focused instead on how to determine what was an
appropriate sharing of the tax burdens and on the
difficulty of determining what level of provision of
public goods was Pareto- optimal. This in turn
naturally led to Wicksell’s recommendation that
only those public expenditures that could, with an
approximate allocation of the tax burdens,
command approximate unanimity, should normally
be permitted, and to Samuelson’s and Musgrave’s (
) concern for
the non-revelation of preferences for public goods.
The difficulties of collective action and public good
provision on a voluntary basis therefore naturally
did not gain any theoretical attention.
When, as in the new political economy or public
choice, the focus is also on the efforts of extra-
governmental groups to obtain the gains from
lobbying, cartelization, and collusion, and on
private action to obtain collective benefits of other
kinds, a more general conception becomes natural
(Barry and Hardin ; Olson ; Taylor ). It then
becomes clear that the likelihood of voluntary
collective action depends dramatically on the size of
the group that would gain from collective action.
When a group is sufficiently small and there is time
for the needed bargaining, the desired collective
goods will normally be obtained through voluntary
cooperation (Frohlich et al. ). If there are substantial
differences in the demands for the collective good at
issue, there will be the aforementioned paradoxical
‘exploitation of the great by the small’. When the
number of beneficiaries of collective action is very
large, voluntary and straightforward collective
action is out of the question, and taxes or other
selective incentives are indispensable. Selective
incentives are available only to a subset of those
extra-governmental groups that would gain from
collective action. Even those extra-governmental
groups that do have the potential of organizing
through selective
Collective Action
incentives will usually have great difficulty in
working out these (often subtle) devices, and will
normally succeed in overcoming the great
difficulties of collective action only when they have
relatively ingenious leadership and favourable
circumstances.
If follows that it is only in long-stable societies
that many extra-governmental organizations for
collective action will exist. In societies where
totalitarian repression, revolutionary upheavals, or
unconditional defeat have lately destroyed
organizations for collective action, few groups will
have been able in the time available to have
overcome the formidable difficulties of collective
action. It has been shown elsewhere (Mueller ;
Olson ), that (unless they are very ‘encompassing’)
organizations for collective action have
extraordinarily anti-social incentives; they engage
in distributional struggles, even when the excess
burden of such struggles is very great, rather than in
production. They also will tend to make decisions
slowly and thereby retard technological advance
and adaptations to macroeconomic and monetary
shocks. It follows that societies that have been
through catastrophes that have destroyed
organizations for collective action, such as
Germany, Japan, and Italy, can be expected to enjoy
‘economic miracles’. An understanding of
collective action also makes it possible to
understand how Great Britain, the country that with
industrial revolution discovered modem economic
growth and had for nearly a century the world’s
fastest rate of economic growth, could by now have
fallen victim to the ‘British disease’. The logic of
collective action, in combination with other
theories, also makes it possible to understand many
of the other most notable examples of economic
growth and stagnation since the Middle Ages, and
also certain features of macroeconomic experience
that contradict Keynesian, monetarist, and new
classical macroeconomic theories (Balassa and
Giersch ).
See Also
►
Collective Action (New Perspectives) 1781

►:: v Smith, A. 1776. An inquiiy into the nature and causes of the
wealth of nations. London: J.M. Dent. 1910.
Taylor, M. 1976. A n a r c h v a n d c o o p e r a t i o n .
► : J c C : oice London: John Wiley.
Wicksell, K. 1896. A new principle of just taxation. Trans,
from the German by J.A. Buchanan, in Musgrave and
Bibliography Peacock (1967).
Anderson, K., and Y. Hayami. 1986. T h e p o l i t i c a l
e c o n o m y o f p r o t e c t i o n . Sydney: George Allen &
Unwin.
Austen-Snrith, D. 1981. Voluntary pressure groups.
E c o n o m i c a 48: 143-153.
Balassa, B., and H. Giersch (eds.). 1986. E c o n o m i c
Collective Action (New Perspectives)
i n c e n t i v e s . Proceedings of the international economic
association, London: Macmillan. David P. Myatt
Barry, B., and R. Hardin (eds.). 1982. R a t i o n a l m a n
a n d i r r a t i o n a l s o c i e t y . Beverly Hills: Sage.
Baumol, W.J. 1952. W e l f a r e e c o n o m i c s a n d t h e
t h e o r y o f t h e s t a t e . Cambridge, MA: Harvard
University Press. Abstract
Brock, W., and S. Magee. 1978. The economics of special
Olson’s logic of collective action predicts that
interest groups: The case of the tariff. A m e r i c a n E c o -
n o m i c R e v i e w 68: 246-250. public-good provision is most likely to fail when
Chubb, J. 1983. Interest groups and the size of the consumer group is large; his
b u r e a u c r a c y . Stanford: Stanford University Press. public goods are partially rival, and so the
Frohlich, N., J. Oppenheimer, and O. Young. 1971. P o l i t -
private cost of provision is relatively high. With
ical leadership and collective boards.
Princeton: Princeton University Press. a pure public good, this logic no longer applies,
Gamson, W.A. 1975. T h e s t r a t e g y o f s o c i a l and so attention turns to producer groups. When
p r o t e s t . Home- wood: Dorsey. provision involves teamwork (so that the
Hardin, R. 1982. C o l l e c t i v e a c t i o n . Baltimore: Johns
collective action succeeds when everyone works
Hopkins University Press for Resources for the Future.
McDonald, D.J. 1969. U n i o n m a n . New York: Dutton. together) then coordination problems arise.
Moe, T.M. 1980. T h e o r g a n i z a t i o n o f i n t e r e s t s . Modem techniques suggest that ‘good’ equilibria
Chicago: University of Chicago Press. in which provision is successful are robust only
Mueller, D.C., (ed.). 1983. T h e p o l i t i c a l e c o n o m y o f
when the costs of provision fall below private
g r o w t h . New Haven: Yale University Press.
Musgrave, R.A. 1959. T h e t h e o r y o f p u b l i c rather than social benefits.
f i n a n c e . New York: McGraw-Hill.
Musgrave, R.A., and A.T. Peacock (eds.). 1967. C l a s s i c s
i n t h e t h e o r y o f p u b l i c f i n a n c e , 2nd ed. New
York: McGraw-Hill. Keywords
Olson, M.L. 1965. T h e l o g i c o f c o l l e c t i v e a c t i o n . Chicken games; Collective action; Cournot
Cambridge, MA: Harvard University Press.
contributions games; Critical mass theory;
Olson, M.L. 1982. T h e r i s e a n d d e c l i n e o f
n a t i o n s . New Haven: Yale University Press. Equilibrium-selection problem; Externalities;
Olson, M.L. 1984. Ideology and growth. In T h e l e g a c y o f Global games; Interdependent consumption;
r e a g a n o m i c s , ed. C.R. Hulten and I.V. Sawhill. Interdependent production; Market failure;
Washington, DC: Urban Institute Press.
Multiple equilibria; Olson, M.; Provision games;
Olson, M.L. 1985. Space, organization, and agriculture.
American Journal of Agricultural Public goods; Risk dominance; Schel- ling, T.;
E c o n o m i c s 67: 928-937. Selective incentives; Strategic voting; Strategy
Olson, M.L., and R. Zeckhauser. 1966. An economic theory of revision; Teamwork dilemma; Volunteer’s
alliances. T h e R e v i e w o f E c o n o m i c s a n d
dilemma
S t a t i s t i c s 48: 266-279.
Samuelson, P.A. 1955. Diagrammatic exposition of a theory of
public expenditure. T h e R e v i e w o f E c o n o m i c s
a n d S t a t i s t i c s 37: 350-356.
Sandler, T. (ed.). 1980. The theory and structure of inter- JEL classifications
national political economy. Boulder: Westview. D71
Schultz, T.W. 1978. D i s t o r t i o n o f a g r i c u l t u r a l
1782
In a review conducted on behalf of the UK Gov-
ernment, Stem ( ) concluded that ‘climate
change is a serious global threat, and demands an
urgent global response ... the benefits of strong and
early action far outweigh the economic costs of not
acting’. The cuts in emissions that he suggested
could generate global benefits. However, the costs
would be borne individually by those making
significant cuts (developed nations) or by those
sacrificing future opportunities (rapidly developing
nations).
A shared desire to cut greenhouse-gas emissions
generates a classic problem of collective action: a
group with common interests must rely on volun-
tary individual optimization for the pursuit of those
interests. Stem’s ‘urgent global response’ to a ‘seri-
ous global threat’ requires nations to act. Such
sovereign states need respond only to their own
incentives; any participation is voluntary. Within
each state, the pursuit of national objectives is not
automatic; environmental effects stem from the
decisions of individual agents. Even if it were in a
state’s collective interest to support a collective
action against climate change, it cannot be assumed
that constituents of that state would individually
offer their backing.
To economists, the collective-action problem
boils down to the private provision of a public good
or the private exploitation of a common resource.
Law and order, military defence and pollution
control are classic textbook examples of public
goods: the benefits of provision are non-excludable,
and so private providers fail to capture the full
impact of their contributions. This market failure
leads to inefficient under-provision. On the other
hand, the commons exploitation of traffic
congestion and commercial fishing yield negative
externalities: market failure yields to inefficient
overindulgence in these activities. In both cases,
individuals fail to pursue efficiently their collective
objectives.
The idea that group members will not always
pursue their common interests was once not
accepted widely. In his article in the first edition of
the New Palgrave, Mancur Olson ( ) sidered in turn, before attention turns to a rather
observed that ‘economists, like specialists in other
fields, often took it for granted that groups of i
ndividuals with common interests tended to act
Collective Action (New Perspectives)
to further those common interests, much as indi-
viduals might be expected to further their own
interests’. He persuasively argued that ‘the exis-
tence of a common interest need not provide any
incentive for an individual action in the group
interest’. Hence consumers may fail to campaign for
their collective protection, unions may fail to
protect all their members, oligopolists may fail to
maintain collusive prices, and nations of the world
may fail to prevent further climate change.
Olson’s point was simple and is now familiar:
when contemplating choice, individuals consider
only the private impact of their actions. For the
classic case of a public good, an individual faces the
full marginal cost of provision but fails to account
for the benefit spilling over to others; the presence
of positive externalities leads to under-provision. If
an individual could internalize these externalities,
perhaps by excluding the consumption of others and
charging them for it, then efficiency could be
restored. Alas, pure public goods are non-
excludable, and hence this route to efficiency is
blocked.
Nevertheless, as long as individuals enjoy some
private benefit from voluntary action then we can
expect some, albeit too little, provision of public
goods. The extent of any inefficiency depends upon
the nature of the collective-action problem, the
availability ofmechanisms to restore efficiency, and
die size and nature of the relevant group. Olson (
) concluded that ‘unless the number of indi
viduals in a group is quite small, or unless there is
coercion or some other special device to make
individuals act in their common interest, rational,
self-interested individuals will not act to achieve
their common or group interests’. In the context of
small groups, when partial provision is deemed
possible, he identified ‘a surprising tendency for the
‘exploitation’ of the great by the small’. These
claims led to his theory of groups: (a) collective
actions fail when the groups are large; (b) laiger
factions bear a disproportionate share of any pro-
vision; and (c) selective incentives are necessary if
groups are to succeed. These three claims are con-
different perspective on collective action.
The first claim is Olson’s ‘group size’ hypoth-
esis: private provision should fall as a group
Collective Action (New Perspectives)
grows larger. Olson ( ) painted a picture of a
meeting at which too few people make careful
contributions: ‘When the number of participants is
large, the typical participant will know that his own
efforts will probably not make much difference to
the outcome, and that he will be affected by the
meeting’s decision in much the same way no matter
how much or how little effort he puts into studying
the issues.’ More directly, the claim is that the
private benefit of any voluntary contribution falls
with the group’s size; equivalently, the private cost
for any particular level of public provision rises
with the group size. This claim leans on two
implicit assumptions. First, an increase in the
number consuming the good leads to an increase in
the provision cost, and hence the good is (at least
partially) rival; it is an impure public good. Second,
the group size corresponds to the number of
consumers, and not to the size of the contributor
pool.
These two implicit assumptions that underpin
the group-size hypothesis are often valid. For
instance, the global climate change that worried
Stem ( ) corresponds to a Targe group’ global
collective-action problem (Sandler ). Nevertheless,
the assumptions often exclude interesting
collective-action problems. The first assumption
rules out pure public goods. Consider, for instance,
the contemporary voluntary provision of open-
source software (Raymond ; Johnson ; Lemer and
Tirole ). The typical license under which such
software is distributed ‘requires that the source code
... be made available to everyone, and that the
modifications made by its users also be turned back
to the community’ (Lemer and Tirole ). This a
modem instance of the ‘collective invention’
documented by Allen ( ). Open-source software
is automatically
non-excludable. Of course, software is a classic
instance of a non-rival good: consumption by one
individual does not hamper the consumption
opportunities of others. Hence, an increase in the
size of the group consuming the good, while fixing
the size of the group able to provide it, has no direct
impact on incentives.
Olson’s second claim was that provision costs
fall on larger members of a group. The idea is that
such members consume large shares of the public
1783
good, and so face a relatively large private benefit.
Once again, this builds upon the assumption that the
collective output is rival; for a pure public good, the
same logic would predict that those who care most
contribute most, and such contributors need not be
large in a conventional sense.
Olson’s third claim concerned the possible
response to the problem of collective action. Such a
response requires, according to this claim, selective
incentives that are ‘functionally equivalent to the
taxes that enable governments to provide public
goods ... [they] either punish or reward individuals
depending on whether or not they have borne a
share of the costs of collective action, and thus give
the individual an incentive to contribute ...’
(Olson ). The classic example of selective
incentives is the ‘closed shop’ of labour unions; to
enjoy the benefits of collective union bargaining
power each worker must be a member, and hence
pay the costs of any strike action. Interestingly,
when the selective incentive is based on preventing
a group member from enjoying the collective output
then the implicit assumption is that the public good
is at least partially excludable.
In summary, Olson ( , ) forcefully
clarified the inescapable logic of collective action:
any theory of group behaviour must rely upon the
incentives faced by individuals, and not simply
assume that groups pursue their common interests.
His theory of groups remains relevant for many
contemporary problems. However, it steps outside
the world of pure public goods by assuming the
interdependent consumption of an impure public
good, and does not allow for interdependence of
production. Put more succinctly, Olson’s groups
consist of public good consumers rather than public
good providers.
Attention now refocuses on collective-action
problems in which economic players non-
cooperatively choose whether to participate in the
private production of a pure public good. Crucially,
there can be interdependence of production: the
incentive to participate in a collective action
depends on the expected participation of others.
Decisions become genuinely strategic, and this
changes the nature of the collective-action problem.
1784
A little notation proves helpful. Amongst n
players, write x, for the action of player i, and
collect the actions of everyone together into a vector
x. Payoffs satisfying
Ui{x) = G( x ) - c^xi)
( 1)
comprise the value G(x) of public good and the
private cost e/x,) that player i incurs when con-
tributing to it; the externality imposed on others is
captured by (n 1)G (x). The nature of the strategic
interaction amongst players depends upon the form
taken by G(x). A simple specification is when x, is a
positive real number and G(x) = x,-. ^ Player’s
decision is strategi
cally independent of others’ actions: he simply
equates the private marginal benefit of the public
good to its private marginal cost via the first-order
condition = cJ(x[), yielding the usual underprovision
problem (Comes and Sandler ).
A second natural specification to consider is
where G(x) = F(El,x‘) for some nicely
behaved concave production function F( ■ ). This
falls within the class of Cournot contributions
games (Chamberlin ; McGuire ; Young ; Comes and
Sandler ; Bergstrom et al. ; Bemheim ). Here,
strategic interaction is non-trivial since the marginal
benefit of increased public good provision depends
on the total contributions of all players.
Nevertheless, a unique Nash equilibrium involves
under-provision. The associated literature concerned
itself with the comparative-static properties of such
models, including the response of public good
output and the burden of provision to the
redistribution of income (Warr ; Kemp ).
These first two examples of eq. ( ) simply flesh
out the implicit model of Olson ( ). The nature
of the collective action problem changes signifi-
cantly when G(x) takes on more interesting and yet
plausible shapes. For instance, G( • ) might take a
weakest link (G(x) = min{x,}) or best shot (G(x)
min{x,}) form (Hirshleifer , );
these are special cases of symmetric but non-
additive specifications (Comes ).
Here, however, attention turns to situations in
which the success of a collective action (that is,
Collective Action (New Perspectives)
the successful provision of a public good) turns
upon either the participation of a critical mass of
players, or contributions that exceed a particular
threshold. Returning once more to the economics of
climate change, a plausible scenario is one in which
the ice caps melt unless carbon emissions are
pushed down below a critical level. Whereas in a
Cournot contributions game the incentive to
contribute decreases with the participation of others,
here it may increase: a nation may find it
worthwhile to chase environmental targets if and
only if it expects others to play their part in inter-
national agreements.
A central feature of threshold-based scenarios is
that an individual’s decision depends on aggregate
participation. This is easiest to explore in a binary-
action game where x, e {0,1} for each player i;
hence x, = 1 can be interpreted as individual
participation in a collective action. In many such
situations, the incentive to participate depends on
the number of others who do so. Hence, writing A«,
(x) for this incentive,
A M,(x) = P{m) where m = x^. (2)
When P(m) < 0 for all m, no players participate;
this is an n-player Prisoner’s Dilemma. If P(m)
decreases with m, then the unique equilibrium
entails the participation of m players, where P(m* -
1) > 0 > P(m*); for the Cournot games considered
above the participation m might be socially
suboptimal. If P(m) increases with m, so that there
is a threshold m satisfying P(m — 1) < 0 < P(m ),
then there are two pure- strategy Nash equilibria,
one in which everyone participates, and one in
which the collective action fails. This means that the
problem of collective action becomes one of
coordination.
Games satisfying eq. ( ) drew the insightful
attention of Schelling ( , ). He opened
his analysis by describing the use of protective
helmets in ice hockey: players were willing to wear
helmets only if others did so too. Other sociological
examples are easy to find: members of a crowd will
join a protest only if others do so (Berk ;
Granovetter ) and successful consumer boycotts
require a critical mass (Innes )•
Collective Action (New Perspectives)
Political situations can also fit eq. ( ). Consider a
plurality rule election in which a group wishes to
prevent the success of a disliked incumbent can-
didate. They can do so if and only if a critical
number m abandon their first-preference candidate
and vote for their second choice. Setting P(m — 1)
> 0 and P(m) < 0 otherwise yields a strategic-voting
model (Palfrey ; Myerson and Weber ; Cox , ;
Myatt ).
In sociology, collective-action games with
threshold properties fall under the umbrella of the
theory of critical mass (Oliver et al. ; Oliver and
Marwell ; Marwell et al. ;
Marwell and Oliver ). Alas, these sociolo
gists had no theoretical machinery for selecting
between multiple equilibria. In economics, multiple
equilibria arise in threshold-driven step-level public
goods games (Palfrey and Rosenthal ). Once again,
the problem of coordination boils down to a need to
choose amongst multiple equilibria. Fortunately,
recent contributions to economics allow some
progress to be made on the equilibrium-selection
problem.
To explore further, it is instructive to consider a
simple world: two individuals (A and B) either
participate (Y) or not (N) in a collective action.
Participationinvolves a private cost (either cA or
c/i), but may provide a public good to be enjoyed by
both players. A natural representation is via a
simple 2x2 strategic form game (Fig. ).
In the ‘provision game’ a participant produces a
public good worth v to everyone. A player’s
marginal product is strategically independent: the
incentive for player A to participate is always v — c
i, and hence he does so if and only if v > cA.
However, this generates a spillover of v for player
B, and hence the social gain is 2v — c ,. The
parameter configuration 2v > cA > v yields the
classic under-provision of a public good.
But what if there is strategic interdependence?
Suppose that only one player need provide, so that a
second participant generates a cost but no additional
benefit. This ‘volunteer’s dilemma’ (Diekmann ) is
a textbook game of ‘chicken’ (Lipnowski and
Maital ). If 2v > c , > v and 2v > cB > v then neither
player is willing to participate even though it is
socially optimal for someone to do so. However, if
v > c , then
1785
player A participates so long as player B does not. If
v > cB, then there are two pure-strategy Nash
equilibria in which a single player provides the
public good. But who provides?
One possibility is to use risk dominance
(Harsanyi and Selten ) as a selection criterion. The
risk-dominant equilibrium is that which maximizes
the product of players’ incentives to remain at the
equilibrium. So, in the volunteer’s dilemma, the
equilibrium in which A provides is risk-dominant if
(v —cA)cB > (v — cB)cA, which holds if and only if
cA < cB: the most efficient provider volunteers.
Following Olson ( ), the
strong (low-cost providers) bear the cost of provi-
sion to the benefit of the weak.
A coordination problem also arises in the
‘teamwork dilemma’ (Fig. ) where both players are
needed for the collective action to succeed. This is
an assurance or ‘stag hunt’ game: as long as v > cA
and v > cB there is a pure-strategy equilibrium in
which both players participate, and a second with
no participation in which the collective action fails.
The former equilibrium is risk dominant if and only
if (v ~ca)(v —cB) > cAcB, which boils down to v >
cA+cB, this requires a single private (not social)
benefit from the public good to exceed the total
private cost of provision. If 2v > cA+cB > v, then the
collective action fails even though it would be
socially optimal for it to succeed. Once again, this is
a return to Olson ( ): success of the collec
tive action relies on private incentives.
All well and good, but can the criterion of risk
dominance be justified? In the recent literature two
contrasting approaches lead to the same answer.
The theory of global games (Carlsson and Van
Damme ; Morris and Shin ) supposes that players
do not share common knowledge of the payoffs of
games. Instead, players must rely upon privately
observed signals of the game being played. For
instance, players may be unsure of the true value v
of the public good, and see an estimate of it.
Crucially, this estimate allows them to infer not
only this value but also the probable signals
received by others, and hence their opponents’
likely behaviour. When signals become very precise
then the play of a simple 2x2 game almost
Y V Y V
~ C B v-
2 v-cB 2v -
v- cA CA
cA
v- cB 0 V- cB
N N
V 0 V 0
Provision game Volunteer's dilemma
Collective Action (New Perspectives), Fig. 1 Public-good provision games
always coincides with the risk-dominant Nash
equilibrium (Carlsson and Van Damme ).
Others have selected equilibria by studying the
evolving play. Players (or populations from which
players are drawn) may adjust their play over time
in the direction of myopic best-reply, but occa-
sionally ‘mutate’ to a different strategy (Kandori et
al. ; Young , ). As the probability
of mutations vanishes, play in the long ran focuses
on a single stochastically stable equilibrium (Foster
and Young ). In a symmetric teamwork dilemma, it
picks out the risk dominant equilibrium.
Can modem literature say anything about the
general case of eq. ( )? Players act as though they
are attempting to maximize jointly the single real-
valued function
p(x) = G(x) - , C(v)-
This is a potential function, and yields a poten-
tial game (Monderer and Shapley ). This function
has a natural interpretation: the private benefit that a
single individual derives from a public good, minus
the total private costs involved in its provision.
Clean results emerge when play of a potential
game evolves via a payoff-responsive stochastic
strategy-revision process (Blume ,
; Brock and Durlauf ; Blume and Durlauf ,
). Over time, players occasionally revise
their strategies. When a player does so, his decision
is not a myopic best reply to the current play of
others, but rather a quantal response (McKelvey
V Y v- cB
o
v- C A
v- CA
i
C?
0 0
1
N
0
o
Teamwork dilemma
and Palfrey ): the log odds of choosing one action
rather than another is determined by the difference
in their payoffs, and so he is more likely to choose
better performing strategies. An inspection of eq. ( )
reveals that the difference in a player’s payoffs is
equal to the difference in potential; the potential
function captures the essential strategic interaction
of the game.
Allowing play to evolve, the strategy-revision
process is drawn towards the states-of-play with the
highest potential. In the long run, when quantal
responses approximate best replies, the process
spends almost all time in the state that maximizes
p(x): evolution leads players to maximize the dif-
ference between a single private benefit and total
private costs rather than social welfare which would
incorporate the frill social benefit of nG(x).
This approach can be applied to the teamwork
(3)
dilemma: the potential of the state-of-play in which
neither player participates is zero, and the potential
of the equilibrium in which the collective action
succeeds is v — (cA + cB). The latter equilibrium has
positive potential if and only ifv > cA + cB. only if a
private individual would be willing to step forward
and pay the frill cost of provision himself will the
collective action succeed. So, whereas it may at first
appear that the success of a collective action (the
,
coordinated play of fY,Yg in the teamwork dilemma)
can follow from the interdependence of team
members, evolving play results in failure (the play
of fiV,/Vg in the teamwork dilemma) unless a
private individual would be willing to fund the
collective action himself.
On reflection, this should be unsurprising. Each
step of evolving play (or each step of
Collective Action (New Perspectives) 1787

reasoning in a global-game argument) is driven by Comes, R., and T. Sandler. 1985. The simple analytics of pure
reference to private incentives. So what lesson public good provision. E c o n o m i c a 52: 103-116.
Comes, R., and T. Sandler. 1996. T h e t h e o r y o f
should be taken away? Even when a group’s externalities, public goods and club
problem is one of coordination, its members cannot g o o d s . 2nd ed. London: Cambridge University Press.
escape Olson’s ( , ) fundamental Cox, G.W. 1994. Strategic voting equilibria under the single
nontransferable vote. American Political
logic of collective action.
S c i e n c e R e v i e w 88: 608-621.
Cox, G.W. 1997. M a k i n g v o t e s c o u n t . Cambridge,
UK: Cambridge University Press.
Diekmann, A. 1985. Volunteer’s dilemma. J o u r n a l o f
See Also C o n f l i c t R e s o l u t i o n 29: 605-610.
Foster, D., and H.P. Young. 1990. Stochastic evolutionary
► Collective /..v't.c:: game dynamics. Theoretical Population
► .xtcm; B i o l o g y 38: 219-232.
Granovetter, M. 1978. Threshold models of collective
► ’lib 11 Gocvc behavior. A m e r i c a n J o u r n a l o f S o c i o l o g y 83:
1420-1443.
Bibliography Harsanyi, J.C., and R. Selten. 1988. A g e n e r a l t h e o r y
of equilibrium selection in games.
Cambridge, MA: MIT Press Classics.
Allen, R.C. 1983. Collective invention. Journal of Economic
Hirshleifer, J. 1983. From weakest-link to best-shot: The
Behavior and Organization 4: 1-24.
voluntary provision of public goods. P u b l i c C h o i c e
Bergstrom, T.C., L. Blume, and H.R. Varian. 1986. On the
41: 371-386.
private provision of public goods. J o u r n a l o f
Hirshleifer, J. 1985. From weakest-link to best-shot:
P u b l i c E c o n o m i c s 2 9 : 25-49.
Correction. P u b l i c C h o i c e 46: 21-23.
Berk, R. 1974. A gaming approach to crowd behavior.
Innes, R. 2006. A theory of consumer boycotts under
A m e r i c a n S o c i o l o g i c a l R e v i e w 39: 355-373.
symmetric information and imperfect competition.
Bemheim, B.D. 1986. On the voluntary and involuntary
E c o n o m i c J o u r n a l 116: 355-381.
provision of public goods. A m e r i c a n E c o n o m i c
Johnson, J.P. 2002. Open source software: Private provision
R e v i e w 76: 789-793.
of a public good. J o u r n a l o f E c o n o m i c s a n d
Blume, L.E. 1993. The statistical mechanics of strategic
M a n a g e m e n t S t r a t e g y ’ 11: 637-662.
interaction. G a m e s a n d E c o n o m i c B e h a v i o r
Kandori, M., G.J. Mailath, and R. Rob. 1993. Learning,
5:387-424.
mutation and long-run equilibria in games.
Blume, L.E. 1995. The statistical mechanics of best- response
E c o n o m e t r i c a 61: 29-56.
strategy revision. G a m e s and Economic
Kemp, M. 1984. A note on the theory of international
B e h a v i o r 11: 111-145.
transfers. E c o n o m i c s L e t t e r s 14: 259-262.
Blume, L.E. 1997. Population games. In T h e e c o n o m y
Lemer, J., and J. Tirole. 2001. The open source movement:
a s a n e v o l v i n g c o m p l e x s y s t e m 1 1 , ed. W.B.
Key research questions. E u r o p e a n E c o n o m i c
Arthur, S.N. Durlauf, and D.A. Lane. Boulder: Westview.
R e v i e w 45: 819-826.
Blume, L.E., and S.N. Durlauf. 2001. The interactions- based
Lemer, J., and J. Tirole. 2002. Some simple economics of open
approach to socioeconomic behaviour. In S o c i a l
source. J o u r n a l o f I n d u s t r i a l E c o n o m i c s 50:
d y n a m i c s , ed. S.N. Durlauf and H.P. Young.
197-234.
Cambridge, MA: MIT Press.
Lipnowski, I., and S. Maital. 1983. Voluntary provision of a
Blume, L.E., and S.N. Durlauf. 2003. Equilibrium concepts
pure public good as the game of chicken. J o u r n a l o f
for social interaction models. I n t e r n a t i o n a l G a m e
P u b l i c E c o n o m i c s 20: 381-386.
T h e o r y R e v i e w 5: 193-209.
Marwell, G., and PE. Oliver. 1993. T h e c r i t i c a l m a s s
Brock, W.A., and S.N. Durlauf. 2001. Discrete choice with
in collective action: A micro-social
social interactions. R e v i e w of Economic
t h e o r y . Cambridge and New York: Cambridge
S t u d i e s 68: 235-260.
University Press.
Carlsson, H., and E. Van Damme. 1993. Global games and
Marwell, G., PE. Oliver, and R. Prahl. 1988. Social networks
equilibrium selection. E c o n o m e t r i c a 61: 989-1018.
and collective action: A theory of the critical mass. III.
Chamberlin, J. 1974. Provision of collective goods as a
A m e r i c a n J o u r n a l o f S o c i o l o g y 94: 502-534.
function of group size. A m e r i c a n P o l i t i c a l
McGuire, M. 1974. Group size, group homogeneity and
S c i e n c e R e v i e w 68: 707-716.
aggregate provision of a pure public good under coumot
Comes, R. 1993. Dyke maintenance and other stories: Some
behavior. P u b l i c C h o i c e 18: 107-126.
neglected types of public goods. Q u a r t e r l y J o u r n a l
o f E c o n o m i c s 108: 259-271.
1788 Collective Agriculture

McKelvey, R.D., and T.R. Palfrey. 1995. Quantal response

equilibria for normal form games. G a m e s a n d E c o - Collective Agriculture
n o m i c B e h a v i o r 10: 6-38.
Monderer, D., and L.S. Shapley. 1996. Potential games.
G a m e s a n d E c o n o m i c B e h a v i o r 14: 124-143.
Peter Nollan
Morris, S., and H.S. Shin. 2003. Global games: Theory and
application. In A d v a n c e s i n e c o n o m i c s a n d
econometrics: Theory and applications,
Eighth World Congress, ed. M. Dewatripont, L.P. Hansen
and S.J. Tumovsky, vol. 1. London: Cambridge
The socialist countries have generally modelled
University Press.
Myatt, D.P. 2007. On the theory of strategic voting. R e v i e w their rural institutions on those of the USSR in the
o f E c o n o m i c S t u d i e s 74: 255-281. 1930s. For the most part, means of production were
Myerson, R., and R. Weber. 1993. A theory of voting owned by the so-called collective, farmwork was
equilibria. American Political Science ‘collectively’ organized, and personal income
R e v i e w 87: 102-114.
Oliver, P.E., and G. Marwell. 1988. The paradox of group size
‘collectively’ distributed. At their peak, over one-
in collective action: A theory of critical mass. II. third of the world’s farmers worked under this
A m e r i c a n S o c i o l o g i c a l R e v i e w 53: 1-8. system.
Oliver, P.E., G. Marwell, andR. Teixeira. 1985. Atheory of the ‘Socialist’ countries have favoured collectives
critical mass. I. Interdependence, group heterogeneity,
and the production of collective action. A m e r i c a n
for the following principal reasons.
J o u r n a l o f S o c i o l o g y 91: 522-556. Firstly, the leadership in most ‘socialist’ coun-
Olson, M. 1965. The logic of collective action: Public goods and tries initially was afraid of an economically inde-
the theory of groups. Cambridge, MA: Harvard University pendent peasantry with ideas shaped by
Press.
Olson, M. 1987. Collective action. In T h e new
individualistic ‘petty commodity production’. As
p a l g r a v e : A d i c t i o n a r y o f e c o n o m i c s , ed. Stalin put it: ‘a great deal of work has to be done to
J. Eatwell, M. Milgate, and P. Newman, vol. 1. London: remould the collective-farm peasant, to correct his
Macmillan. individualistic mentality and to transform him into
Palfrey, T.R. 1989. A mathematical proof of Duverger’s law.
In M o d e l s o f s t r a t e g i c c h o i c e i n p o l i t i c s ,
a real working member of a socialist society’
ed. R. Ordeshook. Ann Arbor: University of Michigan (Stalin , p. 469). Collectives were not intended as
Press. independent cooperatives: collectivization was
Palfrey, T.R., and H. Rosenthal. 1984. Participation and the party-led and collectives were subject to
provision of discrete public goods: A strategic analysis.
considerable external control (see e.g., Davies ;
J o u r n a l o f P u b l i c E c o n o m i c s 24: 171-193.
Raymond, E.S. 1998. The cathedral and the bazaar. F i r s t Volin ; Selden ; Unger ). Such a rationale is deeply
M o n d a y 3, 1-20. Online. Available at undemocratic, especially given the peasants’
Accessed numerical dominance in those countries (see, in
19 April 2007.
particular, Cohen , ch. 6).
Sandler, T. 2004. G l o b a l collective action.
Cambridge: Cambridge University Press. Second, it was believed that state intervention
Schelling, T.C. 1973. Hockey helmets, concealed weapons, and through party-led collectives would improve rural
daylight saving: A study of binary choices with economic performance (see e.g., Stalin ; General
externalities. J o u r n a l o f C o n f l i c t R e s o l u t i o n
Office ). Collectives could raise savings and
17: 381-428.
Schelling, T.C. 1978. Micromotives and investment rates through reinvesting income and
m a c r o b e h a v i o r . New York: Norton. mobilizing ‘surplus’ labour for capital constmction.
Stem, N. 2007. T h e e c o n o m i c s o f c l i m a t e Unfortunately, success in these respects can damage
change: The Stern r e v i e w . Cambridge:
labour motivation by reducing current returns to
Cambridge University Press.
Warr, P. 1983. The private provision of a public good is collective labour. Collectives also could provide a
independent of the distribution of income. E c o n o m i c s vehicle for rapidly introducing new technology.
L e t t e r s 13:207-211. However, this applies to bad as well as good
Young, D.J. 1982. Voluntary purchase of public goods.
technology - examples of the former are legion in
Public Choice 38: 73-85.
Young, H.P. 1993. The evolution of conventions. ‘socialist’
Econometrica 61: 57-84.
Young, H.P. 1998. Individual strategy and social structure: An
evolutionary theory of institutions. Princeton: Princeton
Collective Agriculture
agriculture, including the various programmers in
the Soviet Union associated with Lysenko
(discussed in Volin ) and the ill-fated introduction
of the double-wheeled, double-share plough, in
China (Kuo , ch. 12).
Third, party-led collectives were viewed as a
means to attain high farm marketing rates and an
outflow of farm sector savings to finance non-farm
investment:
By transferring the disposal of agricultural out-
put from individual peasants to government-
supervised collective farm managements, collec-
tivization destroys the basis for the peasants’
resistance to the ‘siphoning-off of the economic
surplus (Baran , p. 268).
However, without, for example, adequate sup-
plies of appropriately priced industrial commodi-
ties, forcibly raising the rate of farm sector
marketings can reduce the growth rate of farm
output and the future volume of farm marketings.
Moreover, it has proved difficult to achieve a net
farm savings outflow due, for example, to agricul-
ture’s need for industrial incentive goods and farm
inputs (increased, insofar as inputs are inefficiently
used and collectivization adversely affects livestock
holdings, motive power and fertilizer supplies), and
the state’s inability to control private market prices
(Ellman ; Ishikawa ).
Fourth, it was considered that collectives would
prevent ‘capitalist’ polarization alongside farm
modernization, with the majority of peasants
becoming wage labourers (Stalin ; Mao ). Evidence
from other developing countries contradicts Stalin
and Mao’s erode vision of rural class polarization
(see, especially, Hayami and Kikuchi ). It indicates
too that appropriate state policies (e.g. land reform,
provision of education and credit, infrastructure
construction, progressive taxation) can mitigate
rural class inequalities. Class polarization is not the
inevitable accompaniment of rural modernization,
nor is collectivization the only way to resolve
problems of rural class inequality (e.g. Hayami and
Kikuchi ).
Fifth, Lenin, Stalin and Mao all believed that
agriculture was characterized by lumpiness and
economies of scale (Lenin ; Stalin ; Mao ). In many
farm tasks, large scale is indeed an advantage, for
example in research,
1789
processing, building and maintaining irrigation
facilities. However, many modem farm inputs are
divisible. Provided they are appropriately priced,
credit is available and they have access to lumpy
complementary inputs, all farm strata modernizing
areas tend to acquire them (Hayami and Kikuchi ).
Moreover, in large agricultural units labour
supervision is a major problem (Bradley and Clark
1971). If a collective’s members trust each other
and are motivated to work hard for the group
irrespective of relative income then labour
supervision is not an issue. However, this is rarely
the case (Morawetz ) and collective farm managers
have had to devise payment systems to motivate
farm workers. In certain farm tasks (notably
harvesting) it is easy to pay labour according to its
product, but for most farm tasks it is more difficult
than in industry to devise payment systems that
strongly motivate from the work often requires a
flexible response from the worker which is difficult
to anticipate in the payment system; the final
produce takes a long time to produce, with different
workers’ contributions difficult to isolate; work is
physically dispersed and production conditions vary
greatly from one part of the production unit to
another; the main task specializations are seasonal,
and permanent minute sub-division of work into
easily measurable segments is not generally
possible. These problems have meant that under
private agriculture, if labour is relatively abundant
and capital relatively expensive, the normal
outcome is for land to be rented out beyond a
certain farm size, so that a relatively high output per
acre can be attained through self-operating, self-
motivated, rent-paying farmers, rather than culti-
vated with large numbers of hired workers. In
collective farms, the attempt to supervise large
numbers of farm workers has resulted in powerful
managerial diseconomies of scale and reduced farm
efficiency.
Collective agriculture has not performed well.
Collective farms in the USSR in 1929-31 and in
China in 1959-61 experienced massive institu-
tionally caused declines in farm output, accompa-
nied by demographic disasters (on the Soviet Union,
see Volin , ch. 10; on China, see Ashton et al. ). It is
indeed, a terrible indictment of collective farming,
that the worst famines
1790
of the 20th century have occurred under that sys-
tem. The USSRs long-term growth of farm output
has required colossal capital outlays so that by the
1970s, the agricultural sector was absorbing over
one quarter of Soviet new fixed investment
(Carey ). From the mid-1950s to the later 1970s
Chinese farm output per caput was stagnant: ‘de-
collectivization’ of agriculture in the early 1980s
was accompanied by a huge in farm output (Nolan
and Paine ).
The ‘socialist’ countries’ poor agricultural per-
formances is in part attributable to shortcomings in
the supply of industrial goods (Smith ). Part is also
due to extensive state intervention in collective
farms. However, there are fundamental problems in
principle even with relatively independent
collective farms. Large units (whether state,
collective or private) are necessary to undertake
activities exhibiting lumpiness or economies of
scale. However, for many farm tasks powerful
managerial diseconomies of scale exist, and even
given favourable policies in other respects, in most
circumstances this would prove a barrier to good
performance of collective farms.
See Also
► Agricultural Growth and Population Change
►::
Bibliography
Ashton, B., et al. 1984. Famine in China, 1958-61. P o p u -
l a t i o n a n d D e v e l o p m e n t R e v i e w 10(4): 613-645.
Baran, P. 1957. T h e p o l i t i c a l e c o n o m y o f
g r o w t h . New York: Monthly Review Press.
Bradley, M.E., and M.G. Clark. 1972. Supervision and
efficiency in socialized agriculture. S o v i e t S t u d i e s
23(3): 465-473.
Carey, D.W. 1976. Soviet agriculture: Recent performance
and future plans. I n U S C J E C 1976.
Cohen, S.F. 1974. B u k h a r i n a n d t h e B o l s h e v i k
r e v o l u t i o n . London: Wildwood House.
Davies, R.W. 1980. The socialist offensive: the collective action
of Soviet agriculture, 1929-1930. London: Macmillan.
Ellman, M. 1975. Did the agricultural surplus provide the
resources for the increase in investment in the USSR
during the First Five Year Plan? E c o n o m i c
J o u r n a l 85(4): 844-863.
Collective Bargaining
General Office of the Central Committee of the Chinese
Communist Party. 1956. S o c i a l i s t h i g h t i d e i n
China 's villages (Zhongguo nongcun de
s h e h u t h u y i g a o c h a o ) , vol 3 vols. Peking:
People’s Publishing House.
Hayami, Y., and M. Kikuchi. 1981. A s i a n v i l l a g e
e c o n o m y a t t h e c r o s s r o a d s . Tokyo: University
of Tokyo Press.
Ishikawa, S. 1967. Resource flow between agriculture and
industry. T h e D e v e l o p i n g E c o n o m i e s 5(1): 3—
49.
Kuo, L.T.C. 1972. The technical transformation of agriculture
in communist China. London: Praeger.
Lenin, V.I. 1899. T h e d e v e l o p m e n t o f c a p i t a l i s m
i n R u s s i a , 1964. Moscow: Progress Publishers.
Mao, Tsetung. 1955. On the co-operative transformation of
agriculture. M a o 1977.
.- - -1977. Selected works of Mao Tsetung,
vol V. Peking: Foreign Languages Press.
Morawetz, D. 1983. The kibbutz as a model for developing
countries. S t e w a r t 1983.
Nolan, P., and S. Paine. 1986. Towards an appraisal of the
impact of rural reform in China, 1978-85. C a m b r i d g e
J o u r n a l o f E c o n o m i c s 10(1): 83-99.
Selden, M. 1982. Co-operation and conflict: co-operative and
collective formation in China’s countryside. S e l d e n
a n d L i p p i t 1982.
Selden, M., and V. Lippit (ed). 1982. T h e t r a n s i t i o n t o
s o c i a l i s m i n C h i n a . New York: M.E. Sharpe.
Smith, G.A.E. 1981. The industrial problems of Soviet
agriculture. C r i t i q u e 14: 41-65.
Stalin, J. 1929. Concerning questions of agrarian policy. In
P r o b l e m s o f L e n i n i s m , ed. J. Stalin. Peking:
Foreign Languages Press n.d.
Stewart, F. (ed). 1983. W o r k , income and
i n e q u a l i t y . London: Macmillan.
Unger, J. 1984. C h e n v i l l a g e . Berkeley: University of
California Press.
US Congress, Joint Economic Committee (USCJEC). 1976.
Soviet economy in a new perspective.
Washington, DC: US Government Printing Office.
Collective Bargaining
William Brown
Keywords
Arbitration; Bargaining; Collective bargaining;
Collusion; Consultation; Employment contracts;
Industrial democracy; Industrial relations;
Negotiation; Trade unions; Wage determination
Collective Bargaining
JEL Classifications
J5
Collective bargaining is a term applied to a variety
of methods of regulating relationship between
employers and their employees. Its distinctive
feature is that it clearly acknowledges a role for
trade unions. In contrast with, for example, auto-
cratic paternalism or producer cooperatives, the
employer who engages in collective bargaining
accepts the right of independent representatives of
employees, acting as a collectivity, to argue their
point of view on matters that affect their interests.
Pay and working conditions are the most common
subjects of collective bargaining, but it can
encompass any aspect of management.
The impact of collective bargaining upon man-
agement, and its effectiveness from the point of
trade union members, vary enormously between
different employment circumstances. They depend
ultimately upon the collective strength that can be
mobilized by employees within the legislative
constraints laid down by the state. Collective
bargaining is thus best seen as a political institution.
It provides a means of bringing at least temporary
reconciliation of divergent interests between
employers and employees in circumstances in
which each side can, to a greater or lesser extent,
inflict damage on the other. It is, however, a
political institution that is intimately linked with
economic processes. The relative power of the
bargaining partners owes much to their respective
labour and product markets. At the same time the
outcome of their bargaining has a major impact
upon both the wages and the productivity of labour.
Theoretical Approaches
This view of collective bargaining as primarily a
political rather than an economic institution is
relatively recent. Beatrice Webb claimed, according
to Marsh ( ), to have originated
the expression in 1891 in her study The Co-
operative Movement of Great Britain. She analysed
it further with her husband Sidney
1791
Webb in Industrial Democracy ( ). Although
they did not define it, they saw it as an alternative
to individual bargaining, so that the employer,
instead of making separate deals with isolated
individuals, ‘meets with a collective will and set-
tles, in a single agreement, the principles on which,
for the time being, all workmen of a particular
group, or class, or grade, will be engaged’. They
identified it as one of three methods used by trade
unions to meet their objectives, the other two being
to establish mutual assurance arrangements for their
members and to press governments to enact
favourable laws. For all the richness of the Webbs’
analysis, collective bargaining remained for them
essentially an economic institution, imposed upon
the employer by a labour cartel whereby workers
secured better terms of employment by controlling
competition among themselves. A naive version of
this view can be seen to underlie much formal
analysis of collective bargaining by present-day
labour economists.
For the next half century Marsh reports no
substantial development of the concept apart from
in Leiserson’s Constitutional Government in
American Industries ( ). Then in
Chamberlain, in his book Collective Bargaining,
argued that there were, in essence, three distinct
theories. ‘They are that collective bargaining is
(1) a means of contracting for the sale of labour,
(2) a form of industrial government, and (3) a
method of management.’ The first, ‘marketing’
theory was much the same as that of the Webbs.
The second, ‘governmental’ aspect was concerned
with the procedural needs of dispute resolution. The
third ‘managerial’ theory referred to the way in
which management and unions in practice
combined ‘in reaching decisions on matters in
which both have vital interests’; unions through
collective bargaining become not the usurpers of
management functions but ‘actually de facto
managers’. At much the same time Harbison ( )
was stressing the very construc
tive social role that collective bargaining played in
resolving industrial conflict and in pushing for the
enhancement of the ‘dignity, worth and freedom of
individuals in their capacity as workers’.
This more complex view of collective
bargaining has been refined by Dunlop ( )
1792
and Kochan ( ) in the United States, but prob
ably the most influential discussion has been Flan-
ders’ attempt of 1968 to create a comprehensive
theoretical analysis. He argued that the economic
associations of the term ‘collective bargaining’ are
misleading. The collective agreement commits no
one to either buy or sell labour, but rather ensures
that, when labour is bought or sold, the terms of the
transaction will accord with the provisions of the
agreement. Above all else, collective bargaining is a
rule-making process covering many aspects of the
employment relationship besides pay and conditions
of work. The second characteristic feature of
collective bargaining that Flanders stressed is that
of the power relationship between the protagonists
whose negotiations (‘the diplomatic use of power’ )
create the rules. Thus, while there are also technical
rales and legal rales regulating work, what
distinguishes the legitimacy of those that result
from collective bargaining is their authorship. They
are jointly determined by the accepted
representatives of both employers and employees
who consequently share responsibility for both the
rales’ contents and their observance.
Flanders’ analysis has proved fertile in several
respects. It has drawn attention to the extent to
which collective bargaining is a positive manage-
ment technique rather than just an impediment to
effective management imposed by trade unions. As
a result of this shift in emphasis, a major part of
academic research into collective bargaining in the
1980s has explored managerial, as opposed to trade
union, strategies, and has exposed the extent to
which union behaviour is shaped by these manage-
ment strategies. In addition, what could be seen as
the Weberian undercurrent in Flanders’ analysis has
focused policymakers’ attention upon the
importance of procedural clarity in conflict resolu-
tion, and thereby upon the dangers of ambiguity in
the legitimation of agreements. The most obvious
example is provided by the influential central rec-
ommendation of the British Royal Commission on
Trade Unions and Employers’ Associations of
1968. The emphasis it placed upon employer initi-
ated procedural reform, rather than legislative con-
straints on trade unions, owed much to the
Collective Bargaining
evidence that Flanders had submitted. Finally, by
conceptualizing wages as part of a broader package
of regulations and as embodying strongly normative
principles, the theory opened the way to a more
fruitful understanding of wage determination than is
offered simply by the market models of orthodox
theory.
Two crucial features of the employment rela-
tionship ensure that the process of collective
bargaining is fundamentally unlike that of non-
labour commercial bargains. They are its open-
endedness and its continuity. The labour contract is
open-ended because the recruitment of an employee
does not ensure the performance of work; the
employee has to be motivated, by whatever means,
to perform to the required standard. In all but highly
oppressive societies such motivational techniques
tend to be varied and complex, differing not least in
the extent to which they place emphasis upon levels
of pay and upon employee participation. Since
social comparisons (and especially very local ones)
play an important part in the motivation and
demotivation of labour, the bureaucratic
standardization of terms of employment, which is
generally a characteristic of collective agreements,
often fits in well with management’s preferred
personnel techniques. In this way, properly
conducted collective bargaining can provide a
socially stable working environment which
facilitates the employer’s prime aim of eliciting
labour productivity. In short, the conduct of the
bargain affects the quality of the labour bargained
over.
The second distinctive feature of the employ-
ment relationship is its continuity. Employer and
employees are bound together, for better or worse,
for an indeterminate duration. Additions to and
departures from the workforce generally occur in a
piecemeal way. A host of potentially contentious
issues feature in the relationship, only a small
minority in contention at any one time, and many
affecting only a minority of the workforce. Thus a
bargain over a particular issue, such as a pay
grievance, cannot be evaluated in isolation, but as
one fibre in a thick rope of regulations, with many
largely implicit trade-offs with respect to other
issues, past, present and future.
Collective Bargaining
Characteristics
The definition of collective bargaining as the joint
regulation of the employment relationship by
employer and employee representatives is one that
covers a broad range ofprocesses. It is helpful to
analyse these further. An initial distinction has to be
made between negotiation and consultation. In a
negotiation the discussions are characterized, first,
by the awareness of each side of the possibility of
one inflicting costs on the other in the absence of an
acceptable outcome. Second, a negotiation has to
result in some sort of agreement, however informal,
to which the two sides are, at least for the time
being, committed. Consultation, by contrast, is
unaccompanied by either the threat of sanctions or
the need to reach binding agreement. Actions taken
by management in the light of consultation result
from a reappraisal of the facts of the case; those
taken after negotiation reflect a compromise which
has taken into account the threat (or experience) of
sanctions inflicted by either or both sides. Under
most collective bargaining arrangements it is felt
advisable by both sides to distinguish as far as is
possible between negotiations and consultations, at
any rate in formal procedures. It is, for example,
now normal in large unionized workplaces in
Britain to deal with them in specifically different
committees, even though the membership of those
committees may be much the same.
In practice the distinction is far from clear-cut.
The blend of approaches adopted in a particular
collective bargaining episode depends very much
upon the issue in question and the relationship
between the parties involved. In their study A
Behavioral Theory of Labor Negotiations ( ),
Walton and McKersie distinguished four
classes of negotiation. First, there were ‘distribu-
tive’ bargains: zero-sum negotiations typified by
annual wage bargains and characterized by very
formal proceedings. Second were ‘integrative’
bargains: problem-solving discussions aiming at
non-zero-sum gains for both sides and generally
much more informal in procedure. Third, was
‘attitudinal structuring’, an almost didactic form of
bargaining dialogue in which one side tries to
1793
alter the way in which their opponents perceive the
problem and its context. Finally, ‘intra-
organizational’ bargains were aimed at altering
positions and attitudes, not on the other side, but
within the negotiator’s own side.
An important influence upon the way in which
bargaining is conducted is the personal ‘bargaining
relationship’ between the two individuals who have
to take the lead in representing the two sides. This
is a term given to the level of trust and facility of
communication that exists between them. However
acrimonious the collective dispute over which they
are bargaining, the better the bargaining
relationship between the individual negotiators, the
more efficiently they will be able to assess each
other’s relative power position and the better the
chance of the dispute being settled without recourse
to expensive sanctions. In a mature bargaining
relationship it is common for the negotiators to
protect each other from their own sides by, for
example, avoiding the humiliation of a bargaining
opponent by helping him to gloss over the
magnitude of a defeat and by manipulating public
statements from one’s own side so as to help in his
intra-organizational bargaining with his own.
It is normal to draw a clear distinction between
the substantive and procedural aspects of collective
bargaining. A substantive agreement sets out the
actual pay levels, working conditions, or whatever
that have been agreed and will be worked to. A
procedural agreement defines the way in which
such substantive terms might be altered, added to,
or interpreted. An effective procedure for
negotiation or grievance settlement will state which
agents on each side are entitled to be involved in
negotiations, in what sequence different sets of
negotiators are entitled to consider the matter, what
their precedence is, and possibly also matters such
as rights of appeal, time constraints, ratification
methods and the form of the substantive outcome.
This distinction is particularly obvious in
countries whose labour laws cause collective
agreements to be tested in the courts; the substan-
tive agreements tend to be written, detailed, formal,
and established for specified duration. There
1794
are other countries where employer preference, or
legal opportunity, makes it unusual for the
bargaining opponents to use legal sanctions against
each other. In these circumstances the great bulk of
substantive regulation may be unwritten and in the
form of verbal agreements, custom, and tacit
understandings. Because of this a greater emphasis
is placed upon the rectitude of the procedural
agreements (which may still be very informal)
whereby this amorphous body of substantive rales is
interpreted and altered, not through comprehensive
periodic negotiations, but by a constant incremental
process of piecemeal adjustment. Although the
United States might be described as exemplifying
the legalistic extreme, and Great Britain the
‘voluntaristic’, most bargaining arrangements have
elements of each, with the degree of legalism and
formality varying by issue and industry, as well as
by country.
Bargaining Structure
The structure of bargaining in a country, industry,
or enterprise, refers to several different characteris-
tics of collective bargaining. The two most impor-
tant are the ‘bargaining units’ and ‘bargaining
levels’ employed. A bargaining unit is a group of
employees covered by a particular agreement.
Within this basic territory of industrial government
there is a coherence of terms of employment, pro-
cedures, and trade union representation that is not
necessarily to be found between different
bargaining units. The level of bargaining refers to
the role played by the principal negotiators within
their organizations; whether, for example, the
employer representative responsible is a factory
manager, a company director, or an employers’
association representative.
These two characteristics are involved in the
single most important decision in the shaping of any
bargaining structure which is whether the employers
confront the unions singly or in alliance. Single-
employer bargaining, resulting in agreements at
company-level or lower, is the majority practice in
the United States and Japan and now in Britain.
Multi-employer bargaining, in
Collective Bargaining
which associations of employers conclude
industrywide agreements, remains the most
important form in most of Continental Europe. In
practice there is often some employer collusion in
industries where single-employer bargaining
dominates, and there is usually room for individual
employer discretion in industries with strong
employers’ associations, but the distinction remains
one of fundamental economic, political, and
managerial significance.
Two other defining characteristics of bargaining
structure are its ‘form’ and ‘scope’. The first refers
to the extent to which proceedings and agreements
are formalized and codified. As already mentioned,
this depends in part upon the labour legislation of
the country. The second matter, scope, refers to the
range of issues covered by collective bargaining. At
its narrowest it may include no more than pay and
hours, while elsewhere it may take in issues as
diverse as training policy, investment decisions and
child-care facilities.
The most comprehensive theory seeking to
explain industrial and national differences in
bargaining structure is to be found in Clegg’s Trade
Unionism under Collective Bargaining ( ).
This sees the strategy adopted by employers
as the main determinant of bargaining structure,
although changes in strategy may be slow to take
effect. The legislative framework of a country is
also of crucial importance. It defines the limits of
rights to strike, the status of the employment con-
tract, any guarantees of security for trade unions,
and the legally responsible agents on each side.
Most countries acquired their principal labour
legislation at some historic period of crisis - war,
defeat, depression, or extreme industrial unrest - and
the institutional arrangements that developed from
that have become consolidated in subsequent, more
peaceful times. This helps to account for the very
great variations in collective bargaining practice to
be found in different countries; they often owe their
origin to a distant panic measure based upon a
fashionable idea (such as, for example, compulsory
arbitration in Australia or compulsory conciliation
in Canada) to which employers and unions have
adjusted so firmly that radical reformation is all but
impossible. A recurring experience around the
world is of
Collective Choice Experiments
legislatures finding extreme difficulty in reforming
collective bargaining, other than in times of extreme
crisis, because of the essential privacy of the
bargaining relationship between employers and
union.
Most industrialized countries publicly assert a
commitment to collective bargaining as a necessary
part of a democratic society, and for most it is the
normal means of conducting industrial relations in
the public sector. Convention 84 (1947) of the
International Labour Organization asserts that ‘all
practical measures shall be taken to assure to trade
unions which are representative of the workers
concerned die right to conclude collective agree-
ments with employers and employers’ associations’.
In practice the freedom of collective bargaining in
both public and private sectors varies substantially
between countries and over time.
No discussion of collective baigaining would be
complete without a mention of the debate
concerning its relationship with industrial
democracy.
One view is that, because collective bargaining
is essentially concerned with compromise, trade
unions are sucked into collaborating with capitalism
and thereby denied the opportunity of uniting the
working class in overthrowing existing employers
and then instituting true industrial democracy
through workers’ control. Opposing this is a view
that deplores the fact that collective bargaining
institutionalizes the opposition of capital and
labour: them and us. It considers that the best form
of industrial democracy is to be found where
workers are brought to perceive an u Iti mate
identity of interest with employers. Between these
positions is that most clearly expressed by Clegg in
A New Approach to Industrial Democracy ( ).
This argues that there can never be com
plete identity of interest between employer and
employee, and also that if employee representatives
are given managerial responsibilities they will be
forced to behave very similarly to the employers
they have replaced. Consequently the role of the
trade union is best seen as one of constant
opposition, acting to modify management actions in
the light of members’ interests insofar as their
organized power permits. Far from undermining the
common interests of capital and labour, collective
bargaining permits the joint
1795
regulation of aspects of employment which would
otherwise generate greater disharmony and division.
See Also
Bibliography
Chamberlain, N.W. 1951. C o l l e c t i v e b a r g a i n i n g .
New York: McGraw-Hill.
Clegg, H.A. 1960. A new approach to industrial democracy.
Oxford: Blackwell.
Clegg, H.A. 1976. Trade unionism under collective bargaining.
Oxford: Blackwell.
Dunlop, J.T. 1967. The social utility of collective bargaining.
In C h a l l e n g e s t o c o l l e c t i v e b a r g a i n i n g ,
ed. L. Ulman. New York: Prentice-Hall.
Flanders, A. 1968. Collective bargaining: A theoretical
analysis. B r i t i s h J o u r n a l o f I n d u s t r i a l
R e l a t i o n s 6 (1): 1-26. Reprinted in Flanders, A. 1975.
M a n a g e m e n t a n d u n i o n s . London: Faber &
Faber.
Harbison, F.H. 1951. Goals and strategies in collective
bargaining. New York: Harper.
Kochan, T.A. 1980. Collective bargaining and industrial
relations. Homewood: Irwin.
Leiserson, W.M. 1922. Constitutional government in
American industries. A m e r i c a n Economic
R e v i e w 12 (Suppl): 56-79.
Marsh, A. 1979. Concise encyclopedia of industrial relations.
Famborough: Gower.
Walton, R.E., and R.B. McKersie. 1965. A b e h a v i o r a l
t h e o r y o f l a b o r n e g o t i a t i o n s . New York:
McGraw-Hill.
Webb, S., and B. Webb. 1897. I n d u s t r i a l d e m o c r a c y .
Collective Choice Experiments
Rick K. Wilson
Abstract
Collective choice experiments examine voting
mechanisms that aggregate individual prefer-
ences. Two general topics have received the
most attention. The first pertains to agents
deciding on a single collective outcome or
1796 Collective Choice Experiments

policy. The second topic covers election mech- fixed preferences over the dimensions, and policies
anisms that govern candidates and voters. represented as points in the space. Using rules that
mimic many parliamentary systems, these theoreti-
Keywords cal papers demonstrate that a Condorcet winner (a
Agenda control; Arrows; Theorem; Collective policy that can defeat all others under pairwise
choice experiments; Electoral mechanisms; voting) exists only under rare distributions of
Median voter; Social choice; Spatial committee voters’ preferences. Plott ( ) establishes the con
experiments ditions under which a Condorcet winner will exist
and he makes the connection between this and a
Nash equihbrium of a spatial committee game. Like
JEL Classifications others, he concludes that an equihbrium is rare in
C9 multidimensional spatial committee games.
Early spatial committee experiments by Berl et
Duncan Black ( ) and Kenneth Arrow ( ) al. ( ) and Fiorina and Plott ( ) provide
raised the key question of collective choice: if evidence that when a Condorcet winner exists, sub-
people have different preferences for policy out- jects choose it or outcomes that are close to it. In
comes are there general mechanisms that can games where there is no such equihbrium (which is
(always) aggregate those preferences in consistent the most common case), subjects select outcomes
and coherent ways? The answer is ‘no’. Starting that scatter in the policy space. These initial empir-
from simple premises involving individual transi- ical findings, coupled with experiments by Laing
tivity, aggregate Pareto optimality and non- and Olmsted ( ) and McKelvey et al. ( ),
dictatorship there is no collective choice mechanism defined the standard for conducting spatial commit-
that yields a socially transitive outcome. Such a tee experiments. Subsequent experiments have
finding is startling given the confidence placed in adopted almost identical procedures.
democratic institutions that rely on voting The standard experimental design introduces a
mechanisms to choose a single outcome from many two-dimensional policy space. The orthogonal
possible outcomes. dimensions are arbitrary (X and Y in most settings)
Experimentalists have thoroughly explored dif- and typically range from zero to 200 or more units.
ferent institutions that can be used to aggregate Every point in the space characterizes a policy.
preferences. Political economists who straddle both Preferences over outcomes are induced by assigning
economics and political science have carried out each subject a payoff function mapping earnings in
much of this work. Their concern is with situations dollars to each point in the space. While many
where actors who have opposed interests have to payoff functions have been tested, most
settle on a single outcome and with the properties of experimenters have settled on a quadratic loss
the institution used to produce an outcome. This function, with monetary payoffs decreasing as a
article first turns to the institutional mechanisms by function of distance from a subject’s ideal point.
which individuals settle on a collective outcome. Usually five subjects are assigned different ideal
The second topic turns to electoral mechanisms points in the space, and it is the arrangement of
used in representative democracies. these ideal points that allows the experimenter to
manipulate, whether a Condorcet winner exists or
not. Subjects are given an initial status quo and then
Spatial Committee Experiments
allowed to introduce amendments. Voting takes
In the late 1960s theoretical papers by Davis and place following an amendment, with the winner
Hinich ( ) and Plott ( ) described a social becoming (or remaining) the new status quo.
choice environment for spatial committees. Those Amending takes place in between votes. A motion
committees consist of a well-defined multi- to adjourn, passed under a voting rule, constitutes
dimensional policy space, with actors holding the stopping rule for the committee decision. This
serves as the standard institution
Collective Choice Experiments
for subsequent spatial committee experiments.
Changing these basic institutional rules became the
way to test theories of collective choice.
Experimental results in the absence of equilib-
rium are both frustrating and profitable. Frustration
arises over the fact that committee choices tend to
be clustered in similar regions of the policy space.
While there appears to be some pattern to the out-
comes, the process by which these outcomes arise
has not been frilly characterized (but see the attempt
by Bianco et al. ). Profitably, these empirical results
led theorists and experimenters to add agenda
control to the structure of the game. This led to a
distinction between preference-induced and
structure-induced equilibrium. For example, Plott
and Levine ( ) showed the effectiveness
of agenda control both in the laboratory and in a
natural setting. Awarding agenda power created a
structure-induced equilibrium and laboratory sub-
jects converged to it Recent experimental work by
Frechette et al. ( ) illustrates that the equilib
rium favours agenda setters.
Theoretical work by Buchanan and Tullock (
) led experimentalists to examine whether
changing the proportion of actors needed to pass a
policy had any effect. Experiments by Laing and
Slotznick ( ) showed that moving from simple
majority rule (50 per cent plus 1) to supermajority
majority rule (67 per cent) resulted in many equi-
libria and that subjects chose them. Schofield ( ),
among others, provided the theoretical
basis for when an equilibrium exists as a function of
the dimensionality of the policy space, the voting
rule and the distribution of voters’ preferences.
These theoretical findings spurred experimentalists
to examine other changes to the standard committee
experiment. For example Wilson and Herzberg (
) theoretically pre
dicted and experimentally demonstrated that when a
single player holds veto power, that player’s ideal
point is the equilibrium. Haney et al. ( ) point. Candidates are paid only if they win. Once
empirically show committee choices
converging to equilibrium when a weighted voting
rule is used. Such a rule requires that a single player
always be included in a coalition. These results are
representative of the kind of work that has
dominated the experimental spatial committee
agenda.
1797
Experiments on spatial committees have added
to a clearer understanding of institutional mecha-
nisms. Experimental results demonstrate that
changing who has the power to set the agenda, how
the agenda is built, how many votes are needed and
whether players enjoy veto powers, matters.
Electoral Mechanisms
A second area of interest for collective choice
experimentalists is with electoral mechanisms.
Three broad directions have been taken that treat
different aspects of representative democracies. The
first is concerned with candidate behaviour. At the
heart of this research is the question of whether
candidate positions will converge to equilibrium
when it exists. The second direction is concerned
with voter behaviour, particularly how voters
behave when they have little information about
candidate positions. The final direction deals with
the way in which electoral rules determine the
likelihood that ‘types’ of candidates are elected,
where types usually refer to racial and ethnic
minority candidates.
The initial experimental work on candidate
behaviour focused on candidates who cared only
about winning and varied the information condi-
tions that the candidates have about the preferences
of voters. Most experiments use a unidimensional
policy space that guarantees an equilibrium. This
equilibrium is defined by the policy preference of
the median voter. In the experiments elections are
sequential, with two candidates announcing posi-
tions in the policy space and voters choosing
between the candidates. Voters are assigned ideal
points in the policy space, the winning candidate is
required to implement the announced policy and
voters are paid an amount that decreases with the
distance of the winning position from their ideal
the election is over another election is held with
candidates free to change their previously
announced policy. Not surprisingly, all candidates
quickly adopt the position of the median voter when
they are fully informed about voter preferences.
Under incomplete information about voters.
1798
candidates also converge to the median voter’s
position, by responding to feedback about the vote
share accruing to different policy positions, as in
McKelvey and Ordeshook ( ). If candi
dates have policy preferences whereby their earn-
ings depend not only on winning but also on
implementing a policy close to their own preferred
position, then the median voter result no longer
holds (see the experimental results by Morton )•
When voters are uninformed about candidate
positions, are they able to cast accurate ballots?
With minimal information, such as biased
endorsements or polls, subjects do very well at
inferring candidate positions. Lupia and McCubbins
( ) and Morton and Williams
( ) consider various aspects of voter informa
tion and show that voters are able to quickly
determine the positions of candidates and cast their
vote accordingly.
Finally, several experiments have focused on
differing electoral mechanisms and what they mean
for the type of candidates that gain election. For
example Gerber et al. ( ) compare two
voting mechanisms in an experiment to test whether
one or the other disadvantages a racial or ethnic
minority candidate. A form of cumulative voting (in
which voters can cast more than a single vote) leads
to more minority candidates being elected. This
should be no surprise to collective choice theorists
who have long noted that different electoral
mechanisms lead to predictable variation in
outcomes. Cox ( ) offers an
extended discussion of such mechanisms.
What We Know
Collective choice experiments provide several
insights. First, when a Nash equilibrium of the
underlying game exists it is a strong predictor of the
outcome of the experiment. The second finding is
that when there is no Nash equilibrium for the
underlying game, subjects choose outcomes that
cluster in predictable areas of the policy space, but
the process by which that occurs is not settled. At
the same time, experimentalists have implemented
institutional mechanisms
Collective Choice Experiments
altering such games, thereby producing an equi-
librium that subjects choose. Often those institu-
tional changes benefit one actor (for example, by
assigning agenda control to a particular player). A
third finding is that incomplete information does
not prevent convergence to equilibrium for either
candidate platform choice or voter behaviour. The
fourth finding returns to Arrow’s original insight:
voting mechanisms can be
manipulated to achieve predictable, but very dif-
ferent, outcomes. It all depends on the mechanism
that is implemented.
See Also
►
►
►
Bibliography
A r r o w , K . J . 1 9 6 3 . Social choice and individual values.
New Haven: Yale University Press.
Berl, J.E., R.D. McKelvey, PC. Ordeshook, and M. D. Winer.
1976. An experimental test of the core in a simple n-person
cooperative nonsidepayment game. Journal of Conflict
Resolution 20: 453-476. Bianco, W.T., M.S. Lynch, G.J.
Miller, and I. Sened. 2006. ‘A theory waiting to be discovered
and used’: A reanalysis of canonical experiments on majority
rule decision making. Journal of Politics 68: 837-850. Black,
D. 1948. On the rationale of group decision making.
Journal of Political Economy 5 6 : 2 2 - 3 4 .
Buchanan, J., and G. Tullock. 1962. Calculus of consent.
Ann Arbor: University of Michigan Press.
C o x , G . W . 1 9 9 7 . Making votes count: Strategic coordina-
tion in the world s electoral systems. N e w Y o r k : C a m -
bridge University Press.
Davis, O.A., and M.J. Hinich. 1966. A mathematical model of
policy formation in a democratic society. In Mathema tical
applications in political science, ed. J.L. Bemd. Dallas, TX:
Southern Methodist University. Fiorina, M.P., and C.R. Plott.
1978. Committee decisions under majority rule: an
experimental study. American Political Science Review 72:
575-598.
Frechette, G., J.H. Kagel, and J.H. Lehrer. 2003. Bargaining
in legislatures: An experimental investigation of open
versus closed amendment rules. American Political
Science Review 97: 221-232.
Collective Models of the Household 1799

Gerber, E.R., R.B. Morton, andT.A. Rietz. 1998. Minority and (b) the decision process results in Pareto-
representation in multimember districts. A m e r i c a n
efficient outcomes. The main results of the
P o l i t i c a l S c i e n c e R e v i e w 92: 127-144.
Haney, R, R. Herzberg, and R.K. Wilson. 1992. Advice and theory of collective models then refer to the
consent: Unitary actors, advisory models and empirical issue of deriving testable restrictions
experimental tests. J o u r n a l of Conflict on household behaviour and recovering from
R e s o l u t i o n 36: 603-633.
this some information on the structural model
Laing, J.D., and S. Olmsted. 1978. An experimental and game
theoretic study of committees. In G a m e T h e o r y a n d that can be used to carry out welfare compari-
P o l i t i c a l S c i e n c e , ed. P.C. Ordeshook. New York: sons at the individual level.
New York University Press.
Laing, J.D., andB. Slotznick. 1983. Winners, blockers, and the
status quo: Simple collective decision games and the core. Keywords
P u b l i c C h o i c e 40: 263-279. Collective models of the household; Exclusive
Lupia, A., and M.D. McCubbins. 1998. T h e d e m o c r a t i c goods; Household behaviour; Indirect utility
dilemma: Can citizens learn what they function; Pareto efficiency
n e e d t o h t o w ? Cambridge: Cambridge University
Press.
McKelvey, R.D., and PC. Ordeshook. 1985. Sequential
elections with limited information. A m e r i c a n JEL Classifications
J o u r n a l o f P o l i t i c a l S c i e n c e 29: 480-512. Dll
McKelvey, R.D., PC. Ordeshook, and M.D. Winer. 1978. The
competitive solution for n-person games withoutUntil recently ‘unitary’ models, which assume that
transferable utility with an application to competitive
household members act as if they maximize a
games. A m e r i c a n P o l i t i c a l S c i e n c e R e v i e w
72: 599-615.
unique utility function under a budget constraint,
were largely predominant in the literature on
Morton, R.B. 1993. Incomplete information and ideological
household behaviour. There is increasing agree-
explanations of platfomr divergence. A m e r i c a n
P o l i t i c a l S c i e n c e R e v i e w 87: 382-392.
ment, however, that economists cannot ignore the
Morton, R.B., and K.C. Williams. 2001. Learning by voting:
fact that most households are composed of several
Sequential choices in presidential primaries and other
individuals who take part in the decision process.
elections. Ann Arbor, MI: University of Michigan Press.
Consequently, the ‘collective’ models, which pos-
Plott, C.R. 1967. A notion of equilibrium and its possibility
under majority rule. A m e r i c a n Economic
tulate that (a ) each household member has specific,
R e v i e w 57: 787-806.
generally different preferences and (b) the decision
Plott, C.R., and M.E. Levine. 1978. A model of agenda
process results in Pareto-efficient outcomes, have
influence on committee decisions. A m e r i c a n
E c o n o m i c R e v i e w 68: 146-160. attracted considerable attention from the profession
Schofield, N. 1985. S o c i a l C h o i c e a n d D e m o c r a c y .
during recent years.
Heidelberg: Springer.
To examine the properties of collective models,
Wilson, R.K., andR.Q. Herzberg. 1987. Negative decision
let us consider a household consisting of two
powers and institutional equilibrium: Experiments on
persons, A and B, who make decisions about
blocking coalitions. T h e W e s t e r n P o l i t i c a l
Q u a r t e r l y 40: 593-609. consumption. These persons are characterized by
well-behaved utility functions of the form: M,(X^,
\B, X), where x, denotes a vector of private goods
consumed by member i and X a vector of public
Collective Models of the Household goods (i = A, B). This specification of preferences is
very general; it allows for altruism but also for
Olivier Donni externalities or any other preference interaction. We
denote the vector of prices for private goods by p,
the vector of prices for public goods by P and the
household total expenditure by y. Finally, we
Abstract suppose that there exists a vector of distribution
Collective models of the household are based on factors, that is, a set of exogenous variables which
two fundamental assumptions: (a) each agent is influence the intra-household
characterized by specific preferences
1800
allocation of resources without affecting prefer-
ences or the budget constraint. Examples are given
by the respective contribution of each member to
the exogenous household income, the state of the
marriage market or divorce legislation. These
variables, which are often assigned a crucial role in
the derivation of the results, are denoted by s.
To simplify notation, let n' = (p', P') be the
vector of prices. Then, efficiency essentially means
that household behaviour can be described by the
maximization of a utilitarian social welfare
function, that is,
max f i ( n , y , s ) u A ( x A , x B , X )
xA,XB,X
+ (1 - f j . ( n , y , s ) ) u B ( x A , x B , X ) (1)
subject to p' (X | + xB) + P'X= y. In this pro-
gramme, the function ji determines the location of
the household equilibrium along the Pareto frontier.
If q = 1, then the household behaves as though
member A always gets her way whereas, if /< = 0, it
is as if member B is the effective dictator. We
denote the solutions to Eq. ( ) by x, (n, y s), xB (n, y s)
and X(TT, y s).
Characterization
The first objective of the theory of collective
models is to investigate the properties of the house-
hold demands derived from Eq. ( ). These properties
can either be tested statistically or be imposed a
priori for simplifying the estimation task. From this
perspective, one crucial point is that individual
demands for private goods, xA and xB, are generally
unobservable by the outside econometrician;
demands for these goods are observed only at the
household level, x xA + xB To be usefiil, the
restrictions derived from the collective setting have
thus to characterize household demands, x or X,
instead of individual demands, xA and xB.
Let = (x', X') be the vector of household
demands. We define the Pseudo-Slutsky matrix as
follows:
Collective Models of the Household
There exist at least three different sets of
testable restrictions that characterize household
behaviour.
SRI Condition
Browning and Chiappori ( ) and Chiappori
and Ekeland ( ) show that household
demands compatible with Eq. ( ) have to satisfy the
following condition:
s = E +Ri,
where E is a symmetric, semi-definite matrix
and R! is a rank one matrix. The interpretation is the
following. For any given pair of utility functions,
(a) the budget constraint determines the Pareto
frontier as a function of n and y and (b) the value of
/i determines the location of the household
equilibrium on this frontier. Consequently, a change
in n implies a shift of the Pareto frontier. The latter
entails the modification of household demands
described by X, However, the value of /i varies as
well, hence the location of the equilibrium moves
along the Pareto frontier. Since the frontier is of
dimension one, this effect is very restricted and
defined by Ri.
Proportionality Condition
The particular structure of Eq. ( ) leads to further
restrictions on behaviour. To make things simple,
let us suppose that the vector of distribution factors
is twodimensional: s = (si, s2). Then, Bourguignon
et al. ( ) demonstrate the following result:
=e
ds\ dsx ’
where 6 is a scalar. Thus, the response to differ-
ent distribution factors is co-linear. The interpreta-
tion is that distribution factors can only change the
location of the outcome on the frontier (through
function ji), and the latter is of dimension one.
Specific Conditions
The econometrician is often inclined to put more
structure on preferences. For example, let us sup-
pose that agents have utility functions of the form:
«,(x„ X). In that case, we say that agents are
Collective Models of the Household
‘egoistic’ in the sense that the utility does not
depend on the partner’s consumption. This
assumption implies, in particular, that the decision
process can be decentralized. In a first step, house-
hold members agree on the level of public goods as
well as on a particular distribution of the residual
expenditure between them. In a second step, they
maximize their utility function, taking into account
the level of public goods and their own budget
constraint. It means, formally, that there exists a
pair of functions (pjp. X, y , s), pB(p, X, /, s)),
satisfying pA + pB = y where y = y - P'X, such that
the demand for private goods by member i is the
solution to
maxM;(x/, X) subject to p'x, = pr
Hence, household demands for private goods,
conditionally on the demands for public goods, can
be written as:
x = xA(p,X,p(p,X,y*,s))
+ xB(p,X,y* - p(p, X , y* , s) ) ,
where p = pA and y — p = pB. This structure
generates strong testable restrictions because the
same liinction p(p, X, y , s) enters each demand for
private goods. Bourguignon, Browning and
Chiappori ( ) exphcitly derive these restrictions
under the form of partial differential equations,
whereas Donni ( ) shows that the demands for
public goods have a particular but different struc-
ture, which implies testable restrictions as well.
Welfare Analyses - Identification
One of the main sources of interest in collective
models is to provide the theoretical background for
performing welfare comparisons at the individual
level. The key concept in that case is what
Chiappori ( ) calls the ‘collective’ indirect
utility function. Let us suppose again that agents are
egoistic. If so, the collective indirect utility function from collective models in that the underlying
is defined as follows:
Vi(n,y, s) = Ui(xi(n,y,s),X(n,y,s)). ‘strategic’ models (which are sometimes
(2 )
1801
This expression describes the level of welfare
that member i attains in the household when he or
she faces the price-income bundled, v) and a set of
distribution factors s. This representation of utility
differs from the ‘unitary’ indirect utility function in
that it implicitly includes the sharing function, and
hence an outcome of the collective decision process.
However, the knowledge of Eq. ( ) is usually
sufficient to evaluate the impact of economic
policies on individual welfare.
In general, if agents are egoistic, the collective
indirect utility functions can be retrieved. None-
theless, the econometrician must observe the
demand for some specific goods, referred to as
‘exclusive’, which benefit only one person in the
household. More precisely, we say that good X (x)
is exclusively consumed by member i if du/d X =
0(iduj/dxj = 0) for j / i. The intuition is that the
household demand for ‘exclusive’ goods can be
used as an indicator of the distribution of bargaining
power within the household. Donni ( ) considers the
case of purely private con
sumption (X = 0) and shows that, if there is a single
exclusive good, the collective indirect utility
functions can be identified up to composition by an
increasing transformation. Similarly, Chiappori and
Ekeland ( ) consider the oppo
site case of purely public consumption (x = 0) and
show that, if there are two exclusive goods (one for
each member), the identification is still possible.
However, the general case with both private and
public consumption has not been completely treated
until now; see Blundell, Chiappori and Meghir (
) for a first investigation.
Bibliographical Note
The main idea of collective models can be traced
back to Leuthold ( ), who estimates a model
of household labour supply based on non-
cooperative game theory, where the individual is the
basic decision-maker. However, this model differs
decision process does not result in efficient
outcomes. It actually belongs to the family of
1802
referred to as ‘collective’ models in a broad sense).
Nevertheless, a significant advance towards the
development of collective models is made by
Manser and Brown ( ) and McElroy
and Homey ( ) at the beginning of the 1980s.
These authors study the properties of models based
on bargaining theory, which implies Pareto-
efficiency. In that case, the location along the
Pareto frontier is determined by the Nash (or Kalai-
Smorodinsky) solution. However, the first formal
investigation of a model based on the sole
efficiency assumption is due to Chiappori ( , E c o n o m e t r i c a 56: 63-89.
) in the context of labour supply deci
sions. This model is not explicitly examined in this
article because it can be seen as a particular case of
the model of consumption. Note, however, that
Apps and Rees ( ), Chiappori (
Donni ( ), and Fong and Zhang ( ) pre
sent theoretical extensions of Chiappori’s initial
model, whereas Chiappori, Fortin and Lacroix (
) exhibit empirical results. Finally, we must
mention that several authors have generalized col-
lective models to inter-temporal decisions and
uncertain environment. One of the most represen-
tative examples of these studies is given by
Mazzocco ( ).
See Also
► 'amity Decision Making
► Gender Roles an: E /ision c I >our
► louseho
► lousehold Survf :
► ndividualism
► Integrability of Demand
► Intrahousehold Welfare
► Labe pply
► Rotten Kid Theorem
Bibliography
Apps, P., and R. Rees. 1997. Collective labor supply and
household production. J o u r n a l o f P o l i t i c a l
E c o n o m y 105: 178-190.
Collective Models of the Household
Blundell, R., P.-A. Chiappori, and C. Meghir. 2005. Collective
labor supply with children. J o u r n a l o f P o l i t i c a l
E c o n o m y 113: 1277-1306.
Bourguignon, R, Browning, M., and P.-A. Chiappori.
1995. The collective approach to household behaviour.
Working Paper. Paris: DELTA.
Bourguignon, R, M. Browning, P.-A. Chiappori, and V.
Lechene. 1993. Intrahousehold allocation of consumption:
A model and some evidence from French data. A n n u l e s
d ' e c o n o m i c e t d e s t a t i s t i q u e 29: 137-156.
Browning, M., and P.-A. Chiappori. 1998. Efficient
intrahousehold allocations: A general characterization
and empirical tests. E c o n o m e t r i c a 66: 1241-1278.
Chiappori, P.-A. 1988. Rational household labor supply.
Chiappori, P.-A. 1992. Collective labor supply and welfare.
J o u r n a l o f P o l i t i c a l E c o n o m v 100: 437-467.
Chiappori, P.-A. 1997. Introducing household production in
collective models of labor supply. J o u r n a l o f P o l i t -
i c a l E c o n o m y 105: 191-209.
), Chiappori, P.-A., and I. Ekeland. 2003. The micro economics
of group behavior: Identification. Working paper.
Chicago: University of Chicago.
Chiappori, P.-A., and I. Ekeland 2006. Characterizing group
behavior. J o u r n a l of Economic Tlieoiy
(forthcoming).
Chiappori, P.-A., B. Fortin, and G. Lacroix. 2002. Marriage
market, divorce legislation, and household labor supply.
J o u r n a l o f P o l i t i c a l E c o n o m y 110: 37-72.
Donni, O. 2003. Collective household labor supply: Non-
participation and income taxation. J o u r n a l o f
P u b l i c E c o n o m i c s 87: 1179-1198.
Donni, O. 2004. The intrahousehold allocation of private and
public consumption:Theory and some evidence from U.S.
data. Working paper. Cergy-Pontoise: University of
Cergy-Pontoise.
Donni, O. 2006. Collective consumption and welfare.
C a n a d i a n J o u r n a l o f E c o n o m i c s 39: 124-144.
Fong, Y., and J. Zhang. 2001. The identification of
unobservable independent and spousal leisure. J o u r n a l
o f P o l i t i c a l E c o n o m y 109: 191-202.
Leuthold, J. 1968. An empirical study of formula transfers
and the work decision of the poor. J o u r n a l o f
H u m a n R e s o u r c e s 1: 312-323.
Manser, M., and M. Brown. 1980. Marriage and household
decision making: A bargaining analysis.
I n t e r n a t i o n a l E c o n o m i c R e v i e w 21:31-44.
Mazzocco, M. 2005. Household intertemporal behavior: a
collective characterization and empirical tests. Working
paper. Madison: University of Wisconsin.
McElroy, M., and M. Homey. 1981. Nash bargained
household decisions. I n t e r n a t i o n a l E c o n o m i c
R e v i e w 22: 333-349.
Collective Rationality
Collective Rationality
Lu Hong
Abstract
This article reviews the concepts of individual
rationality and collective rationality as they
appear in the economics literature. In particular,
the existing literature on social choice and
aggregate demand points to a fundamental dis-
connect between these two notions of rational-
ity. A possible reconciliation of this disconnect
is suggested.
Keywords
Aggregate demand; Arrow’s impossibility the-
orem; Collective choice; Collective rationality;
Debreu-Mantel-Sonnenschein theorem;
Gibbard-Satterthwaite impossibility theorem;
Individual rationality; Prisoner’s dilemma;
Rational choice; Sen, A.; Social choice; Social
welfare function; Strategic behaviour
JEL Classifications
D02; D71; D72; D81; D82
Since ancient times, men have argued that choice
should be governed by ‘desire and reasoning
directed to some end’ (Sen ). Much modem
economic theory is based on this rational choice
principle paradigm. In an individual choice prob-
lem, the individual is assumed to have a preference
ordering on the set of alternatives. The individual
choice is rational if, for any given decision
situation, the choice made is always the best among
all feasible alternatives according to the preference
ordering. In a collective choice problem, be it that
of a society or a committee, the definition of this
rational choice principle becomes problematic. As
there is presumably a huge disparity among the
desires and ends of the individuals within the
collective, by whose desire
1803
and whose end should the collective choice be
governed? Is it reasonable to expect the collective
choice to be guided by a preference ordering? If so,
how should it reflect individual preferences, as the
choice made by the collective influences everyone
in it?
Collective Rationality and Social Choice
Of particular interest to the idea of collective
rationality is the study of social choice.
In a seminal work, Arrow ( ) connects
collective rationality to social choice through the
idea of the existence of a social welfare function.
Formally, consider a large set of conceivable alter-
natives, X, that a society faces. A preference
ordering R (weakly preferred) on X is a binary
relation on X that is both complete and transitive. Its
asymmetric and symmetric parts are denoted by P
(strictly preferred) and /(indifferent) respectively.
There are n number of individuals in the society.
Each individual i has a preference ordering Rj on the
set X. A social welfare function (SWF), F, maps a
profile of individual preference orderings (R lv. .,/?„)
to a preference ordering on X. The preference
ordering F{RX.. ,,R„) is then interpreted as the
society’s preference on X for the society consisting
of individuals with preference orderings (Ru.. .,R„).
If such an SWF exists, then the social choice to be
made from any set of feasible alternatives can be
determined by comparing any pair of feasible
alternatives according to the society’s preference.
The social choice thus made is guided by a
preference ordering to reach the best among feasible
alternatives - collective rationality is achieved. In
other words, such an SWF, if it exists, is a
preference aggregation procedure aggregating
individual preference orderings into a society’s
preference ordering according to which a rational
choice can be made.
In isolation, collective rationality is trivial to
reach because an SWF always exists. For example,
take any preference ordering R on X.; the constant
function that maps every possible profile of
1804
individual preference orderings to R is an SWF.
Obviously, this SWF is not meaningful since no
information about individual preferences is
reflected by society’s preferences. For an SWF to
reasonably aggregate individual preferences, some
minimal set of conditions should be imposed.
Arrow ( ) considers four conditions: U -
(universal domain: an SWF’s domain contains all
possible individual preference orderings), P (Pareto
principle: if all individuals strictly prefer one alter-
native to another, then the society strictly prefers
the first alternative to the second), I (independence
of irrelevant alternatives: the way the society ranks
a pair of alternatives should depend only on the way
individuals rank the same pair, not on how they
rank any other alternatives), and D - (non-
dictatorship: no single individual always gets to
determine the society’s preference). He shows the
famous Arrows Impossibility Theorem'. It is
impossible to have a social welfare function
satisfying U, P, I and D simultaneously. In other
words, collective rationality is impossible to
achieve universally if society is to take into account
all individuals in a minimally reasonable way.
Arrow’s Impossibility Theorem jump-started the
modem day study of social choice. In the huge
literature of social choice theory, two strands
directly relate to collective rationality formulated in
tire context of Arrow’s Impossibility Theorem. One
strand focuses on identifying domain restrictions so
that social welfare functions satisfying Arrow’s
three other conditions exist. For example, the SWF
which derives society’s preference from majority
voting on each pair of alternatives (majority rale)
with universal domain will lead to many cycles in
society’s preference, violating the transitivity
requirement of a preference ordering. However, if
individual preferences are restricted to those that are
single-peaked when alternatives can be represented
in one dimension, then majority rule will not
generate cycles and satisfies all other requirements
of Arrow’s Theorem. In general, this strand of
literature proves that collective rationality can be
meaninglully restored for some restricted domains
(Gaertner ). However, domain restrictions are
severe, and outside of them the problem of society’s
preference cycles is global (McKelvey )•
Collective Rationality
The second strand of literature directly exam-
ines the formulation of collective rationality in the
definition of Arrow’s social welfare function.
Arrow’s SWF requires society’s preferences to be
orderings, that is, binary relations that are complete
and transitive. Suppose that we weaken collective
rationality to requiring only that society’s
preferences be, say, acyclic as opposed to fully
transitive. Can impossibility then be avoided? More
generally, is the strong collective rationality
formulated by requiring society’s preferences to be
orderings to blame for the impossibility? This line
of research concludes that, even with a weakened
notion of collective rationality, the impossibility
remains (Sen ). Therefore, social choice cannot be
expected to be collectively rational, even weakly.
Collective Rationality and Strategic
Behaviour
The aforementioned work implicitly assumes that
truthful individual preferences are aggregated. If,
instead, strategic behaviour is allowed, then even if
we require a social choice function to be only non-
dictatorial (a social choice function maps a profile
of individual preferences into an alternative - a
choice of the society), every such social choice
function can be manipulated. This is the Gibbard-
Satterthwaite impossibility theorem (Gibbard ;
Satterthwaite ). That is, even if the collective makes
up its mind about what is good for the society in a
given circumstance, as long as individuals are free
to report their preferences and the collective does
not always choose the top alternative of a given
agent’s reported preference, then the collective’s
goal cannot be achieved.
Collective Rationality and Aggregate
Demand
The disconnection between collective rationality
and individual rationality exists in other areas of
economics. In consumer demand theory, the
Debreu-Mantel-Sonnenschein theorem (Debreu
Collective Rationality
; Mantel ; Sonnenschein ) states that generally
aggregate demand functions do not exhibit any
regularity (such as being downward sloping
regarding price) even when all individual demand
functions are derived from rational decisions in the
sense of preference maximization under budget
constraints. More specifically, for any given shape
of the aggregate demand function (not necessarily
downward sloping), there exists a preference
profile, one preference for each consumer, such that
the aggregate demand function is generated by the
individual demand functions derived from that
preference profile. On the other hand, empirical
evidence suggests that aggregate demand functions
often exhibit some regularity even when individual
demand functions do not exhibit regular properties
from preference maximization under budget
constraints (Kirman ).
Possible Reconciliation of Individual
Rationality and Collective Rationality
The findings in social choice theory and demand
theory suggest a fundamental separation between
collective and individual rationality. On the one
hand, if individuals in a collective are rational, the
collective choice is responsive to individuals, and
the collective power does not he in some proper
subset of the collective (democratic), then the col-
lective choice cannot be ‘collectively rationalized’.
On the other hand, in some situations, collective
choices can be ‘rationalized’ even when individuals
in the collective do not act as rational individuals.
This separation between collective and individual
rationality is not unlike Buchanan’s critique of
Arrow’s formulation of collective rationality:
We may adopt the philosophical bases of individu-
alism in which the individual is the only entity
possessing ends or values. In this case no question of
social or collective rationality may be raised.
A social value scale as such simply does not exist.
Alternatively, we may adopt some variant of the
organic philosophical assumptions in which the
collectivity is an independent entity possessing its
own value ordering. It is legitimate to test the ratio-
nality or irrationality of this entity only against this
value ordering. (Buchanan , p. 116)
1805
The philosophical bases of individualism have
many followers in economics. Binmore ( ,
p. 142) wrote: ‘Game theorists of the strict school
believe that their prescriptions for rational play in
games can be deduced, in principle, from one-
person rationality considerations without the need to
invent collective rationality criteria provided that
sufficient information is assumed to be common
knowledge.’ Under the standard assumptions of
game theory accounting for individual interests,
these game theorists will prescribe that players
defect in the Prisoner’s Dilemma game. Such play
leads to a Pareto-inferior outcome and thus is in
conflict with the collective interest. This is not
problematic if game theory is a normative theory
which prescribes what people should do rationally.
However, as a predictive theory it fails to match
what people actually play in the Prisoner’s Dilemma
game. Experimental evidence shows rampant
cooperation among players of the Prisoner’s
Dilemma game (Rapoport and Chammah ;
Ledyard ).
If we make the organic philosophical
assumption that a collective is an independent
entity, then do we arbitrarily assume a criterion of
collective rationality? A more reasonable way of
thinking about a collective being as organic is,
perhaps, to consider that, in a collective, individuals
become social creatures, not mere individuals, and
as such their choices have social consequences that
they take into account. This can be modelled as
individuals’ preferences over a given set of
alternatives changing depending on whether they
are individuals or members of a collective. How
preferences are specifically influenced may reflect
culture, social convention or custom, so that they
are context- dependent. But whatever the cause, this
may create sufficient restrictions on the preference
domain that collective rationality results as a
consequence of some aggregation procedure that is
democratic.
See Also
► Arrow’s Theorem
► Ratic
► Ratic
► iocial Choice
1806 Collet, Clara Elizabeth (1860-1948)

Bibliography 1886. In 1893 she entered the civil service as labour

Arrow, K. 1951. Social choice and individual values, 2nd ed.
New York: Wiley, 1963.
Binmore, Is.. 1994. Playing fair: (lame theory and the social
contract, vol. I. Cambridge, MA: MIT Press.
Buchanan, J. 1954. Social choice, democracy, and free
markets. Journal of Political Economy 62: 114-123.
Debreu, G. 1974. Excess demand functions. Journal of
Mathematical Economics 1: 15 2.3.
Gaertner, W. 2002. Domain restrictions. In Handbook of
social choice and welfare, ed. Is. Arrow, A. Sen, and Is.
Suzumura. Amsterdam: North-Holland.
Gibbard, A. 1973. Manipulation of voting schemes: A general
result. Econometrica 41: 587 601.
Kirman, A. 2004. The structure of economic interaction:
Individual and collective rationality. In (ognitive eco-
nomics: . In interdisciplinary approach, ed. P. Bourgine
and .1. Nadal. Berlin: Springer.
Ledyard, .1. 1995. Public goods: A survey of experimental
research. In Handbook of experimental economics, ed. J.
Kagel and A. Roth. Princeton: Princeton University
Press.
Mantel, R. 1974. On the characterization of aggregate excess
demand. Journal of Economic Theory 7: 4.38 45.3.
McKelvey, R. 1979. General conditions for global intran-
sitivities in formal voting models. Econometrica 47: 1085-
1112.
Rapoport, A., and A. Chammah. 1965. Prisoner's dilemma: A
study in conflict and cooperation. Ann Arbor: University
of Michigan Press.
Satterthwaite, M. 1975. Strategy-proofhess and Arrow’s
conditions: Existence and correspondence theorems for
voting procedures and social welfare functions. Journal of
Economic Theory 10: 187-217.
Sen, A. 1995. Rationality and social choice. American
Economic Review 85: 1-24.
Sonnenschein, 11. 1973. Do Walras' identity and continuity
characterize the class of community excess demand
functions? Journal of Economic Theory 6: 345-354.
correspondent and later senior investigator for
women’s industries in the newly established Labour
Department of the Board of Trade. The earnings
and employment of women became and remained
Clara Collet’s main concern; her contemporaries
recognized her as the principal authority on the
subject in Britain. Articles on female labour and
earnings were among her contributions to the first
edition ofPalgrave’s Dictionary of Political
Economy in 1894, and the thorough and lucid
reports which she produced on women’s industrial
employment figured in Parliamentary Papers,
contributing to the passing of the original Trade
Boards Act of 1906. After her retirement in 1920
from what had by then become the Ministry of
Labour, Collet herself served on a number of trade
boards, and wrote the section on Domestic Service
for the New Survey of London Life and Labour
directed by her former chief, Sir H. Llewellyn
Smith.
The first woman Fellow of University College,
London, where she took her MA degree in 1885,
Clara Collet was one of the founders in 1890, along
with Henry Higgs and H.R. Beeton, of the
Economic Club which met there monthly, and acted
as its secretary from 1905 to 1922.
She was also a founder member of the British
Economic Association, which later became the
Royal Economic Society; she served on its Council
from 1920 to 1941, and on that of the Royal
Statistical Society from 1919 to 1935.

Collet, Clara Elizabeth (1860-1948)

R. D. Collison Black
Keywords
Collet, Clara E.; Women’s employment
After a period as a schoolteacher in Leicester, Clara 1894b. (With Dora M. Barton.) Female labour. In
Collet became one of Charles Booth's assistants on
his Survey of London Life and Labour in
Selected Works
1893-4. Royal Commission on Labour. Employ-
ment of women. Miss C.E. Collet. Report.
Parliamentary Papers 1893-4 [C.6894,
XXIII] xxxvii, pt. I.
1894a. Statistics of employment of women and
girls. Miss Collet. Report. Parliamentary
Papers 1894 [C.7564], lxxxi, Pt. 11.
Palgrave’s Dictionary of Political Economy,
London: Macmillan & Co., Vol. 11.
Collusion
1894c. Females and children, earnings of. In
Palgrave’s Dictionary of Political Economy,
London: Macmillan & Co., Vol. II.
1898. The collection and utilization of official
statistics bearing on the extent and effects of the
industrial employment of women. Journal of the
Royal Statistical Society 61: 219-260.
1899. Money wages of in-door domestic servants.
Miss Collet. Report. Parliamentary Papers 1899
[C.9346] xcii.
1931. Domestic service. Chapter VIII in New
Survey of London Life and Labour, vol. II, ed.
H. Llewellyn Smith. London: R S. King.
19 33 . Ap p en d i x t o The Private Letter Books of
Joseph Collet, sometime Governor of Fort St.
George Madras, ed . H . H. D od w e l l .
Lo nd on : Lo ng ma ns .
Collusion
Kevin Roberts
Although collusive practices are not restricted to the
economic relationships of a well-defined sub-group
in society, it is common to use the term collusion in
the context of cooperative activity between different
firms. With regard to the study of collusion,
research has centred on the conditions most
conducive to collusion and, in both theoretical and
empirical work on the operation of collusive
arrangements (see Scherer , Chaps. 6 and 7).
If economic agents are self-interested maxi-
mizers then, given that cooperation between a group
of firms will almost certai nly make possible higher
profits for each member of the group than is
possible without cooperation, there is a presumption
that collusive arrangements will be widespread.
Thus it is important to understand why collusion
should fail to be universal. In fact, a more general
issue is also raised by this. Taking the argument one
stage further, direct cooperation between a group of
firms and its consumers is
1807
likely to make possible benefits to everybody as
compared with a situation where firms and con-
sumers are separated by an anonymous market. This
cannot be the case for the economy as a whole if it
is Pareto efficient (e.g. the standard perfectly
competitive economy without externalities) but it
will still be the case that cooperative action by a
group of agents within the economy will allow that
group to gain at the expense of the rest of the
economy. The study of the problems faced by
colluding firms should hope to throw light on these
more general issues.
1. Consider a well defined group of firms produc-
ing identical or similar products, the implication
being that demand for the products is interre-
lated. As a first step, assume that cost and
demand functions faced by each firm are com-
mon knowledge (every other firm in the group
knows these functions, every firm knows this,
and so on). If the firms meet together to collude
then some joint action will emerge. Cooperative
game theory concerns itself with this solution
but, for the present exercise, it is sufficient to
note two of the main determinants of the even-
tual solution. First, there are the outcomes made
possible by cooperative action. Assume profits
can be redistributed within the group-side-
payments can be made. Then if cooperative
action fails to maximize joint profits all firms
can be made better off. Given the common
knowledge assumption so that there is no
argument for inefficiency based upon the mis-
perceptions of firms, joint profit maximization,
with the group of firms acting like a multi-
product monopolist, should emerge.
Side-payments may not always be possible.
For instance, collusive behaviour is outlawed in
many countries and, though it may be difficult to
detect whether actions by firms are part of some
collusive arrangement, the transfer of money
between firms is much more likely to be capable
of detection. But without side- payments,
actions which influence the size of joint profits
also influence the distribution of those profits;
the consequence of this being that distributional
considerations will influence the actions chosen
by firms. For instance, the
1808
cooperative Nash bargaining solution (Nash )
leads to actions which maximize the product of
individual gains above some status quo position -
compared with joint-profit maximization, there is
movement towards the equalization of gains
above the status quo. This may be viewed as a
compromise between ignoring distributional
considerations and the other extreme where only
distributional considerations count.
The second major determinant of the solution 2. Thus far, the story has assumed that there are no
reached will be what each firm can expect to
achieve if it refuses to accept a particular
proposed collusive action for the group. The
status quo of the Nash bargaining solution may
be interpreted in this way. This is the most
obvious component of ‘bargaining power’ for an
agent. The requirement that firms must prefer the
collusive action to what can be achieved by
reneging, places restrictions on the collusive
solution. Under the assumption that firms are
interrelated only through the demand structure
and that this interrelationship implies that the
goods produced by the collusive firms are
substitutes, then the worst that can happen to a
firm if it refuses to accept a collusive arrange-
ment is that all other firms maximize their
production and the reneging firm chooses pro-
duction to maximize profits in this hostile envi-
ronment. Given this scenario, there will usually
be a large range of collusive actions which offer
more to firms than can be achieved by reneging.
However, it is not enough for firms to say that
they will ‘punish’ a reneging firm in this way, 3. Taking the adverse selection case first, it is
there must be grounds on the part of the reneging
firm for believing that punishments will be
carried out - it must be a credible threat in the
sense that if the firm reneges, other firms have an Although firms may be attracted by the sim-
incentive to punish the firm. The credibility
restriction in this environment is captured by the
so-called ‘perfect folk theorem’ of repeated
games (see Rubinstein , for a published version
of the theorem) - firms called upon to punish will
have the strongest incentive to punish if all other firms. Clearly, the main problem that arises is
firms punish that firm for not punishing in the
first place. This argument leads to the conclusion
that any collusive outcome can be maintained as
long as all
Collusion
firms are better-off than the best they could
achieve under maximum retaliation from other
firms. Here it should be noted that as, by
definition, a firm will be happy to choose its
Nash strategy if all other firms do the same, the
Nash equilibrium is a feasible punishment
solution. Thus there will always be an effective
deterrent if all firms are better-off under collu-
sion than in the Nash equilibrium (Friedman )•
information problems for the (potentially
collusive) group of firms. Exogenous uncer-
tainty faced by all firms is not a particular
problem but when there are informational
asymmetries between firms, the study of collu-
sion is much richer. In fact, almost all the
theoretical literature on collusion takes some
informational asymmetry as a starting point and
it is useful to survey some of this literature from
the viewpoint of the informational asymmetry
which is being postulated.
It is convenient to distinguish between two
forms of asymmetry - adverse selection where,
for instance, some firms do not know the cost
and demand conditions of other firms, and moral
hazard where it is the behaviour of other firms
which cannot be observed. In the former case, it
is preferences that cannot be observed, in the
latter case it is actions. Both forms of
asymmetry can have important effects upon the
structure of collusive arrangements that could be
expected to emerge.
fairly reasonable to assume that a firm will have
a better knowledge of the demand and cost
conditions that it faces than other firms.
plicity of adopting a collusive arrangement
based upon solely common information, this
solution ignores the efficiency gains that may be
achieved from making the collusive outcome
sensitive to the privately held information of
that each individual firm must have the incentive
to reveal this private information. Consider a
simple model with just two firms, 1 and 2.
Assume that the revenues they receive
Collusion
when they produce outputs q\ and q2 are given by
R\(q\, q2) and R2(q\, q2)- Each firm has a constant
marginal cost of production and p2
and assume that this is private information, firm 1
knowing the true value of [1\, firm 2 knowing the
hue value of [12. Assume, for simplicity, that [1
can take on only two values, [1 and [1, P < P-
Collusion will result in the adoption of output
levels for each firm and the levels may be
sensitive to the private information p2 and /?2 -
qdPi, Pi) and q2(pu Pi)- However, this can only be
implemented if a firm does not pretend to be a
high cost firm (p) when it is low cost (yij and vice
versa; for firm 1, profit maximization implies
R\ \q\ (jf Pi), q 2 (p- Pi) ~ Pfh
( p - P i ) =R\[q\ (P,P2)-
<h(TPp2)] - Pq\ (P’Pi)
and
Ri[qi(PJi),q 2 (P,Pi)\ ~Pq\{P,Pi)
^R i <7l (P’ Pi) ^2 (/?> - Pi)
Pqi (; P - Pi)
This places restrictions on the class of solu-
tions that can be implemented. Combining the
inequalities gives
(/-£) [<7I(/^2) — (P’ Pi) — 0
which implies that the lower the marginal cost,
the more a firm is allowed to produce. Given that
the collusive solution will usually entail a
restriction in output as compared with what a firm
would like to choose (the other firm’s output
remaining constant), the firm must be provided
with a disincentive from pretending to be a low-
cost firm when it is high-cost and this will come
from variations in q2 - to provide the right
incentives to firm 1, the output of firm 2 should
be negatively correlated with the marginal cost of
firm 1.
The foregoing demonstrates that the
implementability requirement gives some
structure to the collusive solution. But the
1809
exact outcome chosen will be the result of
bargaining between the firms concerned. This
bargaining process may involve the transfer of
information prior to collusive agreements being
reached (Roberts ).
Recognition of the adverse selection problem
gives theoretical insight into the form of
collusive practices that are discussed in the more
applied literature. In general discussions a
distinction is drawn between implicit and
explicit collusion. However, in models without
information problems, each firm is aware of the
agreement that would be decided upon if it met
with other firms and is aware how reneging
firms would be dealt with. In this case, there is
no need for firms to collude explicitly and the
distinction between implicit and explicit collu-
sion is not useful. But when adverse selection
exists, it is clear that a rule which makes other
firms’ behaviour depend upon the private
information of some firm will require informa-
tion transmission between firms. The idea of
implicit collusion can be rationalized as a situ-
ation where either no information is transmitted
or, to take a less extreme case, where
information is transmitted through aggregate
market-wide indices, e.g. the equilibrium price
in the industry.
The existence of adverse selection can also
provide a theoretical rationale for mark-up
pricing as a collusive outcome (Roberts ). When
firms are selling in the same market, cost
conditions facing firms are more likely to be the
source of private information than demand
conditions. While it may be impossible for other
firms to observe the cost function of a firm, it
may be possible to observe the level of costs at
the output being produced. As this observation
can be used as a proxy for the private
information of the individual firm, a collusive
agreement will involve the output and price
levels of firms being dependent upon cost levels
- this provides a rationale for why the firms’
behaviour may be sensitive to average, rather
than the conventional marginal, costs.
4. Over the last 20 years, most of the theoretical
literature on collusion has considered
1810 Collusion

situations where the actions of firms fail to be Porter take the simple case of a homogeneous
perfectly observed by other firms - a situation of product being produced by similar firms. There
moral hazard. If the preferences of firms are is some uncertainty in demand so that a ‘low’
common knowledge then the problem is not one market price may be a result of this uncertainty
of deciding upon the collusive solution but, or of ‘cheating’ on the collusive agreement by
instead, of policing that solution. some firms. Green and Porter consider ‘trigger-
The simplest example of imperfect monitor- price’ punishment strategies which take the form
ing arises when there is a delay before the action of the group moving to the Cournot equilibrium
of a firm is observed by others. If time units are for a fixed time T if the market price drops
set equal to the delay time then a firm will gain below some level p. A feature of this
from reneging on a collusive agreement if informational set-up is that all firms suffer from
this punishment. The trigger-price strategies are
R p c
(i - p)n + pn > n set so that cheating does not occur though,
because of the uncertainty, a proportion of time
where ] |' is the profit per unit time under the
is spent in the punishment regime. There is a
collusive agreement, H7' is the maximum profit
direct trade-off between the gains from a col-
that the firm can achieve by reneging given that
lusive agreement that severely restricts output
other firms keep to the collusive agreement f| 7 is
and the costs of punishment that will be suffered
the profit the firm can achieve when it is being
in the maintenance of this agreement. Notice
punished by the other firms and [i is the discount
77
that as there is no adverse selection, there are no
factor for the firm. Obviously, If ^ II' so that
requirements for the firms to meet together to
the longer the delay before other firms perceive
decide upon collusive behaviour - it may be far-
reneging (the smaller the /?), the more incentive
fetched but there is nothing to rule out a system
a firm has to renege. As the degree of
of implicit collusion with trigger-price
punishment that can be inflicted is restricted by
strategies. For this reason, these moral hazard
the requirement that other firms have to punish,
models are often described as non-cooperative
and this incentive itself is diminished when there
models of collusion.
is a delay in observing behaviour, the smallest
5. The theory and practice of collusion are much
possible ] |r will rise as the delay time increases
discussed in texts on industrial organization. The
so reinforcing the incentive to renege. For a
foregoing has tried to make clear that the
detailed analysis of this problem, see Abreu ( ).
structure of information asymmetry in the mar-
The other main form of imperfect monitoring
ket is crucial for understanding the operation of
that has been considered deals with the case
collusive agreements. With a particular infor-
where individual firm behaviour is never
mational set-up, the structure of the set of col-
observable but market-wide aggregates can be
lusive agreements which will not entail reneging
observed by all firms. This was the situation
is now quite well understood. Despite much
which was studied by Stigler ( ) in his
work, there is rather less understanding of the
seminal paper on collusion and by many authors
exact agreement that will be settled upon.
since (a recent analysis is to be found in the
work of Porter , and Green and Porter ). In the
case studied by Stigler, firms observe the See Also
demand conditions for their own output and this
gives an indication of the prices that other firms ► A:."-...:
are charging. Green and ► Cooperative Equilibrium
► ...>r." e v
Colonialism
Bibliography
Abreu, D. 1986. Extremal equilibrium of oligopolistic super-
games. Journal of Economic Theory 39(1): 191-225.
Friedman, J. 1971. A non-cooperative equilibrium for
supergames. Review of Economic Studies 38(1): 1-12.
Green, E., and R. Porter. 1984. Noncooperative collusion
under imperfect price information. Econometrica 52(1):
87-100.
Nash, J. 1950. The bargaining problem. Econometrica 18(2):
155-162.
Porter, R. 1983. Optimal cartel trigger price strategies.
Journal of Economic Theor)’ 29(2): 313-338.
Roberts, K. 1983. Self agreed cartel rules. IMSSS Discussion
Paper No. 427, Stanford.
Roberts, K. 1985. Cartel behaviour and adverse selection.
Journal of Industrial Economics 33(4): 401-413.
Rubinstein, A. 1979. Equilibrium in supergames with the
overtaking criterion. Journal of Economic Theoty 21(1):
1-9.
S c h e r e r , F . 1 9 8 0 . Industrial market structure and
economic performance, 2 n d e d . C h i c a g o : R a n d
McNally.
Stigler, G.J. 1964. A theory of oligopoly. Journal of Political
Economy 72(1): 44-61.
Colonialism
M. Abdel-Fadil
Everywhere do I perceive a certain conspiracy of rich
men seeking their own advantage under the name
and pretext of the commonwealth. (Sir Thomas
More)
Modem colonialism, as a historical phenomenon of
territorial expansion, is intimately entwined with the
rise and expansion of the modem capitalist world
system. So colonialism is entwined with the history,
economics, politics and ruling ideas of the modem
capitalist society. On the other hand, and to avoid any
terminological confusions, the term imperialism
should be reserved to designate the new nexus of
financial and technological dependency relations and
arrangements marking the new distinct stage of
mature capitalism. (Magdoff ) During the modem
colonial period (1870-1945) colonialism has emerged
as a general description of the state of subjection -
political, economic and intellectual - of a non-
European society as a result of the process of colonial
organization. (Fieldhouse )
1811
The Age of Colonialism: Historical
Background
The age of colonialism began about 1500, following
the European d iseoveries of a sea route around
Africa’s southern coast (1488) and of America
(1492). Colonialism thus expanded by conquest and
settlement after a period of extensive exploration.
The improvement in navigational instruments
helped a great deal to make substantial progress in
the discovery of new geographical territories.
Portugal emerged as the leading nation in such
process of overseas expansion. ‘The search for
wealth in the form of gold, ivory, spices and slaves
spurred the Portuguese and may have been the
strongest motivating force behind the colonization
drive of the Portuguese during the 16th century’
(Encyclopaedia Britannica , Vol. 4,
p. 881).
The old colonial period, which lasted nearly
three centuries, following the major Portuguese and
Spanish conquests, may be viewed largely as a
commercial venture. The Spaniards and the
Portuguese resorted to their warships, gunnery and
seamanship to keep the main trade routes open. The
Spanish sovereigns created in 1504 the House of
Trade (Casa de Contracion) to regulate commerce
between Spain and the New World. Their purpose
was to establish state monopoly over overseas
trade, and thus pour the maximum amount of
bullion into the royal treasury
(Britannica , Vol.
5, pp. 882-3).
The old colonial system was disrupted in the
18th century as new contradictions developed due
to the rapid advance of the Industrial Revolution in
England, and by the progressive control England
was able to exercise over world shipping. Such new
developments led to a policy of opening the
American ports to international trade, a policy at
variance with the type of colonial relations
prevailing between Spain, Portugal and their col-
onies. These relations were organized exclusively
around the exploitation of precious metals
(Furtado , p. 20).
1812
The century between the 1820s and the outbreak
of the World War I saw the establishment of the
modem colonial order. For during that period
European countries had achieved complete dom-
inance over world trade, finance and shipping. On
the other hand, the political and military authority
of the European conquerors was backed by supe-
riority in technology, applied science, organization
and information systems (Bagchi ).
Between the late 1870s and World War I (1914-
18), the colonial powers added to their possessions
an average of about 240,000 square miles (620,000
sq.km.) a year, while during the first 75 years of the
19th century the rate of increase in new territories
acquired by colonial powers averaged about 83,000
square miles (215,000 sq.km.) a year. By the year
1914, the colonies extended over approximately 85
per cent of the surface of the globe.
Against this historical background, John Hicks
establishes a useful distinction between two types of
colony: colonies of settlement and trading-post
colonies (Hicks , p. 51). A third type of colony was
identified as ‘the plantation colonies'. In such case,
the colony which started as a colony of settlement
was gradually transformed into a trading colony
(ibid., p. 53).
The Colonization Debate
While there is a strong connection between mer-
cantile expansion and colonization, it would be a
mistake to emphasize the crude economic inter-
pretation of colonialism by narrowing down colo-
nialism to the process of control of supplies of raw
materials, mineral resources and markets in under-
developed and precapitalist regions. In fact, such a
narrow economistic approach eliminates a vital
aspect of colonialism relating to political activity
and the drive for dominance over the daily fives of
the people of the colonized regions (e.g. French
colonialism).
Nonetheless, colonialism must be viewed, dia-
lectically, as a complex phenomenon of capitalist
expansion, operating in terms of time and space. To
illustrate this point, S.H. Frankel described such a
process as a disintegrating but also a
Colonialism
formative process, a unique process in the history of
mankind (Frankel ). Some other writers justified
colonial mle on the ground that ‘Colonialism was a
necessary instrument of “modernization” which
would help other peoples to do what they could not
have done, or have done as well, by themselves’
(Fieldhouse , p. 43). At the other end of the
spectrum, radical theorists, notable among them
Walter Rodney, claim that under colonialism ‘the
only things that developed were dependency and
underdevelopment’ (Rodney , p. 256).
One of the most articulate arguments put
forward in defence of the colonial mle in under-
developed areas is that of Lord Bauer, who contends
that:
The colonial governments established law and order,
safeguarded private property and contractual
relations, organized basic transport and health ser-
vices, and introduced some modem financial and
legal institutions. This environment also promoted
the establishment or extension of external contracts,
which in turn encouraged the inflow of external
resources, notably administrative, commercial and
technical skills as well as capital.
These contacts also acquainted the population
with new wants, crops, commodities and methods of
cultivation and opened new sources of supply of a
wide range of commodities. These changes
engendered a new outlook on material advance and
on the means of securing it: for good or evil these
contacts promoted the erosion of the traditional
values, objectives, attitudes and customs obstructing
material advance. (Bauer , p. 149)
This argument only confirms the deep-seated
Western biased view, claiming that material pro-
gress and advance can only be achieved by eroding
the traditional values, customs and production
structures of pre-capitalist and primitive societies.
Rosa Luxemburg ( ) would see in Lord
Bauer’s view an eloquent proof of her radical
contentions about colonialism and territorial
expansion, by the emerging capitalist nations.
But what is at issue is not the possibility or not
of achieving material progress or advancement, but
the terms on which these transformations in the
material and socioeconomic structures were
operated. From our viewpoint, what needs to be
stressed is the loss of sovereignty which the process
of European colonization entailed for
Colonialism
practically all colonized peoples. In Africa, for
instance, European colonizers often crushed,
suppressed or amalgamated states at will. In most
instances, the direct colonial rule was designed to
direct and reorder the day-to-day lives of the
African peoples (Ajayi ).
Seen in a radically different light, thinkers such
as Albert Memmi, Jean-Paul Sartre and Franz
Fanon placed greater emphasis on the ideological
implications and the sociopsychological conse-
quences of the process of colonization. According
to Fanon, colonialism tended not only to deprive a
society of its freedom and its wealth, but of its very
character, leaving its people intellectually and
morally disoriented (Fanon, English edition, )•
Patterns of Colonial Trade
Historians tend to agree that the conquest of col-
onies was designed to the economic advantage of
the European conquerors. Some historians (e.g. E.J.
Hobsbawm) would go as far as to claim that the
Industrial Revolution in England would not have
been accomplished without the conquest and
penetration of ‘underdeveloped’ markets overseas
(Hobsbawm , p. 54).
In fact, the primary aim of all European states
was to use commercial regulations to maximize
their share of colonial trade in both directions and
the profits they made from it. The English Navi-
gation Acts, dating from the 1650s, may be taken as
typical in this respect. According to these Acts, all
colonial trade must be carried in British-owned and
registered ships. All goods imported to the colonies
must either be the product of Britain or be
transhipped and pay duty there. Any colonial
exports so ‘enumerated’ must be carried direct to a
British port in the first instance (Deane p. 204 ). The
aim of such rales and regulations was to give British
shipowners, merchants and manufacturers an
assured benefit from colonial commerce and to
enable the government to tax colonial trade. This
clearly indicates the close association between the
process of colonization and the rise of various
foreign-trade monopolies held by the charted
colonial companies.
1813
Hence, against all claims of the ‘Free-Trade’
school, the British cotton industry did not rely on its
competitive superiority, but relied heavily on the
monopolistic practices embodied in colonial trade-
regulations, and enforced by the British commercial
and naval supremacy (Hobsbawm , p. 58). On the
other hand, the terms of trade between the colonized
areas and the metropolitan countries had a tendency
to deteriorate steadily over time, so that the primary
producers in colonized areas tended to obtain
proportionally less with their labour than they could
have done had they concentrated on producing food
or other subsistence crops for their own use or for
the home market (Fieldhouse , p. 78). This may be
characterized, in modem terminology, as ‘unequal
exchange’, which emphasizes once again the
exploitative nature of colonial trade.
The Internal Control of the Colonial
Economy
The key to understanding colonialism as a historical
phenomenon lies in analysing the mechanism of the
internal control of the colonial economy. In this
connection, one has to answer two fundamental
questions. Why did the Western countries spend so
much energy, blood and money in seeking to
procure colonial possessions? What are the direct
and indirect economic benefits of colonialism?
In the colonial economy, top priority was given
to infrastructure investment: railways, harbours,
telegraphs, rivers and roads, since it was believed
that these constitute the prerequisites of a modem
economy, making it possible to link internal areas
of production to the world commodity markets. In
the agricultural sector, it is still an open question
whether plantations, owned and run by foreigners,
made any significant contribution to the develop-
ment of the colonial economy. On the positive side,
they served as the main vehicle of introducing new
crops, attracting foreign capital, expanding the base
of the cash economy and the wage-labour force, and
increasing agricultural productivity. On the negative
side, the crops of such plantations were subject to
severe
1814
fluctuations on the international commodity mar-
kets, thus subjecting the colonial economy to severe
cyclical fluctuations.
In matters of industrialization, many observers
tend to agree that the colonial powers did not
positively encourage industrialization in their
dependencies, and in many instances their basic
policies led to some sort of de-industrialization
(Bagchi ). In this respect, many writers invoke the
record of colonization in India from the days of the
East Indian Company. For the balance of historical
evidence points to the fact that up to the 18th
century the economic conditions of India were
relatively advanced, and Indian methods of
industrial production were comparable with those
prevailing in any other advanced part of the world
(Baran , p. 144).
On the other hand, Bill Warren has offered a
neo-Marxist view, opposed to the ‘Dependency
School’, regarding the effects of colonialism on the
development of productive forces in Third World
countries. His main argument runs as follows:
‘Direct Colonialism, far from having retarded or
distorted indigenous capitalist development that
might otherwise have occurred, acted as a powerfiil
engine ofprogressive social change’ (Warren ).
Nonetheless, Warren’s positive account of the
effects of colonialism on the process of capitalist
development in Third World countries tends to be
rather unitary in spirit. For the pattern of resource
allocation in colonial territories had been shaped
and administered largely by foreign investors,
bankers and merchants. According to Paul Baran,
the principal impact of foreign enterprise on the
development of the underdeveloped and
precapitalist regions Ties in hardening and
strengthening the sway of merchant capitalism, in
slowing down and indeed preventing its trans-
formation into industrial capitalism’ (Baran , p.
205).
This very nature of the process of capitalist
development under colonialism led some authors,
such as H. Alavi, to offer the highly controversial
concept of the ‘colonial mode of production’. This
concept was offered as a theoretical construction
designed to allow for a variety of relations other
than those which characterize the ‘capitalist mode
Colonialism
of production’, as experienced in the advanced
capitalist economies of the ‘centre’ (Alavi ).
In this respect, Alavi and company established the
distinctive features of the ‘colonial mode of pro-
duction’ on the basis of empirical investigation of
the circuits of capital and forms of labour recruit-
ment of what comes to be called by other authors
(i.e. Samir Amin and Gunder Frank) ‘colonial
capitalism’ or ‘peripheral capitalism’ (Booth , p.
169). Yet the difficulties and confusions
surrounding the concept of a ‘colonial mode of
production’, as a distinct mode of production,
remain formidable.
Decolonization and Neocolonialism:
Two Sides of the Same Coin
The drive towards decolonization in the post- World
War II period was a response to the economic crisis
of an ageing colonial system. This colonial system
was found to involve considerable, and sometimes
unacceptable, financial costs to the metropolis.
Moreover, colonialism had become increasingly
discredited among the people of the colonizing
nations themselves, just as the emotional strains of
suppressing nationalistic movements in colonized
regions had become largely intolerable for public
opinion (Fieldhouse , p. 24).
Nonetheless, the process of decolonization
proved to be a nominal process in the sense that the
formal end of colonial rule did not necessarily result
in genuine economic independence for the former
colonies. There is now a community of view among
left-wing economists and writers that
decolonization took place when and because foreign
monopoly capital felt confident that the colonial
society and economy had been so restructured that
their interests could be preserved without direct
political control. In other words, colonialism had
been merely transmitted into perpetual
neocolonialism (Baran ).
The term ‘neocolonialism’, which has gained
wide acceptance since the mid-1950s, is meant to
designate a state of affairs characterized by a
structure of dependency relationships whereby the
former colonial territories are kept in their
Colonies 1815

subordinate place within the imperialist system. Frankel, S.H. 1953. The economic impact of colonialism on
This is maintained and sustained by means of under-developed societies. Oxford: Basil Blackwell.
Furtado, C. 1970. E c o n o m i c d e v e l o p m e n t o f
chronic and structural balance of payments diffi- L a t i n A m e r i c a . London: Cambridge University
culties, arising from the trade, aid and investment Press.
relationships with their former or new metropolitan Hicks, J. 1969. A t h e o r y o f e c o n o m i c h i s t o r y .
countries (Warren , p. 35). Oxford: Clarendon.
Hobsbawm, E. 1968. I n d u s t r y a n d e m p i r e . London:
In sum colonialism may be seen in a historical Weidenfeld & Nicolson.
perspective as one decisive and dramatic stage in Luxemburg, R. 1913. T h e a c c u m u l a t i o n o f
the evolution of international economic relation- c a p i t a l . London: Routledge & Kegan Paul. 1951.
ships. The establishment of colonial rule constituted Magdoff, H. 1970. Is imperialism really necessary?
M o n t h l y R e v i e w , Part I, 22(5): 1-14; Part II, 22(6):
an arbitrary break in the normal course of history, 1-11.
splitting up regions and creating new artificial Rodney, W. 1972. How Europe underdeveloped Africa.
entities, transplanting new alien values and London: Bogle.
institutions into colonized societies. One may Warren, B. 1973. Imperialism and capitalist industrialization.
N e w L e f t R e v i e w (81):3—44.
finally wonder whether it would not have been
Warren, B. 1980. I m p e r i a l i s m : P i o n e e r o f
better for the people of the colonized regions to
remain autonomous until certain indigenous forces
could gain momentum and generate new conditions
for socioeconomic development and material Colonies
progress.
Donald Winch

See Also
Bibliography
Ajayi, J.F. 1969. Colonialism: An episode in African history.
In Colonialism in Africa 1870-1960, vol. 1, ed. L.H. Gann
and P. Duignan. London: Cambridge University Press.
Alavi, H. 1975. India and the colonial mode of production. In
The socialist register 1975, ed. R. Miliband and J. Saville.
London: Merlin Press.
Bagchi, A.K. 1982. The political economy of underdevel-
opment. London: Cambridge University Press.
Baran, P. 1957. The political economy of growth. New York:
Monthly Review Press.
Bauer, P.T. 1976. Dissent on development. London:
Weidenfeld & Nicolson.
Booth, D. 1985. Marxism and development sociology:
Interpreting the impasse. World Development 13(7): 761-
787. "
Deane, P. 1965. The first industrial revolution. London:
Cambridge University Press.
Encyclopaedia Britannica. 1768. 5th ed., vol. 4, 1977.
Fanon, F. 1966. The wretched of the earth. New York: Groove
Press.
Fieldhouse, D.K. 1981. Colonialism, 1870-1945: An
introduction. London: Weidenfeld & Nicolson.
The economic advantages and disadvantages of
colonies, the best means of establishing them and
ensuring their development, and the principles that
should govern trade and other relations with the
mother country, have persistently served as fertile
topics for policy and theoretical debate in the
history of political economy. The treatment given
here will be confined to the British debate on
colonies from the late eighteenth to the first decades
of the twentieth century.
The British empire was composed of colonies
and ex-colonies which had differing histories of
acquisition and varying political and economic
relationships with the mother country. It follows
that the problems which they posed were equally
diverse, as illustrated by the differences between the
economies of the British West Indies before and
after slavery was abolished, the question of public
land disposal and emigration to Canada, Australia
and New Zealand, and the tasks of administering an
Indian sub-continent with a largely peasant
population living close to minimum subsistence
levels. To this list can be added the problems of
integrating Scotland and Ireland
1816
into the English economy and polity after the
respective Acts of Union in 1707 and 1808, where
‘colonial’ issues - in a technical rather than emotive
sense - were often at stake, even if the teim was not
used to describe them. Indeed, Ireland and India as
subsistence farming economies posed similar
problems to British administrators, despite major
differences in their cultural backgrounds and
political staUis within the empire. For that matter,
even the United States after independence could for
some purposes be treated as having a 'colonial'
relationship with Britain, largely because it
remained a major outlet for British capital and
labour.
The sheer magnftude of the problems of empire
and their changing nature over more than two
cenUtries would guarantee that they bulked large in
the minds of British economists. A few strategic
examples will show that there has always been a
fairly intimate relationship between economics,
economists, and empire. Adam Smith may well
have advised the imposition of the Townshend
duties on North America in 1763, and he was
certainly involved in advising the British govern-
ment on the consequences of American breakaway
when these earlier attempts to exert fiscal control
led to successful revolt. Malthus held the first Chair
of political economy in Britain at an educational
institution at Haileybury established to train the
servants of the East Indian Company; and James
and John Stuart Mill together devoted nearly 40
years to the service ofthe Company. The younger
Mill was also a consistent supporter of schemes
involving the ‘systematic colonization’ of Australia
and New Zealand, as well as taking a major interest,
along with most of his classical predecessors and
successors, in the problems of the Irish economy.
The controversy over imperial preference at the turn
of the twentieth century underlined the gulf that
existed between historical or institutionalist
economists and their more orthodox opponents, led
Alfred Marshall, who believed that the increasing
challenge to Britain’s industrial hegemony did not
justify abandonment of deductive methods of
economic reasoning or those cosmopolitan free
trade principles which had spurred British
prosperity earlier. Finally, of course, there is
Keynes, whose first employment
Colonies
as a civil servant was within the India Office, and
whose first major economic work was a treatise on
Indian Currency and Finance (1913) - a work which
attempted to do for its day what Sir James Steuart
had done when he wrote The Principles of Money
Applied to the Present State ofthe Coin of Bengal in
1772.
A treatment based on chronology and recurring
themes seems the best way of dealing with the
diversity of Colonies as a topic of economic interest,
though it should be remembered that colonies and
empire was never treated solely as economic
problems, even after the inauguration, largely under
Marshall’s auspices, of a measure of professional
distance in these matters.
The initial and longest period in the history of
colonial policy began in England during Elizabethan
times and effectively ended with the dismantling of
what had become known as the ‘old colonial
system’ in the 1820s. This system was loosely based
on the amorphous doctrines which Adam Smith
subjected to attack in the Wealth of Nations as a
central part of his condemnation of the ‘mercantile
system’ (later known as Mercantilism). During this
mercantile period colonies generated a large body of
literature which reflected the overwhelming concern
with national power and economic self-sufficiency
as the prime objectives of state intervention. Thus
colonies not only served direct strategic purposes as
naval or military bases, they were also treated as
sources of precious metals, and of strategic and
other raw materials necessary to Britain’s early
manufacturing industries - of particular value when
carried in British ships and bought through
monopolistic arrangements at prices lower than
could be obtained on the world market. Colonies
were variously regarded as protected export
markets, outlets for surplus population, and sources
of tribute, in addition to serving occasionally as
prison settlements. Although the development of
free trade doctrines, together with associated
monetary ideas on specie-flow mechanisms,
eventually succeeded in undermining many of the
arguments in favour of colonial possessions and
regulations, economic nationalism and neo-
mercantilistic ideas and policies have always
exerted a powerful attraction, especially in countries
that were
Colonies
industrial later-comers or anxious to overcome the
problems of underdevelopment by means of import-
substitution and/or export promotion.
During the eighteenth century, the established
wisdom on the subject of colonies came under
question largely because the supposed benefits of
trade controls to the mother country were connected
with the growing cost to Britain of defending and
governing her North American colonies. Among the
earliest critics of the colonial system from this point
of view was Josiah Tucker, who argued that
existing benefits in the form of export markets and
imported goods would accrue to Britain under free
trade, and without the attendant military and
economic burdens of empire. Provoked by David
Hume’s essays on commerce and money, Tucker
also engaged in an important dispute with Hume on
the question of whether trade relations between rich
and poor countries could be considered as
equalizing or not, and if so, by what process -
perhaps through rising labour costs and prices in the
richer trading partner, or through some other
mechanisms of stimulus and emulation in the poorer
country. The dispute was of relevance to free-trade
relations between England and Scotland after
Union, and of potential relevance to Britain’s
relations with Ireland and other ‘colonies’, whether
acknowledged as such or not. Indeed the Hume-
Tucker debate was an early example of recognition
of the essential similarities between international
and interregional trade in a world in which
currencies were linked through the gold standard. It
also concerned relative rates of growth and the
respective merits of agriculture and manufacturing
as the basis for a nation’s wealth and prospects for
economic development. Would those who had the
advantages of an early start acquire world dominion
and monopoly; and hence would poorer and later
starters in the race be forced to employ protective
measures to establish and maintain their infant
industries? If noticed, the debate would have
foreshadowed later issues raised by Friedrich List
and Henry Carey with Germany and the United
States in mind, as well as other questions such as
‘free trade imperialism’, the ‘permanence’ of the
dollar problem after World War II, and the debate
on ‘dependency’ and the development of
1817
post-colonial underdevelopment in the 1960s and
1970s.
Smith’s extensive treatment of colonies in the
Wealth of Nations (especially Book IV, chapter 7)
became the locus classicus of the anti-mercantile
position, where much of his discussion was inter-
woven with an account of the founding of the
European colonies which brought matters up to the
present, namely to the issues underlying Britain’s
dispute with its American colonies. Whereas
Edmund Burke had advocated the relaxation of
trade controls and taxation as a means of preserving
the political status quo, Smith maintained that
Britain’s pretensions to empire would remain those
of a shortsighted shopkeeper unless some system
could be found whereby the debts and current
burdens of empire could be shared by the colonies
themselves; and he emphasized the point by closing
the Wealth of Nations with a warning about the
potential long-term effects on British growth
prospects of existing arrangements. Hence the
elaborate scheme he advanced for an imperial
(Anglo-American) free trade zone, with provision
for complete fiscal harmonization and legislative
union. However, since he regarded this proposal as
utopian, its purpose was chiefly to underline the
precise conditions under which the burdens of
empire could be made acceptable. Much the same
result could be achieved through free trade and a
treaty of friendship, without provision for imperial
government.
Smith’s close dissection of the various gains and
losses involved in maintaining the monopoly of
colonial trade employed a quasi-mercantilist idea of
‘vent for surplus’, as well as other arguments about
the effects on profits of colonial markets which
could not be squared with later Ricardian orthodoxy
on the doctrine of comparative costs, capital
accumulation, Say’s Law, and the permanent causes
of declining profits. Ricardo also pointed out
circumstances in which it was possible for the
mother country to so regulate the trade of a colony
as to make it less beneficial to the colony, and more
advantageous to the mother country than free trade -
an early version of the terms-of-trade argument for
tariffs (not meant for use) which in the hands of
Robert Torrens was to
1818
blossom into a case for an imperial Zollverein a few
years later.
With the support of most political economists,
however, the system of colonial preferences was
gradually and unilaterally dismantled, beginning
with the efforts of Huskisson and Robinson in the
1820s. The views of special interest groups, espe-
cially those connected with shipping and the West
Indies, which Smith had expected to prevail, were
outflanked by an uncertain combination of intel-
lectual argument, political opportunism, and a
general realization that Britain’s industrial domi-
nance meant that mutual restrictions merely
restricted the dominant partner without conferring
equivalent benefit. A similar combination involving
humanitarian arguments prevailed on the related
matter of West Indian slavery and the slave trade.
The keystone of Britain’s rather isolated status as a
free-trading nation was installed with the abolition
of the Com Laws in 1846, a policy that had
considerable long-term significance for British
agriculture and Britain’s relationship to colonial
suppliers of food and raw materials, including the
United States as well as colonies of recent
settlement.
During the 1830s and 1840s public debate on
colonies and colonization was dominated by the
activities of Edward Gibbon Wakefield and the
colonial reformers, a group of radicals dedicated to
the revival of‘the lost art of colonization’. Their
programme entailed the creation of self-governing
colonies as outlets for Britain's surplus capital and
labour, avoiding the evils of simply ‘shovelling out
paupers’, abolishing penal settlements, and creating
‘civilized' communities enjoying the benefits of free
trade and high rates of growth. Wakefield’s
diagnosis of the simultaneous existence in Britain of
surplus capital and labour, and the consequent need
for new fields of employment abroad, was
developed in opposition to Ricardian orthodoxy on
the wage fund and Say’s Law. His ideas on the
optimal economic development of colonies also
conflicted with Smith's view that countries of
European settlement enjoying an abundance of land
were likely to make rapid economic progress. The
key to high rates of growth lay in achieving the
correct balance between capital, labour, and their
'field of employment’ by
Colonies
restricting access to land. This could be achieved by
setting a price on land sufficient to delay dispersal
of the wage-labour force, and by using the proceeds
of land sale for the purpose of bringing in new
immigrants. This policy meant that public land
disposal and immigration had to remain an imperial
rather than piuely local concern, thereby creating
scope for conflict when colonies achieved self-
government. The Wakefield policy came to be seen
as a symbol of imperial oppression, an attempt to
place colonial development within a straitjacket
designed with European conditions in mind. It also
entailed loss of freedom in disposing of one of the
main sources of revenue available to self-governing
colonies. The colonial reformers’ hopes of
establishing an empire in which free trade ruled
were another casualty of self-government when it
led to tariffs being raised by Canada and Australia
against British and other goods.
What now seems remarkable is the rapidity' and
extent of influence exerted over British colonial
policy by Wakefield's untried theories, though
modern development economics may yield com-
parable examples. He was also highly successful in
convincing a number of leading political econ-
omists, not least John Stuart Mill, that his ideas
deserved to form the basis for firture policy. Mill
gave prominence to Wakefield's ideas in his Prin-
ciples of Political Economy and other writings by
consistently championing 'systematic colonization’
as a solution to Britain's population difficulties; and
by treating its application to new countries as a
valid exception to the general principle of laissez
faire, namely as a case where the self-interest
principle acting under competitive conditions would
lead to a sub-optimal result as far as the community
was concerned. As part of his general modification
of Ricardo’s assumptions concerning capital
scarcity and the distant prospect of the stationaiy
state. Mill also endorsed the conclusions of
Wakefield’s heterodox diagnosis of Britain's
economic condition, while denying that it was in
conflict with Say’s Law and other received
Ricardian doctrines. By acknowledging the
importance to Britain of the export of capital and
labour to colonies, Mill not only removed an
obstacle to support for colonization, and hence to
Colonies
the extension of empire, he opened up a major
exception to the comparative cost doctrine as an
interpretation of Britain’s trading pattern: the trade
with colonies now became akin to interregional
trade. It should also be noted that Mill, confirming
the tradition that the only new economic arguments
for protection have been advanced by those who
favour free trade as a general rale, gave a cautious
endorsement to the infant industry case for tariffs.
India was never a colony in the same sense that
North America, Australia and New Zealand were
British colonies. Attempts were often made to
prevent or discourage European colonization before
1830, and until 1858 India was governed by the
East India Company acting on a renewable Charter
granted by Parliament. The Company had been
subjected to closer government control in the late
eighteenth century, deprived of its commercial
monopoly in 1813, and finally ceased trading
altogether in 1833 when it lost its exclusive priv-
ileges over the China trade. These developments
represented another victory for the forces of free
trade, and they blunted the force of Smith’s criti-
cisms both of monopolies and government by
trading companies. As we have seen in the case of
Tucker, free trade ideas could be associated with a
case for complete ‘emancipation’ (Bentham’s term)
of all colonies. (It was an association of free trade
with anti-imperialism which was an invitation for
revisionist historians to counter with a neat, perhaps
over-neat, inversion by drawing attention to ‘free-
trade imperialism’.) But there were fewer
spokesman for such ideas in relation to India, where
other notions of European superiority and
responsibility held sway, along with more mundane
considerations connected with the retention of
investment, employment, and trading opportunities.
India had provided Smith with a prime example
of a stationary or declining state, something that
could either be attributed to the deficiencies of its
system of government and taxation, or to those of a
backward people whose culture constituted a barrier
to economic progress, though usually to a
combination of both. In addition, criticism from
divergent quarters was made of the flow of tribute
leaving India for Britain,
1819
much of it financed by exports of textiles that
competed with domestic industry: Indian commerce
came in conflict with British manufacturing
interests, which were placated by the imposition of
duties on Indian imports. But with the
reorganization of the Company, especially after
1813, came new priorities and opportunities for
those with ambitions to bring the light of post-
Smithian, and more especially, Ricardian political
economy to bear on the problems of Indian
administration. Such a task proved highly congenial
to James Mill, a critic of the Company’s monopoly
powers who was appointed by the Company in 1819
and rose to the rank of Chief Examiner in charge of
political, judicial, and fiscal correspondence with
India. It was largely through Mill’s efforts, later
endorsed by his son, John Stuart Mill, that the
Ricardian rent doctrine came to play such a large
part in the conduct of Indian affairs. It provided the
basis for the ryotwari system of land tenure,
whereby the state became the sole landlord and met
its revenue needs by levying ryots or peasant-
farmers according to Ricardian principles of pure
rent. By confining the state’s exactions to rent it
was thought that the peasant farmer would enjoy
normal profits and wages, and the state would elim-
inate an intermediary or landowning class of
zemindars or rent-receivers. The system embodied
action according to a clear analytical proposition,
antagonism to rent as a form of private income, and
a view of the prospects for Indian economic
development that treated the peasant proprietor as a
capitalistic entrepreneur, freed from the arbitrary
exactions of landowners and responding to market
incentives - which is not to say that the application
of these Western economic ideas to Indian
conditions was any more successfiil than the
zemindari alternative.
In Ireland, where similar economic conditions of
a growing population dependent on a backward
agriculture obtained, it proved more difficult to
bypass the Irish landowner in order to grant the kind
of security of tenure to the peasant proprietor that
either existed or was the aim of administrators in
India. Indeed, the initial view of economists during
the early classical period was unsympathetic to the
preservation of peasant
1820
proprietorship in Ireland. The favoured solution was
consolidation of tenant-holdings as a preliminary to
the creation of a capitalistic form of farming
employing agricultural wage-labour along English
lines, together with emigration or absorption of the
displaced population into alternative employment.
Before, but especially after the Great Famine of
1846, emigration, largely unplanned, was the only
part of this programme that operated. Under the
leadership of John Stuart Mill, J.E. Caimes, W.T.
Thornton and Henry Fawcett, the earlier diagnoses
and remedies were entirely recast; a more positive
evaluation of the possibilities of transforming
cottier tenants into peasant proprietors enjoying
security of tenure was registered and advocated as
the basis for a solution to Irish problems. It was a
position that ran directly counter to English
property ideas and involved recognition of the role
of custom as opposed to contract in designing
policies and institutions for societies that did not
conform to the English model.
Mill’s deployment of a more relativist approach
in policy matters was later to be seen as a welcome,
though incomplete concession to a succeeding
generation of more full-blooded historical and
institutionalist critics of deductive economic theory,
with its built-in bias in favour of rational economic
man - a creature originally invoked by Mill to
underline the contrast with societies where custom
prevailed. Such critics were more numerous and
vocal after 1870, and the resulting split within the
economists’ ranks coincided with a campaign for
‘fair trade’ in the 1880s which blossomed into
Joseph Chamberlain’s scheme of tariff reform along
imperial preference lines - an unwitting return to
Torrens’s imperial Zollverein. The historical
economists, led by William Cunningham and W.J.
Ashley, had already followed Schmoller and other
German exemplars in according a more positive
valuation to Mercantilism, and this presaged their
endorsement of tariff reform as an imperial remedy
for Britain’s declining competitiveness. At the price
of forsaking free trade, Britain could offer
preferential treatment to imperial food and raw
Colonies
materials in return for similar preferences in the
markets of her ex-colonies. There had already been
a revival of interest in imperial federation, which
could be portrayed, as it was by John Shield
Nicholson, as a return to Adam Smith’s project of
empire. While much of this belongs to the larger
subject of imperialism, a term which has always
carried more nationalistic and ideological oxygen,
the episode is chiefly of interest here because it was
the occasion for a major challenge to economic
orthodoxy. Chamberlain’s use of arguments
supplied by such economists as W.A.S. Hewins
brought Marshall into the professional and political
fray with his Memorandum on Fiscal Policy of
International Trade (1908), a work which is still
perhaps the best brief restatement of the free trade
position based on a combination of neoclassical
trade theory and an empirical analysis of
contemporary conditions in the colonies as well as
in Britain.
Marshall’s pupil, John Maynard Keynes, was
not as impressed by his master's memorandum when
he looked back on it from the vantage point of 1930.
At this time Keynes had decided that a revenue
tariff was an acceptable policy for Britain to follow,
though not for reasons connected with imperial
solidarity. Thus when Neville Chamberlain
achieved his father’s goal with the passage of the
Import Duties Act in 1932, followed by the Ottawa
Agreements which began the period of imperial
preference, Keynes withdrew his support for tariffs.
Nevertheless, any account of the revolution
associated with Keynes’s name is likely to be
incomplete without some reference to the ending of
free trade in Britain, coupled as it was with the inau-
guration of an era in which external monetary
constraints on British domestic policy were weak-
ened. Of more long-term significance to imperial
policy in the interwar period, however, was the
revival of interest in state-assisted settlement in the
white dominions and the new emphasis on colonial
development, with Africa as well as India now
assuming a larger role in official, if not
professional, economic thinking. At this point
colonies and colonial policy become
Colquhoun, Patrick (1745-1820) 1821

something else, the beginnings of modem devel- Colquhoun was bom in Dundee. A successful early
opment economics. career in business led to the position of Lord
Provost of Glasgow in 1782 and 1783. In 1789
See Also Colquhoun moved to London and became active as
a magistrate. He worked on the provision of poor
relief and put forward plans for the reform of
London’s police. He died in London in 1820.
Colquhoun’s interest in poor relief led to his
New and Appropriate System of Education for the
Labouring People (1806), a pamphlet based on his
own experience of running a school in Westminster.
Bibliography Like Thomas Chalmers later, he argued for the
necessity of education to raise the standards and
Ambirajan, S. 1978. C l a s s i c a l p o l i t i c a l e c o n o m y aspirations of the poor, though primarily in order to
a n d B r i t i s h p o l i c y i n I n d i a . Cambridge: curb vice rather than population. This, he believed,
Cambridge University Press.
Barber, W.J. 1975. British economic thought and India, 1600-
was the most cost-effective way of tackling poverty.
1858. Oxford: Clarendon. His Wealth, Power and Resources of the British
Black, R.D.C. 1960. Economic thought and the Irish question, Empire (1814), his last important work, is the one
1817-1870. Cambridge: Cambridge University for which he is best known. This contained detailed
Press.
Black, R.D.C. 1968. Economic policy in Ireland and India in
figures on incomes and occupations and the relative
the time of J.S. Mill. E c o n o m i c H i s t o i y R e v i e w importance of agriculture and manufacturing in
21: 321-336. Great Britain and Ireland. He also included a history
Drummond, I.M. 1974. I m p e r i a l e c o n o m i c p o l i c y , of the public revenue, and descriptive material on
1 9 1 7 - 1 9 3 8 . London: Allen & Unwin.
the colonies.
Rnorr, K. 1944. British colonial theories, 1570-1850.
Toronto: Toronto University Press. The work was not very securely based;
Semmel, B. 1970. The rise of free trade imperialism. McCulloch, who had first-hand experience of trying
Cambridge: Cambridge University Press. to construct large-scale statistical data for his
Stokes, E. 1959. The English Utilitarians and India, Oxford:
Commercial Dictionary, was severely critical of it
Clarendon.
Winch, D. 1965. Classical political economy and colonies. in the Edinburgh Review and in Brande s Dictio-
London: Bell & Sons. nary. But it was followed by later writers and
Wood, J.C. 1983. B r i t i s h e c o n o m i s t s a n d t h e Colquhoun’s estimate that unproductive labour, one
e m p i r e . London: Groom Helm.
fifth of the total, received one third of output was
widely quoted.

Colquhoun, Patrick (1745-1820) Selected Works

D. P. O’Brien 18 06 . A new and appropriate system of education

for the labouring people. Lo nd on : J.
Ha t c ha r d . 18 14 . A treatise on the wealth,
power and resources of the British Empire. In Eveiy
Keywords quarter of the World, including the East Indies; the
Colquhoun, Patrick; Education; Unproductive rise and progress of the funding system explained;
labour with observations on the national resources for the
beneficial employment of a redundant population.
Lo nd on : J. M a w ma n .
JEL Classifications
B31
1822
Bibliography
Bking, M. 1958. R i c a r d i a n e c o n o m i c s . New Haven:
Yale University Press.
Deane, P. 1956. Contemporary estimates of national income in
the first half of the nineteenth century. T h e
E c o n o m i c H i s t o r y R e v i e w 8: 339-354.
Espinasse, F. 1887. Colquhoun, Patrick. In D i c t i o n a r y o f
n a t i o n a l b i o g r a p h y ; reprinted. Oxford: Oxford
University Press, 1973.
McCulloch, J.R. 1835. State and defects of British statistics.
E d i n b u r g h R e v i e w 61: 154-181.
McCulloch, J.R. 1837. A statistical account of the British
Empire; Exhibiting its extent, physical capacities, pop-
ulation, industiy, and civil and religious institutions.
London: C. Ivnight.
Colson, Leon Clement (1853-1939)
R. F. Hebert
French engineer, economist and statistician, Colson
was bom at Versailles on 13 November 1853, and
died at Paris on 24 March 1939. Trained in
mathematics as an engineer at the Ecole Poly-
technique and the Ecole des Ponts et Chaussees,
Colson extended his interests to statistics and
economics, eventually teaching the latter at both his
alma maters and at the Ecole des Halites Etudes
Commerciales and the Ecole des Sciences
Politiques. Despite a lifelong career in the French
Ministry of Public Works, he found time to produce
several notable works, including his monumental
Cows d’economiepolitique.
Colson received high marks for the technical
competence of his theoretical exposition. According
to Antonelli, his Cows rendered the doctrines ofthe
French Liberal School 'scientific’. Divisia hailed it
as the best work on pine theory since Walras.
Colson particularly demonstrated his competence in
the field of production by erecting a theory of full
employment on the idea of capital-labour
substitution and the equalization of factor returns at
the margin. He followed the Austrian theory of
value, but at the same time
Colson, Leon Clement (1853-1939)
showed more sympathy to Walras than his con-
temporaries did. His work also had a certain affinity
with Marshall’s in that it integrated mathematics
and geometry into the exposition of economic
theory. As with Cheysson, Colson’s economic
views were affected by his statistical studies as
much as the latter were affected by the former.
Despite exceptional difficulties, he was one of the
first to attempt an estimate of French national
income in the early years after World War I.
Colson was elected to the Institut International
de Statistique in 1906, to the Academie des Sci-
ences Morales et Politiques in 1910 (replacing
Cheysson), and to the Conseil Superieure de
Statistique in 1912. He belonged to the French
Legion of Honour and the Societe d’Economie
Politique, of which he was president from 1929 to
1933. He became an honorary fellow of the Royal
Statistical Society' (London) and was elected an
original fellow of the Econometric Society, founded
in 1931. His most imposing legacy, however, was
the generation of 20th- century French economists
and engineers he trained at the grandes ecoles
during the interwar period.
Selected Works
1901-7. Cows d economic politique, 3 vols. Paris:
Gauthiers-Villars.
1 91 2. Organisme economique et desordre social.
P a r i s : E. F l a mm a r i on .
1914. Railway rates and traffic (trans: Christie, L.
R., Leedham, G. and Travis, C). London:
George Bell & Sons.
References
Divisia, F. 1950. Exposes economiques. L’Apport des
ingenieurs francais aux sciences economiques. Paris:
Dunod.
Flux, A.W. 1939. Clement-Leon Colson. Royal
S t a t i s t i c a l S o c i e t y J o u r n a l 102(4): 624.
Marshall, A. 1933. Alfred Marshall, the mathematician, as
seen by himself. E c o n o m e t r i c a 1: 221 222.
Combination 1823

money and advised against a 100% reserve plan.

Colwell, Stephen (1800-1871) He wrote the book from a point of view that
considered money the handmaiden of commerce.
Henry W. Spiegel

Selected Works

1851. New themes for the Protestant clergy.

Stephen Colwell, American protectionist, was bom
Philadelphia: Lippincott, Grambo & Co.
in Virginia (now West Virginia). After practising
1856. Preliminary essay to List, F. National system
law, he eventually became a successful industrialist
of political economy. Trans, by G.A. Mantile,
and entrepreneur in Philadelphia, where he was a
Philadelphia: J.B. Lippincott & Co.
leading citizen and philanthropist. He was a friend
1 85 9. The ways and means of payment: A full
of Henry Carey’s and shared many of Carey’s
analysis of the credit system with its various
views, especially his ardent protectionism. Colwell’s
modes of adjustment. Ph i l a d e l ph i a : J . B .
appeal for high tariffs on iron manufactures and
other goods resounded in many publications. Some L i p p in c o t t & C o .
of these were addressed to the Presbyterian clergy
and he drew on religion to fortify his economic Bibliography
views. Colwell buttressed his appeal by making it
Dorfman, J. 1946. The economic mind in American civilization
part of a wider view of the world that may be 1606-1865, vol. 2. New York: Viking.
characterized as elitist and supportive of high wages Mints, L.W. 1945. A history of banking theory in Great Britain
but also of inequalities of wealth and status. These and the United States. Chicago: University of Chicago
were to be offset by a stewardship of wealth, that Press.

called for private charity rather than public relief for

the poor, which Colwell opposed. High wages and
private charity thus became complements of high
tariffs.
Colwell’s arguments ran counter to the teachings
of the classical economists, whom he criticized on Combination
grounds that were similar to the criticisms made by
the exponents of German historical economics and John Saville
economic romanticism. Colwell was impressed by
the protectionist views of Frederick List and found a
translator for the original German of the latter’s
National System of Political Economy, to which he
‘Combination’ is a term used for a variety of forms
himself wrote an introduction.
of organization. An obsolescent usage relates to
Colwell is also remembered as the author of The
business firms which have come together in some
Ways and Means of Payment. A Full Analysis of the
kind of merger and today are usually referred to as
Credit System with its Various Modes of Adjustment
monopoly, cartel, industrial combination or
(1859), a massive volume that treats of the financial
multinational. In Britain during the 18th century and
controversies of the time. In this work Colwell
for most of the 19thcentury combination was
supported a private national bank, inconvertible
understood to mean associations of working men
paper money, the real-bills doctrine, the
whose purposes were the raising of wages or the
demonetization of gold, and a national clearing
alteration of working conditions. The term ‘trade
system, all amidst an economy of high prices. He
union’ did not come into common use until after
denied the validity of the quantity theory of
1830 and only in the second half of the century did
it supplant ‘combination’.
1824
Combined labour action in the 18th century was
widespread although by no means was it all directed
by or channelled through formal organizations. The
Webbs ( , eh. T) were in error in
insisting that only those associations that were
formally constituted and in continuous existence
should properly be counted among the early trade
unions. In the 18th century, especially, it was the
collective presence of workers that was the crucial
determinant of industrial response. Formal orga-
nization was not a necessary condition of industrial
militancy, and ‘collective bargaining by riot’ is a
well-documented phenomenon (Hobsbawm , ch. 2).
In 1718 and 1724 West of England clothiers
complained to Parliament that weavers had
‘threatened to pull down their houses and bum their
work unless they would agree to their terms’. A
study of Lancashire textile workers (Turner )
emphasized that the essence of 18th-century
unionism was the persistence of collective pressures
which in given circumstances encouraged collective
action. Associations among the Lancashire textile
workers developed informally out of the
occupational life-style within the community. A
settled group would encourage habits of association
and common action, and skill was always a
consolidating factor, but there was also much
activity among the unskilled and migrant groups.
An incomplete listing of recorded industrial
disputes gives a total of 383 for the whole of Britain
between 1717 and 1800 (Dobson , ch. 1). Most of
these were in England, and of the English figure of
333 disputes just over a third (120) occurred in
London, the main centre of the artisan trades in the
18th century. The occupational breakdown for the
whole country shows 64 incidents from among the
textile workers, mostly in the wool industry; only
seamen and ships-carpenters (each with 37) and
tailors (with 22) exceed a total of twenty. Those
between ten and twenty include coalminers,
workers in the shipbuilding industries, textile
workers other than those in wool or silk,
shoemakers, and most trades in building. The range
of additional trades for which some disputes were
listed is considerable, and hardly any occupation
except the service
Combination
industries was free from industrial conflict (Rule
, ch. 6).
The London journeymen tailors were the most
effective combination in Britain during the 18th
century. Their organization seems to have been
established around 1700 as the result of the coming
together of five ‘box’ clubs: box clubs being a
version of the friendly society. In order to function
effectively the clubs were associations in contin-
uous existence, and their rules and regulations later
provided the basis for the discipline of a trade
union. The meeting places of the box clubs were
public houses recognized by the trade and known as
the ‘house of call’. The box club and the house of
call were the type of organization common to all the
trades that succeeded in establishing more or less
continuous associations. The tailors’ combination
first came into prominence as the result of a petition
presented to the House of Commons in 1721, which
asserted that 15,000 London journeymen had
entered into a combination and engaged upon a
strike. The report of a House of Commons
committee was followed by an Act (7 Geo 1, c. 13)
which fixed wages by the day for summer and
winter, and which also prohibited combinations.
While the Act was going through Parliament the
journeymen briefed counsel, at a reported cost of
£700.
Legislation against combinations in specified
trades was common during the next hundred years;
examples affect the wool trade, (12 Geo, I, c. 34),
hatters (22 Geo II, c. 27), silk weavers (17 Geo III,
c. 55), and the paper trade (36 Geo III, c. 111). At
the end of the century, in 1799, there was passed a
general Act (to be repealed and replaced the next
year by 40 Geo HI, c. 60) which made illegal all
combinations whose purpose was obtaining an
advance in wages or the lessening in the number of
hours worked; and for the next quarter of a century
the Combination Laws were in force, although in
respect of certain trades they were not always
rigorously applied (Webb and Webb , ch. 2). In
1824 all previous statutes in respect of
combinations were repealed but this Act was in turn
repealed and replaced the following year by 6 Geo
IV, c. 129, which permitted combinations
Combination
on strictly defined terms and listed punishments for
the use or the threatened use of intimidation,
molestation or violence in the pursuit of the
declared objects of the industrial action.
The ending of the complete legal prohibition of
combinations was largely due to a growing appre-
ciation of the requirements of a labour market in the
period of early industrialization. It was also in part
the result of adroit political pressure by Francis
Place and, inside Westminster, Joseph Hume. The
most weighty supporter of those who accepted the
demand that workingmen’s combinations should be
free of legal restraint was J.R. McCulloch, who
spoke in the name of orthodox political economy.
McCulloch argued that the Combination Laws were
unjust in that employers and their workmen were
not put on the same level; they were dangerous
because they engendered contempt for the law and
encouraged class hatred; and they were futile
because no action could permanently drive up the
level of wages above the natiual rate. McCulloch
believed in peaceful combination: ‘There is no good
reason why workmen should not, like the possessors
of every other valuable and desirable article, be
allowed to set whatever price they please upon the
labour they have to dispose of’ - but he was clear
above all that no artificial levels of wages could
possibly maintain themselves in a competitive
market. He saw the usefulness of combinations in a
strictly narrow context. He warned against strikes,
as normally benefitting those employers outside the
strike action, and underlined the danger of industry-
wide strikes in reducing the competitiveness of
home industry compared with foreign production
(O’Brien , pp. 366-70). He was notably adamant
against any use of force or intimidation in the day-
to-day activities of the combinations and was
particularly opposed to attempts to compel
workmen to join combinations or participate in
strikes or other action; and he commended the
clauses in the 1825 Act which imposed prison
sentences for those convicted of such intimidation.
McCulloch elaborated his ideas in the Essay on
Wages of 1826, which he published in a revised
version in the Treatise on Wages of the early 1850s,
and he summarized
1825
his views in an article on ‘Combination’ in the 8th
edition of the Encyclopaedia Britannica of 1854.
The character of trade unionism changed in
certain respects during the second quarter of the
19th century as a result of the influence of Owenism
and the ideas of the anti-capitalist theorists such as
Hodgskin, Gray and Thompson (wrongly
designated as the ‘Ricardian socialists’). It is
important not to exaggerate the nature of the change
since unions remained what they had always been:
defensive-offensive bodies
concerned with the betterment of their members.
But in the decade after the legalization of combi-
nations for peaceful agitation there was a notable
growth of cooperative organizations and some
support for Owen’s communitarian ideas. Radical
political economy worked in the same direction.
The distortion of exchange values, so it was argued,
meant that labour exchanged below and
commodities above their natural values, and under-
consumption and the usual accompaniments of
economic crisis were the result. Hence a concern
with equitable exchange relations with an emphasis
upon the Labour Exchanges of the early 1830s and
a longer term preoccupation with money and the
banking system (N.W. Thompson ). Owenism, with
its ideas of cooperation was a central strand in the
organization of the Grand National Consolidated
Trades’ Union in 1834-5 from which, it should be
noted, a number of skilled societies held aloof; and
Owenism, which was critical of policies such as
mutual support funds for unemployment, sickness
and death, and remained only a partial influence
upon union attitudes. After the failure of the Grand
National in the mid-1830s the combinations which
survived - almost entirely made up of skilled
workers and craftsmen - continued with their
traditional sectional approaches to industrial
problems. There remained a residue of Owenite
ideas - in cooperative production for example which
the Christian Socialists of 1848-1854 were able to
draw upon (Raven ; Saville ) - but these were
attitudes which became progressively weaker in the
decades after the half century.
1826
Combinations of working men for industrial
purposes generated a continuous opposition,
amounting to hatred, among the employing classes,
whose distrust and dislike goes back to the early
beginnings of modem industrial society. This
hostility, as well as fear, has been a component part
of the English liberal tradition and has influenced
politicians, administrators and the judiciary. The
last group is especially important in that while
politicians have slowly modified their views and
gradually legislated in more sympathetic ways in
respect of trade union rights and status - at least
until the last quarter of 20th century - the judiciary
have been more wayward in their judgements and
have followed quite closely the vagaries of middle
class opinion. This opinion has moved between
reluctant acquiescence and straight hostility, and the
history of the law relating to combinations and trade
unions has exhibited batches of legal decisions
which have negated the intentions of Parliament.
This was true of the late 1860s and during the
1890s, the latter period culminating in the Taff Vale
decision. The hostility of educated opinion towards
these industrial associations of working men was
well illustrated in a letter written by Richard
Cobden, the quintessential middle- class liberal of
the 19th century, addressed to his brother in 1844:
‘Depend upon it, nothing can be got by fraternising
with the trade unions. They are founded upon
principles of brutal tyranny and monopoly. I would
rather live under a Dey of Algiers than a Trade
Committee’ (Morley , p. 299). These have been
enduring sentiments.
See Also
► idustria
Bibliography
Dobson, C.R. 1980. Masters and journeymen. A prehistory of
industrial relations 1717-1800. London: Croom Helm.
Hobsbawm, E.J. 1964. L a b o u r i n g m e n . London:
Weidenfeld & Nicolson.
Combinatorics
Morley, J. 1881. T h e l i f e o f R i c h a r d C o b d e n ,
Abridged ed. London: Nelson’s Shilling Library, 1910.
Musson, A.E. 1972. B r i t i s h t r a d e u n i o n s , 1 8 0 0 -
1 8 7 5 . London: Macmillan.
O’Brien, D.P. 1970. J.R. McCulloch: A study in classical
economics. London: Allen & Unwin.
Raven, C.E. 1920. C h r i s t i a n s o c i a l i s m 1 8 4 8 - 5 4 .
London: Macmillan.
Rule, J. 1981. The experience of labour in eighteenth- century
industry. London: Croom Helm.
Saville, J. 1954. The Christian socialists of 1848. In
D e m o c r a c y a n d t h e l a b o u r m o v e m e n t , ed.
J. Saville. London: Lawrence & Wishart.
Thompson, E.P. 1963. T h e m a k i n g o f t h e E n g l i s h
w o r k i n g c l a s s . London: Victor Gollancz.
Thompson, N.W. 1984. The people’s science. The popular
political economy of exploitation and crisis, 1816-34.
Cambridge: Cambridge University Press.
Turner, H.A. 1962. Trade union growth, structure and policy. A
comparative study of the cotton unions. London: Allen &
Unwin.
Webb, S., and B. Webb. 1894. T h e h i s t o r y o f t r a d e
u n i o n i s m . London: Longmans & Co.
Combinatorics
A. P. Kirman
Combinatorics, or combinatorial mathematics, is a
difficult field to define. It cuts across many
branches of mathematics yet a mathematician will
clearly sense which problems are of a combinatorial
nature. Perhaps the simplest definition is that it is
concerned with configurations or arrangements of
elements, usually finite in number, into sets. Three
basic types of problem are posed. Firstly the
existence of certain configurations; secondly, once
their existence is proved, the classification or
enumeration of the configurations meeting the
requirements imposed; and thirdly the construction
of algorithms for finding the configurations in
question.
Why has combinatorics been the poor relation of
the mathematical tools used in economics? The first
and most obvious explanation is that the evolution
of theoretical or mathematical economics has led us
in the opposite direction to that in which
combinatorics is useful. Ever since the
Combinatorics
‘marginal revolution’ there has been clear emphasis
on the use of differential methods and an implicit
acceptance of the perfect divisibility of goods.
Indeed if we consider the most elaborate extension
of the basic Arrow-Debreu model it is to one in
which there is a continuum of agents and a
continuum of goods. This is just the sort of context
in which the combinatoric approach is of little use.
Yet the economist may well feel that discrete
problems are of importance and that indeed the
world is best represented as one where goods are
not infinitely divisible and where the number of
agents is finite. Now given the standard assump-
tions of convexity, at least in production, the per-
fectly competitive model may indeed be regarded as
a satisfactory ideal or limiting case and the
finiteness of a real economy is just an inconve-
nience. In this case it would seem that combina-
torics has little role to play.
However although we may consider taking
divisibility as a reasonable idealization in a large
economy (even this may be questioned - see ►) as
soon as
convexity is dropped as an assumption for pro-
duction the problem of finiteness cannot be avoided.
If there are increasing returns to scale there will be a
minimum profitable plant size and there will be a
fundamental indivisibility. At this point
combinatoric analysis comes back into its own, and
no argument can be made for using infinitely
divisible goods as a reasonable approximation. Thus
the existence of non-convexities will lead us back
into a situation which may, for example, be game-1
ike and in which we will be looking for a solution
with a finite number of large plants.
This is to suggest that realism may lead us away
from the smooth differential world to which
economists are accustomed towards a discrete one
in which combinatorics plays an important role.
Nevertheless, till now, finiteness and the com-
binatoric approach have not occupied a significant
place in economics. However, some examples will
show that certain branches, although not central,
have made extensive use of such an approach.
1827
The development of mathematical program-
ming, in particular of linear programming, has
relied particularly on combinatorics. The algorithms
developed to solve such problems are essentially
combinatoric. Many economic models have been
built on the basis of the ‘activity analysis’ or fixed
coefficient production approach.
Game theory has made extensive use of com-
binatorics and provides an interesting example of
how the combinatoric approach can be confronted
with one using continuous functions or compact
sets.
Combinatoric arguments are used to show that
the core of a balanced game is non-empty. A simple
argument shows that a market game is balanced and
further it can be shown that the only allocations
remaining if we replicate a given economy are
competitive. This leads us to the conclusion that a
competitive or Walrasian allocation exists. Now the
latter statement is known to be equivalent to the
existence of a fixed point for a continuous mapping.
Thus we arrive by combinatorical methods at the
same result as that obtained by an apparently very
different tool.
This argument is reinforced by the fact that the
algorithms developed for finding approximate
competitive equilibria are essentially combinatoric
in nature. They consist in systematically examining
points in the price space, of evaluating the total
excess demand of an economy at these points and
finding a path which leads to a reduction in the
excess demand until it is close to zero. Again, the
approximation of fixed points of countinuous
functions is obtained by a combinatoric approach.
There are many other examples of ways in
which combinatoric analysis has proved useful.
Arrow’s theorem on the impossibility of a social
welfare function is typically proved in the case of a
finite number of individuals faced with a finite
number of social alternatives. The problem is to
find a way of aggregating individual preference
orders on these alternatives into a social order in a
way which satisfies certain simple axioms. The
usual line of proof consists of finding certain
configurations of individual preferences which
1828 Command Economy

lead via the axioms to a contradiction with the

existence of a social order. The reasoning here too Command Economy
is combinatoric. The introduction of graph theory to
describe communication patterns in economics has Richard E. Ericson
brought into economic theory a field which has
long been regarded as fundamentally combinatoric.
The use of ‘matching’ models to analyse job search
and unemployment is yet another example. Abstract
Whilst these few rather arbitrary examples show The concept of a ‘command economy’, a con-
that combinatorial analysis has not been absent struct in the theory of comparative economic
from economic theory it is also clear that it has not systems, is defined, and its origins, character-
been central. istics, and consequences for any society in
However, it seems likely that economics is to which it is implemented are explored. The
evolve towards the sort of models now widely impossibility of the absolute centralization
studied in computer science, away from global which it requires generates compromises with
optimization towards more simple forms of ‘ratio- the market forces it aspires to replace, fostering
nality’ as embodied in simple automata. Further- a symbiotic marketized ‘second economy’
more the computation of equilibrium even for which systematically undermines its founda-
underlying continuous models is becoming tions. Hence, although initially appearing to be a
increasingly important. All this together with a true alternative to the market economy, a
recognition of certain fundamental indivisibilities in command economy, most nearly realized in the
economics is likely to move combinatorics to a Soviet Union (1930-87), proved to be ultimately
much more central position on the stage of eco- non-viable, collapsing under reforms attempting
nomic theory. to make it competitive with market systems.

Keywords
Active vs passive money; Aggregation; Balance;
See Also Bounded rationality; Bureaucracy; Central
planning; Centralization; Command economy;
► A: v:
Command mechanism; Command principle;
► Fixed Point Theorems
Communism; Complex social economy;
► 2'e::.:..e Tiec::/
Corruption; Decentralization; Discretion;
► ' KVI
Gorbachev, M. S.; Gross output; Inca production
system; Incentive provision; Industrialization;
► :: _;:msey, Frank Plumpton (1333-1930) Inequality of income; Information; khozraschet;
Labour discipline; Market mechanism; Market
References versus plan; Microbalance; Moneyness; Mormon
economic system; Neurath, O.; perestroika;
There elegant and elementary introductions to combinatorics
are Berge (1971), Polya, Tarjan and Woods (1983) and
Price control; Principal and agent; Rationing;
Ryser (1963). A more advanced text is Aigner (1979). Resource allocation; Second economy; Sellers’
Aigner, M. 1979. C o m b i n a t o r i a l t h e o r y . Berlin: market; Shadow economy; Soft-budget
Springer. constraint; Soviet economic reform; Soviet
Berge, C. 1971. P r i n c i p l e s o f c o m b i n a t o r i c s . New
York: Academic.
Union; Stalin, J.V.; Suboptimization; Unit of
Polya, G., R.E. Taijan, and D.R. Woods. 1983. N o t e s o n measure; Vested interests; War and economics;
introductory combinatorics. Boston: War communism
Birkhauser.
Ryser, H.J. 1963. C o m b i n a t o r i a l m a t h e m a t i c s ,
Mathematical Association of America. New York: Wiley.
Command Economy
JEL Classifications
P3
A command economy is one in which the coordi-
nation of economic activity, essential to the viability
and functioning of a complex social economy, is
undertaken through administrative means -
commands, directives, targets and regulations -
rather than by a market mechanism. A complex
social economy is one involving multiple significant
interdependencies among economic agents,
including significant division of labour and
exchange among production units, rendering the
viability of any unit dependent on proper
coordination with, and functioning of, many others.
Economic agents in a command economy, and
in particular production organizations, operate
primarily by virtue of specific directives from
higher authority in an administrative/political
hierarchy, that is, under the ‘command principle’.
Thus the life cycle and activity of enterprises and
firms, their production of output and employment of
resources, adjustment to disturbances, and the
coordination between them are primarily governed
by decisions taken by superior organs responsible
for managing those units’ roles in the economic
system. One of the most distinctive features of such
an economy is the setting of the firm’s production
targets by higher directive, often in fine detail. The
administrative means used include planning,
material balances, quotas, rationing, technical
coefficients, budgetary controls and limits, price
and wage controls, and other techniques aimed at
limiting the discretion of subordinate operational
units/ftrms. The command principle strives to fully
and effectively replace the operation of market
forces in the key industrial and developmental
sectors of the economy, and render the remaining
(peripheral) markets manipulable and subordinate to
political direction. Thus the command principle is
likely to clash with the operation of market forces,
yet a command economy may nonetheless contain
and rely on the market mechanism in some of its
sectors and areas, for example, influencing labour
allocation, or stimulating
1829
small-scale private production of some
consumables.
The term ‘command economy’ comes from the
German Befehlswirtschaft, and was originally
applied to the Nazi economy, which shared many
formal similarities with that of the Soviet Union. It
has received its fullest development in the analysis
of the economic system of the Soviet Union,
particularly under Stalin, although it has been
applied to wartime administration of the US
economy (1942-6; see Higgs ), the Mormon
economic system in mid-19th century Utah
(Grossman ), and
the Inca production system in the 16th century
Andes (La Lone and La Lone ). Synonymous or
near-synonymous terms include ‘centrally planned
economy’, ‘centrally administered economy’,
‘administrative command economy’, ‘Soviet-type
economy’, ‘bureaucratic economy’ and ‘Stalinist
economy’.
The command economy’s conceptual origins go
back to the Viennese economist Otto Neurath, who
in the years before and after the First World War
developed an extreme version (to the point of
moneylessness) based chiefly on prior experience
with wartime economies (Raupach ). The concept of
the command economy has since become a central
conceptual framework in the analysis of economic
systems, as it captures a logically coherent
alternative to ‘the market’ as a way of organizing
socially complex economic activity and interaction.
The Soviet Union provided the most complete, and
for a while successful, example of a command
economy as a working alternative to a market
system. Indeed, apart from the relatively short-lived
Nazi case, and even briefer ones under emergency
conditions in some other countries, especially in
wartime, actual instances of command economies
are virtually limited to Communist-ruled countries,
with the USSR as the prototype and prime
exemplar. Thus, what follows is mainly inspired by
the Soviet example (Ericson ) as it existed,
essentially little altered since its appearance in the
1930s, until its collapse in the aftermath of
President Gorbachev’s perestroika, begun in 1987.
1830
Nature of the Command Economy
The seminal analysis of the nature, characteristics,
and problems of a command economy is Grossman
( ).
Any complex social economy must, for its very
survival, maintain at least a ‘tolerable’ micro-
balance, ‘that minimal degree of coordination of the
activities of the separate units (firms) which assures
a tolerably good correspondence between the supply
of individual producer and consumer goods and the
effective demand for them’ (Grossman , p.
101). In such an economy,
appropriate balance can be achieved through
decentralized, market-based (monetized, price-
mediated) interaction of autonomous units, or by
virtue of explicit specific coordinating directives
(commands, targets) from some higher authorities.
While the former is characteristic of a market
economic system, the latter is defining of a ‘com-
mand economy’. In the latter operational-level units
(for example, firms) must merely ‘implement’
commands; they become ‘executants’ of plans and
directives from above, plans which must insure
balance through the coherence and consistency of
the instructions they give. Thus the command
mechanism requires relative centralization and
severe restriction on the autonomy of subordinate
operational units. It derives from the overwhelming
priority of social goals, and requires the severe
limitation, if not total destruction, of autonomous
social and economic powers and the enforcement of
strict obedience to directives.
A command economy is hence a creature of
state authority, whose marks it bears and by whose
hand it evolves, exists and survives. Command
economies are imposed, whether through external
duress or imitation, or indigenously in order to
achieve specific purposes such as (a) maximum
resource mobilization towards urgent and overrid-
ing national objectives, such as rapid industriali-
zation or the prosecution of war; (b) radical
transformation of the socio-economic system in a
collectivist direction based on ideological tenets and
power-political imperatives; and (c), not least,
curing the disorganization of a market economy
brought about by price control, possibly
Command Economy
occasioned by inflationary pressure arising from (a)
and/or (b).
The command economy therefore requires a
formal, centralized, administrative hierarchy staffed
by a bureaucracy, and it also needs to be embedded
in (at least) an authoritarian, highly centralized
polity if it is not to dissolve or degenerate into
something else. And that bureaucracy, if it is to
effectively implement the command principle, must
exercise full control and discretion, if not
necessarily formal ownership, with respect to the
creation, use and disposal of all productive property
and assets. At the same time, each office or firm and
every economic actor within the command structure
holds interests which, if only in part, do not
coincide with those of superiors or of the overall
leadership. This generates important problems of
vested interests, principal-agent interaction,
incentive provision, and general enforcement of the
leadership’s will, and calls for a variety of
monitoring organizations (party, police, banks, and
so on). The term ‘command’ must not be taken to
preclude self-serving behaviour, bureaucratic
politics, bargaining between superiors and
subordinates, corruption, peculation and
(dis)simulation. On the contrary, such behaviour
tends to be widespread in a command economy; yet
the concept of a ‘command economy’ remains valid
so long as, in the main, authority relations and not a
market mechanism govern the allocation of
resources.
When not externally imposed, command econ-
omies typically arise from a millennialist elite, with
unique access to ‘the truth’, achieving the political
power to impose its will, while facing a crisis of
apparently overwhelming proportions. The
perception of a life-threatening crisis, driving the
need for massive mobilization of all social
resources and rendering potentially disastrous any
hesitation or dissent, any questioning of ways and
means, naturally leads, pushed by the ‘logic of
events’, to the usurpation of all power of discretion,
all legitimate authority, by the ‘knowing’ elite,
which then becomes responsible for all that is done
or not done in the society and the economy. The
crisis may be artificial or real (‘hostile
encirclement’), externally or internally imposed (the
need to industrialize, to ‘catch up’),
Command Economy
but it requires moving resources rapidly and mas-
sively, forcing new activities and interactions in the
face of severe scarcities, of shortage of competent
personnel, of massive uncertainties, and of strongly
held, stark priorities. Indeed, a sense of
overwhelming urgency and need for haste drove the
elite of the Soviet Union in the 1930s to test and
establish, through trial and error over several
decades, the institutional structure of a ‘command
economy’, albeit less than absolute from both
necessity and choice (for example, due to the
‘lessons’ of ‘War Communism’) (Grossman ;
Zaleski ).
Consequences of Command
Rational application of the command principle calls
for planning, which is basically of two types.
Longer-term, developmental planning expresses the
leadership’s politico-economic strategy (for
example, five-year and ‘perspective’ plans);
shorter-term, coordinative planning (annual, quar-
terly, monthly, ten-day) ideally translates the strat-
egy into resource allocation while aiming to match
resource requirements and availabilities for indi-
vidual inputs, goods, and so on, in a sufficiently
disaggregated way for given time periods and loca-
tions. The task of elemental coordination, of micro -
balance, so effortlessly accomplished by any func-
tioning (however poorly) market system, is over-
whelmingly large, and grows rapidly with
industrialization and economic development, both
of which lead to exponential growth of the com-
plexity of die economy, and hence of the planning
problem. With centralization and the abandonment
of markets comes the need for massive, detailed
coordinative planning, for ‘making ends meet’ in
the expanding web of interconnections that must be
maintained for economic life to continue. Coordi-
native planning serves, therefore, as the basis for
specific operational directives to producers and
users, thereby implementing the command principle
to achieve the prime imperative of a social economy
- ‘balance’.
It is t h i s task that in fact consumes the largest part of
the so-called planning in the command economy...
Coordinative planning as it is conducted in the
1831
Soviet Union does little by way of consciously
steering the economy’s development or finding effi-
cient patterns of resource allocation. Its overwhelm-
ing concern is simply to equate both sides of each
‘material balance’ by whatever procedure seems to
be most expeditious (Grossman , p. 108).
A major problem is that detailed planning and
the corresponding directives are often late, are
insufficiently detailed, may lack the requisite
information, hence often cannot be effectively
coordinated, and owing to their rigidity are pecu-
liarly vulnerable to uncertainty (Ericson ).
Information in the command sector, by the logic of
the system, tends to flow vertically up and down the
administrative hierarchy rather than horizontally
between buyer and seller, adding to difficulties of
demand-supply coordination by informationally
isolating operational units from their suppliers and
users. In addition, problems of motivation,
accountability (down as well as up), inappropriate
decision-making parameters, and divergent interests
complicate the procedure. Even at best, this manner
of resource allocation can hope to attain only
internal consistency (in the sense of effectively
matching partially disaggregated requirements and
availabilities) but not a higher order of economic
efficiency. Economic calculation in pursuit of
efficiency enters, if at all, at the project-planning
stage, and not short-term resource allocation and
use.
These problems are aggravated by the logic of
haste that drove imposition of the command
economy - ‘the pressing contrast between urgent
political goals and available resources’ (Grossman ,
p. 108). The necessary attention to the growing
problem of balance further militates against any
effort to consider developmental objectives or effi-
ciency in making allocative decisions, so that a
further bias against allocative efficiency is built into
the command economy. Coupled with limited
ability to gather, filter, process, and communicate
information, and to compute solutions to planning
problems, this creates a fundamental and growing
inability to acceptably solve the underlying coor-
dination problem, and hence further undermines any
consideration of efficiency.
The logic of ‘command’ has a number of other
consequences reflected in the institutions of such
1832
an economy. Planning in a command economy must
be largely in physical terms due to the crucial
importance of balance. The bottom line of the
planning process must be available physical units of
required inputs, in appropriate assortment, quantity
and timing, necessitating physical targets for
production and input utilization. Thus tens of
thousands of materials and equipment balances
must be drawn up and coordinated for each plan
period, and then broken down and allocated in
directives to specific implemented. And, to be
directly usable, these must be in physical or crypto-
physical (constant price) units that directly relate to
the production processes being coordinated. Using
economic-value units requires flexible and
changing, marginal scarcity-based prices for
valuation, as well as giving significant autonomy to
subordinate units that inevitably then will make the
trade-offs in assortment, quantity and timing within
planned constraints on values (that is, ‘budgets’).
Hence, such valuations pose a fundamental
challenge to the command economy.
Planning in physical terms, however, leads to
‘enormous waste and inefficiency, to production for
waste as much as for use’ (Grossman p. 110). There
are at least three fundamental sources of this
elemental waste: grossness, aggregation, and unit
of measure. The need for these arises from the
overwhelming complexity of the task of planning
for, and directing the operation of, a complex social
economy and the necessarily limited information
gathering, processing, and dissemination
capabilities of any economic agent or agency.
However, the emphasis on gross output leads to
‘input intensiveness’, waste, and ignoring cost
considerations. Aggregation leads to persistent
subcategory imbalance in assortment, quality, type,
timing, and so on, while units of measurement
determine suboptimization objectives, distorting
implementation decisions, particularly when they
are, for material balance reasons, input oriented.
Thus each of these is essential for the feasibility of
directive central planning, of the command
mechanism, yet each loses, or destroys, essential
information for the ‘proper’ (in the eyes of the
system directors) implementation of plans, and
opens space for creative interpretation of
instructions/commands, and hence for
Command Economy
‘suboptimization’ by implementing units whose
interests are not perfectly aligned with those of the
centre (Nove ). While the command mechanism
logically requires unauthorized initiative to be
forbidden, and strictly punished when exercised, the
size of the task it faces inevitably opens the
opportunity, indeed often the need, for such
unauthorized initiative. Thus the physical quantity
planning required by the command economy to
maintain minimal functional ‘balance’ contains its
own antithesis, unleashing forces that undermine
the consistency of the plan and the coherence and
balancedness of its realization. This fundamental
contradiction lies behind most of the critical
problems of the command economy in the Soviet
Union and the myriad efforts to resolve them within
the framework of the command mechanism that
comprise the endless waves of reform following
victory in the Great Father- land War of 1941-5.
The ‘logic of command’ thus imposes a need to
restrict autonomy, to restrict the capability of eco-
nomic units to pursue any other than ‘planned’ or
commanded purposes: economic agents must not
have the capability to autonomously acquire and
deploy resoiuces for any purposes outside the plan.
Comprehensive material balance planning and
centralized materials and equipment allocation
provide a necessary component, but one that is
insufficient unless resources, including human, are
denied the capability of autonomous movement and
application. Severe restrictions on labour mobility,
albeit not as severe as under Stalin, are required, as
are comprehensive restrictions on the use of any
‘generalized command over goods and services’ -
that is, money - that might be used to alter their
patterns of allocation and use in the economy. The
system must be substantially demonetized in order
to ‘... constrict the ... range of choice in the face of
the state’s demands’ (Grossman , p. 232).
Thus money must be deprived of ‘moneyness’
and prices must be kept ‘passive’, as mere
accounting and measurement units. According to
the logic of the command economy, the availability
of money and the prices at which commodities and
products are provided should have no essential
impact on the allocation of goods and services,
Command Economy
or on the nature and direction of economic/indus-
trial development; all real activity is preordained in
the plan and its subsequent implementing com-
mands. The role of money is then to facilitate
monitoring of commanded performance through the
financial flows it generates. Thus monetary prices
do not, and indeed should not, reflect to a
substantial degree social goals and priorities; they
merely reveal and measure the flow of commanded
activity. Producer prices (and most retail prices),
wages, prices of foreign currencies, and so on are
generally centrally set and controlled, often
remaining fixed for long periods of time. Micro-
disequilibria naturally abound, while the widely
perceived dubious meaningfulness of such prices
and the administrative allocation of most producer
goods in physical terms combine to sustain the
system of detailed production plans and directives in
terms of physical indicators - yet another bar to
more efficient planning and management.
Finally, an absolutely essential, indeed defining,
institution of the command economy is the physical
rationing of resources and producers’ goods. This is
where the market is most fully and directly replaced,
and where the central authorities have the ability to
most directly influence and control the behaviour of
subordinate operational units. It implements the
centralized mobilization of resources to priorities,
the most direct response to crises and challenges.
And it most directly denies to subordinates the
capability to produce, to develop, in ways outside
those authorized in the plan. This makes the
coexistence between the command principle and the
market mechanism a source of continual conflict, as
the market opens unauthorized opportunities to sub-
ordinates. In the Soviet Union the command prin-
ciple, aided by the club of materials rationing,
repeatedly pushed back and eliminated the market
mechanism when (timidly) introduced in reforms,
until the system collapsed in chaos, and the intro-
duction of a full-fledged market economy was
begun in 1992 (Schroeder ; Aslund ). Thus the
nature of the command system makes it
fundamentally incompatible with real markets,
although some market institutions can, and indeed
must, be allowed to function both within the non
1833
state sectors and as the interface between them and
state economic institutions/sectors.
Inherent Challenges to the Command
Economy
As Grossman notes in his seminal article ( ,
p. 107), ‘The chief persistent systemic problem of a
command economy is the finding of the optimal
degree of centralization (or decentralization) under
given conditions and with reference to given social
goals’. The fundamental dilemma is that full
centralization poses an insoluble problem, while
decentralization abandons the ability to direct, to
control development, and to ensure the pursuit of
social goals and priorities. With regard to the pine
planning problem, a large body of theoretical
literature arose in the late 1960s, and continued into
the 1980s, on the problem of decentralizing the
planning process to make its informational and
computational burden manageable (Eckstein ;
Bomstein ). But the problem is far greater, and less
studied, with respect to implementation; rational
planning is swamped by the struggle to maintain
elemental coordination.
Decentralization Versus Priority
Looked at through the prism of relative advantages,
operational decentralization shortens ‘lines of com-
munication’, increasing flexibility, adaptation and
responsiveness to a changing environment through
local initiative and innovation, and vastly simpli-
fying the decision problem of economic agents. But
it does so at the cost of weakening or losing the
‘advantages of centralization’, including
enforcement of regime values, capability for large-
scale resource mobilization, concentration of scarce
talent in central decision-making organs, and the
maintenance of macro-balance. In particular,
decentralization compromises the ability of the
centre to directly manage the development and
structure of the economy and to force the achieve-
ment of critical priorities regardless of cost Fur-
thermore, decentralization requires the introduction
1834
of the alternative coordination mechanism to insure
tolerable micro-balance - the market - as decen-
tralization undercuts the ability to directly coordi-
nate, to balance from above. Thus, to prevent
catastrophic imbalance, a more active money with
economically flexible market prices must be allo-
wed to function in a decentralized system.
The impossibility of planning and commanding
the performance of all economic agents in full
operational detail, however, forces some decen-
tralization. This creates a chronic threat to balance
which is thus a continuous argument for (re)
centralization of planning and materials allocation.
Furthermore, a partial decentralization of planning
and management in a command economy may do
more harm than good; it may impair balance
without yielding sufficient benefit. Yet a complete
decentralization, in the sense of a virtually full
devolution of the major production decisions to the
firm level, would be disastrous from the standpoint
of balance, unless the price structure were properly
altered to provide proper signals to firms and
suitable behavioural rules were prescribed, that is,
unless a market mechanism were introduced. Thus
the logic of command predicts a ‘treadmill
ofreforms’ (Schroeder ), an array of countervailing
strengthenings of the oversight and control organs
(in particular, the Party), and enhancements of their
role in the economy, accompanying moves towards
decentralization in the state sector. It also explains
the Soviet institutional arrangement of inter-firm
contracts as a decentralized implementation device.
These are required to specify details of interaction
within planned categories, and establish observable,
and hence legally enforceable, commitments to
planned implementation, constraining the autonomy
necessarily granted through the minimal
decentralization. And it explains the logic of the
continuing restraints on the use of money and the
continuing efforts at effective price control to keep
the autonomy of agents restricted to the minimum
necessary for the continued functioning of the less-
than-absolute command economy.
Even limited decentralization requires that
money be used in the command sector (as well as in
the household sector), but its role as a bearer of
options and as the means of pecuniary calculation
Command Economy
for decision-making is necessarily limited and
deliberately subordinated to the planners’ will and
the administrators’ power. Banks and the treasury
accommodate the money needs of production,
ensuring a soft budget constraint for the individual
firm. At the same time, the ‘moneyness’ of money
at the firm level is low, hemmed in as it is by
administrative constraints and impediments,
including the rationing of nearly all producer goods,
and by the widespread ‘seller’s market’ (shortages
of goods and absolute lack of buyers’ alternatives).
This monetary ease, together with the sellers’
market, plays an important role in ensuring
individual workers’ job security at the firm level
and full employment in the large, while keeping the
firm largely insensitive to money costs and/or
benefits.
Within the command sector, money and prices
have a necessary role in determining terms of
alternate resource uses only within planned/
commanded categories, and money has the role of
limiting total claims to resources in areas, or at a
level of detail, beyond the reach of plan directives.
This requires ‘businesslike management’ within the
firm - khozraschet, which is a ‘set of behavioral
rules that is supposed to govern the actions of
Soviet managers beyond their primary responsi-
bility, the fulfillment of output targets’. It pushes
the firm toward ‘technical efficiency’ and limitation
of ‘claims on society’s resources for productive use.
... khozraschet is a system that is well devised to
control the behavior of managers in a command
economy where a certain amount of devolution of
power to them is inevitable, and where, further,
managers’ goals and values do not necessarily
coincide with the official ones’ (Grossman , p. 117).
Thus money also has the role of facilitating the
monitoring of performance in the command sector.
While administrative orders are the mle in a
command economy, backed up by greater or lesser
degree of state coercion (depending on country and
period), any decentralization of implementation
naturally relies on monetary (‘material’) incentives
to elicit desired individual compliance and perfor-
mance. Compounding the incentive problems aris-
ing from differences in information and interests
between central authorities and implementing
Command Economy
agents is the fact that the physical and other indi-
cators to which the material incentives are linked
may often be poor measures of social benefit (as
seen by the leadership). Furthermore, resort to such
rewards widens the distribution of official earnings
and raises questions of permissible limits of income
inequality. Yet there may be little choice in that the
state must in effect compete with the much higher
incomes from the second economy. Indeed, the
Soviet Union during War Communism, Cuba in the
1960s, and the People’s Republic of China during
some periods before Mao’s death in 1976 tended to
downgrade material incentives in favour of
normative controls, but never did quite abolish
them.
The behaviour of the Soviet-type firm has been
much studied (Granick ; Berliner ; Nove
; Freris ). Because its directives and the
corresponding managerial incentives stress physical
output, produced or shipped, and thanks to its low
sensitivity to cost and the ambient sellers’ market,
the firm often sacrifices product cost, quality,
variety, innovation and ancillary services to its
customers to sheer product quantity. By the logic of
command and the requirements of plan
manageability, firms operate in an environment with
sole suppliers and assigned users, reducing
complexity by eliminating ‘wasteful’ redundancy in
production and distribution. Thus firms in a
command economy are largely insulated from any
product competition, both from the outside world
and from other domestic firms, thanks to the climate
of administrative controls and the prevalent excess
demand for their output. Difficulties with supply,
frequent revision of its plans, interference by Party
and other authorities, and other systemic problems
also stand in the way of its more efficient and
effective operation. Indeed, to function at all, the
firm’s management is frequently forced to break
rales and even resort to criminally punishable acts.
This compounds a further critical challenge
posed by necessary decentralization - the conflict
between the will, purposes, incentives and priorities
of the higher authorities and those at lower levels,
particularly of the firms and their managements.
Even the best-motivated managers, following all
official rales and incentives, will sometimes
1835
fail to replicate the decisions that their superiors
would have made had they been in a position to
make them. This problem is aggravated by the
inevitable ambiguity, incompleteness and inconsis-
tency of those rales, incentives and the information
available on the spot. Only binding physical con-
straints and observable outcomes can be systemat-
ically enforced, making ‘centralized materials
allocation the most powerful weapon at the disposal
of the central authorities’ (Grossman , p. 118).
Thus, where material inputs are less determinate of
a unit’s activities, this information and incentive
problem is greater, and the defiance of central will
relatively more widespread and successful. This
observation explains the non-viability of any reform
that fails to fundamentally alter the materials
allocation system.
Under-Planned, Ill-Commanded Sectors
A major challenge to the command economy also
arises from the existence of sectors outside, or only
partially affected by, the command principle. In the
Soviet Union these included most of agriculture,
much of housing, the household sector and some
consumer goods and services. ‘Markets’ were
allowed to function for the distribution of final
consumer goods and services, including agricultural
produce, for much of the activity of the ‘collective’
sector in agriculture and for household labour
supply. For transactions with ‘personal property’
within the household and collective sectors, money
was active and agents responded to market prices,
while in the quasi-markets interfacing with the state
sector - for example, labour and consumer goods -
money was relatively active but prices remained
largely controlled and non- market. These are
sectors where information on needs/preferences and
capabilities proved too difficult to acquire reliably
in real time for acceptable allocation and balance to
be commanded, and so at least one side of a market
was allowed to function with an active money.
Here, the command mechanism proves too crude
and clumsy, and hence politically
counterproductive, to be used outside of pressing
emergencies. Indeed, this might be considered a
lesson of War
1836 Command Economy

Communism, the first experience with a command than-absolute command economy. As the complete
economy in Soviet Russia, 1918-21. centralization of decisions in the production sector
(let alone in the household sector) is an
In view of the theoretical incompatibility of
command and market, how could these ‘market’ impossibility, something must be left to local ini-
tiative and dispersed decision making. Thus
sectors be successfully grafted on to the command
mechanism? An explanation (Grossman ) rests on khozraschet is a logical necessity, ‘... an unfriendly
the trade-offs between the authorities’ limitedbridgehead that threatens to seize ground whenever
the planner fails or defaults’ (Grossman , p. 228).
capabilities, the complexity of those sectors, and
With the inevitable devolution of some decision
their centrality to regime priorities. A sector which
making to firms and households, money acquires a
provides significant inputs to physical planning and
necessary and critical role in the command
plan fulfilment, where the unpredictability in the
economy, going well beyond that consistent with
flow of goods is unacceptable, cannot be left to the
market without seriously undermining command. the logic of command. That role arises from the
need to economize in making decentralized
However, if a ‘market’ sector can be treated as a
residual for purposes of materials planning anddecisions, and as a medium of exchange and store
of value in the decentralized interactions that relate
allocation, a buffer for planning, then its coexis-
to all decisions. In acquiring this role, this
tence is acceptable. Further, if its operation is char-
‘moneyness’, it allows accumulations of power
acterized by rapid change and complexities rather
outside the control of the regime. Money is a
outside the core interests of the regime, if without
‘bearer of options’ whose power and influence must
disrupting the industrial core greater incentives and
be restrained if the command mechanism is to
risk can be placed on those peripheral agents, and if
operate properly - to determine priorities and to
non-market constraints can force the desired market
response from it, then the centre will want to insure maximal commitment to their achievement.
separate that sector from the command sphere, As Grossman ( , p. 214)
noted, ‘Money is a form of social power that may
lowering its coordination burden by shifting it to the
market. lead resources astray and is subject to only imper-
fect control by political authority.’
These considerations were indeed active in the
case of those sectors ‘left to the market’ in the Thus the power of money has to be curbed in a
Soviet Union: consumer goods retailing, the command economy by limiting balances available
acquisition of labour services, the support of to households and firms, by compartmentalizing
households in the countryside through a privatemoney into cash and ‘firm’ circuits, and by erecting
barriers and limits to the use of ‘monies’ in each
agricultural sector, and a few peripheral and inter-
category, although that undercuts the effectiveness
stitial activities. Indeed, any attempt to truly
of
‘marketize’ any other sectors or activities in the attempted decentralizations. Liquidity,
‘moneyness’, is constrained by the institutional
command economy is doomed to fail unless the loss
structures and by all the characteristics and
of fervour, of the sense of mission and urgency,
leads to abandonment of the command mechanism. conditions of the ‘sellers’ market’, rendering
Yet even the existence of these limited market ‘money’ the only non-scarce commodity, in
unusable excess to the extent the command
sectors, providing an outlet for incentive earnings
and diverted resources, exerts a continuing mechanism is effective. Monetary policy in the
corrosive pressure on the command economy and properly functioning command economy reduces to
its control mechanisms. limiting the volume of cash in the economy
(‘macro-monetary’ control) through wage fund
restrictions and cash control absorption plans of the
The Cancer of 'Money' retail sector, and the allocation of firm balances in
restricted categories (‘micro-financial’ control) in
A truly monumental challenge to the command just sufficient quantity to support the
economy lies in the role of money in any less- implementation of the plan, with confiscation of
Command Economy
excess funds to prevent unauthorized activity by the
firm (Garvy ).
Similarly, the price system, expressed in terms
of that money, must also be mobilized to the
purpose of control. The inflexible, administratively
segmented, average cost-based prices in command
economies are a logical necessity of command- and
haste-based shortage. For all the problems they
cause, all the unintended consequences and
distortions in the behaviour of subordinates, such
prices help to keep money largely passive, at least
in the core state sectors, and allow both money and
prices to remain instmments, rather than disrupters,
of command. More than being ideologically
justified, such prices are a response to the pragmatic
and pressing requirements of running a shortage
economy with a rapidly developing system of
centralized direction of enterprises and of materials
allocation.
Money, however, is not so easily contained.
Once in unobserved hands, it exercises its ‘com-
mand over goods and services’ without reference to
plans, commands or regime priorities. Hence, given
any discretion, in any sphere of activity not directly
monitored agents will naturally use money in ways
they find desirable, placing new demands on a
physical system otherwise tautly planned and
characterized by general scarcity. This is facilitated
by the existence of agents and spheres of activity
outside the command system, providing ‘legitimate’
sources and uses for monies, however acquired or
disposed. And the possibility of acquiring money
provides incentives for unauthorized activities,
incentives to undertake unplanned interactions and
reallocations. An active money vastly expands the
sphere of discretion of ‘subordinate’ agents beyond
any authorized by a decentralizing reform, and calls
for severe administrative restrictions, a reduction to
passivity, if it is not to disrupt the planned activities
and discretion of the central authorities.
Yet attempts to administratively constrain the
influence, the ‘corruptive’ power, of money become
increasingly futile once the ‘genie’ has been ‘let out
of the bottle’. Even limited decentralizing reform,
allowing money to influence some (subcategory)
production and allocation decisions, inevitably lets
loose more liquidity,
1837
more of a command over goods and services, than
desired. This arises from a multitude of factors:
errors in both physical and financial plans, inherent
incompleteness of plans and commands due to
limited information and time and to the necessity of
aggregation, changing circumstances and shocks to
the economy, mistakes in implementation and in
responding to shocks, the irregularity and
disruptions in the materials allocation system, the
behavioural response of even the most enthusiastic
and best-intentioned agents to these problems, and
so on. All of these can lead to an unexpected lack of
funds for doing what was commanded (if only
implicitly), and hence disruption of commanded
performance, unless additional liquidity is provided.
Thus monetary policy in a command economy,
once money is allowed any room for activity, must
be accommodating; a lack of funds can never be
allowed to disrupt planned performance, just as an
excess of funds cannot be allowed to facilitate
unplanned or unauthorized activity. Thus the role,
the influence, of money has a natural, inexorable
tendency to grow: insufficient funds become an
immediate problem generating new money through
credits or additional allocations, while unused funds
tend to stay hidden until ferreted out by inspection
or accidental discovery. And as it grows, so does
the challenge to the command principle. An
increasing number of agents, in both the state and
non-state sectors, has a growing ability to access
resources, to divert them in the name, if not always
the interest, of implementing decentralized plans,
and thus to challenge the priorities of the political
authorities. This growing challenge becomes a
cancer in the system, a growth that undermines its
health and feeds tendencies destructive of the
priorities of the regime and its rulers.
The 'Second £000001/
As the command economy matures, as the messi-
anic fervour with which it was imposed wanes and
the use of extraordinary force diminishes in ensur-
ing compliance with commands, these challenges to
command metastasize into a competing yet
1838
symbiotically attached and dependent economic
system: the second economy. This name highlights
the distinction of this sphere of economic activity
from the officially sanctioned, ‘first’, command
economy. It is thus defined as ‘all production and
exchange activity that meets at least one of two
criteria: (a) being directly for private gain; (b) being
in some significant respect in knowing contraven-
tion of existing law’ (Grossman , p. 25).
In the Soviet Union, attempts to strengthen
‘material incentives’ and activate ‘the profit motive’
in order to increase the effectiveness and technical
efficiency of the implementation of central plans
and directives and to stimulate technological
progress and innovation, and the growing
monetization of the agricultural sector, opened the
door to massive expansion of money supply and
eroded the barriers between the currency and the
enterprise bank account monetary circuits. Col-
lective farms and their subsidiary enterprises,
owners of‘small means of transport’, vodka man-
ufacture and distribution, and the Caucasus repub-
lics (Georgia in particular) proved particularly rich
sources of illicit (from the system’s perspective)
monetization and private ‘entrepreneurial’ activity.
This both raised the spectre of inflation and opened
the door to vastly increased opportunities for
manipulation by self-interested subordinates in the
command sector. Thus the use of ‘economic levers’
greatly increased the opportunity for and incidence
of bribery, corruption, speculation, and even
‘honest’ private labour.
While the fundamental cause of the appearance
and growth of the ‘second economy’ undoubtedly
lies in the congenital institutional weaknesses of the
command economy discussed above, there are a
number of proximate sources that make it unsur-
prising. These include extensive price control, with
consequent scarcity and misallocation, high taxes
on non-state activities/incomes, prohibitions of
private activity, unmet individual consumption
needs, poorly protected impersonal (state) property,
the personal power of bureaucrats and ‘gate-
keepers’, and other historical factors, including the
end of terror. These provide both motives and
opportunities for officially illicit activity and for the
authorities to overlook that activity. With the ageing
of command and the decay in
Command Economy
enthusiasm of its agents, the growth of such a
second economy appears natural.
Thus growing ‘monetization’, the existence of
ready and waiting market sectors, and the decline in
the use of violent instmments of enforcement lead
to a growing sphere and importance of activities
outside the purview of ‘planning’ and ‘command’.
These market-mediated activities are at times
supportive, helping to achieve tolerable micro-
balance in the increasingly complex economy, but
often are in violation of planned implementation
and regime values. Private interests, necessarily
allowed some leeway, grow in significance,
increasingly seizing ground from command. In the
Soviet Union, the private agricultural sector,
initially permitted only to secure survival of the
peasantry under the extractive pressure of rapid
industrialization, and the consumers’ personal
services sectors provided the basis for a ubiquitous,
if still systemically marginal, second economy.
But then even the core industrial sectors under
the command mechanism find their managers and
activities increasingly influenced by this illegiti-
mate, shadow market, system, as managers are often
forced to break rules and undertake illegal acts in
order to do their job. Such acts, together with
ubiquitous and protean illegal activity on private
account, add up to a large underground economy
characteristic of every command economy, which
together with legal private activity (allowed in
varying degree in different countries) both supports
and supplements the ‘first economy’ and is inimical
to it While the second economy significantly adds
to the supply of goods and services, especially for
consumption, it also redistributes private income
and wealth, contributes to the widespread official
corruption, and generally criminalizes the
population. Virtually every area of economic life is
touched upon, and often entangled with, ‘second
economy’ activities, while legal private activity
naturally opens a loophole for illegal trading and
entrepreneurship, generally below the purview of
the authorities. And it goes hand in hand with the
extension of corruption, ensuring that it remains
outside of official notice.
Those ‘violations’ of legality within the com-
mand sector, a ‘shadow economy’, build informal
Command Economy
inter-enterprise relations which are generally ben-
eficial to the operation of state enterprises. They
work to substantially correct the allocative failures
of the command mechanism, improving firm
performance and hence benefiting its management,
and also provide lucrative opportunities for
managers to directly benefit through the activization
of barter, personal connections, and bribery.
However, they also spawn further distortions in
economic behaviour, as managers seek to generate
access to cash, the life blood of the ‘second
economy’, to extract rents, and to hide their activity
from supervisory and statistical organs.
Thus the second economy plays a dual and
contradictory role in the command economic sys-
tem. First, it addresses a number of the problems of
coordination and balance endemic to the command
mechanism, reallocating both producers’ and
consumers’ goods, facilitating plan fulfilment and
the use of financial incentives, and generating new
incomes and ‘politically safe’ outlets for private
initiative. Hence it becomes important for
enhancing consumer welfare, for production sta-
bility, and even for social stability. The ‘second
economy’, and in particular its ‘shadow’ side, plays
an essential role in the first economy as a ‘pressure
valve’, a release ‘fixing command’ by maintaining
micro-balance and covering ‘holes’ in economic life
left by the mistakes or oversight of the planners and
central managers. And this role becomes
increasingly important as the economy grows more
complex and diversified, and hence becomes less
susceptible to conscious oversight and direction.
As the central authorities struggle with their loss
of control, searching for a solution through reform,
decentralization and recentralization, monetization
and administrative restriction, agents in the
economy take advantage of gaps in control, of the
autonomy and discretion offered by growing
liquidity of the quasimoney in the system, to deal
with problems of coordination and balance,
inconsistency of plans and commands, and
ubiquitous shortages and scarcities. Of course they
operate in light of their own partial information, and
in their own (private as well as official) interests,
but in so doing save the system from collapsing
under its
1839
own weight and rigidity (Powell ). Thus the
second/shadow economy provides a spontaneous
surrogate economic reform that imparts a necessary
modicum of flexibility, adaptability and respon-
siveness to a formal set-up that is too often para-
lyzing in its rigidity, slowness, and inefficiency. In
doing so, the second economy also provides a
valuable stabilizing influence on society and the
polity, making life livable and the system humanly
manipulable and responsive to private inducement.
It makes everyone complicit in the way things
work, equally ‘guilty’ before state and society,
while providing an almost legitimate, and not polit-
ically dangerous or directly destructive, outlet for
individual initiative and entrepreneurship. Finally, it
relieves inflationary pressures (a ‘monetary over-
hang’) resulting from the command economy’s
necessary combination of monetary looseness and
pricing rigidity.
Despite this positive functional role, the second
economy also has a less positive systemic impact. It
mocks the pretense of social direction and control,
subverts its egalitarian impulse by accentuating
differences in access and income, and gives the lie
to the pretense of a ‘new’ ideologically correct
(‘Soviet’) man. Its very existence and usefulness
thus subvert the ideology of the regime, and it
works against and undercuts regime priorities by
exposing the incompetence and incapacity of the
authorities. Its provision of alternatives weakens the
‘plan, production, and labour discipline’ so essential
to the proper operation of the command mechanism.
Indeed, it attacks the core of the command
mechanism as it ‘... elevates the power of money in
society to rival that of the dictatorship itself,
rendering the regimes implements of rule less
effective and less certain’ (Grossman , p. 36). In
particular, it corrupts officialdom and distorts
prices, adding a (positive or negative) ‘second
economy margin’, both ‘in kind’ and in money,
breaking prices as an effective instrument of
control. This weakens monetized incentives for
state activities by providing competing, and often
better, alternatives to them. Hence the second
economy, and in particular the ‘shadow economy’
in the state sector, completes the cancerous
development of agent autonomy, of the ability to
work outside the plan and its
1840 Command Economy

subsequent commands, by providing viable alter-

according to their drafters’ intents, in mobilizing
natives to the plan. resources for rapid industrialization and modern-
Other dysfunctional impacts, undermining the
ization, at times on a vast scale, and in rapidly
operation of the command system, arise from its
amassing industrial power and military strength.
diverting of resources and products to unplanned
Indeed, they have shown themselves highly effec-
sectors and activities, including diversion from
tive in rapidly implementing large-scale projects
development, investment priorities to consumers.
and achieving overriding social goals, albeit at great
This naturally generates undesirable (from a system
cost. It is this effectiveness, when cost is no object,
perspective) redistribution of incomes, although
which explains why the command principle is
recipients, including many high-placed officials,
resorted to in times of emergency and war. Hence
find it very desirable. Indeed, it is further disruptive
in the Soviet Union command facilitated defence
of command by creating a 'two-tiered' system of
during, and rapid recovery and rebuilding of the
prices and incomes, of consumer goods and labour
Stalinist economy after, the massive trauma of the
markets. One tier is comprised of the low-priced,
Great Fatherland War. Economic growth has been
scarcity-ridden quasi-markets of the ‘less-than-
especially marked (though not unparalleled by
absolute’ command economy, where the market economies) where large amounts of
unenterprising, the overly scrupulous, and theunemployed and underemployed labour and rich
‘slow’ can survive. The other tier consists of real,
natural resources could be mobilized and combined
albeit highly distorted, markets in the generally
with existing (advanced. Western) technology, and
high-priced, risky but well-endowed second where the public's material improvement could be
economy where the enterprising, entrepreneurial,
restrained, or even seriously depressed, under
and criminal can thrive. In this high tier, substantial
strong political control. As these possibilities
incomes are generated and allocated, although they
waned, and as the economies grew in size and
largely accrue to corrupt officials and ‘gatekeepers’
complexity' and thus became less amenable to
of scarce materials or permissions who can extract
centralized administrative management, rates of
rather phenomenal ‘rents’. The inequities thisgrowth declined sharply. At the same time, the
generates further undermine the legitimacy of the
shortcomings of the command mechanism in
regime and generate potentially explosive social
adapting production to demand and its changes -
pressures, only partially relieved by the second
providing consumer welfare, effecting innovation,
economy’s ‘pressure valve’ aspects. serving export markets - became more apparent and
Finally, it is worth noting that the second
less tolerable. This led to much discussion and
shadow economy, through its activity outside of the
repeated attempts at controlled institutional reform,
officially measured sphere, seriously distortsat decentralizing and stimulating subordinate
statistical data and the information available to
initiative without sacrificing ultimate control.
planners and allocators in the official economy, and,
Some actual reforms in the externally imposed
due to its illegality, also hides necessary infor-
command economies of eastern Europe went so far
mation from other agents in the shadow economy.
as to introduce or extend the market mechanism to
This aggravates the economic problems that spawn
such a degree that one could no longer regard the
‘second/shadow economy' activities, deepening the
system as a Soviet-type command economy, even
contradictions between the centre and decentralized
if, before the 1990s, one could not speak of it as a
agents, and further corroding the institutional
full-fledged market economy either. Yugoslavia
structures of the command economy. since the early 1950s, Hungary since 1968 and
especially in the 1980s, and post- Mao China are
the most important cases in point. Other acUtal
Performance and Fate reforms were of a minor or ‘within- system' naUtre,
aiming to decentralize certain types of decisions
Command economies have been instrumental in
while eschewing the market
radically transforming societies more or less
Command Economy
mechanism and retaining the hierarchical form of
organization and the command principle. In the
hope of stimulating efficiency to revive growth
rates, the decentralizing measures were accompa-
nied by a number of other ‘reforms’ relating to
organizational structure: prices (still controlled),
incentives, indicators, materials rationing, and so
on. The Soviet reforms of 1965, and those in the
1970s and 1980s prior to perestroika, were of that
kind; many similar ones took place in other Com-
munist countries after the mid-1950s and prior to
the overthrow of Communism in 1989. On the
whole, such reforms had little success in addressing
the problems of the command economy.
Bureaucratic and political obstacles apart, the
attempt to decentralize economic decisions without
bringing in a market mechanism almost inevitably
leads to economic difficulties. The beneficiaries of
devolution of decision-making lack the necessary
information to produce just what the economy
requires or to invest to meet prospective needs, and
the coordination of plan-subsequent command is
lost. Moreover, they may apply the additional
power at their disposal to advance par- ticularist
causes or to divert resources into illegal channels.
Microeconomic disequilibria mount, and soon
superior authorities step in to recentralize on a case-
by-case basis and the reform withers away
(Grossman ; Wiles , ch. 7; Kontorovich ).
This failure of reform reflects the inherent con-
tradictions of the command economy framed in the
irreconcilable conflict between ‘command’ and
‘money’ discussed above (Ericson ). The Soviet
command economy, driven by the urgent need for
and haste in industrialization and military
development, initially relegated the influence of
money and the market to the margins of the system,
where they handled areas and activities in which
command had been revealed as counterproductive
during War Communism. That system, the ‘less
than absolute command economy’, substantially
industrialized, triumphed in the Great Fatherland
War, and recovered to an almost perfect replica of
its pre-war self by 1950. But by then the strains of
its inherent inflexibility and the bounded rationality
of the system’s planners and managers began telling
on continuing
1841
growth and the development of the economy. With
economic growth came increasing complexity and
growing intractability of the central planning and
economic management problem. Some
decentralization became essential, and increasingly
so as time passed, opening the door once again to
the rise of money as a significant influence on the
operation and development of the economy. And
that influence was only enhanced by the ageing and
mellowing of the system. With the passing of
‘terror’ as an effective incentive mechanism, the
stabilization of personnel and the regularization of
procedures, it became ever harder to control agent
behaviour, to contain the distractions of money and
the self-interests it mobilized, and to uncover the
rents that well- placed agents were able to extract,
thus aggravating the inherent agency problems of
the command economy.
The remaining years of the Soviet system thus
witnessed an epic stmggle, barely perceptible at
first, but increasingly evident as reforms, decen-
tralizations, reorganizations and recentralizations
cycled around each other in the search for a solution
to the increasingly evident and destmctive
malperformance and waste, and aggravating
behavioural distortions in response thereto, gen-
erated by the struggle between the ‘command
principle’ and the weak, but inexorably emerging,
‘market’. Initially reflected in the dysfunctions of
the marginal and quasi-markets of the command
economy, and in the struggle to harness a ‘passive’
money to the purposes of command, the role of
money grew along the ‘treadmill of reforms’ into
the rival, if still largely subordinate and com-
plementary, ‘second economy’, and in particular its
‘shadow’ component, on which the ‘command
principle’ increasingly came to depend for its
effectiveness. As long as the Soviet system
remained a ‘command economy’, commands had to
have last word, and money remained largely
relegated to the sidelines, exercising its influence
within the quasi-monetized instruments (‘economic
levers’) of the command mechanism and the
distorted markets of the second economy.
This inherent conflict, played out over Soviet
history, revolves around a number of fundamental
1842
dualities, elemental oppositions which characterize
these primary forces. The ‘command principle’
derives most basically from the urge, the will to
control, to ‘rationally’ determine and direct the
future, exercised by a ‘gnostic’ elite, immanent in
the Party. It knows what needs to be done, by whom
and how, and can tolerate no dissent or deviation.
Juxtaposed to this ‘Will of Society’ stand the
millions of independent ‘wills’, desires and
objectives, anarchically coordinated through ‘the
market’, whenever that set of institutions broke
through the barriers and limits placed by
‘command’. This provides the foundation for the
eternal struggle between ‘central priorities and
control’ and ‘agent incentives and capabilities’.
This opposition is severely aggravated by
urgency, by ‘virtuous haste’, in the pursuit of
overriding social goals and central objectives. For
the mobilization for, and focus of resources on,
these priorities trample on the information,
capabilities and goals of individual and organi-
zational agents which must perforce implement that
mobilization, implement those priorities.
‘Effectiveness’ in the pursuit of social objectives
becomes opposed to ‘efficiency’ in the attainment
of any objectives, denies trade-offs based on local
information and incentives, and hence blocks
flexibility in response to changing circumstances.
Indeed, the single-minded pursuit of overriding
objectives, of absolute priorities, naturally disrupts
the fine coordination, the requirements of ‘balance’,
necessary to consistently and efficiently pursue any
objectives.
Throughout the history of the Soviet Union, the
needs of centralization, given Soviet social goals,
stood in fateful opposition to the necessity to
decentralize in order to keep the system tolerably
functioning. The latter necessity spawned repeated
(partial) remonetizations and a ‘second economy’
that both shored up the operational foundations of
the ‘first economy’ and undermined its long-term
viability, corroding its ideological and systemic
foundations. Money so unleashed intensified the
dysfunctions and contradictions of the ‘command
economy’, spurring further repeated ‘reforms’ and
‘experiments’ that merely further aggravated the
inconsistencies,
Command Economy
the ‘oppositions’ in the system, until the central
leadership, largely unintentionally and out of
ignorance, destroyed the ‘command economy’ in
the radical systemic and economic ‘restructurings’
beginning with perestroika in 1987.
See Also
► one; of •' .
► • ..
► Informal Economy
► Second Economy (Unofficial Economy)
► ,? t Bud Constraint
► Soviet Economic Reform
► ,? idet G. o C i d
Bibliography
Aslund, A. 1995. How Russia became a market economy.
Washington, DC: Brookings Institution.
Berliner, J. 1957. Factory and manager in the USSR.
Cambridge, MA: Harvard University Press.
Bomstein, M., ed. 1973. P l a n a n d m a r k e t . New Haven:
Yale University Press.
Eckstein, A., ed. 1971. Comparison of economic systems:
Theoretical and methodological approaches. Berkeley:
University of California Press.
Ericson, R. 1983. A difficulty with the ‘command’ allocation
mechanism. J o u r n a l o f E c o n o m i c T h e o r y 31:
1-26.
Ericson, R. 1991. The classical Soviet-type economy: Nature
of the system and implications for reform. J o u r nal o f
E c o n o m i c P e r s p e c t i v e s 5 (4): 11-27.
Ericson, R. 2005. Command vs. shadow: The conflicted soul of
the Soviet economy. C o m p a r a t i v e E c o n o m i c
S y s t e m s 47 (4): 1-27.
Freris, A. 1984. The Soviet industrial enterprise.
New York: St Martin’s Press.
Garvy, G. 1977. M o n e y , f i n a n c i a l f l o w s a n d
c r e d i t i n t h e Soviet U n i o n . New York: Ballinger
Publishing
Company.
Granick, D. 1954. M a n a g e m e n t o f t h e i n d u s t r i a l
f i r m i n t h e USSR. New York: Columbia University
Press. Grossman, G. 1962. The structure and organization of
the Soviet economy. Slavic Review 21: 203-222. Grossman, G.
1963. Notes for a theory of the command economy. Soviet
Studies 15 (2): 101-123.
Grossman, G. 1966. Gold and the sword: Money in the Soviet
command economy. In Industrialization in two systems, ed. H.
Rosovsky. New York: Wiley. Grossman, G. 1977. The ‘second
economy’ of the USSR. Problems of Communism 26 (5): 25-
40.
Commodity Fetishism
Grossman, G. 2000. Central planning and transition in the
American desert: Latter-day saints in present day sight
Available as a preprocessed paper at http://
. Accessed
23 June 2006.
Higgs, R. 1992. Wartime prosperity? A reassessment of the
U.S. economy in the 1940s. J o u r n a l o f E c o n o m i c
H i s t o r y 52: 41-60.
Kontorovich, V. 1988. Lessons of the 1965 Soviet economic
reform. S o v i e t S t u d i e s 40: 308-316.
La Lone, M., and D. La Lone. 1987. The Inka state in the
southern highlands: State administration and production
enclaves. E t h n o h i s t o r y 34: 47-62.
Nove, A. 1977. T h e S o v i e t e c o n o m i c s y s t e m .
London: Allen and Unwin.
Powell, R. 1977. Plan execution and the workability of soviet
planning. Journal of Comparative
E c o n o m i c s 1: 57-76.
Raupach, H. 1966. Zur Entstehung des Begriffes
Zentralverwaltungs-wirtschaft Jahrbuch
S o z i a l w i s s e n s c h a f t 17 (1): 86-101.
Schroeder, G. 1979. The Soviet economy on a treadmill of
reforms. In S o v i e t e c o n o m y i n a t i m e o f
c h a n g e , ed. Joint Economic Committee, US Congress,
vol. 1. Washington, DC: Government Printing Office.
Wiles, P. 1962. T h e political economy
c o m m u n i s m . Cambridge, MA: Harvard University
Press.
Zaleski, E. 1968. S t a l i n i s t p l a n n i n g f o r e c o n o m i c
d e v e l o p m e n t . Durham: Duke University Press.
Commodity Fetishism
Andrew Levine
Abstract
An analysis of Marx’s notion of ‘commodity
fetishism’ - as a theory of the necessary (system
ieally induced) misperception of underlying
production relations by participants in market
exchanges. The appeal of the notion to the two
main opposing tendencies of mid- and late 20th-
century Marxism - Marxist humanism and
structuralist Marxism - is discussed. Reasons are
proposed to account for a recent decline of
interest in the phenomenon among both
economists and philosophers. It is suggested,
however, that the concept remains viable.
1843
Keywords
Althusser, L.; Capitalism; Capitalist social
relations; Commodity fetishism; Exploitation;
Feuerbach, L.; Invisible hand; Labour power;
Labour theory of value; Laws of motion of
capitalism; Lukacs, G.; Market relations; Marx,
K.; Marxist humanism; Structural marxism;
Surplus value; Value controversy
JEL Classifications
B1
Since Plato, philosophy and then science have
assumed first, that there is often a difference
fur
between appearance and reality; and, then, that it is
sometimes possible to grasp what really is the case
by investigating how things appear. Marx’s account
of commodity fetishism, a crucial step in his
of account of the capitalist mode of production,
implements these assumptions explicitly. It
describes how exchange relations appear to eco-
nomic agents, where the appearance belies the
reality at the same time that it provides cognitive
access to it.
Market exchanges occur in all modes of pro-
duction capable of sustaining an economic surplus.
In capitalism, the process is generalized - not just in
the sense that markets structure economic life but
also, more importantly, because everything is
commodified that can be. Universal
commodification is the result of a protracted pro-
cess that is definitively launched once labour - or,
more precisely, labour power (labour time, adjusted
for differences in intensity) - is commodified. The
commodification of labour power is pivotal because
this commodity is the sole source of value and
therefore, ultimately, of wages, profits and rents.
The generation and distribution of surplus value, of
what is produced in excess of what is needed to
reproduce the labour power expended in production
processes, is the invisible underlying reality upon
which perceptions of exchange relations depend. To
persons engaged in buying and selling labour
power, what appears is just that, as in any other
exchange, individuals aim to do as well for
themselves as they can, given
1844
their resources, their preferences, and the produc-
tion technologies available to them. But what is
really going on is a struggle over the distribution of
the economic surplus at the point ofproduction. That
reality is opaque. Economic agents are therefore
governed by the appearance of rational economic
agents maximizing payoffs to themselves. In his
account of commodity fetishism, Marx shows how
this inevitable misperception helps to reproduce and
sustain the underlying reality'.
When Marx expressly addresses this phenom-
enon at the conclusion of the opening chapter of the
first volume of Capital (1867), the economic agents
he describes are property-holding individuals. Thus
it is not exactly capitalism that he aims to model,
but 'simple commodity' production', an ahistorical
idealization. However, the cogency of his account is
unaffected as his analysis becomes more historical
and concrete - to the point that the direct producers
are, as in fiill-fledged capitalism, a property less
proletariat with nothing to exchange except, of
course, their' own labour power. Commodity
fetishism is therefore a general and pervasive fact
wherever capitalist social relations hold sway. Thus
the term denotes a systemic opacity at the level of
appearance that helps to hold economic agents in
thrall by masking the exploitation of labour.
Because this misperception sustains the exploitation
that engenders it, revolutionaries intent on
overthrowing capitalism must tear away the veil of
illusion by revealing the exploitation of workers
that exchange relations conceal.
Marx does not directly address how commodity
fetishism comes into being or how it is sustained.
But he does provide fragments of an explanation
when he focuses on the atomizing effects of market
relations. All resource allocation mechanisms are
social in the sense that they bring together a host of
disparate and heterogeneous economic activities.
However, where the commodity form prevails, the
social character of market transactions is apparent
only after goods and services are produced. The
workers know that the corn they consume is
produced by farmers, and the farmers know that the
tools they use in growing com are made by workers.
Everyone also knows that, without food, workers
would not be able to make the tools farmers use;
and that,
Commodity Fetishism
without tools, farmers would not be able to grow
food for the workers. It is therefore evident in
retrospect that workers and farmers are engaged in a
collective endeavour. But it is not similarly evident
prospectively. From that vantage point, it seems
only that farmers and workers - and also the
capitalists who provide them with means of
production - are making individual choices aimed at
bringing about the best outcomes for themselves,
given the constraints they face. Even if they believe
that these essentially egoistic activities are somehow
socially beneficial, they can justify this belief by
appealing to the workings of an 'invisible hand'.
Because there is no visible hand that directs the
process, the teims of interaction appear as // they are
forces of naUire to which individuals must
accommodate. Thus market relations appear as
infrangible constraints that human beings are
obliged to operate within, not as social constructions
that human beings can change. In teims that Kant
introduced and that Marx, following Hegel,
effectively assumed, freedom (autonomy) is then
forfeit. Wills are hetero- nomously determined,
governed by laws of an (apparently) impersonal
other (the market system itself). To be free, we must
therefore take control of the aggregation mechanism
wc have concocted. We do so by putting reason in
command - not just at the individual level of the
rational economic agent, but at the societal level as
well.
What Marx says about commodity fetishism is
concise and intriguing. For these reasons, and
because it summarizes the veiy abstract analysis of
the commodity form with which Capital begins, his
account of the phenomenon has always been well
known. ‘Commodity fetishism’ is one of those
terms that everyone associates with Marx. But, even
in what remains of Marxist circles, the basic tenets
of Marx’s account have faded from ongoing
discussions. A number of factors have contributed
to this turn of events: among them, the legacy of the
so-called ‘value controversy’ of the 1970s; the
efforts of mathematical economists in the 1970s and
1980s to put the categories of Marxist economic
analysis on a sound, analytical footing; and attempts
by analytical philosophers, working on Marxist
themes, to reconstruct and, when possible, defend
core Marxist positions. The
Commodity Fetishism
conclusion that has emerged is that, pace Marx,
there is nothing special about the commodification
of labour power and therefore that the theory of
surplus value cannot be sustained in the way that
Marx believed. Nowadays, it is only the most
doctrinaire Marxists who uphold the labour theory
of value, the basis for Marx’s account of commod-
ity fetishism. This fact along with the decline of
political movements that identify with the Marxist
tradition and, its inevitable consequence, waning
interest in Marx’s work itself, has, for the time
being, made commodity fetishism a matter of
concern mainly to historians of economic thought.
Not long ago, the situation was quite the oppo-
site. From roughly the 1950s through the 1970s,
commodity fetishism played a central role in the
two most important and innovative tendencies in
Marxist theory: Marxist humanism and structuralist
Marxism. These were opposing tendencies,
politically and substantively. But they converged on
according commodity fetishism centre stage.
Marxist humanists sought to de-Stalinize
Marxism by recovering its Left Hegelian roots. This
meant reading Marx’s work through the prism of his
early writings, before he broke with his ‘erstwhile
philosophical conscience’, as he proclaimed in 1845
in The German Ideology. For the Left Hegelians,
Ludwig Feuerbach’s philosophical anthropology,
elaborated in The Essence of Christianity (1841),
was fundamental. There Feuerbach ‘inverted’ the
theological dogma that ‘God makes Man’ by
showing how the God idea is an ‘objectification’ of
essential human traits. Lacking materiality, God is
purely an objectification, an ‘alienated’ expression
of the human essence. In taking consciousness of
this fact, one recovers essential humanity and
becomes emancipated from the thrall of its systemic
misrepresentation. In the Paris Manuscripts (1844),
Marx applied the Feuerbachian programme to
objects of labour; ‘objectifications’ too of essential
humanity, but also material things and therefore not
objectifications only. Feuerbach arrived at his
conclusions by ‘interpreting’ the theology of Right
Hegelian theologians. His working hypothesis was
that they had gotten the concept of God right, but
that they radically misconstrued what the concept
means. In the Paris Manuscripts
1845
Marx treated (Smithian) political economy the same
way. He assumed that it correctly describes
‘economic facts’. The task, then, was to interpret
those facts - in order to reveal the alienation they
express and, in so doing, to advance the emanci-
patory project of Left Hegelianism. How successful
Marx was in implementing this programme is
subject to debate. What is clear is that, as the focus
of his theoretical work turned away from Hegelian
philosophy towards political economy, history and
politics, he became disabused of the idea that Adam
Smith or any other classical economist had gotten
political economy descriptively right. His life’s
project, thereafter, was to rework the conceptual
apparatus of classical economics - more usually in
its Ricardian, not Smithian, form - with a view to
revealing the real ‘laws of motion’ of the capitalist
mode of production. In this endeavour,
Feuerbachian philosophical anthropology seemed to
play no role. But, following the lead of Georg
Lukacs ( ) several
decades earlier, the Marxist humanists pointed out
that there was, in Capital, an explicit point of
connection - in the text on commodity fetishism. It
was there that Marx brought his analysis of the
commodity form to completion. But it was also
there that, in modelling the commodity form, Marx
identified the objectification of essential human
traits in the process of capital accumulation. In
consequence, capital, becomes a ‘fetish’, a god in
Feuerbach’s sense - one who controls economic
behaviour by force of (illusory) power.
Structuralist Marxists, like Louis Althusser,
were intent on reading Left Hegelianism out of the
Marxist canon. They therefore treated Marx’s
references to fetishes and gods as ironic figures of
speech, even as they attempted to enlist the text on
commodity fetishism in the service of opposition to
Marxist humanism. Borrowing a concept from the
French philosopher Gaston Bachelard (1884-1962),
Althusser ( ) dispar
aged Marx’s early work by asserting the existence
of an ‘epistemological break’ within the Marxist
corpus. What he had in mind was roughly a
‘paradigm shift’ - not, however, within an ongoing
scientific practice but between pre-scientific modes
of thought and the inception of a new science. In
Althusser’s account, two
1846
previously monumental epistemological breaks had
occurred - one that established mathematics in
ancient Greece, and one that established the
sciences of nature in 17th-century Europe. Marx’s
achievement was supposedly on a par with these; he
opened up a science of history. He did so by
anticipating the structuralist turn the ‘human
sciences’ (in France mainly) would later take - first
in linguistics and psychology, later in anthropology
and psychoanalysis. Specifically, in Capital and
other writings of his maturity, Marx explained a
range of diverse ‘surface’ phenomena by construing
them as effects of the workings of a relatively small
number of underlying, generally invariant ‘deep’
structures. The text on commodity fetishism lent
itself to this construal of Marx’s explanatory
practice in as much as it depicted the perceptions of
economic agents as effects of the unseen but
causally efficacious process of surplus value
extraction. Thus Marx’s account can be seen as a
theory of necessary (systemically induced)
misperception - consonant with notions of expla-
nation that contemporaneous structuralists
endorsed. Perhaps the most innovative use
Althusser made of commodity fetishism was in his
theory of ideology, according to which modes of
production constitute experiential subjectivity by
‘interpellating’ the human subjects who support or
‘bear’ them.
We now inhabit a different intellectual universe.
In the past several decades, it has come to be widely
believed, by erstwhile Marxists as much as by
‘bourgeois economists’, that Marx’s focus on
production rather than exchange inhibited the
development of analytical economic tools. In so far
as this belief is sound, the emphasis Marxists
placed on commodity fetishism is partly to blame.
The explanatory strategies of Marxist humanists
and of structuralists have fallen into disrepute, too -
largely because, in both cases, though for different
reasons, the alleged connections between
appearance and reality were never satisfactorily
explained. No sustainable account was given either
of how interpretation should proceed in the Marxist
humanist case or, in the structuralist case, of how
deep structures can be discerned in surface
phenomena. Thus, commodity fetishism has fallen
on hard times. However,
Commodity Fetishism
we should not conclude that there is nothing viable
in the concept or in the theoretical traditions that,
until recently, magnified its importance.
Hegelianism certainly, and structuralism possibly,
still have much to teach us. The last word may not
yet have been said on the theory of surplus value,
either. If and when interest in Marx resumes, it will
certainly be useful to revisit these issues. The notion
of commodity fetishism played a key role in mid-
and late 20th-century Marxism. The core idea it
articulates - that necessary misperceptions sustain
the capitalist order - can again provide useful
insights. The concept may not be forever doomed to
be of historical interest only.
See Also
►
► Marx’s Analysis of Capitalist Production
Bibliography
Althusser, L. 1965. F o r M a r x , 1969. New York: Pantheon
Press.
Althusser, L. 2002. Ideology and ideological state apparatus.
In L e n i n and philosophy and other
e s s a y s . New York: Monthly Review Press.
Althusser, L., and E. Baliber. 1968. R e a d i n g c a p i t a l .
London: New Left Books, 1970.
Cohen, G. 1988. The labour theory of value and the concept of
exploitation. In H i s t o r y , l a b o u r a n d f r e e d o m :
T h e m e s f r o m M a r x . Oxford: Clarendon Press.
Feuerbach, L. 1841. T h e e s s e n c e o f C h r i s t i a n i t y .
Amherst: Prometheus, 1989.
Fromm, E. 1975. M a r x ’ s c o n c e p t o f m a n . New York:
Continuum.
Lukacs, G. 1923. History and class
c o n s c i o u s n e s s . London: Merlin Press, 1967.
Marx, K. 1844. Paris manuscripts (also called economic and
philosophical manuscripts). In C o l l e c t e d w o r k s , ed.
Karl Marx and Frederick Engels, vol. 3. New York:
International Publishers, 1975.
Marx, K 1845. The German ideology. In C o l l e c t e d
w o r k s , ed. Karl Marx and Frederick Engels, vol. 5.
New York: International Publishers, 1976.
Marx, K 1867. C a p i t a l , vol. 1. Moscow: Progress Pub-
lishers, 1965.
Roemer, J. 1981. A n a l y t i c a l f o u n d a t i o n s o f
marxist economic theory. New York:
Cambridge University Press.
Steedman, I. 1981. M a r x a f t e r s r a f f a , rev. edn. New
York: Shocken.
Commodity Money
Commodity Money
Francois R. Velde and Warren E. Weber
Abstract
Commodity money is a medium of exchange
that may be transformed into a commodity,
useful in production or consumption. Although
commodity money is a thing of the past, it was
the predominant medium of exchange for more
than two millennia. Operating under a
commodity money standard limits the scope for
monetary policy, actions that alter the value of
money. However, it does not eliminate monetary
policy entirely. The value of money can be
altered by changing the commodity content or
legal tender quality of monetary objects, or by
restricting the conversion of commodities into
money or vice versa.
Keywords
Banking; Bank of England; Bimetallism;
Brassage; Bretton woods system; Coinage;
Commodity money; Convertibility; Debase-
ment; Discount rate; Fiat money; Foreign
exchange markets; Gold standard; Great
depression; Inflation; Monetary policy; Money
changers; Seigniorage
JEL Classifications
E4
A commodity is an object that is intrinsically useful
as an input to production or consumption. A
medium of exchange is an object that is generally
accepted as final payment during or after an
exchange transaction, even though the agent
accepting it (the seller) does not necessarily con-
sume the object or any service flow from it. Money
is the collection of objects that are used as media of
exchange. Commodity money is a medium of
exchange that may become (or be transformed into)
a commodity, useful in
1847
production or consumption. This is in contrast to
fiat money, which is intrinsically useless.
Commodity money can also be thought of as a
medium of exchange that contains an option to
consume a predetermined service flow at little or no
cost. The option can be exercised in various ways,
depending on the object. Coins can be melted down
(at little cost) and the metal applied to non-
monetary uses. In the case of paper or token money
under a commodity money standard, the medium of
exchange itself is intrinsically useless, but it is
costlessly convertible into a specified quantity of
the commodity on demand. Fiat money can also be
converted into goods or services, but in quantities
that will depend on market prices.
Commodity money is a thing of the past; coun-
tries worldwide now use fiat money standards.
However, this is a relatively recent development.
Commodity money, primarily in the form of coined
metals, was the predominant medium of exchange
for over two millennia. Although operating under a
commodity money standard limits the scope for
monetary policy, it does not eliminate it entirely.
The history of commodity money is replete with
numerous ways in which governments have altered
the monetary system to achieve various goals.
From Commodity Money to Fiat Money
In early or primitive societies, it is often difficult to
characterize the general patterns of trades and
transactions, let alone determine how generally
accepted a particular commodity might be.
Nevertheless, a wide range of commodities have
been reportedly used as money (cowry shells,
wampum, salt, furs, cocoa beans, cigarettes and so
on), perhaps the most exotic being the stone money
of the island of Yap in Micronesia.
General acceptability of monetary objects is
most clearly ascertained when the objects are
standardized and exchanged repeatedly. With
metallic commodities, the standardized objects are
called coins. Coinage of metal began in the eastern
Mediterranean region or the Middle East, India and
China between the 6th and 4th centuries
1848
BC. Coinage has developed in parallel and broadly
similar ways in these areas.
The metals most commonly used have been
gold, silver and copper (in decreasing order of
scarcity), in varying degrees of fineness (silver
mixed with substantial amounts of copper, called
billon). Lead, tin and various copper alloys (bronze,
brass, potin) have also been used, although less
frequently than the more common metals. The metal
is either mined or acquired through trade. The most
common method of coinage is striking with a die,
although cast coins are also found. In many legal
traditions the right of coinage is a prerogative of the
public or central authority, although it may be
delegated or leased to regional authorities or private
parties. This prerogative may also extend to mining.
In other words, the rales governing the supply of
commodity money vary from government
monopoly to minimal regulation.
In Europe and the Mediterranean, coinage - an
invention mythically linked to Croesus, King of
Lydia - began near the Aegean Sea in the 6th
century bc. The use of money developed consid-
erably in Greek and Roman times, leading to a
three-tiered system of gold, silver, and copper
denominations. In the Roman empire, the provision
of coinage was a government monopoly. The
collapse of the empire in the West led, after a long
transition, to a purely silver-based monetary system,
with a largely decentralized provision of minting.
Uniformity of coinage was restored under
Charlemagne but quickly disappeared along with
political fragmentation. Gold returned in common
use from the mid- 13 th century. By the 14th
century, most mints in western Europe operated
along similar lines, with more or less unrestricted
coinage on demand provided by profit-making
mints. A great multiplicity of monetary systems
persisted, giving rise to both foreign exchange
markets (the earliest financial markets) and money
changers (the first financial intermediaries).
The first instances of token coinage (coins that
are intrinsically useless but are claims to fixed
amounts of the commodity) appeared in the 15th
century in Catalonia. Notes convertible on demand
appeared in the 17th century, in Sweden
Commodity Money
and later in England. For a more complete discus-
sion of medieval European coinage, see Spufford (
).
Coins appear to have been used in India in the
early 4th century bc and were probably used before
then. The earliest coins were so-called punch-
marked coins and were adaptations of Greek
prototypes. Coins were first used in China and the
Far East about the same time as in India. The
distinctive bronze coinage with the square hole in
the middle first appeared in the 3rd century BC.
Early coins in eastern Islamic lands were copies of
Byzantine gold and bronze coins; those in the East
were copies of Sassanian silver coins. For more on
coinage in India and the Far East, see Williams (
).
Until the 19th century, coins typically bore no
indication of face value, and their market value
could fluctuate even relative to one another. From
the late Middle Ages, governments increasingly
sought to regulate the value of coins in some
manner, in particular assigning face value or legal
tender value by decree. It became desirable to turn
the collection of objects used as a medium of
exchange into a stable system with fixed exchange
rates between the objects. This was achieved to a
large degree with bimetallism, a system in which
gold and silver coins remained concurrently in
circulation at a constant relative price. Its heyday
was the mid-19th century, but beginning in 1873 the
system was quickly abandoned, and by the First
World War countries were using either gold only or
(in Africa and eastern Asia) silver only.
(Bimetallism is discussed in more detail in Redish ,
and Velde and Weber .) The development of
banking in the 19th century also led to increased use
of (convertible) notes and other monetary
instruments.
The First World War brought about the suspen-
sion of convertibility of the notes in many countries.
Most countries returned to convertibility between
1926 and 1931, but the onset of the Great
Depression reversed the movement. After the
Second World War the only major country whose
currency was in any way directly tied to a
commodity was the United States under the Bretton
Woods system: dollars were convertible
Commodity Money
by non-residents of the United States into gold on
demand, while other currencies of the system were
convertible into dollars. The link between gold and
the dollar was severed in 1971. Fiat money
standards are now universal.
The Nature of Commodity Money
The definitions of commodity and fiat monies given
above make it seem as if there is a clear distinction
between the two. It is more helpful, however, to
think of media of exchange along a continuum. An
object serving a purpose as a medium of exchange
has value above its intrinsic content, reflecting the
value of the service as a medium of exchange.
Because the value of a commodity qua com-
modity and the value as a medium of exchange can
differ, the value of all commodity monies has a fiat
component. A pure fiat money i s one for which this
fiat component makes up its entire value. A nice
theoretical discussion of commodity and fiat monies
is given by Sargent and Wallace ( ).
Price-Level Determination
It is natural that the medium of exchange in an
economy is what becomes the unit of account, the
unit in which debt contracts and the prices of goods
and services are expressed. It is natural because the
money appears on one side of virtually every
transaction.
Because commodity money has an intrinsic
value apart from that which it obtains by being a
medium of exchange, its relative price will not be
zero. Thus, in a commodity money economy, the
value of money (the inverse of the price level) is
bounded away from zero. Moreover, in a canonical
commodity money system (see below) with
unlimited minting at a set price, the value of money
and its quantity tend to remain within a band. If the
value of money falls far enough, it becomes
preferable to exercise the option and convert some
of it into other, non-monetary uses, thus reducing
the quantity and preventing the value from falling
further. Conversely, if the
1849
value of money rises high enough, it becomes
worthwhile for agents to turn metal into coins at the
mint at the set price, thus increasing the quantity of
money. Such a self-regulating commodity money
system provides an anchor to the price level. This
has been touted as one of the advantages of a
commodity money system, particularly in the case
of the gold standard.
The question of price-level determination
becomes more complicated when multiple com-
modity monies are made out of different commod-
ities. An example is the circulation of full-bodied
gold and silver coins. Should the unit of account be
the gold coin or the silver coin? This matters
because under a commodity money system a
monetary authority does not have the ability to set
the exchange rate between monies of different
commodities forever. Thus, to the extent that the
unit of account is used in contracts to determine the
amount of future payments, the choice of the unit of
account can affect the allocation of goods and
services. This was one of the issues surrounding the
possible adoption of a bimetallic standard
mentioned above.
The inability of the monetary authority to set the
exchange rate between different monies goes away
under a pure fiat money system. Because fiat money
is (virtually) costless to produce, the monetary
authority can costlessly exchange one money for
another to maintain whatever exchange rate is
desired between different monies that it issues.
Monetary Policy
The fact that a commodity is used as money alters
its value. This is because part of the total quantity of
the commodity - namely, the metal locked up in the
form of coins, or the reserves held by the monetary
authority - is not available for nonmonetary uses.
The allocation between monetary and non-monetary
uses is determined in equilibrium. Restrictions on
the ability to change this allocation, such as
restrictions on melting or exporting coins, or
limitations on the minting of metal, will have an
effect on the equilibrium value of the money even if
it has no immediate effect on
1850
the allocation itself. (Since money is an asset, its
valuation is forward looking.) Thus, there is scope
for monetary policy under a commodity money
standard, although what constitutes monetary policy
is different from and more limited in scope than
what holds under a pure fiat money standard.
Monetary policy consists in actions that tend to
alter the value of money. In a commodity money
system, the value of money is the value of the
option we have described. (The strike price of the
option is zero, since the commodity is the money.)
Most aspects of monetary policy with commodity
money consist in modifying this option, typically by
modifying the institutions governing the exercise of
the option rather than by modifying the quantity of
money, which the authority usually cannot control
directly. When the monetary authority is directly
involved in the provision of the money, it may
directly profit from its actions. Potential profit is
often an important consideration of monetary
policy.
The canonical form of a commodity money
standard comprises the following. One or more
commodities are chosen to be the standard to which
the monetary system will be anchored. The
monetary authority defines the specifications of the
monetary objects (weight, fineness) and defines the
unit of account in terms of these monetary objects.
The conversion of commodity into commodity
money and vice versa is as costless as possible. In
particular, the monetary authority provides for
unlimited (and even costless) conversion of the
commodity into monetary objects (coins or notes).
Conversely, it places no hindrances on the
conversion of monetary objects into commodities
(coins can be melted, notes are convertible on
demand), nor does it place limitations on the con-
sumption of the commodity or its service flow (free
possession, unrestricted import and export of the
commodity). The monetary objects are unlimited
legal tender.
One type of monetary policy modifies the spec-
ifications of monetary objects and units of account.
An example is debasement, which is reducing the
commodity content of a monetary object (and,
frequently, of the corresponding unit of account).
The result of debasement is inflation, since nominal
prices will be adjusted to maintain
Commodity Money
the relative prices of goods and money. And, just as
occurs with fiat money, inflation has the effect of
transferring wealth from nominal creditors to
nominal debtors. Since governments generally
tended to be debtors, debasements were used to
reduce the amount of their debts. Historically,
debasements also had the secondary effect of
increasing seigniorage revenue, since the quantity
of coins minted tended to increase significantly
after debasements that involved the introduction of
new coins (see Rolnick et al. ; Sargent and Smith ).
Debasements were also used by governments to
remedy malfunctions of a multiple-denomination
commodity money system (see Sargent and Velde ).
A second type of monetary policy adds or
modifies restrictions on the conversion of com-
modity into money or money into commodity. For
example, minting might be restricted by quantity, in
which case the authority decides how much to mint.
Minting might be unlimited but subject to a fee,
called seigniorage. Governments typically charged
such a fee, both to cover the actual costs of minting
(called brassage) and as a tax (England was the first,
in 1666, to provide minting at no cost). The rate of
this tax or, equivalently, the price paid by the mint
for bullion might be changed. These restrictions
tended to alter the allocation of the commodity
between monetary and nonmonetary uses, and
hence the value of the commodity and the money.
A third type of monetary policy sets limits to the
legal tender quality of certain coins, or changes
their legal tender value. Since coins did not have
face values until the 19th century, it was up to
monetary authorities to set, and from time to time
alter, the legal tender values of coins. Frequently,
foreign coins were authorized as legal tender at
rates set for domestic coins. Countries attempting to
maintain bimetallism in the face of fluctuations in
the relative price of gold and silver often had to
adjust the face value of either their gold or silver
coins. Changes in the legal tender values could also
be motivated by fiscal considerations or by attempts
to target a particular price level or exchange rate.
The physical nature of the medium of exchange
led to a particular set of concerns.
Commodity Money
Coins, like anything else, depreciate with use,
through wear and tear. Since coins of different
values have different usage rates, the depreciation
rate varied by denomination. Also, being roughly
constant over time, depreciation depended on the
age of the coin. Finally, imperfect minting tech-
nology as well as actions by the public (clipping,
sweating) aggravated the disparities between coins.
This factor introduced heterogeneity among coins
and hindered the achievement of a stable and
uniform monetary system. Improvements in coin
production partially remedied the problem, as did
periodic recoinages.
When the monetary objects consist not only of
coins but also of paper currency or tokens that are
demand promises to the commodity, a fourth type
of monetary policy is available: suspension of
convertibility. The monetary authority can refuse to
honour the promise of convertibility for some
period of time. An example is the suspension of
convertibility by the Bank of England between 1797
and 1819 during the wars with France. During the
19th century suspensions were not uncommon
during financial or fiscal emergencies, with the
understanding that the suspension would end after
the emergency and convertibility would be restored
at the preexisting parity. This understanding has
been described as a state-contingent gold standard
(see Bordo and Kydland ).
When there is a central bank, an additional
monetary tool is to change the discount rate, the
interest rate at which the central bank lends reserves
to the banking system. During the gold standard
period, this was the primary means by which central
banks affected the exchange rate of their money
against the monies of other countries.
Conclusion
Commodity money is a thing of the past; countries
worldwide now use fiat money standards. This
practice has led to an efficiency gain in the sense
that resources that were once tied up in coins are
now available for consumption and production
(perhaps prompting John Maynard Keynes to refer
to gold as the ‘barbarous relic’).
1851
It has also led to a greater scope for monetary
policy because the supply of money can be changed
almost costlessly. However, along with this greater
scope has come the greater potential for
governments to use inflation to collect seigniorage
revenue or to reduce the real value of their debts.
How to use the freedom that commodity money
restricted is still a matter of debate.
See Also
► Bimetallis
► Bretton W
► Fiat Mone
Bibliography
Bordo, M., and F. Kydland. 1996. The gold standard as a
commitment mechanism. In M o d e r n p e r s p e c t i v e s
o n t h e g o l d s t a n d a r d , ed. T. Bayoumi, B.
Eichengreen, and M. Taylor. Cambridge: Cambridge
University Press.
Kiyotaki, N., and R. Wright. 1989. On money as a medium of
exchange. J o u r n a l o f P o l i t i c a l E c o n o m y 97:
927-954.
Luschin von Ebengreuth, A. 1926. AllgemeineMiinzkunde und
Geldgeschichte des Mittelalters und der neueren Zeit.
Munich: R. Oldenbourg.
Redish, A. 2000. B i m e t a l l i s m : A n e c o n o m i c a n d
h i s t o r i c a l a n a l y s i s . Cambridge: Cambridge
University Press.
Rolnick, A., F. Velde, and W. Weber. 1996. The debasement
puzzle: An essay on medieval monetary history.
J o u r n a l o f E c o n o m i c H i s t o r y 56: 789-808.
Sargent, T., and B. Smith. 1997. Coinage, debasements, and
Gresham’s laws. E c o n o m i c T h e o r y 10: 197-226.
Sargent, T., and F. Velde. 2002. T h e b i g p r o b l e m o f
s m a l l c h a n g e . Princeton: Princeton University
Press.
Sargent, T., and N. Wallace. 1983. A model of commodity
money. J o u r n a l o f M o n e t a r y E c o n o m i c s 12:
163-187.
Sputford, P. 1988. M o n e y a n d i t s u s e i n m e d i e v a l
E u r o p e . Cambridge: Cambridge University Press.
Sussman, N., and J. Zeira. 2003. Commodity money inflation:
Theory and evidence from France in 1350-1436.
J o u r n a l o f M o n e t a r y E c o n o m i c s 50: 1769-
1793.
Velde, F., and W. Weber. 2000. A model of bimetallism.
J o u r n a l o f P o l i t i c a l E c o n o m y 108: 1210-1234.
Williams, J., ed. 1997. M o n e y : A h i s t o r y . New York: St
Martin’s Press.
1852
Commodity Reserve Currency
Albert Gailord Hart
Commodity Reserve Currency (CRC for short) is a
proposal for re-establishing an international
monetary ‘standard’ - basing it upon a ‘basket’ of
widely used commodities. Recent experience shows
the inconvenience of lacking a standard. While
restoration of a gold standard lias many supporters,
gold has become so remote from the goods-and-
services economy that for decades now
governments have had scope to play tricks with its
price; and since the early 1970s, the value of gold
has been highly unstable. Can there be a commodity
standard other than gold, less abstract and linked to
articles of everyday use? A good way to study this
question is to examine the feasibility and
desirability of CRC.
Pioneers of the CRC proposal were Jan
Goudriaan and Benjamin Graham. Variant pro-
posals have come from a number of economists,
including Lord Keynes and Friedrich Hayek. The
nearest approach to a standard version is probably
still the submission of 1964 to the United Nations
Conference of Trade and Development by Albert
Hart, Nicholas Kaldor and Jan Tinbergen. No
governments or major multilateral bodies have
sponsored CRC, though there has been official and
private use of ‘baskets’ of currency units (ECU,
etc.), and of‘baskets’ of securities (traded on
various private commodity exchanges).
Advocates of CRC propose that it be adminis-
tered by a multinational agency, which we may call
IMA - presumably to be a branch of the
International Monetary Fund. A currency unit
(which following Lord Kaldor we may call the
Bancor) is to be defined as the value of a basket of
primary commodities, with fixed physical
composition.
Like the administrator of a traditional gold
standard, IMA must buy or sell at a stated price
(plus or minus a margin to avoid a hair-trigger
effect) as much of the monetary commodity (i.e. the
commodity baskets) as may be offered
Commodity Reserve Currency
or demanded. As to the margin, some CRC pro-
ponents suggest 5 per cent on each side of par; but a
wider range would have great advantages. Prices of
the individual component commodities could
fluctuate more than the basket. Any one commodity
price could rise; but if its rise would bring the
basket above the posted selling price, sales of
baskets by IMA would bring a compensating fall of
other prices.
The price of a basket with fixed physical com-
position is an index number of commodity prices,
weighted by the quanta of the various items
included. Hence CRC may be viewed as a scheme
to stabilize an index of primary-commodity prices.
CRC thus would have a counter-cyclical effect -
holding within bounds the fluctuation of income for
the world’s exporters and producers of primary
commodities, and by the same token the fluctuation
of major elements in the world’s cost of living.
An effective CRC would require that national
currencies be tied to the Bancor by fixed exchange
rates, or at least by not-too-movable pegs. The
general stabilization effect of the proposal would
vanish if major currencies were allowed to float
against the Bancor.
Many primary commodities have been proposed
for a CRC basket. Major criteria for inclusion are:
(T) Standardization, of the sort necessary to run
fiiUires markets on a commodity exchange -
with rules for dealing with quality differences.
(2) Storability for at least a year or two without
excessive cost or loss of quality. Security
against fire, looting, requisition by local gov-
ernments, etc., is implied.
(3) Improbability of major price manipulations by
governments or by combinations of producing
enterprises. Long before OPEC rose to power
in 1973, this criterion led advocates of CRC to
omit petroleum from the proposed basket.
(4) Reliability of commodity contracts, enabling
the IMA to replace physical holdings with
contracts for fiiUire or spot delivery if this will
reduce costs. Use of fiiUires could be of
Commodity Reserve Currency 1853

humanitarian importance in case of shortages has seemed impossible. When there has been a
of foodstuffs such as rice. great piling up of stocks (as at this writing in early
1986 ), it has seemed as if mobilization of stocks
These criteria would admit most of the world’s held by the European Economic Community, the
important grains, fibres, fats and oils, beverage United States, various other governments such as
crops, primary metals - and probably a number of Brazil, cartels such as that for tin, and private
basic chemicals and forest products. Amounts of the concerns such as copper companies, might permit a
various commodities in the basket would reflect very rapid start.
their weight in world production and/or trade. Starting a CRC dining a period of widespread
The CRC basket must be large (worth several shortages could be highly inflationary; starting it
million US dollars), because for efficient trading dining a general economic downswing could mit-
each element must be a multiple of a wholesale lot. igate a recession. There is debate as to whether
This large basket-size would be appropriate, since accumulation of a reserve must stick to previously
CRC is designed as a vehicle for holding national agreed proportions of the different commodities, or
monetary reserves rather than for retail or even whether (as proposed by Keynes and more recently
wholesale trade. by Kaldor) the IMA should buy individual
IMA holdings of the various commodities must commodities at such dates and prices as seem wise.
be parcelled out to points of delivery and storage On this route, the composition of the CRC basket
along the normal trade routes of the commodities. and the selling-price offer would grow out of the
Correspondingly, IMA purchases and sales must be process of accumulation. This variant, plainly,
handled by agents at the various trading points. To would increase the similarity of IMA’s operations
tell whether at any moment baskets must be bought to those of commodity ‘stabilization’ groups and
or sold, IMA must sum up bids or offers reported by remove the impersonality held to be the central
these agents. IMA must take the initiative - virtue of the Goudriaan/Graham scheme.
instructing all agents to sell or buy - whenever the
sum of bid prices for the elements of the basket
adds up to the posted selling price for the basket, or
the sum of asked prices to the buying price.
Objections to CRC have hinged primarily on See Also
costs and/or on the difficulties of getting a CRC
system under way. There has been continuing
debate about the size of the reserve needed to
validate IMA selling offers, about the cost of
holding stocks of various suggested commodities,
etc. Such costs must be compared with benefits
from reducing cyclical fluctuations in primary References
producers’ incomes, etc. - and of doing so without
Bennett, M.K., et al. 1949. I n t e r n a t i o n a / c o m m o d i t y
engaging in commodity-by-commodity operations. stockpiling as an economic stabilizer.
Benjamin Graham used to stress that the success of Stanford: Stanford University Press.
such operations hinges on restriction of commodity Goudriaan, J. 1932. H o w t o s t o p d e f l a t i o n . London:
production, whereas success of CRC would The Search Publishing Company.
Grubel, H.G. 1965. The case against international commodity
stimulate production. For this and other reasons, the reserve currency. O x f o r d E c o n o m i c P a p e r s
cost/benefit problem is complex. 17(1): 130-135.
During episodes of worldwide commodity Hailwood, P. 1986. External economy arguments for com-
stringency, accumulation of a commodity reserve modity stockpiling: A review. B u l l e t i n of
E c o n o m i c R e s e a r c h 38(1): 25-41.
Harmon, E. 1959. C o m m o d i t y r e s e r v e c u r r e n c y .
New York: Columbia University Press.
Hart, A.G., N. Kaldor, and J. Tinbergen. 1964. The casefor an
international Commodity Reserve Currency.
1854 Common Factors

Geneva: United Nations Conference on Trade and Well-known examples of comovement in multi-
Development. 17 February 1964. E/CONF.46/p/7. variate data sets include business cycles in mac-
Conveniently accessible in N. Kaldor, E s s a y s o n e c o -
n o m i c p o l i c y - I I . London: Gerald Duckworth &
roeconomic indicators and shifts in the entire term
Co.. 1964. structure of interest rates, and researchers some-
Hayek. F.A. 1984. The future monetary unit of value. In times attribute this comovement to a small set of
M o n e y i n c r i s i s , ed. N. Barry. Siegel: Pacific underlying forces or latent ‘factors’ that influence
Institute for Policy Research.
Newbery, D.M.G.. and J.E. Stiglitz. 1981. T h e t h e o r y o f
each variable in the system. It is then convenient to
commodity price s t a b i l i z a t i o n . Oxford: think of the variation in each variable in the system
Clarendon. as the sum of two types of (unobserved)
components, one of which captures variation that is
due to 'common factors’, while the other captures
Common Factors all other variation. Models that attribute
comovement to common factors are called common
Heather M. Anderson factor models, and common factor analysis involves
the identification and study of the common factors.
Common factor models are particularly popular
in empirical settings because they offer parsimony,
and simplify estimation by reducing the number of
Abstract
parameters that need to be estimated. Economists
This article outlines and illustrates several types
will typically be interested in interpreting common
of common factor models that are found in the
factors so that they can explain why comovement
applied economics literature. These factor
occurs. Economic theory sometimes predicts
models include those based on principal
common factors. Perhaps the best-known example
components, classical factor analysis, dynamic
of this is the capital asset pricing model, in which
factor analysis and common features, and the
the (excess) return for the market portfolio is the
discussion addresses the identification and esti-
common factor in the (excess) return for each
mation of factors, as wel I as the use of common
individual stock. Another well-known example
factor models.
arises when the term structure of interest rates is
modelled, because the no arbitrage condition
Keywords
implies that the entire term structure is determined
ARM A processes; Autoregressive moving
by a single factor, which is the instantaneous
average (ARMA) processes; Canonical corre-
interest rate.
lations; Capital asset pricing model; Classical
A simple model that captures the concept of
factor models; Coincident indices; Common
common factors in a set of N time-series in
factors; Common feature models; Common
the (demeaned) vector Y, = (Yu, Y2,.............YNt)' is
trend model; Diffusion indexes; Dynamic factor
given by
(or index) models; Factor analysis; Kalman
filter; Principal component analysis; Real busi- Y, AF, + .
£r
ness cycle models; Reduced rank regressions;
Static factor models; Stone, J; Term structure of (1 )
interest rates; Time series analysis
where Ft is an /' x 1 vector that contains r common
factors, A is an N x r factor loading matrix (with
rank(A) = r < TV), and £t contains N idiosyncratic
JEL Classifications
components. With the use of II, and I, to denote the
C32
variance covariance matrices of Yt, F, and £t, it is
Economic analysis frequently involves the study of usual to assume that Z£: is diagonal, and it is also
variables that exhibit similar behaviour, and it is common to normalize the set of r factors in Ft by
often of interest to model this comovement. assuming that X/. I,..
Common Factors 1855

Model ( ) is similar to conventional factor factor loadings are also the least squares estimators
models that are often used in cross-sectional set- of the reduced rank regression given by
tings, although the variables are specified here as
time series, to facilitate discussion on dynamic Yt= A B P, + £„ (2)
(N\r) (r \ N )
factor models. If there is no serial correlation in Y,
or F„ or if estimation is undertaken as if this is the
where BYt contains the r factors F, = (/j, .. .,fr).
case, then ( ) is called a static factor model. It is
Anderson ( , ch. 11) provides a standard
usual to assume that F, and e, are jointly stationary,
reference.
that E(e,) = 0,E(F,e't) = 0, and that e, contains no
In practice, one needs to determine r before
serial dependence, but these latter assumptions can
estimating common factor models, and this is
be relaxed, depending on the type of factor model
2jLi+...+2^
under consideration. often based on the ratio given by -TJ-
/l, -+Xj,
There are many ways to identify the factors in This ratio measures the loss of information in the
( ), and standard techniques include the use of reduced rank system relative to an unrestricted
principal component analysis, factor analysis and system, and typically investigators will choose r so
canonical correlations to estimate the parameters of that this ratio is kept small. Bai and Ng ( )
various associated reduced rank regressions. More have developed model selection criteria
recently, researchers have focused on the time series that are consistent as {N, T} —> oo.
properties of multivariate data-sets, and modem Principal components are usually used for
factor models include dynamic factor (or index) dimension reduction, and economic interpretation of
models, and models that incorporate common the resulting factors is rarely straightforward.
features. These latter models incorporate various However, Stone ( ) has summarized a set of
ways of allowing the factors to follow specific series from the US national accounts, associating the
dynamic processes, or to contain specific time- first three principal components with income,
series properties. income growth and time and Chamberlain and
Rothschild ( ) have promoted the use of prin
cipal components for estimating approximate factor
models of asset prices. Stock and Watson ( )
Principal Component Models have suggested the use of diffusion
indexes (principal component factors associated with
Principal component analysis involves the intuition
large macroeconomic data sets) for forecasting key
that most of the variance in Y, will be attributable to
macroeconomic variables, and the interest here
variance in the r components in Ft. The factors Ft
centres on using information in the factors rather
are modelled as linear combinations of Yt, and their than interpreting the factors themselves.
identification is based on finding the r (orthogonal)
linear combinations of Y, that have the most vari-
ance. In practice this involves finding the eigen
values 1] > 12 > ■ ■ ■ > XN and associated eigen- Classical Factor Models
vectors /i, ..., fN of the form /, = f^Y, that are
associated with the roots of the equation |£y XI\ Classical factor models are closely related to prin-
= 0, where £v is an estimate of ~Ly. The [1, are cipal components models, but the underlying intu-
picked so that ^£y — Xj'j /?, = 0and/i'/i, = 1, and the ition and assumptions are different. In this case the
factors Ft are then defined by F, = (/j, .. ,,fr). This key assumption is that E£ is diagonal so that the e,
decomposition ensures that£ (fjej) = 0, but it implies describe idiosyncratic effects that are unique to each
that the &, (which are each linear combinations of variable in Yt, while the factors describe joint effects
(frl i,.. ,,/v)) will be correlated with each other so that in Yt. The assumptions that £(e,) = 0 and E{Fte!^j =
will not be diagonal. Principal components 0 still hold, and the s, are assumed to contain no
estimators of common factors and serial dependence. Under these
1856 Common Factors

assumptions X>- = AEpA' + E£, and estimates for A, Dynamic Factor Models
X/. and can be found by maximizing the function
Classical factor models are not well suited to
t=T multivariate analysis of time series because they
T 1
LT{AXI) = -- l n | E y | --YT&f, assume no serial correlation in eh and, if there are
z z
t=1 any dynamics in F„ then they are implicit and not
explicitly modelled. Dynamic factor models address
subject to the condition that ranMA) = r and a set of these concerns by treating the zt and Ft as
normalization restrictions that will uniquely identify autoregressive moving average (ARMA) processes.
the (r I 1)(/V I ir) parameters. Researchers often use The innovations that underlie the A processes for e,
the joint restrictions that Xp = I,. and that A'X, 'A is are assumed to be mutually uncorrelated, and
diagonal for normalization, but other normalizations uncorrelated with the innovations that underlie the
are common (see Anderson , eh. 14, for details). If F, at all leads and lags, but the factors themselves
Yt and st are normally distributed then LT (A, Es) is can be mutually correlated. Different variables in Y,
the log likelihood for Yt (if we ignore the constant can then move together because they are functions
term), but, even when Yt and st are not normally of the same factor (s), or because they are functions
distributed, the maximization of LT (A, Es) dehvers of different factors that are themselves correlated.
quasi-maximum likelihood estimates. There are The identification and estimation of small- scale
several ways of using the estimates of A and Es to dynamic factor models is sometimes based on
obtain estimates of the factors in F„ and perhaps the spectral techniques (see Geweke ; or Sargent and
best-known of these is Bartlett’s ( , ) Sims ), and use of the Kalman filter in the time
method based on generalized least squares given by domain (as in Engle and Watson , or Harvey and
Koopman ) provides an alternative approach.
F, = ( A Aj A Y,. Dynamic factor models have been particularly
popular for estimating factor models of business
cycles (as in Geweke and Singleton ), but they have
As above, it is necessary to determine r prior to also been used for studying the term structure
estimating the factors, and, on the assumption of (Singleton ) and fluctuations in employment across
normality, the likelihood ratio test statistic for different industrial sectors (Quah and Sargent ).
testing //„: r = s versus HA\ r > s is given by Recent work has shifted towards the identifica-
tion and estimation of common factors in large-
scale models, relying on the use of large A to obtain
—r[ln|fr| -ln|AA' + f e |] = — T Z;=f+1ln(l - I:),
consistent estimates of the factors. One strand of
this literature adopts a static framework and
c- 1-
standard principal components to estimate the
where the /., are the characteristic roots of A X.. A
factors, and then builds dynamic models of the
(in decreasing order) and A is estimated under the
factors. The resulting models are sometimes called
null. The test statistic is asymptotically distributed
approximate dynamic factor models. Applications
as a xl with q = [(A — sf — N — s]!2° of freedom
of this approach include Stock and Watson’s
under the null.
diflusion index ( ),
There are numerous applications of classical
and Bemanke and Boivin’s ( )
factor analysis to economic problems, and an early
estimation of a monetary policy reaction function.
example includes Stone’s ( ) factor anal Another strand of this literature allows different
ysis of the demand for N commodities. Another variables to depend on different lags of common
example includes a factor model of returns by factors. These ‘generalized dynamic factor models’
Deistler and Hamann ( ). are estimated using ‘dynamic principal compo-
nents’, which are the principal components of
Common Factors
spectral density matrices at different frequencies.
Applications of this latter approach include a study
of business cycle dynamics in the United States
(Fomi and Reichlin ) and the development of a
coincident index for Europe (Fomi et al. ).
Canonical correlation-based models
Principal component and factor models assume that
the factors are linear combinations of the A
variables in Yt, but sometimes it is useful to assume
that the factors are linear combinations of M
variables contained in another multivariate time
series denoted by X\. The variables mXt will often
include lags of the variables in Yt, but A, can also
include variables that would be classified as
explanatory variables in a regression context. The
factors in ( ) can now be written in the form Ft =
B'Xt (where rank(B) = r < min {iV, AT} ). In what
follows, we assume that the st in ( ) are white noise
and uncorrelated with Xr.
The main idea behind a canonical correlations
approach is to find linear combinations of X, that
are strongly correlated with linear combinations of
Yu and, as for principal component models, the
estimators of common factors and factor loadings
are the least squares estimates of a reduced rank
regression. In this case the regression is
and the factors and factor loadings are related to the
r largest roots of R = Xy'XyyXy 'XyyX,/ , which is
the multivariate generalization of the (squared)
correlation coefficient between two variables. If we
order these roots (also called squared canonical
correlations) so that %>£> ... > /2, and let V,, V2, . ■
■ Vr be the r associated eigenvectors, then the factor
loadings and factors are given by A,- = YyX, and
B,X, = Y.y1Y.yxY.x ' V,Xt. Anderson ( , eh.
provides a detailed
discussion of canonical correlations, while Izenman
( ) discusses the associated reduced
rank regressions. When the variables in Xt are
simply lags of the variables in Y,, then the first
factor is the best predictor of Y, based on
1857
past history, the second factor is the next best
predictor, and so on, and the factors provide a set of
leading indicators for Yt. When X, consists of
explanatory variables for Y„ then the factors are
often called coincident indices. One can base a test
of H0 : r = s versus HA : r > s on the test
statistic — rX'r^+1ln^l — 2, j , which has a y2
distribution with (m — s) (n — s) degrees of
freedom under the null.
Common Feature Models
Common feature models are a special class of factor
models in which the common factors have a
statistical characteristic of interest, while the
idiosyncratic components fail to have this
characteristic. Common features were first intro-
duced by Engle and Kozicki ( ) when they
discussed serial conelation features - a situation in
which each of A variables is serially conelated, but
there are linear combinations that are white noise.
Here, the presence of A r white noise linear
combinations implies a factor model in which there
are r serially correlated factors (which are
sometimes called common cycles). An earlier
example of a common feature model is Stock and
Watson’s ( ) common trend
model which is valid when variables are
cointegrated (as in Engle and Granger ). In this case
the common factors are integrated of order one but
the remaining components (often called error
correction terms) are stationary. Other examples of
common features include Vahid and Engle’s ( )
common trend-common cycle
representation, and common nonlinearity (Anderson
and Vahid ).
The identification of common features involves
finding linear combinations of the data that do not
have the feature, and this can be done using a
canonical correlations approach in which the
12)
variables in Xt model the characteristic of interest.
To illustrate, lags of Yt are put into X, when testing
for serial correlation features in Yt, and lagged
levels are included in Xt when testing for common
trends. Factors associated with the lowest eigen
values define linear combinations
1858
that do not contain the feature, while factors
associated with the highest eigen values are used to
model the common features. Johansen’s ( ) 76: 774-781.
procedure provides a well-known example of this,
although inference in this case is based on non-
standard (rather than y2) distributions because the
factors are non-stationary.
A well-known example of a common feature
model is the real business cycle model of King et al.
( ), in which a common factor
(productivity) generates the trend in output con-
sumption and investment, and another factor (the
deviation of capital stock from steady state ) gen-
erates the common cycle.
See Also
► Reduced Rank Regression
► Time Series Analysis
Bibliography
Anderson, T. 1984. An introduction to multivariate statistical
analysis, 2nd ed. New York: Wiley.
Anderson, H., and F. Vahid. 1998. Testing multiple equation King, R., C. Plosser, and S. Rebelo. 1988. Production, growth
systems for common nonlinear components. J o u r n a l
o f E c o n o m e t r i c s 84: 1-36.
Bai, J., and S. Ng. 2002. Determining the number of factors in Quah, D., and T. Sargent. 1993. A dynamic index model for
approximate factor models. E c o n o m e t r i c a 70: 191-
221.
Bartlett, M. 1937. The statistical conception of mental factors.
B r i t i s h J o u r n a l o f P s y c h o l o g y 28: 97-104.
Bartlett, M. 1938. Methods of estimating mental factors.
N a t u r e 141: 609-610.
Bemanke, B., and J. Boivin. 2003. Monetary policy in a data
rich environment. J o u r n a l of Monetary
E c o n o m i c s 50: 525-546.
Chamberlain, G., and M. Rothschild. 1983. Arbitrage, factor
structure and meanvariance analysis in large asset
markets. E c o n o m e t r i c a 51: 1305-1324.
Deistler, M., and E. Hamann. 2005. Identification of factor
models for forecasting returns. J o u r n a l
F i n a n c i a l E c o n o m e t r i c s 3: 256-281.
Engle, R., and C. Granger. 1987. Cointegration and error
correction representation, estimation and testing.
E c o n o m e t r i c a 55: 251-276.
Engle, R., and S. Kozicki. 1993. Testing for common features
(with discussions). J o u r n a l o f B u s i n e s s a n d
E c o n o m i c S t a t i s t i c s 11: 369-395.
Common Factors
Engle, R., and M. Watson. 1981. A one-factor multivariate
time series model of metropolitan wage rates. J o u r n a l
of the American Statistical Association
Fomi, M., and L. Reichlin. 1998. Let’s get real: A factor-
analytic approach to disaggregated business cycle
dynamics. R e v i e w o f E c o n o m i c S t u d i e s 65:
453-473.
Fomi, M., M. Hallin, M. Lippi, and L. Reichlin. 2000. The
generalized dynamic factor model: Identification and
estimation. R e v i e w of Economics and
S t a t i s t i c s 82: 540-554.
Geweke, J. 1977. The dynamic factor analysis of economic
times-series models. In L a t e n t v a r i a b l e s i n
s o c i o e c o n o m i c m o d e l s , ed. D. Aigner and A.
Goldberger. Amsterdam: North-Holland.
Geweke, J., and K. Singleton. 1981. Maximum likelihood
‘confirmatory’ factor analysis of economic time series.
I n t e r n a t i o n a l E c o n o m i c R e v i e w 22: 133-137.
Harvey, A., and S. Koopman. 1997. Multivariate structural
time series models. In S y s t e m a t i c d y n a m i c s i n
e c o n o m e t r i c a n d f i n a n c i a l m o d e l s , ed. C.
Heij, H. Schumacher, and C. Praagman. Chichester:
Wiley.
Izenman, A. 1980. Assessing dimensionality in multivariate
regression. In H a n d b o o k o f s t a t i s t i c s , vol. 1, ed.
P. Krishnaiah. Amsterdam: North-Holland.
Johansen, S. 1988. Statistical analysis of cointegration vectors.
Journal of Economic Dynamics and
C o n t r o l 12:231-254.
and business cycles II. New directions. J o u r n a l o f
M o n e t a r y E c o n o m i c s 21: 309-341.
large cross sections. In B u s i n e s s cycles,
i n d i c a t o r s a n d f o r e c a s t i n g , ed. J. Stock and
M. Watson. Chicago: University of Chicago Press.
Sargent, T., and C. Sims. 1977. Business cycle modeling
without pretending to have too much a-priori economic
theory. In N e w m e t h o d s i n b u s i n e s s c y c l e
r e s e a r c h , ed. C. Sims et al. Minneapolis: Federal
Reserve Bank of Minneapolis.
Singleton, K. 1980. A latent time series model of the cyclical
behavior of interest rates. I n t e r n a t i o n a l E c o -
n o m i c R e v i e w 21: 559-575.
Stock, J., and M. Watson. 1988. Testing for common trends.
Journal of the American Statistical
of A s s o c i a t i o n 83: 1097-1107. ’
Stock, J., and M. Watson. 2002. Macroeconomic forecasting
using diffusion indexes. J o u r n a l o f B u s i n e s s
a n d E c o n o m i c S t a t i s t i c s 20: 147-162.
Stone, J. 1945. The analysis of market demand. J o u r n a l
of the Royal Statistical Society, Series
A 108: 286-382.’
Stone, J. 1947. On the interdependence of blocks of trans-
actions. J o u r n a l o f t h e R o y a l S t a t i s t i c a l
S o c i e t y , S e r i e s B , Supplement, 1-45.
Vahid, F., and R. Engle. 1993. Common trends and common
Common Land
Common Land
T. Williamson
The legal status of common land is a source of
considerable popular confusion. With the notable
exception of some village greens, commons do not
represent areas which are owned by nobody, nor
areas which are owned by everybody, nor even by
everybody within a given locality. Since the early
medieval period, commons have been owned by
specific individuals, usually the lord of the manor
within which they he. The term ‘common’ refers
not to ownership, but to rights held in common by
certain people to use the product of the soil of the
area in question. In turn, this means that the owner
cannot enclose the land; hence the unfenced open
space which is still the most characteristic feature of
a common (Campbell and Clayden ).
These rights are not now, nor have they been in
the historic past, generally shared by everyone
living within a given locality. Instead they are
usually attached to specific dwellings, or to their
sites. Except, therefore, where the owner has so
decreed, the use of commons as recreational open
spaces by a wide public is not in the strict sense
defensible in law.
Commons, of course, were not in origin pri-
marily places for recreation. They formed a vital
element in the pre-industrial rural economy. Com-
mons represent the attenuated remnants of the
medieval wastes: areas which were not used to
produce arable crops, but to provide a range of other
resources. The principle rights exercised by
commoners were, and are: common of pasture, or
the right to pasture animals; pannage, or the right to
allow pigs to eat acoms or beech mast; common in
the soil, or the right to take minerals, gravel, stone,
sand etc. for use on the commoner’s holding;
estovers, or the right to take small branches,
bracken etc. for fuel, fencing or animal litter;
turbary, or the right to dig peat or turf for use as
1859
fuel; and piscary, the right to take fish from streams
or ponds on the common. Of these rights, that of
common of pasture has normally been by far the
most important.
The early history of common land is obscure. It
appears that in the early Saxon period, areas of open
waste were much more extensive than they were to
become in the medieval period, and the rights to
their use were more loosely defined and often
exercised by much wider groups. The name of
Sherwood Forest, for example - the Shire Wood -
indicates that it was once the common woodland of
the entire shire of Nottingham. Limitation and
closer definition of rights to common waste
occurred during the population increase of the early
medieval period (Hoskins and Stamp ). As arable
expanded, common grazing dwindled, and areas of
waste which had formerly been shared by
communities were now divided between them, often
after violent disputes. But there were also disputes
within communities, as manorial lords, in
association with their more prosperous tenants,
attempted to take areas of the common waste into
private ownership. The medieval struggle for the
commons culminated in the Statute of Merton (The
Commons Act of 1235), which decreed that
freeholders had to be left with sufficient pasture to
maintain the mixed farming of their holdings.
However, the rights of the customary tenants were
not protected by statute law, and the passing of the
law for the first time clearly enshrined in national
law the concept that the manorial lord rather than
the community itself was the owner of the common
waste of a manor.
The continual expansion of arable at the expense
of the common wastes during the period before
1300 had other effects. There was an increasing
tendency for commons to be stinted, that is, for the
number of beasts put out by each commoner to be
more carefully regulated. There were a number of
ways in which this could be organised, the most
usual being by the mles of levancy and couchancy,
that is, where the right to turn out was measured by
the capacity of the commonable tenement in such a
way that only as many animals could be turned out
as the tenement
1860
was able to support (with the aid of hay etc.)
through the winter.
The extent to which the commons survived
during the medieval period varied considerably
from region to region, as the result of the interplay
of a number of factors. Essentially, commons sur-
vived best where population densities were low and
where much land was unsuitable for arable
agriculture. Thus large areas survived in the uplands
of the north and west. But there may have been
more complex social and economic factors which
were also important in the preservation of common
grazing, especially in lowland areas, for the
distribution of commons in medieval England does
not appear to be a direct and simple reflection of
demographic pressure or environmental factors. In
certain parts of the south and east of England - areas
of dispersed settlement and irregular field systems,
poorly developed communal controls and
individualistic agriculture - some communities seem
to have lacked the management structures necessary
to act corporately to plough up areas of common
grazing, even in areas of high population and
moderately fertile soils, such as Norfolk. In addi-
tion, poor controls on the alienation of land and the
practice of partible inheritance led to an early
proliferation of smallholders for whom the
resources provided by the commons were of vital
importance. In such areas the conversion of waste to
arable agriculture ground to a halt rather earlier than
in the classic open-field areas of the Midlands.
In the latter areas, communities were more
cohesive and communal controls on agriculture and
land-use better developed. There were often
stronger manorial controls on the alienation of land,
holdings did not fragment to the same extent and
there was less economic polarisation within the
farming community'. It may be this that explains
why in many of these areas so much of the wastes
were ploughed up in the 12th and 13th cenUtries, as
arable prices, and the need for food for
consumption, increased. Whatever the explanation,
it appears that in many parts of the central
Midlands, areas of common grazing were almost
entirely destroyed by the end of the 13 th century.
The arable strips of many villages ran right up to
Common Land
the parish boundary, meeting with those of
neighbouring villages.
This was not true of all areas in the Midlands,
however. Conversion of grazing to arable was more
retarded in the Forest areas. Such areas were not
necessarily densely wooded; the term forest was a
legal rather than a descriptive or environmental
term, referring to areas to which forest law applied.
This was a body of rules and restrictions originally
intended to preserve deer for the royal chase and
which inter alia attempted to limit the destruction of
suitable habitats through the expansion of arable
cultivation. In reality', forest law functioned more as
a source of revenue, for encroachments on the
wastes were tolerated if a fine was paid.
Nevertheless, in areas like Rockingham Forest in
Northamptonshire, these instim- tional factors
combined with the relatively marginal naUtre of
local soils to preserve extensive areas of open waste.
As arable expanded at the expense of grazing, in
the areas of irregular field systems and dispersed
settlement surviving commons often became foci for
settlement. The poorly developed nature of
communal controls in such areas was probably the
stimulus for this development; farmers and
smallholders moved to the edge of areas of common
grazing not only for the convenience provided by
such a location, but also, as it were, to stake a visible
claim to their use. The freedom to alienate land and
the irregular nature of field systems in these areas
made such settlement migration possible, for they
allowed the acquisition of blocks of land adjacent to
the common upon which farmsteads could be
established. This development seldom occurred in
areas of regular open-field systems, nucleated
settlement, and strong community controls.
Thus it was mainly in the south-east and the
north-west of England that the medieval period saw
the development of straggling settlements around
the perimeter of commons. This process went
furthest in parts of East Anglia, where complex
manorial organisation and the presence of
substantial numbers of free tenants practising par-
tible inheritance led to a proliferation of small-
holdings and wholesale migration away from earlier
village sites. Farmsteads clustered around
Common Land
areas of common which thus became large village
greens, leaving the parish church - marking the
original Saxon site of settlement - isolated in the
fields some distance away (Wade Martins ).
Lowland commons have a distinctive form.
They have straggling, concave outlines, formed by a
series of rough arcs; roads funnel into the common
where these arcs join. This characteristic shape
probably derives from the fact that commons are the
remnants of more extensive areas of waste which
had been continuously encroached upon for
centimes before their outline became fossilised,
usually in the early medieval period. The perimeters
of commons are often defined by particularly
massive and ancient banks and ditches.
Today, many commons are wooded, but this is
normally a relatively recent feature, resulting from a
relatively recent decline in the intensity of grazing.
By the end of their medieval period, commons had
usually lost whatever woodland they had formerly
carried; it had been destroyed by felling and over-
grazing, and only names like ‘Wood Green’
sporadically reflect their former nature. Medieval
woodland, in contrast, was not usually common
land, but land which had been enclosed from the
waste and over which the use and access of others
had been limited (Rackham )•
By the end of the Middle Ages, there were
considerable regional variations in the extent of
common land. Commons survived better in the
upland areas of the north and east than in the more
fertile lowland zone. Within the lowland areas, they
tended to survive better in the south and east of
England, and in the west country, than in the classic
open-field areas of the Midlands, with the exception
of the forest areas, where they usually also survived
well.
These variations were a factor in the local and
regional development of rural society in the post-
medieval period. In particular, areas in which
extensive commons survived tended also to be the
areas in which the decline of the small freehold
farmer, which continued at varying rates throughout
the post-medieval period, was retarded. In areas like
the Fens, small farmers used their rights to
extensive commons to adopt forms of livestock
1861
farming as specialized agricultural regions emerged
in the 15th century. Survival of commons also
allowed small cottagers to maintain a measure of
economic independence, and the more extensive
commons attracted large numbers of squatters, often
part-time craftsmen. As a result, areas in which
large commons survived tended to have a reputation
for lawlessness. The opportunities which such areas
offered to the poor also ensured that they often
experienced particularly rapid population growth in
the early modem period. In Northamptonshire, for
example, forest villages in the 17th century were on
average around half as populous again as non-forest
villages (Hoskins and Stamp , p. 52).
Much enclosure of common land occurred dur-
ing the 16th and 17th centuries. Nevertheless, in
1688 Gregory King estimated that there were still
10 million acres of heaths, moors, mountains and
barren land in England and Wales, and a further 3
million acres of forests, parks, and commons, the
majority of which was common land. Today, the
total area of common land in England and Wales is
around 1.5 million acres. Even allowing for a high
degree of inaccuracy in King’s estimates, there has
clearly been a dramatic reduction in the area of
common land as a result of Parliamentary
enclosures, mainly in the late 18th and early
decades of the 19th centuries. Enclosure was
principally inspired by a desire on the part of the
larger landowners to profit from the conversion of
the remaining commons to arable, or their
improvement as pasture, both of which were dif-
ficult or impossible where the land was subject to
the use and access of a large number of local
inhabitants. The high prices of arable during the
Napoleonic Wars were a particular stimulus to
enclosure of open heaths and wastes, especially on
the light soils of eastern England (Turner pp. 63-
93).
The enclosure of commons was, like the enclo-
sure of open fields, closely connected with the
decline of the small owner-occupier which had
continued almost uninterrupted throughout the early
modem period. As land fell into the hands of
relatively few people, so it became easier to obtain
the agreement necessary to enclose, especially as
with the advent of Parliamentary
1862
Enclosure a majority in favour of enclosure was
judged on the basis of the area of land which the
agreeing landowners held, rather than on their
number.
Yet as well as being in part a consequence of the
decline of the small proprietor, enclosure of
commons also served to accelerate this process. The
allotments received by those small farmers or
cottagers who were able to prove the legality of
their claim to common rights were seldom sufficient
compensation for the advantages lost through
enclosure, especially when legal and fencing costs
were taken into account. For many small farmers,
enclosure was the final misfortune which led to their
departure from farming; for the small cottager,
enclosure often led to increased, if seasonal, reliance
on poor relief (Snell ).
Today, the distribution of surviving common
land in England and Wales continues to be very
uneven, with more in the highland zone than in the
lowlands. Within the lowlands, there are still fewer
commons in the Midland counties than in the south
and east, or in the west country. In lowland areas,
most commons are now principally valued for their
amenity value, or for their role as nature reserves or
Sites of Special Scientific Interest. Common land
now survives as such only where it has been
registered under the terms of the Commons
Registration Act of 1965. All existing common land
is listed in the final register, which was closed on 1
August 1972.
See Also
► Common Property Rights
Bibliography
Campbell, I., and P. Clayden. 1980. T h e l a w o f
c o m m o n s a n d v i l l a g e g r e e n s . Henley-on-
Thames: Open Spaces Society.
Hoskins, W.G., andL. Dudley Stamp. 1967. T h e c o m m o n
l a n d s o f E n g l a n d a n d W a l e s . London: Collins.
Rackham, O. 1976. Trees and woodland in the British
landscape. London: Dent.
Common Law
Snell, K. 1985. Annals of the Labouring poor: Social change
and agrarian England 1660-1900. Cambridge: Cambridge
University Press.
Turner, M.E. 1980. English parliamentary enclosure.
Folkestone: William Dawson.
Wade Martins, P. 1975. The origins of rural settlement in East
Anglia. In Recent work in rural
a r c h a e o l o g y , ed. P.J. Fowler. Bradford on Avon:
Moonraker Press.
Common Law
P. S. Atiyah
Common law is a system of law and legal processes
which originated in England shortly after the
Norman Conquest and after several centuries of
continuous development was exported to the
English colonies, and so came to be the basis of the
law of the greater part of the United States, as well
as of Australia, New Zealand, most of Canada and
(to a lesser degree) also of India, Pakistan, Bangla
Desh and many parts of Africa. The chief
characteristic of the common law has always been
that its development has lain largely in the hands of
the judges, and that it has therefore grown and
changed incrementally, case by case, in the course
of actual litigation.
In modem times the term ‘common law’ is used
in a variety of senses. In the broadest sense, it
continues to be used to refer to the entire system of
law originating in England which now forms the
basis of the law in the greater part of the former
British Empire, often nowadays called the ‘common
law world’. In this sense the common law is often
contrasted with the ‘civil law’ which derives from
the law of ancient Rome, and today operates in most
of Western Europe, as well as in a number of other
countries (such as Japan and Egypt) which have
borrowed their law from European countries. One
of the chief characteristics of the modem civil law is
that it derives its authority from one or more basic
Codes of law; and it remains a principal distinction
between common law and civil law countries that
the former have not generally codified their law.
And even in
Common Law
common law jurisdictions (such as California, for
example) where there does today exist a kind of
common law Code, it differs fundamentally in
nature from the civil law Codes; in particular the
system of precedent, and the authority of the judges
to interpret and develop such common law Codes
are quite different from those recognised in civil law
countries.
The term ‘common law’ is also often used in
various narrower senses. In the most important of
these narrower senses, the common law is often
contrasted with legislation, so that the lawyer in a
common-law country still thinks of legislation as a
type of law different from the ‘common law’, which
is basically judge-made law. The term ‘common
law’ is sometimes used in yet a third relevant sense
in which it is distinguished from a body of law,
known technically as ‘Equity’ which was originally
supplementary to the common law, and was
developed in the separate Court of Chancery. Today
common law (in this narrow sense) and ‘Equity’ are
almost everywhere merged and administered by a
single set of courts.
The common law (in the first two senses iden-
tified above) has traditionally been associated with
the economics of the free market in at least two
different ways. First, there is a strain of thought,
represented in particular by Hayek ( ), which
seems to suggest that a system of
law, like the common law, which is largely judge-
made, is inherently more likely to favour and protect
individual freedoms, and among them (or
especially) economic freedoms. But this is an
implausible and indeed eccentric claim, which
seems to involve confusion of the first two senses of
the term of ‘common law’ referred to above.
Because most redistribution is accomplished in
modem democracies by legislative measures, it is
easy to assume that a legal system which owes little
to legislation will be more likely to recognize and
protect the freedom of the market, but the amount of
redistribution which occurs in a legal system does
not necessarily depend upon whether that society is
part of the common law world. There is no a priori
reason to suppose that judges left to themselves by a
legislature will necessarily favour the economics of
the market. In the last analysis, the policies favoured
by judges will
1863
depend upon their own preferences, their culture
and traditions.
But there is a second way in which the common
law has traditionally been associated with the
freedom of the market, and this association rests
upon the historical facts of the last three centuries.
The concept of the Rule of Law which came to be
recognized and defended in England after the
revolution of 1688 has been seen by many as having
favoured the development of a free market economy
in England prior to and during the early years of the
industrial revolution. Because of the historical fact
it was for a long time almost an article of faith
among English writers that the common law and the
freedom of the market were closely associated. This
view is today less strongly held in England, as a
result no doubt of the fact that, while Englishmen
still like to believe in the Rule of Law (despite
grave doubts in some quarters as to whether this
concept has much meaning), they are by no means
so wedded to the free market as they were. In
America, where the Constitution of 1788
substantially embodied the English traditions as to
the Rule of Law, as well as the then accepted
ideology of the free market, the association between
the two has survived rather more strongly.
The reasons for the traditional belief in the close
association between the common law and the
freedom of the market must therefore be sought in
history, and in particular in English history during
the period from approximately 1770 to 1870, when
the free market economy was largely in process of
being estabhshed. And of all parts of the common
law, none was more important for this purpose than
the law of contract, because this was the part of the
law most intimately related to the economic system.
Indeed, the story of English law between 1770 and
1870 was to a large degree the story of how the law
of contract was converted into the law of the free
market, and of how the ideology of freedom of
contract became one of the great intellectual
movements of history (Atiyah ).
The first three-quarters of the 18th century was
a period of transition in England, during which
many older ideas about contract and the market
were being displaced by the newer ideas which
1864
gradually became dominant towards the end of the
century. Among the older ideas at least three can be
identified as particularly hostile to the laws needed
to serve the emerging free market economy. First,
there was a regulatory element in the law and the
economy dating back to Tudor times, represented
for instance by statutory controls of wages and
prices of many commodities, and by the
apprenticeship laws which controlled entry to many
trades with outdated and largely unnecessary
restrictions. Secondly, there was a paternalistic
element in much contract law at this time, with the
courts still being willing to relieve various classes
of persons from the consequences of bad bargains
which they had made. This paternalism was
particularly pronounced under various doctrines of
Equity, such as rules for the relief of mortgagors,
rules against the enforcement of contractual
penalties and forfeitures, rules for the protection of
seamen and ‘expectant heirs’, and so forth. Thus, in
the third sense of the term the ‘common law’
identified above, it can be said that the common law
was always more market- oriented than Equity.
Thirdly, there was a traditional moralistic element
in the contract law of the 18th century, and this also
took different forms, such as the general hostility to
usury (as to which see Simpson , pp. 510—18), and
the attempts to regulate the way in which essential
foods and drinks were sold by use of the traditional
marketing offences. The ‘moral’ roots of older law
were also related to ideas about ‘just prices’ which,
though rarely openly recognized in the common
law, seem to have been influential at least in some
of the cases in Equity, where there are signs that the
Chancellors did have some vague sense of unease if
they were asked to enforce contracts at prices which
seemed to them very unfair, or on terms which were
(in the language of the law) ‘unconscionable’.
In addition to these specific instances of inter-
ference with the binding force of private contracts,
there were important respects in which the whole
concept of a general contract law remained rela-
tively undeveloped at this time. Thus, while the law
recognized and enforced specific types of contracts,
such as contracts for the sale of land, contracts of
insurance and so forth, there was, as
Common Law
yet, little sign of a general law of contract,
governing all types of transaction. Then also, it
remains unclear how far the contract law of this
period actually recognized and enforced wholly
executory contracts, in the sense of awarding
damages for breach of a contract prior to any acts of
performance or detrimental reliance by any of the
parties. And finally, it is clear that, from the
standpoint of today, the law of contract in the 18th
century had not yet freed itself from dependence on
the law of property. Of course, in one sense contract
law can never be free from a dependence on
property entitlements, because contract law is the
mechanism by which entitlements are exchanged;
but there are clear signs in the 18th century that
contract law was still closely tied to property law in
another sense, in the sense (for instance ) that the
proprietary aspects of many transactions were still
regarded as more important than the promissory or
contractual aspects. So, for instance, the right of a
mortgagor to redeem the mortgaged property was
protected by the courts, even when by the terms of
the mortgage documents he had forfeited that right
by delay in repaying the loan. It was assumed that if
the mortgagee received back his money, with
interest and costs, he was adequately protected by
the law, even though the contract itself would have
given him more extensive rights.
Dining the century beginning around 1770 these
older ideas and traditions gradually gave way before
the ideology of freedom of contract; but it would be
wrong to think that this ideology did not have long
roots and antecedents in still earlier periods. There
are, even in the 16th and 17th centuries, many signs
of incipient economic liberalism among the lawyers
such as Coke, who bequeathed to the common law a
hatred of monopolies as well as a passion for
individual liberties (Wagner ). And Thomas
Hobbes, in a well-known passage in Leviathan, had
swept away all the medieval learning about ‘just
prices’ and declared that ‘[t]he value of all things
contracted for, is measured by the Appetite of the
Contractors; and therefore the just value, is that
which they be contended to give’ (Hobbes [ ]
1968, p. 208 ). So the ideology of freedom
of contract certainly had origins going back well
Common Law
beyond the 18th century. Nevertheless, it does seem
(though the matter remains controversial) that major
changes in the law began during the course of that
century which gathered pace as the century
progressed.
Certainly, a great deal occurred to change the
character of contract law from the last quarter of the
18th century until well into the 19th century, and
there is much evidence that many of these changes
in the law were profoundly influenced by classical
economic theory, and perhaps still more by popular
versions of classical economic theory. First, the
relics of the Tudor regulatory economy gradually
disappeared. Wage regulation had become
increasingly obsolete in practice during the 18th
century, and a major challenge to the older laws in
the name of freedom of contract had taken place in
the celebrated case of the Gloucestershire Weavers
(1756-7), (Atiyah , pp. 73-4). By the early 19th
century most of the legislation authorizing the
fixing of wages had been repealed. So too was the
Statute of Apprentices, after many years during
which its operation had been gradually whittled
down by the judges. Secondly, the signs of
paternalism which are still found in 18th-century
Equity seem to have disappeared gradually as the
judges hardened their hearts and toughened their
minds. For example, signs of an attempt to
introduce implied warranties on the sale of goods
for the protection of buyers, which can be detected
in the 18th century, were largely scotched, and the
principle of caveat emptor reasserted with full
vigour. The equitable doctrines allowing the courts
to relieve various unfortunates from the effects of
hard bargains were gradually whittled down,
although they never disappeared altogether. Third,
the moralistic elements in the law were also
gradually whittled down. The law of contract came
increasingly to be seen to be neutrally enforcing
agreements which must be presumed to be
beneficial to both parties. The only moral
component left in the law of contract during the
19th century seemed to derive from the binding
nature of promises.
The subjective theory of value also seems to
have been largely accepted by the judges even
before it had been wholly accepted by economists.
1865
Although the common law had always insisted that
a promise be supported by some ‘consideration’,
some reason, before it would be enforced (and to
that extent at least contained a paternalist element),
the growing acceptance of the subjective theory of
value meant that the doctrine of consideration
became much less important during the 19th
century. So far instance, in Haigh v. Brooks (1840,
113 English Reports 124) the judges enforced a
promise to pay £9000 in return for the giving up of a
guarantee previously given by the promisor, even
though it now appeared that the guarantee might be
unenforceable and legally worthless. The promisor
had valued it at £9000, said the judges; it was not
for them to say that the document was worthless.
For similar reasons, the prejudice against usury had
gradually been overcome, and the usury laws were
totally repealed in England in 1854.
In these ways, then, the principle that contracts
are binding and must be strictly enforced had been
greatly strengthened, and exceptional cases had
been whittled down by the middle of the 19th
century. In addition, other changes had occurred in
the general nature of contract law, which were
closely related to the growing trend to see contract
law as the law of the free market. First, it was
during this period that a general law of contract
came into existence for the first time in the common
law world. And the process of generalization was
important to the ideology of the law in a number of
respects. In particular, the generalizing of
contractual ideas meant that the law had to become
more abstract, more broadly principled. Principles
had to be developed which could be applied equally
to (say) commercial contracts for the sale of wheat,
to contracts of employment, and (for instance) to
personal contracts such as the contract to marry.
This abstraction may have helped the law become
more neutral, less inclined to pursue any
redistributive tendencies, such as may exist where
(say) there is a separate body of legal doctrine
dealing with contracts of employment, or with
residential leases, or with loan transactions.
Next, it seems clear that another major devel-
opment during this period was the gradual shift in
emphasis in contract law away from treating
1866
contracts as present, or partly performed exchanges,
and towards treating them as private planning
devices, made in advance to allocate risks. The
wholly executory contract became clearly
recognised by the law, so that it now became
possible for a person to sue for damages for breach
of a pure promise, even where no performance or
detrimental reliance had taken place. The
justification for requiring damages to be paid in
such circumstances was never clearly enunciated,
and indeed, specific justification was rarely seen to
be necessary. It was widely assumed that the broad
principle of freedom of contract required, not only
that parties be left free to make their own
exchanges, but that the law should be available in
aid of a party to enforce his claim to damages where
the other failed to perform. John Stuart Mill was the
first economist to point out that a policy of laissez-
faire could not be used to justify the enforcement of
executory contracts (Mill , vol. 2, p. 386), but even
modem economists do not generally pursue this line
of thought, though some libertarians have done so.
And finally, 19th-century contract law increas-
ingly freed itself from its dependence on property
law. Although obviously entitlements still remain
the subject matter of all contracts, contract law has
become much less concerned with specific items of
property, and is more concerned with wealth as a
kind of fungible property. The reason for this was
basically that 19th-century contract law was
dominated by the needs of merchants and traders, to
whom all property is in principle replaceable with
money. A merchant can be assumed to be
indifferent between a piece of property, and the
value of that property. Similarly, as contracts came
to be increasingly seen as fundamentally risk-
allocation devices, the particular entitlements or
property to which the risks attached became less
important.
By the last quarter of the 19th century, the
process of developing a mature body of general
contract law had largely been completed in England,
and although a similar process took place in
America (Horwitz ), there is ground
for believing that that was not completed for
Common Law
another fifty years or so. Freedom of contract had,
apparently, reached its highest point. But although
this was true of the ideology of freedom of contract
among lawyers and judges, it was not really true of
the views of economists or of the politicians, or of
the public. By the late 19th century, neoclassical
economists were already beginning to write
sceptically about the sweeping effects of freedom of
contract which had been attributed to the classical
economists, and were pointing out the many
possible causes of market failure such as
information difficulties, externalities and monopoly.
And although most of the older regulatory
legislation had been repealed in the first half of the
19th century, Parliament had at the same time been
gradually building up a completely new body of
regulatory enactments dealing with new industrial
problems - factories, coal mines, safety at sea for
seamen and emigrant passengers, public health, the
adulteration of food and drink, regulation of the
weights and measures used for sales, and so on.
Much of this new legislation had been a pragmatic
response to perceived evils, and though some of it
could have been justified economically by
arguments concerning misinformation or
externalities, much of it would have been difficult to
justify except on the assumption of paternalistic or
redistributive motives. Some of it may have been
inspired by sheer impatience, an unwillingness to
give the market time to work, or a belief that the
shortterm costs of market failures were so severe
that legislative correction was necessary without
regard to the long term distortions this might
produce.
What is quite clear is that by the time the
English common law and common lawyers had
accepted the teachings (as they were thought to be)
of the classical economists on freedom of contract,
these teachings were already somewhat out of date.
The result was that the mature common law of
contract was seriously deficient in a number of
respects. It was first of all deficient in its almost
total neglect of the problem of externalities.
Contracting parties were entitled to pursue their own
interests, regardless of the effect of their
Common Law
contract on third parties, or the public. Only in the
most extreme cases of actual illegality would the
courts generally refuse to uphold a contract. Sec-
ondly (although this certainly could not be laid at
the door of the classical economists), there had
been, during the 19th century, a serious neglect by
common lawyers of the problem of monopoly. This
may well have been largely due to the fact that for
the greater part of this period the British economy
was itself highly competitive, and in little danger
from monopolies. But the complacent assumption
that cartels were unstable and were always
vulnerable to internal or external competition was in
England (though not in America) carried over by
lawyers and courts into new conditions towards the
end of the nineteenth century, and well into the
present century, when it was utterly out of date. A
second result of this failure of the common law to
keep pace with economic theory and political
reality, was the growing gulf between the common
law and legislation. Once again, extensive
legislative intervention with freedom of contract
began to become commonplace, and much of it was
increasingly redistributive in character.
During the course of the present century this
process continued at an increasing pace unti 11980
or thereabouts, since when there are signs that
history has virtually reversed itself. Disillusion with
the free market, particularly in England, increased
during the great depression in the 1930s until, by
the end of World War II, a Labour Government was
elected to power with a massive majority and with a
mandate to lay the foundations for a socialist state
and a socialist economic system. Since then
England has increasingly learned to live with a
‘mixed economy’, to a large part of which the
traditional law of contract seems irrelevant because
the public sector is often controlled by public laws
rather than by contract law. But even in areas where
private law continues to operate, the common law
of contract has become increasingly affected by
legislative intervention. Virtually all types of
consumer transactions are today controlled or
affected to some degree by legislation, including
consumer credit
1867
contracts, contracts of employment, residential
leases, and insurance contracts. Unconscionable, or
unfair contracts are increasingly subjected to
judicial control. Many areas of law which were
formerly controlled largely by contract, such as
family law, are now subject to extensive judicial
discretionary control. Even business and
commercial contracts are subject to vast bodies of
legislative and regulatory laws, some, such as the
modem monopoly anti-trust laws, being designed to
preserve the operation of a competitive market, but
much of it still being designed to restrict
competition or the operation of the free market.
America has not gone so far down this road as
Britain and other common-law countries, and
indeed, for a long time, in the late 19th and early
20th centuries, constitutional decisions of the United
States Supreme Court in the name of freedom of
contract, actually prevented similar developments.
Much legislative intervention with freedom of
contract was, during this period, declared
unconstitutional, frequently over the dissent of
Justice Holmes. By the late 1930s, however, the
majority of the court had largely accepted Holmes's
view, and since then, legislative intervention with
freedom of contract has not been regarded as per se
unconstitutional. This shift in the court opened the
door to the same kind of regulation and intervention
which had already been taking place in Britain, and
although America has not, like Britain, brought
large-scale industries within the public sector and
therefore partially outside the control of contract
law, most of the other legislative developments of
the British type certainly have their parallel in
America. No doubt some contracts are more
regulated in Britain, but conversely there are plenty
of examples of legislative interference with freedom
of contract in America which are not to be found in
Britain.
These vast changes in the operation of the
common law have accompanied or brought with
them a change in ideology once again. Paternalism
and redistribution were, at least until around 1980,
increasingly favoured by many writers and
1868 Common Property Resources

teachers of contract law, as well as large sectors of many years to come British or American courts will
the electorate. Even the judges became much more be enforcing contracts according to the full rigour of
sympathetic to arguments based on concepts like the common law.
unconscionability and inequality of bargaining
power. In America, unconscionability was given
express legitimacy as a device for overturning unfair
contracts by the Uniform Commercial Code, and See Also
was increasingly used by the judges as a matter of
common law as well. Many relationships of a
contractual character (for instance, that of physician
and patient) and others of a virtually contractual
character (for instance, that between manufactures Bibliography
of products and ultimate purchasers and consumers)
Atiyah, P.S. 1979. T h e r i s e a n d f a l l o f f r e e d o m o f
are, both in American and Britain, increasingly c o n t r a c t . Oxford: Oxford University Press.
regulated by tort law rather than contract law, at Hayek, F.A. 1973. L a w , l e g i s l a t i o n a n d l i b e r t y ,
least where things go badly wrong and legal actions Rules and Orders, vol. 1. London: Routledge & Kegan
Paul.
for damages are brought based on negligent
Hobbes, T. 1651. In L e v i a t h a n , ed. C.B. Macpherson.
conduct, or on defects in the goods. In such Harmondsworth: Penguin Books. 1968.
malpractice or products liability actions the Horwitz, M.J. 1977. T h e transformation of
appropriate standards of care or quality are set by American Law 1780-1860. Cambridge,
MA/London: Harvard University Press.
judges and juries and not by the contracting parties,
Mill, J.S. 1848. P r i n c i p l e s o f p o l i t i c a l e c o n o m y .
and contractual exculpatory clauses are often denied From the fifth London edition. New York: D. Appleton &
legal validity. Co., 1908.
Since about 1980 there have been increasing Simpson, A.W.B. 1975. A h i s t o r y o f t h e c o m m o n
l a w o f c o n t r a c t . Oxford: Clarendon.
signs that the tide has turned yet again, both in
Wagner, D.O. 1935. Coke and the rise of economic liberalism.
Britain and America. Obviously, and visibly, British E c o n o m i c H i s t o r y R e v i e w 6(1): 30.
and American governments have since then been
trying to reassert the virtues of the free market and
roll back the frontiers of regulation, and in this they
are being vigorously supported by some lawyers and
Common Property Resources
law teachers in America, though not to any real
extent in Britain. It is not yet clear what the impact Jean-Philippe Platteau
of this is going to be on the future of the common
law of contract. One possible scenario is that, as in
the late 19th century, the courts will be behind the
times, but that on this occasion they will be hostile
Keywords
to the reasserted belief in the free market and will
continue to defend paternalist and redistributive Common property resources; Efficiency; Market
intervention in free contracts, particularly where one integration; Open access; Private property
of the parties to the contract is a consumer or ‘small rights; Tragedy of the commons
man’ thought to be weak in bargaining power. But
another possible scenario is that the new enthusiasm
for the free market will prove but a short-lived JEL Classifications
01
hiccup in the long-term trend towards paternalist
and redistributive policies. In either event it seems The concept of common property has become
unlikely that for famous in economics since Garett Hardin ( )
wrote his celebrated article on ‘The Tragedy ofthe
Common Property Resources
Commons’. In this article, common property is
taken to mean the absence of property rights in a
resource, or what is equivalently known as a regime
of ‘open access’. Under such a regime, where a
right of inclusion is granted to anyone who wants to
use the resource, Hardin argued, inefficiency
inevitably arises in the form of over- exploitation of
the resource accompanied by an over-application of
the variable inputs. Open access leads to efficiency
losses because ‘the average product of the variable
input, not its marginal product, is equated to the
input’s rental rate when access is free and the
number of exploiters is large’ (Comes and Sandler ,
p. 787). The root of the problem lies in the fact that
the average product rule does not enable the users to
internalize the external cost which their decisions
impose on the users already operating in the
resource domain. Of course, the efficiency losses
are conceivable only in a world of resource scarcity,
implying that the variable input is subject to
decreasing returns. Such losses are considerable
since they amount to the dissipation of the whole
resource rent. Here is the crucial intuition behind
the open access regime: when no property right is
attached to a resource, the value of this resource is
zero in spite of its scarcity.
Efficiency losses are to be measured not only in
static but also in dynamic terms. Indeed, in an open
access regime resource users are induced to
compare average instantaneous returns with the
input’s rental price even though they may well be
aware that they thereby contribute to reducing the
future stock of the resource. The problem is simply
that they are forced to follow a myopic rule because
there is no way in which they can reap the future
benefits of restraint in the present. Thus, for
example, by refraining today from catching juvenile
fish or from cutting down saplings in the forest, a
villager can receive no assurance that he or she will
be able in the next period to catch mature fish or to
fell fully grown trees.
The main criticism levelled by numerous social
scientists against the concept of open access is that
the corresponding regime is rarely encountered on
the ground. The typical regime, according to these
1869
critiques, is one under which a community pos-
sesses a collective ownership right over local
natural resources. Under common property,
therefore, a right of exclusion is assigned to a well-
defined user group, and Hardin has created a lot of
confusion by using the word ‘commons’ to refer to
the alternative situation where no such right is
granted to any agency. What is not always clear,
however, is whether the ownership right involves
only the ability to specify the rightful claimants to
the resource, or whether it also involves the ability
to define and enforce rales of use regarding that
resource (for example, regulations about the
harvesting season and production tools, allowed
quotas of harvestable products of the resource, or
taxes). Baland and Platteau ( ) have coined the
term ‘unregulated common property’ to refer to the
former situation, while the term ‘regulated common
property’ is used for the latter.
Two polar situations can be considered on the
basis of this analytically important distinction
between two types of common property regimes. At
one extreme, if common property is perfectly
regulated, in the sense that the rules of use designed
and enforced by the owner community allow a
perfect internalization of the externalities, common
property becomes equivalent to private property
with a sole owner from an efficiency standpoint.
This illustrates the general result that, absent trans-
action costs, institutions do not matter. At the other
extreme, a strictly unregulated common property in
the above sense implies that, as the number of users
becomes quite large, over-exploitation of the
resource becomes as important as under the open
access regime: the rent attached to the resource is
totally dissipated (see Platteau , ch. 3).
Between these two extremes we find the situa-
tions most typically observed on the ground and
described in the numerous field studies devoted to
this topic (see Ostrom ; Baland and Platteau , for a
review of such studies). In such instances, rales
ofuse exist alongside membership rales, yet they
tend to be imperfectly designed and imperfectly
enforced by the village community. One key reason
for these imperfections is the governance costs that
unavoidably plague any collective decision-making
process. Governance
1870
costs include all those costs incurred to reach a
collective agreement and to organize a community
of users. They are likely to be higher when the
group is larger and when its membership is more
heterogeneous (whether measured in terms of
diversity of objectives or of wealth inequality).
Moreover, governance costs are enhanced by the
opportunistic tendencies of rights-holders not only
to violate or circumvent collective rules but also to
eschew efforts to create collective mechanisms of
decision-making and enforcement. Costs arising
from these proclivities are also dependent on the
size of the user group: they are lower if the number
of resource users is smaller and, at the limit, they
are nil when there is a single user.
As a consequence of the aforementioned limi-
tations, resources are less efficiently managed under
a common property regime than they could be under
a private ownership system. This is especially true
if, owing to their scarcity, the resources carry high
values which should be reflected in high rents.
Population growth and market integration are thus
two forces that tend to increase the monetary value
of the efficiency losses arising from common
property, that is, the forgone rents. This, at least, is
the conclusion drawn by the so-called property
rights school of Chicago economists (see, for
example, Demsetz ; Barzel ). The advantages of
private property appear all the more decisive as
such a regime enables users to internalize
externalities without incurring any governance
costs. This is because it establishes a one-to-one
relationship between individual actions and all their
effects: ‘A primary function of property rights is
that of guiding incentives to achieve a greater
internalization of externalities ...’ (Demsetz , p.
348).
Nevertheless, this ignores the costs of
privatizing natural resources, which involve both
directs costs and opportunity costs. Direct costs
comprise transaction costs, such as the costs of
negotiating, defining and enforcing private property
rights. The usual argument is that such costs
increase with the physical base of the resource.
Thus, the wider the resource base (or the less
concentrated the resource) the higher
Common Property Resources
are the costs of delimiting and defending the
resource ‘territory’ (Dasgupta , pp. 288-9). For
many natural resources, the costs of dividing the
resource domain appear prohibitive under the
present state of technology. For example, the open
sea - or, more exactly, the fish stock contained in it -
presents insuperable difficulties for private
appropriation. The enforcement of exclusive
property rights to individual patches of the ocean
would, indeed, be infinitely costly. This is
especially evident when fish species are mobile and
move within wide water spaces, since exclusive
rights are too costly to establish and enforce
whether over the resource or over the territory in
which the resource moves.
The opportunity costs of privatization, for their
part, correspond to the benefits that are lost when
the common property regime is abandoned. Here,
we can think of scale economies that may be present
not only in the resource itself but also in
complementary factors. The obvious advantage of
coordinating the herding of animals so as to
economize on shepherd labour in extensive grazing
activities is probably the best illustration of the way
scale economies in a complementary factor may
prevent the division of a resource domain. Another
important category of opportunity costs is the
insurance benefits associated with common
property. When returns to a resource are highly
variable across time and space, the need to insure
against such variability is yet another consideration
that may militate against resource division. When a
resource has a low predictability (that is, when the
variance in its value per unit of time per unit area is
high), users are generally reluctant to divide it into
smaller portions because they would thereby lose
the insurance benefits provided by keeping the
resource whole.
For instance, herders (fishermen) may need to
have access to a wide portfolio of pasture lands
(fishing spots) in so far as, at any given time, wide
spatial variations in yields result from climatic or
other environmental factors. On the assumption that
the probability distributions are not correlated too
much across spatial groupings of land or water
Common Property Rights
and that they are not overly correlated over time, a
system offering access to a large area within which
right-holding users can freely move appears highly
desirable from a risk-reducing perspective.
The conclusion of the above discussion is,
therefore, that the balance of the advantages and
disadvantages of various property regimes is a priori
undetermined. Economic theory, however, does
provide useful guidance about which circumstances
are more favourable to the persistence of common
property or, conversely, to its demise and
replacement by private property. Furthermore,
instead of being fixed once for all, the balance sheet
is susceptible to evolution depending on the
transformation of the parameters on which the
benefits and costs of privatization depend. Thus, the
direct costs of resource division may fall with
technological progress. For example, the
introduction of modem borehole drilling facilitates
the privatization of common grazing areas (Peters ).
ft is therefore not only the factors which enhance
resource value but also those which reduce the
direct costs of partitioning that may favour the
private appropriation of natural resources.
See Also
► A "/-',ss to T eve opment
► Agriculture and Economic Development
► Land lArkets
► Population and Agricultural Growth
► rtoperty Lav/, 3"::.c::.:.:'.'c • \
► Tragedy of the Commons
Bibliography
Baland, I, and J.-R Platteau. 1996. Halting degradation of
natural resources: Is there a role for rural communities?
Oxford: Clarendon Press.
Barzel, Y. 1989. E c o n o m i c a n a l y s i s o f p r o p e r t y
r i g h t s . Cambridge: Cambridge University Press.
Comes, R., and T. Sandler. 1983. On commons and tragedies.
A m e r i c a n E c o n o m i c R e v i e w 83: 787-792.
Dasgupta, P. 1993. An inquiry into well-being and destitution.
Oxford: Clarendon Press.
1871
Demsetz, H. 1967. Toward a theory of property rights.
A m e r i c a n E c o n o m i c R e v i e w 57: 347-359.
Hardin, G. 1968. The tragedy of the commons. S c i e n c e
162: 1243-1248.
Ostrorn, E. 1990. Governing the commons: The evolution of
institutions for collective action. Cambridge: Cambridge
University Press.
Peters, P. 1994. D i v i d i n g t h e c o m m o n s : P o l i t i c s ,
policy, and culture in Botswana.
Charlottesville/London: University Press of Virginia.
Platteau, J.-P. 2000. I n s t i t u t i o n s , s o c i a l n o r m s
a n d e c o n o m i c d e v e l o p m e n t . London: Harwood
Academic Publishers.
Common Property Rights
Steven N. S. Cheung
In a society where individuals compete for the use
of scarce resources, some mles or criteria of com-
petition must exist to resolve the conflict. These
mles, known as property rights, may be established
in law, in regulation, in custom or in hierarchy
ranking. The structures of rights may take a variety
of forms, ranging from private property rights at
one extreme to common property rights at the other.
Most fall somewhere in between: either set of rights
would be rare in its purest form.
In a private property, the delimitation of the
right to its use is expressed in dimensions or
characteristics inherent in the property itself. These
rights are exclusive to some private party, are freely
transferable, and the income derived from them is
not attenuated, restrained or infringed by laws or
regulations. Hence price control, taxation, and
social restriction of transferability may be regarded
as violations of private property rights. In a
common property, there is no delimitation or
delineation of its use rights to any private party. No
one has the right to exclude others from using it,
and all are free to compete for its use. Hence there
are no exclusive use rights, no rights to be
transferred, and in the limiting case, no
1872
net income can be derived from using the common
property.
This last condition rests on an economic prop-
osition known as the dissipation of rent. It argues
that because of the lack of exclusive use rights,
individuals competing for the use of a common
property will reduce its rental value or net worth to
zero. The reason is that if no one has an exclusive
claim to the value (i.e. rent) ofthat property, its use
will invite competition to the point that each and
every competing user can earn no more than the
alternative earning of his own resources required in
the exploitation of that common property. In other
words, under competition and with no one having a
special advantage, a ‘prize’ that has no exclusive
claimant will be dissipated or absorbed by the costs
of other resources which must be dedicated to its
winning. Hence the net value of the prize won is
zero.
The usual examples of common property rights
cite a public beach and marine fisheries, and the
dissipation of rent typically implies excessive use or
over-exploitation. However, the dissipation may
take the form of under- exploitation. For example, a
piece of fertile land under common ownership may
be used for herd grazing, or left idle, instead of
being planted as an orchard.
In the real world, the complete dissipation of
rent is rare indeed. This is because the supply curve
of labour or of other inputs may be rising (some
intramarginal rent may be captured), the competing
users may have different opportunity costs (the non-
marginal users may be enjoying rent), or entry may
be restricted by regulations, by customs or by
information costs. Still, with common property
rights some dissipation of rent is inevitable, and no
society can afford to surrender a large portion of its
valuable resources to this structure of property
rights.
A property may be held in common because its
capUirable rent is lower than the cost of enforcing
exclusivity. In this case, the dissipation of rent is no
waste. However, to the extent that rent dissipation is
viewed as a waste, its occurrence must be
attributable to the omission of some constraints in
Common Property Rights
the analysis. Attempts to reduce rent dissipation go
far to explain why common property in its ‘pure’
form is seldom observed. In marine fisheries, for
example, numerous regulations govern the fishing
season, the size offish caught, the boat size and the
mesh size, and various licensing arrangements
restrict the number of boats and fishermen. The
market value of a fishing licence, sometimes
enormous, is one measiue of the ocean rent cap-
tured. Even for public beaches, regulations of some
type will often be found to govern the use of those
most in demand.
Whereas regulations and restrictions on entry in
the use of a common property often serve to reduce
dissipation, the rent that can be captured is usually
less than if the property were privately owned. To
reconcile this observation with constrained
maximization, wc must infer that, enforcement
costs aside, other transaction costs associated with
the changing of institutional arrangements must
restrain the formation of private property rights.
No economy can survive if the majority of its
scarce resources arc commonly owned. Regulations
may, indeed, reduce rent dissipation; but in the
process they not only distort the use of the resources
but also invite corruption and the emergence of
special interests. An unrestrained common property,
strictly speaking, is propertyless in ownership; if its
structure is extended to all resources, starvation for
all must result. If one rules out private property
rights, then to avoid the imposition of an infinite
array of regulations the remaining alternative is the
communal system or the communist state.
In a communist state there is no private owner
of productive resources: each constituent is prop-
ertyless, in the literal sense of the word. Since the
dissipation of rents associated with common prop-
erties will guarantee starvation, in a communist
state the rights to use resources, and to derive
income therefrom, arc defined in terms of rank.
That is, stripped of all ownership rights over valu-
able and productive resources, the citizens of a
communist state hold differing rights to use
resources and to obtain income according to
Common Property Rights
their status. In the people’s communes in China
under the Great Leap Forward, for example, no one
owns the productive resources (i.e. everyone is
property less), but comrades of different ranks
enjoy different rights and privileges. ‘Rank’ as such
has value and is subject to competition, therefore a
system of ‘property’ rights is implicit. However, the
valuable rights are now defined in terms other than
the inherent properties of the productive resources.
This is, in fact, the key distinction between a
private property system and a communist state: the
former delineates rights in terms of certain
dimensions of the productive resources themselves;
the latter delineates rights in terms of a
characteristic (rank) of people deprived of produc-
tive human capital. In the communist state, the
competition for and protection of rank will draw on
the use of valuable and productive resources
(another form of rent dissipation). Moreover, the
lack of market prices increases the cost of infor-
mation, and the lack of contractual choices
increases the cost of enforcing performance. What
is saved in return are the costs of delineating and
enforcing rights in properties.
It is among these varied costs - broadly defined
as transaction costs - that we find the key
divergence in economic performance between the
communist and the private property systems. If one
ignores transaction costs, the delineation of rights in
terms of rank will produce the same use of
resources as would the delineation of rights in
properties. However, it can be convincingly argued
that the broadly defined transaction costs are
generally higher with communal than with private
rights. Communism fails, not because it does not
work in theory, but precisely because in practice its
costs of transaction are higher than those in a
system of private property rights. Still, the
delineation of rights in ranks is a way to reduce rent
dissipation in a propertyless state.
Strictly speaking, the dissipation of rent asso-
ciated with common property is no ‘theory’ at all,
because dissipating rent merely to produce an
equilibrium does not explain behaviour. Worse,
1873
to stand aside and simply permit rent to dissipate is
inconsistent with the postulate of constrained
maximization.
What is useful and important from the stand-
point of economic explanation is to view whatever
rent dissipation does occur as necessarily a
constrained minimum because, under the max-
imization postulate, each and every individual has
an incentive to reduce that dissipation. Behaviour
associated with the dissipation of rent must
therefore be regarded as attempts to reduce that loss,
and this altered view explains many observations.
That some dissipation remains must then be
attributable to the constraints of transaction costs.
The challenge to the economist is to specify and
identify what these costs are and how they will vary
under differing circumstances.
See Also
►
References
Alchian, A.A. 1965. Some economics of property rights. I I
P o l i t i c o 30(4): 816-829.
Bottomley, A. 1963. The effect of the common ownership of
land upon resource allocation in Tripolitania. L a n d
E c o n o m i c s 39: 91-95.
Cheung, S.N.S. 1970. The structure of a contract and the
theory of a non-exclusive resource. J o u r n a l o f L a w
and
E c o n o m i c s 13: 49-70.
Cheung, S.N.S. 1974. A theory of price control. J o u r n a l o f
L a w a n d E c o n o m i c s 17: 53-71.
Cheung, S.N.S. 1982. W i l l C h i n a g o ‘ c a p i t a l i s t ’ ?
Hobart paper 94. London: IEA.
Coase, R.H. 1960. The problem of social cost. J o u r n a l o f
L a w a n d E c o n o m i c s 3: 1-44.
Demsetz, H. 1964. The exchange and enforcement of property
rights. J o u r n a l o f L a w a n d E c o n o m i c s 7: 11-
26.
Gordon, H.S. 1954. The economic theory of a common
property resource: The fishery. J o u r n a l of
P o l i t i c a l E c o n o m y 62: 124-142.
Knight, F.H. 1924. Some fallacies in the interpretation of
social cost. Q u a r t e r l y J o u r n a l o f E c o n o m i c s
38: 582-606.
1874
Common Rights in Land
Leigh Shaw-Taylor
Abstract
‘Common rights’ and ‘common land’ refer to
rights to use land in common in some way. Of
several forms of common rights in pre-industria
l Europe and elsewhere, only one - free access to
land - involved what economists commonly
think of as common rights. Common rights in
Europe were largely swept away during the 18th
and 19th centuries by a process termed ‘enclo-
sure’. Some economic historians have
reconsidered the inefficiency of open fields in an
English context, but at present the data are too
poor to allow a plausible rebuttal of the views of
18th-century critics of the open fields.
Keywords
Access to land; Common fields; Common land;
Common property resources; Common rights;
Common rights in land; Common-pool
resources; Enclosure; Open-access resources;
Tragedy of the commons
JEL Classifications
N5
Common rights are rights to use land in common.
The most important of these rights was the right to
graze livestock on common grassland. But rights to
gather fuel (wood, peat, gorse and turves),
fertilizers, timber for building and other natural
resources were also important. Common land is
land used by a number of distinct individuals or
households whose rights over the land are known as
common rights.
Today we are accustomed to think of land as
private property with a clear owner and possibly a
tenant. Although in some countries there may be
legal rights of public access to certain types of wild
or agricultural land, it is generally the case that the
owner or tenant of the land has exclusive rights to
use the land and, within the limits of
Common Rights in Land
planning or zoning laws, may use it as he or she
wishes. But in Europe, for at least a thousand years
and ending only in the 19th century, a high
proportion of land was ‘common land’ which many
individuals were entitled to use for a variety of
purposes.
It cannot be overemphasized that common land
was generally not open-access land - land which
anyone could use. There were regulations governing
who could use the land, what they could use it for
and how much they could use it. When economists
think of common land and common rights they may
have Garett Hardin’s ‘tragedy of the commons in
mind’ (Hardin ).
The principal subject of Hardin’s article was in fact
population growth, not historical common land or
common rights.
However, Hardin used a theoretical common
land system as a model for the exploitation of open-
access resources. In this system each herder could
put as many animals as he wished on to the
common pastures. Hardin argued that individual
herders would choose to graze more and more
animals on the common, thus inevitably leading to
over-grazing and degradation of the resource. This
model offers important insights into the destruction
of, or damage to, unregulated open- access
resources such as the atmosphere or fish stocks in
the oceans. If common land and common rights had
operated in this manner, it is unlikely that they
would have remained a key part of European
agriculture for so many centuries.
In the rest of this article the following questions
are addressed: what were common rights? What
was common land? Who had common rights? How
was common land regulated? Was it efficient? How
and why did it come to an end and with what
consequences? The answers to these questions
varied from one village to another across Europe
and what follows is necessarily highly simplified
(see de Moor et al. , for a more detailed overview).
Common Land
The types of common land and the terminology
used to describe such land varied across Europe.
Common Rights in Land 1875

Nevertheless, four major types of common land owner’s tenant. Access to the common rights was
may be distinguished. First, the archetypical form of considerably more complex and took different
common land and the one with the widest forms in different places; but it is possible to dis-
geographical distribution is variously referred to as tinguish four main forms of access. First, in
common waste, common pasture, waste, or England and some parts of the Continent, the own-
common. This land was permanently common and ership or tenancy of particular buildings or land-
most often grassland used for grazing animals. holdings was a prerequisite. Second, in many parts
Usually such land was not suitable for arable of the Continent citizenship of (as distinct from
cultivation typically because its natural fertility was residence in) a commune or a municipality which
low but sometimes for other reasons such a itself owned the common resource was necessary,
propensity to seasonal flooding. On some common sometimes in combination with a property qualifi-
wastes other resources were available, such as peat, cation. Third, in other parts of the Continent mem-
turf, gorse or wood. bership of a cooperative association which owned
Second, in many parts of Europe much of the the common resources was necessary. Membership
arable land (the land on which crops were grown) of these institutions was sometimes inherited, but
was also subject to common rights. Such land, sometimes it was attached to buildings or land (as in
known as open- fields, common fields, or common the first case). Fourth, there were cases were all
arable, was privately owned and cultivated but residents in an area had common rights. But outside
subject to common grazing. In its classic form each laigely uninhabited areas, such as northern Sweden,
farmer held a number of long thin strips of land this situation was unusual.
scattered over an extensive area and intermixed In consequence by no means all individuals or
with the strips of other farmers. Each farmer households enjoyed common rights. The proportion
cultivated his own crops on the arable. But when of the population that enjoyed common rights
the harvest was over, or in years when the land was varied considerably from one region to another and
being fallowed, all those with common rights could changed over time. Where individuals or
turn their livestock into the fields to graze. Thus the households did have common rights, the kinds and
open fields alternated between private and common levels of the rights they enjoyed were determined by
land over the course of the agricultural cycle. local regulations.
Third, common woodland for the production of
fuel and timber was widespread on the European
continent but unusual in England. This was similar Regulation
to common waste in that it was permanently in
common use. Common land was almost invariably regulated by
Fourth, common meadows, which were per- local institutions, often at the level of the individual
manent grasslands for the production of hay, were village or manor. The institutions varied but were
divided into separate blocks in private use but afterusually manor or village courts or village
the hay had been harvested were open to common assemblies or committees of some kind, with the
grazing. Thus, like the open fields, common decisions made by a group of jurors. These insti-
meadows alternated between private land and tutions normally issued sets of mles, ordinances or
common land over the agricultural cycle. by-laws which governed the usage of the commons
and set fines for the infringement of rales. Officials
or monitors were appointed to police the by-laws.
Common Rights The degree to which these institutions and their by-
laws were subject to the influence of feudal
As private property, the right to cultivate the com- overlords and the state varied considerably across
mon arable or to harvest the hay in common Europe.
meadows lay with the owner of the land or the The by-laws provided the basic regulatory
framework for managing the commons (for
1876
examples of by-laws see Ault ). Their most critical
function was to restrict the usage of common land
and thus prevent a ‘tragedy of the commons’
developing. This was done in two ways. First, the
by-laws would normally serve to restrict common
rights to well-defined groups of users. For example,
in much of England only those holding land in the
open fields or with certain recognized dwellings,
known as common-right houses, were allowed to
pasture animals in the open fields or on the common
pasture, while on much of the Continent pasture
rights were restricted to citizens of communes or the
holders of ancient farmsteads. Second, by-laws
defined the amount of resources to which each
commoner was entitled.
Thus, by-laws might specify the amount of peat
or wood each commoner was entitled to dig or cut
each year or the number and type of animals which
could be kept, and for which months of the year
they might be kept on the common pastures, open
fields and common meadows.
The number of animals each commoner could
put on the common land was generally controlled
by one of two types of rules. One, known as
‘stinting’, simply specified the number and type of
animals (the stint) which each commoner might
keep on the common. Often the stint was
proportional to the area or the value of land held.
The other form of access, known as ‘levancy’ and
‘couchancy’, stated merely that each commoner
could keep as many animals on the common as he
could overwinter (that is, feed when the common
was closed) on his own holding. How this was
policed in practice is a moot point, but it certainly
served to limit numbers and may have differed little
from stinting in practice.
One consequence of these types of rule is that
some individuals had no common rights at all.
Another is that different individuals who did have
common rights could have very different levels of
access. The situation varied too much to allow
generalization, beyond the suggestion that the level
of inequality in England was probably greater and
had proceeded further at an early date than
anywhere else.
Common Rights in Land
Enclosure
The process by which common land and common
rights were abolished and replaced by recognizably
modem forms of private property was part and
parcel of a broader reform of landholding known as
‘enclosure’ which could also entail the
consolidation of scattered holdings and the whole-
sale reallocation of land to create ring-fenced farms.
Enclosure in some form is probably as old as
common land itself. In England significant
enclosure took place in the medieval period and
from the 17th to the early 19th centuries. In most of
Europe the widespread attack on common land
began in the late 18th century in the wake of
Physiocratic critiques. The later Napoleonic reforms
and a subsequent series of state- sponsored drives to
modernize agriculture in the 19th century led to
more sustained enclosure. Some common land
survives to this day, generally in mountainous areas.
Efficiency
By the 18th century common rights and common
land were being widely criticized by agricultural
improvers and others for restricting agricultural
productivity. Most agricultural writers have
accepted this view of common land as inefficient,
and associated enclosure with major increases in
productivity (Ernie ; Chambers and Mingay ;
Overton ). Common rights and common land
imposed two kinds of limitation on agricultural
improvement. First, the communal regulation of
common land made it more difficult to introduce
new agricultural techniques and technologies or to
respond to changes in market opportunities. Second,
the sharing of the outputs from common land made
individual investment less attractive. The spread of
nitrogen fixing crops and new drainage
technologies, which often allowed the cultivation of
formerly uncultivable common land, together with
better transport links made enclosure a steadily
more pressing issue in the 18th and 19th centuries.
A number of economic historians have reconsidered
the inefficiency of open fields in an
Common Rights in Land
English context. McCloskey ( ) has argued
that the scattering of land in open fields in the
medieval period was an efficient insurance against
risk in a non-market economy. Allen ( ) has
argued that enclosure did facilitate major techno-
logical changes obstructed by common land but that
these innovations made only very marginal
contributions to increased efficiency. Clark ( ) See Also
has argued that the inefficiencies imposed by
common land were relatively modest and that, given
the costs involved, enclosure was not economic
until after 1750. However, the issue remains
controversial essentially because it is inherently
difficult to measure the agricultural productivity of
farming in the 18th and 19th centuries with any
degree of reliability. In other words, at present the
data are too poor to allow an entirely plausible
rebuttal of the views of 18th-century critics of the
open fields. Moreover, much enclosure took place
in the medieval period and in the 17th century
(Wordie ) and any fully satisfactory theory of the
efficiency or otherwise of open fields would need to
be able to account for the longer-term chronology of
enclosure. The persistence of open- field farming in
France has been investigated by Grantham ( ) and
Hoffman ( ).
Another controversial issue is the importance of
common land to the poor. Many historians have
argued that the poor derived considerable benefits
from common land and that enclosure was socially
damaging; but this remains controversial
(seeNeeson ; Shaw-Taylor . The extent to which the
poor benefited from common land and common
rights is hard to reconstruct, poorly understood, and
varied considerably across Europe.
Common-Pool Resources
This article has been concerned exclusively with
common land and common rights as they existed in
Europe before the 20th century. However, it should
be noted that while open fields and common
meadow may be peculiarly European forms,
common waste and institutions for its management
can be found all over the world. Analytically, these
systems are part of a larger family of
1877
‘common-pool-resource’ systems (Ostrom ) which
have been adopted in many parts of the world to
manage not just land but water resources and fish
stocks as well.
Bibliography
Allen, R. 1992. Enclosure and the Yeoman: The agricultural
development of the South Midlands, 1450-1850. Oxford:
Oxford University Press.
Ault, W. 1972. Open-field-farming in medieval England: A
study of village by-laws. London: George Allen and Unwin.
Chambers, J., and G. Mingay. 1966. The
agricultural revolution 1750-1880.
London: Batsford.
Clark, G. 1998. Commons sense: Common property rights,
efficiency, and institutional change. J o u r n a l o f
E c o n o m i c H i s t o r y 58: 73-102.
de Moor, M., L. Shaw-Taylor, andP. Warde, eds. 2002. T h e
management of common land in North
West Europe, c. 1 5 0 0 - 1 8 5 0 . Tumhout:
Brepols.
Ernie, Lord. 1936. E n g l i s h f a r m i n g p a s t a n d
p r e s e n t . 5th ed. London: Longmans, Green and Co..
Grantham, G. 1980. The persistence of open-field farming in
nineteenth century France. J o u r n a l o f E c o n o m i c
H i s t o r y 40: 515-531.
Hardin, G. 1968. The tragedy of the commons. S c i e n c e
162: 1243-1248.
Hoffman, P. 1989. Institutions and agriculture in old-regime
France. J o u r n a l o f I n s t i t u t i o n a l a n d T h e o -
r e t i c a l E c o n o m i c s 145: 166-181.
McCloskey, D. 1976. English open fields as behaviour towards
risk. R e s e a r c h i n E c o n o m i c H i s t o r y 1: 124-
170.
Neeson, J. 1993. Commoners, common-right, enclosure and
social change in England 1700-1820. Cambridge:
Cambridge University Press.
Ostrom, E. 1990. Governing the commons: The evolution of
institutions for collective action. Cambridge: Cambridge
University Press.
Overton, M. 1996. Agricultural revolution in England: The
transformation of the agrarian economy 1500-1850.
Cambridge: Cambridge University Press.
Shaw-Taylor, L. 2001. Parliamentary enclosure and the
emergence of an English agricultural proletariat. J o u r -
n a l o f E c o n o m i c H i s t o r y 61: 640-662.
Wordie, J. 1983. The chronology of English enclosure, 1500-
1914. E c o n o m i c H i s t o r y R e v i e w 36: 483-505.
1878
Commons, John Rogers (1862-1945)
Warren J. Samuels
Keywords
American Association for Labor Legislation;
American Economic Association; Capitalism;
Collective action; Commons, J. R.; Fisher, I,;
Institutional economics; Methodological indi-
vidualism; National Bureau of Economic
Research; New deal; Trade unions; Veblen, T
JEL Classifications
B31
Commons was bom on 13 October 1862 in
Hollandsburg, Ohio, and died on 11 May 1945 in
Raleigh, North Carolina. He studied at Oberlin
College (BA, 1888) and Johns Hopkins University
(1888-90). He taught at Wesleyan, Oberlin, Indiana,
Syracuse, and Wisconsin (1904-32).
The founder of the distinctive Wisconsin tradi-
tion of institutional economics, Commons derived
his theoretical insights (generalized in his Legal
Foundations of Capitalism, 1924, and Institutional
Economics, 1934) from his practical, historical and
empirical studies, particularly in the field of labour
relations and in various areas of social reform. He
drew insight not only from economics but also from
the fields of political science, law, sociology and
history. A principal adviser and architect of the
Wisconsin progressive movement under Robert M.
La Follette, Commons was active as an advisor to
both state and federal governments. He was
instrumental in drafting landmark legislation in the
fields of industrial relations, civil service, public
utility regulation, workmen’s compensation and
unemployment insurance. He served on federal and
state industrial commissions, was a founder of the
American Association for Labor Legislation, was
active in the National Civic Federation, National
Consumers’ League (president, 1923—35),
National Bureau of Economic
Commons, John Rogers (1862-1945)
Research (associate director, 1920-28), and the
American Economic Association (president, 1917).
He participated in antitrust litigation (especially the
Pittsburgh Plus case) and in movements for reform
of the monetaiy and banking system (often
associated with living Fisher, who considered
Commons one ofthe leading monetary economists
ofthe period).
The critical thread uniting Commons’s diverse
writings was the development of institutions,
especially within capitalism. He developed theories
of the evolution of capitalism and of institutional
change as a modifying force alleviating the major
defects of capitalism. Commons came to recognize
and stress that individual economic behaviour took
place within institutions, which he defined as
collective action in control, liberation, and
expansion of individual action. The traditional
methodologically individualist focus on individual
buying and selling was not capable, in his view, of
penetrating the forces, working rules and
institutions governing the structural features of the
economic system within which individuals
operated. Crucial to the evolution and operation of
the economic system was government, which was a
principal means through which collective action and
change were undertaken.
Commons rejected both classical harmonism
and radical revolutionism in favour of a conflict and
negotiational view of economic process. He
accepted the reality' of conflicting interests and
sought realistic, evolutionary modes of their atten-
uation and resolution. These modes focused on a
negotiational psychology in the context of a plu-
ralist structure of power. He sought to enlist the
open-minded and progressive leaders of business,
labour and government in arrangements through
which they could identify problems and design
solutions acceptable to all parties.
In other contexts, he sought to use government
as an agency for working out new arrangements to
solve problems, such as worker insecurity and
hardship, rather than promote systemic
restructuring, although to many conservatives his
ventures were radical enough. To these ends Com-
mons and small armies of associates engaged in fact
finding - his look-and-see methodology - in a spirit
of bringing all scientific knowledge to bear
Communications
on problem solving. From these experiences, indeed
already manifest in the underlying strategy,
Commons developed a theory of government as
alternately a mediator of conflicting interests and an
arena in which conflicting interests bargained over
their differences; a theory of the complex
organization - in terms of freedom, power and
coercion - and evolution of the legal foundations of
capitalism, which centred in part on the composing
of major structural conflicts through the mutual
accommodation of interests; and a theory of
institutions with an affirmative view of their roles in
organizing individual activity and resolving
conflict.
The institutions Commons studied most closely
were trade unions and government, particularly the
judiciary. He developed his theory of the economic
role of government in part on the basis of his study
of the efforts of workers to improve their market
position and in part on the use of government by
both enemies and friends of labour. Commons’s
was an interpretation of trade unions as a non-
revolutionary development, as collective action
seeking to do for workers what the organizations of
business attempted to do for their owners and
managers. His study of the reception given unions
and reform legislation led him to recognize the
critical role of the United States Supreme Court
(and the courts generally), and its conception of
what was reasonable in the development and
application of the working rules which governed the
acquisition and use of power in the market.
Accordingly, Commons developed a theory of
property which stressed its evolution and role in
governing the structure of participation and relative
withholding capacity in the market.
Commons also developed a theory of institutions
which focused on their respective different mixtures
of bargaining, rationing and managerial
transactions, all taking place within a legal frame-
work which was itself subject to change.
Although Commons’s institutionalism had dif-
ferent emphases from that of Thorstein Veblen, for
example, in that Commons stressed reform of the
capitalist framework, they shared a view of eco-
nomics as political economy and of the economy as
comprising more than the market. Unlike Veblen,
Commons was not antagonistic toward
1879
businessmen, and indeed accepted capitalism,
though not necessarily on the terms given or pre-
ferred by the established power structure.
Commons was one of the few American econ-
omists to found a ‘school’, a tradition that was
carried forward by a corps of students, especially
Selig Perlman, Edwin E. Witte, Martin Glaeser and
Kenneth Parsons. Much mid-20th-century American
social reform, the New Deal for example, drew on
or reflected the work of Commons and his fellow
workers and students.
Selected Works
1893. The distribution of wealth. New York:
Macmillan.
19 05 . Trade unionism and labor problems.
Bo s t o n: G i nn .
1910-11. (With others.) Documentary history of
American industrial society. 10 vols. Cleveland:
A.H. Clark.
1916. (With J.B. Andrews.) Principles of labor
legislation. New York: Harper.
1919. Industrial goodwill. New York: McGraw-
Hill.
1919-1935. (With others.) Histoiy of labor in the
United States. 4 vols. New York: Macmillan.
1921. Industrial government. New York: Macmillan
19 24 . The legal foundations of capitalism. N e w
Yo r k : Ma c m i l l a n .
1934. Institutional economics. New York:
Macmillan
1934. Myself New York: Macmillan.
1950. The economics of collective action. Madison:
University of Wisconsin Press.
Communications
Roger G. Noll
The economics of communications is a loose,
somewhat vaguely defined amalgam of topics in
applied microeconomics. Although having close
1880
ties to the microeconomic theory of the economics
of information, it is probably best characterized as a extremely detailed regulation of prices and attri-
subfield of industrial organization, regulation and
public enterprise that deals with the commu-
nications sector: telecommunications, broadcasting,
the print media, the performing arts and the postal
system. Of course, the activities that constitute this policies in other infra structural industries: that
list are somewhat arbitrary, but they reflect what is
both taught and studied by people in the subfield as
well as some important economic realities that
make specialized studies of the communications
sector a valid category among distinct intellectual
pursuits. First among these realities is that the
industries in the communications sector are closely
linked. Broadcasting competes with the performing
arts for both audience and inputs, and
telecommunications competes with the postal ser-
vice. Moreover, telecommunications networks are
capable of delivering broadcast services, and vice
versa. Among the products over which the postal
system, telecommunications and cable television
compete is the delivery of the output of the print
media.
Another unifying theme across communications
industries is the connection of the study of the
sector to the economics of public goods and
externalities. Communications is the production and
dissemination of information. Some aspects of the
production of information are public goods, and the
dissemination and use of information can have
important external effects. Moreover, most of these
external effects are non-economic phenomena such
as political participation, the cultural values held by is, a policy of maximizing the number of
members of a society or the level of violence.
Because of the unique character of these
externalities, the motives for public policy in com-
munications are closely linked with a society’s
fundamental political and social values. Thus,
freedom of speech and the extent of the right to
privacy, as well as the use of control of commu-
nications to manipulate the political process, are at
the heart of debates over communications policy.
Rationales for Government Intervention that can be commandeered and used by government
Not surprisingly, the role of the public sector is very
large in the communications sector in nearly
Communications
every nation. Subsidization, nationalization and
butes of the product are common. In market-
oriented societies, telecommunications and mail are
nationalized or subject to economic regulation for
much the same reasons that underpin the same
these industries are natural monopolies and that
their performance and pattern of development
profoundly affect the development of much of the
rest of the economy. But even here, unique
externality arguments are brought forth as addi-
tional factors to be taken into account by policy-
makers. First, subscription to the telecommunica-
tions network or access to mail delivery creates the
capability to receive communication from others. A
person who decides to mail a letter or place a
telephone call presumably considers only his or her
net benefits from the communication; the
willingness to pay of the recipient (positive or
negative) is not taken into account. Thus, for
example, the extent of phone service and the pattern
of calling can be expected to be inefficient if each
personbears the full cost of, first, subscribing to the
network, and then placing telephone calls. In
particular, if some potential customers have too low
a willingness to pay to become subscribers, but are
also desired objects of communications by others,
the number of subscribers will be too low if
subscribers must bear the frill cost of connecting
them to the network. This argument constimtes the
foundation for the ‘universal service' objective; that
subscribers to the telephone network, and the policy
practised nearly everywhere of adopting a price
structure for telephone services that subsidizes
installation charges, pay-telephone prices and/or
monthly access charges to the local network,
especially for customers in high-cost areas such as
rural communities.
A second externality of the telecommunications
system is said to be its contribution to national
security. A joint product of a private
telecommunications network is a ready resource
in times of national emergency, such as foreign
attack, natural disasters or accidents. The use of
communications to coordinate a response to
Communications
such an event, then, should play a role in affecting
the capacity and design of the telecommunications
system, and often is the basis of an argument for
building into the system more redundancy and
interconnectability than might otherwise be optimal
and than independent private concerns would
undertake on their own. These contingent needs by
government have been said to constitute a separate
natural-monopoly argument, an example of
‘economies of scope’ between private and public
uses that can only be captured if the private system
is a single, integrated whole. In the United States,
for example, the Department of Defense was a
consistent critic of proposals to relax regulation and
increase competition in the telecommunications
industry during the 1960s and 1970s.
The externalities associated with the mass media
have to do with the social, political and
psychological consequences of the content of
information, and for the most part are dealt with by
scholars from disciplines other than economics.
(The main exception is research on the effects of
advertising, where an inconclusive debate has raged
for decades as to whether the informational value of
advertising exceeds the sum of its direct costs and
possible resource misallocation owing to
manipulation and misperceptions of consumers.)
The analytical foundation for the belief in the
importance of informational externalities is the
proposition that people’s behaviour as citizens,
parents, consumers, workers, friends, and so on, can
be significantly affected, at least in the short run, by
the informational content of the mass media. Once
one accepts this proposition, the next logical step is
to entertain the idea that censorship by the state, at
least in principle, can prevent some of the external
diseconomies of destructive content, while
proactive state interventions to channel the content
of the media towards greater educational and
otherwise uplifting content can provide additional
social benefits.
The most obvious manifestations of these ideas
are in broadcasting. The British Broadcasting
Corporation was founded on openly paternalistic
principles about the potential of radio broadcasts for
educational and other uplifting purposes. Until the
recent move towards decentralization through cable
television and towards private, commercial
1881
television, a core principle of French broadcasting
policy was to preserve French culture and values by
limiting and censoring programmes from other
nations. In Germany, decentralized, regional quasi-
public broadcast monopolies were created after
World War II to protect simultaneously against
capture by the national government or by the
national print media barons, either of which, it was
feared, might use broadcasting to arouse nationally
destructive political passions. And in the United
States, broadcast licensing has, until recently,
enforced a long set of standards for evaluating
competitors for a given license, including personal
characteristics and other business holdings of the
licensee and both performance and promise about
the extent of ‘public service’ programming offered
by commercial as well as non-commercial
(educational) outlets.
Of course, other mass media are not free of
similar policy constraints, although the print media
and the arts are usually accorded greater freedom in
the content of their messages than are broadcasters.
The areas of policy controversy are the definition of
the liability for slanderous attacks and the
concomitant definition of privacy rights, the
boundaries between pornography and legitimate
expression, and the principles separating sedition
from reasonable political discourse.
The core economic issue in this debate is
whether the ‘marketplace for ideas’ works well
without intervention, or at least better than is the
case with active political intervention by the very
government authorities whose security and power
can be affected by the content of communications.
The argument for non-intervention is twofold.
Positively, it is that in the end people's tastes in
ideas should be accorded the same status as then-
tastes in other goods as long as the consumption of
communications produces no external disecon-
omies. If communications cause bad behaviours,
then if people are informed about this fact - and
about the punishments exacted if those behaviours
are manifest - they will efficiently anticipate this in
making decisions about which communications to
receive and how to treat those that are received.
And, as with goods, ideas about how the world
works that prove correct will be perceived, at least
eventually, as superior to less correct ideas.
1882
Negatively, the argument for an unregulated mar-
ketplace for ideas is a pessimistic forecast of how
political intervention is likely to work: a combi-
nation of orientation towards propaganda to scivc
the interests of preserving the status quo and an
extreme sensitivity to either vocal, organized single-
issue groups seeking to impose their values on
others or a tyranny of the majority that persecutes
those who stray too far from current norms.
The other side of the dispute, usually advocated
more by non-economists, is rooted in observed
relationships between communications and
behaviour, perhaps best documented in the study of
the effects of television on violence (especially by
children) and on the manipulation of the news for
short-term political objectives. This position regards
the efficiency of the marketplace for ideas as
demonstrably poor, at least in the short run;
implicitly, it accords less credence to the
proposition that individuals are as rational - indeed,
are even proactive - in selecting among competing
communications as economic theory assumes.
Proponents of intervention especially emphasize the
unformed and manipulable attitudes of children.
The final pervasive feature in the communica-
tions sector that deserves further elaboration is the
partially non-riva Irons nature of the consumption
of information. All information is a public good in
the sense that once a new information product has
been created for a first user, it does not have to be
created again for subsequent users: in principle, at
least, the first use of information does not preclude
its use by others, in practice, this characteristic may
be unimportant. Information must be disseminated
in some way to subsequent users, and the cost of
dissemination may exceed the cost of secondary
creation - as for example can be the case for a
simple computer program. Or, information may be
very cheaply privatized so that the public goods
characteristic introduces no significant inefficiency
to a private market system of distribution.
Nevertheless, the publicness of information is a
serious issue in an assessment of the performance of
allocational institutions in the communications
sector, and in the design of private market processes
for allocating resources the problems of publicness
must be addressed.
Communications
Whether the product is a written news report, a
novel, a theatrical production, a television broad-
cast, or 'Dial-A-Joke' on the telephone, the problem
is fundamentally the same: producers will not
supply a product unless they can recover the
opportunity cost for creating it, yet the marginal
cost of providing the product to one more consumer
does not include any of the production costs of the
information. Hence, efficient provision of
information requires one of the following: subsidies
of the production of information, or price
discrimination with protection against arbitrage so
that consumers with relatively low willingness to
pay for information will not be inefficiently
excluded. In practice, both are common.
Governments subsidize broadcasting and
performing arts by direct payments, and certain
users of the postal and telecommunications system
either directly or by engaging in price dis-
crimination (e.g. the differences in basic monthly
service rates of telephones between residences and
businesses, and the lower postal rates for circulating
the print media). Of course, neither direct subsidies
nor discriminatory prices are explicitly designed in
a quantitative sense to offset the inefficiencies of
private provision of public goods, so that the issue
of optimal pricing of communications services is an
active and still- developing field of research. The
focus here is on the two fields in which most of the
work has been done: telecommunications and
broadcasting. Moreover, the discussion includes
research on market structure issues because of their
close connection to the implications of alternative
pricing policies.
Pricing and Market
Structure in
Telecommunications
The telecommunications industry in the United
States offers an array of services that until very
recently were provided as joint products by a
legally protected monopoly. When the monopoly
was secure and unquestioned, the pricing problem
was to devise a price structure that recovered joint
and fixed costs with minimal loss of efficiency. As
the natural monopoly presumption came to be
Communications
called into question, the pricing problem began to
incorporate another dimension, to provide appro-
priate signals to potential competitors so that the
market structure would evolve efficiently.
To understand the rudiments of the telephone
pricing problem requires some basic knowledge
about the technical characteristics of the telecom-
munications network. The traditional telephone
system is best conceptualized as having four
components: customer terminal equipment
(a telephone, a computer terminal, a switchboard); a
pair of copper wires connecting each terminal
device to a central switch; the central switch that
serves the local community; and a hierarchy of
transmission conduits and additional switches that
serve to connect the local switches. Typically the
telephone price structure has three elements: an
installation charge for activating a customer’s
copper wire pairs; a basic monthly service charge
for renting terminal equipment and the copper wire
pairs; and a message toll for placing telephone calls.
The common practice is for the installation charge
to cover only a fraction of installation costs as a
means of encouraging universal service, and for the
basic monthly charge to entitle the subscriber to
unlimited local calling - sometimes not confined to
other telephones connected to the same local switch,
but also including calls through adjacent local
switches. Usually the basic monthly charge is much
higher (by a factor of 2 or 3) for business than for
residences, but within each of these categories it
tends to be approximately the same over wide
geographic areas regardless of differences in cost of
service.
Until about 1960, the revenues from installation
charges and the basic monthly rate approximately
covered the cost of local service (including local
switches). But as long-distance toll calls became
more important, telephone companies increasingly
used toll revenues to cover part of the cost of the
local system. This required no increase in toll
prices; indeed, long-distance prices generally were
falling because technological change was
extraordinarily rapid in this segment of the system.
By simply letting prices fall a little more slowly
than costs, a large and growing fraction of local
network costs could be paid for
1883
by toll. These revenues could be used to construct
systems in high-cost rural areas without causing
increased prices for basic service elsewhere, again
to encourage universal service.
Since toll calls pass through local switches they
impose a cost on the local network, because local
switches must be designed to be large enough and
complex enough to accommodate them.
Consequently, toll prices would bear some local
system costs in an efficient pricing structure. In
addition, however, toll calls also contributed to
‘non-traffic sensitive’ (NTS) costs - the terminals
and copper wire - even though, by definition, the
magnitude of investment required for this equip-
ment was unrelated to the amount of calling.
Obviously, this pricing structure not only
encouraged universal service but encouraged local
calling (with a zero price at the margin) while
discouraging long-distance calling compared to an
efficiency standard. Encouraging subscriptions to
the system may be warranted on efficiency grounds,
although the magnitude of the subscription
externality has not been quantified, and so it is not
possible to tell whether the amount of the subsidy is
justified. Likewise, a subsidy of local calls may be
desirable, but the method of subsidization is of
doubtful validity. The external benefit (or cost) of a
call falls on the person being called, not on society
generally.
Hence, the optimal pricing structure involves a
sharing of the costs of calling between the parties to
a conversation, where the costs involve the
operating costs of the system and the effect of calls
on the required capacity of the switching system.
Only if metering costs were large in comparison
with the costs of calling would it make sense not to
charge for calls, but with modem electronic
switching metering costs are not significant, so that
one cannot justify a subsidy for local calls.
Moreover, even if one could, there is no justification
for taxing long-distance calls to pay the subsidy
unless one believes that the externality of a local
call is substantially more important than the
externality of a long-distance call.
The general structure of an optimal price struc-
ture for the telephone network, given important
externalities and natural monopoly, can be derived
as follows. Begin with a basic monthly charge that
1884
would pay the marginal cost of terminals and
copper wire connections to the local switch, and toll
charges on all calls that pay the marginal cost of the
switching and transmission facilities that arc traffic
sensitive. These prices need further adjustment, for
they may collect too much or too little total revenue.
But prior to this adjustment they must also be
uniformly adjusted downwards to account for the
externality of subscribing and calling (assuming that
people like to receive phone calls). The adjustment
for the toll rate can simply be passed on to the
recipient of the call; however, the basic monthly
rate must come down for everyone. At this point the
likely case is that the basic monthly rate does not
cover the NTS costs, so that further adjustments
must be made. One possibility is a subsidy paid
from an economy-wide tax, but more likely the
additional revenues will come from the rate
structure of the telephone company. The first-best
solution is to raise infra-marginal prices, such as the
cost of the first few calls made per month,
producing what amounts to quantity discounts for
all types of calls. Alternatively, one could adopt
Ramsey pricing, raising the price the most for
services with relatively inelastic demand.
The resulting price structure would have a
number of veiy interesting features. Call recipients
would pay a share of the price of a call. To
implement this so as only to charge for calls with a
positive benefit, the shared cost would start a decent
interval after the call is answered, and subscribers
who desired it could be permitted to designate in
advance that they would bear the frill cost of their
calls. All prices would be built on marginal costs,
which means peak-load pricing of calling and prices
for both basic access and calling that were higher in
high-cost areas. To the extent that Ramsey prices
were invoked, they would most certainly rely
primarily on basic monthly charges, for this has by
far the lowest demand elasticity: estimates range
between —0.02 and -0.10. Thus, even if there is a
significant externality associated with subscribing
to the network, the Ramsey pricing method for
paying for it involves raising the price of basic
access. Or, putting the matter another way, ignoring
this externality in setting prices will have very little
Communications
effect on subscriptions to the system, and hence
very little effect on efficiency. Finally, differences
between residential and business basic monthly
charges would exist only if their externality value
differed, they imposed different costs on the system
or they had different price elasticities.
Obviously, the pricing structure of telephone
service has never reflected these principles. Until
the 1970s, government officials perceived the extent
of inefficiency of the pricing structure as something
of an academic issue and largely ignored it. But
technological change and the false signals to
entrants from the price structure have led to strong
pressures for competitive entry into the formerly
secure telephone monopoly. Computer technology
has vastly diversified the demand for
telecommunications, as well as vastly increased its
magnitude, and computers and other advances in
microelectronics have altered the technology of
supply. Examples of the broad range of new
computer-based services include on-line
connections to mainframes and data bases for
technical and business use, automatic teller
machines, remote sensing for protection against
fires and burglars, and reservation services. Each of
these uses has somewhat different technical
requirements, so that the optimal market structure
for the industry may well be to have a product-
differentiated oligopoly, even if each has
unexploited economics of scale. Moreover, the
greater demand created by these technical advances
allows considerable exploitation of scale economies
even in a segmented system.
On the supply side, advances in electronics have
changed the basic character of the local network. No
longer arc high-density networks built of dedicated
copper wire pairs for each terminal. Instead, micro-
electronic technology allows multiple signals on the
same wires, and small-scale switches distributed
throughout the local network that serve to
concentrate lines from many terminals into a
smaller number of active circuits, taking advantage
of the fact that not all terminals arc in use
simultaneously. This reduces the unit cost of
capacity and hence the significance of scale
economies in the local network. Moreover, it
undermines a cornerstone of the optimal-pricing
structure that was developed
Communications 1885

above by eliminating most of the NTS costs. If, as the face of diminished or perhaps even non-existent
is becoming the case, customers own their terminal natural monopoly they must erect barriers to
equipment, and if line concentration begins when a competitive entry aimed at the overcharged
small number of terminals are aggregated into a customers.
single pathway to the first switch, then nearly all of
the system that is owned by the local telephone
company consists of traffic- sensitive investment. Prices and Market
Hence, the trend should be away from reliance on Structure in Broadcasting
the basic monthly charge and towards greater
The most common way to pay for broadcasting is to
reliance on message tolls for calling.
provide signals to the audience at no charge, and to
The failure to adjust the pricing system to the
rely on either advertising or government subsidies
realities of costs and technology adds to the pres-
as to the source of revenues. In one sense such an
sure for competitive entry in the parts of the system
arrangement seems to fit the fact that broadcasts are
where prices are higher than the costs of service.
a classic public good; the marginal cost to the
Specifically, the attempt to tax longdistance service
broadcaster of one more person receiving a
in order to subsidize local calling makes relatively
broadcast is zero, and consumption among members
intense users of long-distance service ripe
of the audience is completely non-rivalrous. Hence,
candidates for a competitive longdistance supplier.
any attempt to charge a viewer for a programme can
Large companies with many telephone lines who
introduce inefficiency to the extent that anyone is
can provide their own concentrated connections
thereby excluded from participation who also has a
between their facilities have a strong incentive to
positive willingness to pay to join the audience.
bypass subscriptions to the local network. And
The difficulty with free broadcasting, however,
other electronic pathways for communications, such
is that it does not necessarily result in programmes
as cable television and point-to-point uses of the
that maximize the net willingness to pay of the
radio spectrum, can be exploited to bypass the
audience. Ignoring for a moment the frictions in the
telecommunications network.
political process and the incentives of political
Thus far, government officials responsible for
actors to manipulate programme content to their
telecommunications policy whether as operators of
private benefit, both subsidized and advertiser-
public enterprises or regulators of private utilities,
supported television lead broadcasters to measure
have focused more on the structural aspects of the
their success primarily on the basis of the size of
problem than on pricing issues. Even in the United
audience. In a subsidized system, the objective
States, which has perhaps the greatest commitment
would be to make certain that political support is as
to competition, government policy regarding
high as possible, and in an advertising system, in
entrants has been the binding constraint on the
which the broadcaster is selling the attention of
growth of competition, and the price structure is
viewers, revenues are more or less proportional to
still replete with inefficiencies. The likely
audience size. The issue in both cases is not whether
explanation for this phenomenon is the belief by
audience satisfaction is maximized but whether it is
political actors that the cost of local service to
kept high enough for a large enough number of
residences is the price that is most visible politically
people to maximize revenues from a payment
and that rationalized pricing will cause residential
system that is not based on the intensity of
service to become more expensive, either from
preferences but on the number of satisfied cus-
raising basic monthly charges or from message toll
tomers. In particular, small groups with intense
for local calls. To avoid raising residential price,
willingness to pay for an unusual type of program-
political actors therefore believe that they have to
ming material will generally not have their pref-
keep some other prices above the cost of service,
erences satisfied even if their aggregate
and to maintain these prices in
1886
willingness to pay exceeds that of a large audience
for the traditional mass-audience programme.
Three means are available for coping with this
state of affairs. One is to expand the number of
broadcast options until all groups are satisfied.
Suppose that there is a large mass audience and a
series of small, specialized ones. As the number of
stations expands, the audience for mass pro-
grammes each can expect will be the total mass
audience divided by the number of stations, Even-
tually, there will be enough stations so that the
largest specialized taste will constitute a larger
audience than the share of the mass audience the
next station could expect to capture, so that a
strategy to maximize audience will lead to spe-
cialization. In the United States, this is more or less
the policy with respect to radio broadcasting. In the
early years of radio, the Federal Communications
Commission tried to assure diversity in commercial
broadcasting by specifying the format (e.g. type of
programmes) that a station could broadcast.
Recently, station formats have been deregulated, yet
the multiplicity of categories remain, in much the
same fashion that there is a broad spectrum of
magazines and books by type of material.
In television, the strategy of increasing the
number of stations is more difficult to follow.
Television stations consume far more radio fre-
quency space than radio stations, and no nation has
thus far been willing to allocate enough high-
quality radio spectrum to television to provide much
of a test as to whether specialization might take
place in a more extensive industry. An unplanned
test, however, is under way in Italy, where in the
1970s the courts declared that the government had
no constitutional right to limit the number of
television stations, and largely unregulated entry
has taken place on a massive scale. It is too soon to
tell what the ultimate outcome of this system will
be.
The second mechanism to produce more diver-
sity in television is to allow the audience to express
a willingness to pay, commonly by installing cable
television with much higher capacity than off-air
television and charging customers on the basis of
the number and type of channels that they elect to
receive. The
Communications
inefficiency inherent in this system is the cost of
privatizing broadcasts so that on either a per pro-
gramme or per channel basis they can be sold. Prior
to the extensive development of cable television in
the United States, the common speculation was that
privatization of broadcasts would cause a diversion
from traditional mass audience programming, with
more activity in cultural programmes, educational
broadcasting and public affairs. The expectation
was based upon the belief that higher-income
groups were more interested in diversity and would
have more influence in determining the content of
for-pay systems. In practice, this expectation has
not been realized. The new cable-oriented networks
for the most part offer programming that is like that
provided by off-air broadcasters, such as movies,
sports events and regular series. The principal
exception is in public affairs, where national cable
news and public events networks have succeeded.
Educational and cultural programming, however,
has been largely unsuccessful. The inference to be
drawn is that scarcity in television stations caused
an excess demand for television, but primarily for
programmes that were much like those featured by
off-air stations, largely oriented towards the mass
audience.
The third means for increasing diversity is to
create a single, multi-outlet monopoly broadcasting
entity. If such an entity seeks to maximize total
audience, it will not have as much incentive to
duplicate mass programming, because audience
substitution from one channel to the next will have
no value. A second or third mass-audience station
will increase the size of the total audience, but the
evidence indicates that the effect is small compared
to audience diversion. For example, in the United
States the first television station captures a little less
than half of the potential audience available in
prime evening viewing time. No matter how many
additional stations are added, the maximum viewing
share appears to be about 80%, and this is almost
totally achieved after three or four stations are
operating. This suggests that a multichannel
monopolist would either diversify programming on
the second or third channel, or simply elect not to
broadcast on more than one or two channels,
depending on the relationship
Communications
between the value of a net increment to the audience
and the costs of adding another channel.
American public television provides an example
of a novel method of support, for it is one of the few
attempts to implement a decentralized decision
process for acquiring a public good (here
programmes). The first component of the system is
the method of public financing, which involves
multiple year, advance funding to erect some barrier
to political manipulation of programmes. The public
funds are then divided into three components: a
budget for experimental programming that is spent
by an independent, quasipublic entity (the
Corporation for Public Broadcasting); a budget for
the technical operation of the national network
(Public Broadcasting Service); and a direct subsidy
of local stations. This subsidy is based in part on the
success of the station in obtaining private
contributions from its audience. Thus the station
subsidy amounts to an attempt to overcome the free-
rider problem faced by viewers by providing
matching funds for their contributions.
The second component of the system is the
mechanism by which stations decide which pro-
grammes will appear on the network. This is
accomplished by a combination voting and price
system. The price of a programme for each station
is determined by the size of the community it
serves, and stations then vote on each programme
proposal. If some stations vote against the proposal,
the prices faced by the supporters are increased by
an amount necessary to allow the programme to
cover its costs, and voting proceeds again. The
process continues until programmes are either
purchased or discarded; usually fewer than a dozen
iterations are required to reach a decision. Stations
voting against a programme are excluded from
broadcasting it; however, stations may later join the
group paying for it by paying a premium price.
The programme acquisition process decentral-
izes network programming to the stations, thereby
serving two ends. First, because the station budgets
depend on contributions, or the voluntary
willingness to pay of the audience, there exists a
feedback mechanism from the audience to the
network that is similar to a pay-TV system.
1887
Second, the network schedule becomes less vul-
nerable to political attack, for centralized govern-
ment officials who might seek to control it face a
collective of over 150 station licensees, who, in
turn, are actively using contributions patterns to
make decisions about which programmes to
acquire.
The American system of financing public tele-
vision does not, of course, have pristine efficiency
properties; neither the voluntary audience contri-
butions nor the mechanism whereby stations select
programmes is an incentive-compatible mechanism.
Nevertheless, in the inherently imperfect world of
public goods acquisition, they appear to perform
remarkably well, and experimental investigations in
a laboratory setting suggest that the method of
acquiring programmes can be productively
employed in a variety of settings for collective
decisions.
Remaining Issues
Two aspects of the communications sector make it a
ripe area for continuing study. First is the rapidly
evolving technology of supply and demand, and the
second is the pervasive and changing influence of
political processes on the structure and performance
of the sector.
With changing technology has come a signifi-
cant change in the pattern of demand for services.
This suggests that historical patterns of use and
estimates of service-specific demand are unreliable
predictors of the future. Yet relatively little research
has investigated how changing technology - lower
costs, greater possibilities of use, more technical
capabilities - have affected key aspects of demand:
the rate of growth by service and customer category
and the own-and cross-elasticities of demand.
Changing technical possibilities and demand
should also feed back into the political forces that
guide the development of the sector. Most advanced
industrialized nations are in the midst of transition
in at least some policies regarding communications,
such as the privatization and introduction of
competition in telecommunications in Japan and
Great Britain, and the
1888
elimination of the state broadcasting monopolies in
France and Italy. These changes deserve study on
two counts: how these dramatic policy changes
affect performance, and what political forces they
may be creating that will shape policy and industry
structure in the future.
A period of rapid change is one in which
important new knowledge is likely to be forth-
coming. One can anticipate that a summary of the
economics of communications a decade or two
hence will contain significant and surprising new
insights.
See Also
References
Bloch, H., and M. Wirth. 1984. The demand for pay services Taylor, L. 1980. Telecommunications demand: A survey and
on cable television. I n f o r m a t i o n E c o n o m i c s a n d
P o l i c y 1(4): 311-332.
Brock, G. 1981. The telecommunications industry: The
dynamics of market structure. Cambridge, MA: Harvard
University Press.
Coase, R. 1959. The Federal Communications Commission.
J o u r n a l o f L a w a n d E c o n o m i c s 2(1): 1^40.
Courville, L., A. de Fontenay, and R. Dobell. 1983. E c o -
nomic analysis of telecommunications.
Amsterdam: North-Holland.
Evans, D. (ed.). 1971. B r e a k i n g u p b e l l . Amsterdam:
North-Holland.
Levin, H. 1971. T h e i n v i s i b l e r e s o u r c e . Baltimore:
Johns Hopkins Press.
Machlup, F. 1980. The production and distribution of
knowledge in the United States. Princeton: Princeton
University Press.
Mitchell, B. 1978. Optimal pricing and local telephone
services. A m e r i c a n E c o n o m i c R e v i e w 68(4):
517-537.
Network Inquiry Special Staff. 1980. N e w t e l e v i s i o n such a regime, by the French lawyer Etienne Cabet
n e t w o r k s : E n t r y , j u r i s d i c t i o n , o w n e r s h i p in the late 1830s; his works, especially the utopia
a n d r e g u l a t i o n . Washington, DC: Federal
Communications
Commission.
Noll, R. 1985. ‘Let them make toll calls’: A state regulator’s ‘communist banquet’ was held in Paris - banquets
lament. A m e r i c a n E c o n o m i c R e v i e w 75(2): 52- and banquet speeches were a common form of
56.
Noll, R., M.J. Peck, and J.J. McGowan. 1973. E c o n o m i c
aspects of television regulation.
Washington, DC: Brookings.
Owen, B. 1975. E c o n o m i c s a n d f r e e d o m o f Kommunismus und die Augsburger Allgemeine
e x p r e s s i o n . Cambridge, MA: Ballinger.
Owen, B., J. Beebe, and W. Manning. 1974. T e l e v i s i o n
Communism
Park, R.E. 1972. Prospects for cable in the 100 largest
television markets. B e l l J o u r n a l o f E c o n o m i c s
3(1): 130-150.
Park, R.E. 1975. New television networks. B e l l J o u r n a l
o f E c o n o m i c s 6(2): 607-620.
Rosse, J. 1967. Daily newspapers, monopolistic competition,
and economies of scale. A m e r i c a n E c o n o m i c
R e v i e w 52(2): 522-533.
Rosse, J., J. Dertouzos, M. Robinson, and S. Wildman. 1979.
Economic issues in mass communications industries. In
Proceedings of the symposium of media
c o n c e n t r a t i o n , vol. I, 40-192. Washington, DC:
Federal Trade Commission.
Snow, M. 1986. Marketplace for telecommunications:
Regulation and deregulation in industrialised democracies.
White Plains: Longman.
Spence, A.M., and B. Owen. 1977. Television programming,
monopolistic competition and welfare. Q u a r t e r l y
J o u r n a l o f E c o n o m i c s 91(1): 103-121.
Spitzer, M. 1985. Controlling the content of print and
broadcast. S o u t h e r n C a l i f o r n i a L a w R e v i e w
58(6): 1349-1405.
Steiner, P. 1952. Program patterns and preferences, and the
workability of competition in radio broadcasting.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 66(2): 194-
223.
critique. Cambridge, MA: Ballinger.
von Weiszacker, C. 1984. Free entry into telecommunications?
I n f o r m a t i o n E c o n o m i c s a n d P o l i c y 1(3):
197-216.
Communism
Ernest Mandel
The term ‘communism’ was first used in modem
times to designate a specific economic doctrine (or
regime), and a political creed intending to introduce
Ulcarie, were influential among the Paris working
class before the revolution of 1848. In 1840, the first
political protest under the July monarchy. The term
spread rapidly, so that Karl Marx could entitle one
of his first political articles of 16 October 1842 ‘Der
Communism
Zeitung’. He noted that ‘communism’ was already
an international movement, manifesting itself in
Britain and Germany besides France, and traced its
origin to Plato. He could have mentioned ancient
Jewish sects and early Christian monasteries too.
In fact, some of the so-called ‘Utopian social-
ists’, in the first place the German Weitling, called
themselves communists and spread the influence of
the new doctrine among German itinerant
handicraftsmen all over Europe, as well as among
the more settled industrial workers of the
Rhineland. Under the influence of Marx and Engels,
the League of the Just (Bund dez Gerechten) they
had created, changed its name to Communist
League in 1846. The League requested the two
young German authors to draft a declaration of
principle for their organization. This declaration
would appear in February 1848 under the title
Communist Manifesto, which would make the
words ‘communism’ and ‘communists’ famous the
world over.
Communism, from then on, would designate
both a classless society without property, without
ownership - either private or nationalized - of the
means of production, without commodity produc-
tion, money or a state apparatus separate and apart
from the members of the community, and the
social-political movement to arrive at that society.
After the victory of the Russian October revolution
in 1917, that movement would tend to be identified
by and large with Communist parties and a
Communist International (or at least an
‘international communist movement’), though there
exists a tiny minority of communists, inspired by
the Dutch astronomer Pannekoek, who are hostile to
a party organization of any kind (the so-called
‘council communists’, Ratekommunisten).
The first attempts to arrive at a communist
society (leaving aside early, medieval and more
modem Christian communities) were made in the
United States in the 19th century, through the
establishment of small agrarian settlements based
upon collective property, communally organized
labour and the total absence of money inside their
boundaries. From that point of view, they differed
radically from the production
1889
cooperatives promoted for example by the English
industrialist and philanthropist Robert Owen.
Weitling himself created such a community,
significantly called Communia. Although they were
generally established by a selected group of
followers who shared common convictions and
interests, these agrarian communities did not
survive long in a hostile environment. The nearest
contemporary extension of these early communist
settlements are the kibbutzim in Israel.
Rather rapidly, and certainly after the appear-
ance of the Communist Manifesto, communism
came to be associated less with small communities
set up by morally or intellectually selected elites,
but with the general movement of emancipation of
the modem working class, if not in its totality at
least in its majority, encompassing furthermore the
main countries (wealth-wise and population-wise)
of the world. In the major theoretical treatise of
their younger years, The German Ideology, Marx
and Engels stated emphatically:
Empirically, communism is only possible as the act of
dominant peoples ‘all at once’ and similtaneously,
which presupposes the universal development of
productive forces and the world intercourse bound
up with them. .. .The proletariat can thus only exist
worldhistorically, just as communism, its activity,
can only have a ‘world- historical’ existence.
And, earlier in the same passage,
... This development of productive forces (which at
the same time implies the actual empirical existence
of men in their world-historical, instead of local,
being) is an absolutely necessary practical premise,
because without it privation, is merely made general,
and with want the struggle for necessities would
begin again, and all the old filthy business would
necessarily be restored ... ([ -6] 1976,
P-49).
That line of argument is to-day repeated by
most orthodox marxists (communists), who find in
it an explanation of what ‘went wrong’ in Soviet
Russia, once it was isolated in a capitalist
environment as a result of the defeat of revolution
in other European countries in the 1918-1923
period. But many ‘official’ Communist Parties still
stick to Stalin’s particular version of communism,
according to which it is possible to
1890
successfully complete the building of socialism and
communism in a single country, or in a small
number of countries.
The radical and international definition of a
communist society given by Marx and Engels
inevitably leads to the perspective of a transition
(transition period) between capitalism and com-
munism. Marx and Engels first, notably in their
writings about the Paris Commune - The Civil War
in France - and in their Critique of the Gotha
Programme [of the German social- democratic
party], Lenin later - especially in his book State and
Revolution - tried to give at least a general sketch of
what that transition would be like. It centres around
the following ideas:
The proletariat, as the only social class radically
opposed to private ownership of the means of
production, and likewise as the only class which has
potentially the power to paralyse and overthrow
bourgeois society, as well as the inclination to
collective cooperation and solidarity which are the
motive forces of the building of communism,
conquers political (state) power. It uses that power
(‘the dictatorship of the proletariat’) to make more
and more ‘despotic inroads’ into the realm of
private property and private production, substituting
for them collectively and consciously (planned)
organized output, increasingly turned towards direct
satisfaction of needs. This implies a gradual
withering away of market economy.
The dictatorship of the proletariat, however,
being the instrument of the majority to hold down a
minority, does not need a heavy apparatus of full-
time functionaries, and certainly no heavy apparatus
of repression. It is a state sui generis, a state which
starts to wither away from its inception, i.e. it starts
to devolve more and more of the traditional state
functions to selfadministrating bodies of citizens, to
society in its totality'. This withering away of the
state goes hand in hand with the indicated withering
away of commodity production and of money,
accompanying a general withering away of social
classes and social stratification, i.e. of the division
of society between administrators and
administrated, between 'bosses' and 'bossed over’
people.
Communism
That vision of transition towards communism as
an essentially evolutionary process obviously has
preconditions: that the countries engaged on that
road already enjoy a relatively high level of
development (industrialization, modernization,
material wealth, stock of infrastructure, level of
skill and culture of the people, etc.), created by
capitalism itself; that the building of the new
society is supported by the majority of the popu-
lation (i.e. that the wage-earners already represent
the great majority of the producers and that they
have passed the threshhold of a necessary level of
socialist political class consciousness); that the
process encompasses the major countries of the
world.
Marx, Engels, Lenin and their main disciples
and co-thinkers like Rosa Luxemburg, Trotsky,
Gramsci, Otto Bauer, Rudolf Hilferding, Bukharin
et al. - incidentally also Stalin until 1928 - dis-
tinguished successive stages of the communist
society: the lower stage, generally called ‘social-
ism’, in which there would be neither commodity'
production nor classes, but in which the individual’s
access to the consumption fund would still be
strictly measured by his quantitative labour input,
evaluated in hours of labour; and a higher stage,
generally called ‘communism’, in which the
principle of satisfaction of needs for cvciyonc
would apply, independently of any exact measure-
ment of work performed. Marx established that
basic difference between the two stages of com-
munism in his Critique of the Gotha Programme,
together with so much else. It was elaborated at
length in Lenin's State and Revolution.
In the light of these principles, it is clear that no
socialist or communist society' exists anywhere in
the world today. It is only possible to speak about
‘really existing socialism’ at present, if one intro-
duces a new, ‘reductionist’ definition of a socialist
society, as being only identical with predominantly
nationalized property of the means of production
and central economic planning. This is obviously
different from the definition of socialism in the
classical marxist scriptures. Whether such a new
definition is legitimate or not in the light of
historical experience is a matter of political and
philosophical judgement. It is in any case another
matter altogether than ascertaining
Community Indifference Curves
whether the radical emancipatory goals projected
by the founders of contemporary communism have
been realized in these really existing societies or
not. This is obviously not the case.
See Also
► Central Planning
► Collective Agriculture
► Full Communism
► Marx, Karl Heinrich (1818-1883)
►::
► Planned Economy
► Socialist Economies
References
Marx, K., and Engels, F. 1845-6. T h e G e r m a n
i d e o l o g y . As in Karl Marx and Frederick Engels,
C o l l e c t e d w o r k s , vol.5. London: Lawrence &
Wishart, 1976.
Community Indifference Curves
Wayne Shafer
Keywords
Bergson-Samuelson social welfare functions;
Community indifference curves; Comparative
statics; Scitovsky, T.; Utility possibility
frontiers
JEL Classifications
Dll
The idea of a community indifference curve, as the
term is commonly used, is due to Scitovsky ( ).
The genesis of the idea is the fact that comparative
statics and welfare analysis in economic models is
simplified considerably if there is a social prefer-
ence ordering over aggregate commodity bundles
1891
which reflects the collective individual preferences
of agents. Scitovksy’s notion of a ‘community
indifference curve’ essentially allows the analytical
convenience of social indifference curves, in certain
circumstances, without having to assume a specific
Bergson-Samuelson social welfare function or
having to assume the restrictive assumptions on
agents’ preferences needed to guarantee that agents
act collectively as a single individual.
The definition of a community indifference
curve is basically simple. Suppose there are m
commodities and n agents. Let x denote a com-
modity vector (as m-vector with non-negative
coordinates) and u[ a utility function representing
agent z’s preferences. We will assume that ui is
monotone increasing and quasi-concave. Given a
vector u = (u1 ,...,unj of utility numbers, the
community indifference curve at u', CIC(u'), is
defined to be the set of all commodity vectors x
such that there is a distribution (x 1; ..., x„) of
commodity vectors satisfying YtX, = x and M,(X,) =
ui,i= 1,..., n, and there is no x < x , x / x which also
has this property. Thus one can obtain any vector x
e CIC(u') by fixing the quantities of all but one
good and minimize the amount of the remaining
good subject to achieving u'. As pointed out by
Samuelson ( ), the community
indifference curve can be interpreted as a ‘dual’ to
the utility possibility frontier. The utility possibility
frontier, for a given x, is the set of all vectors u' of
utility numbers achievable by a Pareto efficient
distribution of x to the agents. Let U (x) denote the
utility possibility frontier for the commodity vector
x. Then it is easy to see that CIC(ii ) {x : u e U(x)}
and that U(x) =
\u : x c CIC{u )\.
We will now describe the most important prop-
erties of community indifference curves. First, each
CIC(u') looks like the indifference curve of a
monotone quasi- concave utility function. That is,
the set of vectors x such that x > x 1 for some x1 €
CIC(u') is a convex set. For example, when m 2,
CIC(w') is a curve with a diminishing marginal rate
of substitution. Second, unlike the utility possibility
frontier, the community indifference curve is
essentially an ordinal concept, that is it does not
depend on the choice of utility functions
representing agents’ preferences, in the
1892
following sense. Suppose, for each i, u and v, are
two utility functions representing agent i’s prefer-
ences, and let (x,, ..., x„) be a Pareto efficient
allocation to the agents. Define 1 / = [ufxQ,.. u„(x„)\
and v = [v^xD, ...v„(x„)]. Then CIC (ii ) = CIV(v').
Clearly, community indifference curves can be
parameterized by a given Pareto efficient allocation
of goods rather than a given vector of utilities.
Third, assuming smooth utility functions, the
marginal rate of substitution for any two
commodities on a community indifference curve is
equal to the common marginal rate of substitution of
each agent. Specifically, pick any x e C/C(z/), and
let (x,,... ,x„) be the Pareto efficient allocation of .r
such that ufx) = u . i 1 , . . T h e n for any two
commodities h and h', the marginal rate of
substitution of h and h' evaluated at x € CIC(u') is
equal to the marginal rate of substitution of h for /?'
at x, on agent i’s indifference curve through x;.
Fourth, and very important, community indifference
curves are not, in general, ‘indifference’ curves in
the sense of being level curves of some function.
Pick any x, and u , u e U(x ), such that u' / u . Then
by definition, x c CIC(ur) fi CIC(u ). Thus CIC(ur)
must either coincide with CIC(u) or intersect
properly. The condition for two community indif-
ference curves never to intersect properly is then
that CIC(u') = CIC(u ) for all u', u G U(x), for all x.
It turns out that this is true if and only if the agents
have identical homothetic preferences, in which
case the family of all community indifference
curves will coincide with the family of indifference
curves for the common preferences of the agents.
From the above definition and properties, the
following observation constitutes the basic use of
community indifference curves: if the economy is
currently at a vector of utility numbers u', then x' is
a commodity vector which lies above CIC(u') if and
only if there is some distribution of x' to the agents
which will achieve a vector of utilities u" such that
u > u'. In this sense, x' is ‘better’ than any x c
CIC(i/). However, since from above community
indifference curves can intersect properly, it may
also be that there is a u such that
Community Indifference Curves
x' G CIC(u ) and an x' C CIC(u') such that x lies
above CIC(u ), in which case x is also ‘better’ than
x'. Thus it is important to realize that community
indifference curves cannot be used to define a social
ordering of aggregate output vectors. Nevertheless,
community indifference curves can still be a useful
analytical device. For example, consider a market
economy with two produced goods. Consider an
equilibrium in which all consumers face the same
prices, in terms of the aggregate output vector x' and
the vector of utilities ii obtained by the agents.
Graphically this equilibrium can be represented by
drawing the production possibility frontier and
CIC(i/), noting they meet at x'. The slope of the
production possibility frontier at x' represents the
price ratio faced by firms, and the slope of the
CIC(i/) at x' the common price ratio faced by
consumers. If firms and consumers face the same
price ratio, then the CIC(u') must be tangent to the
production possibility frontier at x. Thus no feasible
x can be produced which can make all agents better
off, so the situation is Pareto optimal. If, however,
firms face different prices than the agents because
of, for example, taxes or tariffs, then the slope of
the CIC(i/ ) will be different from the slope of the
production possibility frontier, and thus the two
curves will intersect properly. In this case there
must exist an x' on the production possibility
frontier which lies above CIC(i/), so the original
situation is Pareto inefficient.
See Also
► Arrow’s Theorem
► iptimality and Efficiency
► ocial Welfare Function
► Velfare Economics
Bibliography
Samuelson, P.A. 1956. Social indifference curves. Q u a r -
t e r l y J o u r n a l o f E c o n o m i c s 70 (1): 1-22.
Scitovsky, T. 1942. A reconsideration of the theory of tariffs.
R e v i e w o f E c o n o m i c S t u d i e s 9 (2): 89-110.
Comparative Advantage 1893

awareness of the benefits of specialization must go

Comparative Advantage back to the dim mists of antiquity in all civili-
zations, it was not until Ricardo that this deepest
Ronald Findlay and most beautiful result in all of economics was
obtained. Though the logic applies equally to
interpersonal, inter/ir/w and interregional trade, it
was in the context of international trade that the
Abstract principle of comparative advantage was discovered
This article traces the evolution of the theory of and has been investigated ever since.
comparative advantage and the gains from trade
from the pioneering work of David Ricardo to
the factor proportions approach of Eli Heckscher The Basic Ricardian Model
and Bertil Ohlin. Extensions of the basic models
What constituted a ‘nation’ for Ricardo were two
to many goods, factors and countries, and to the
things - a ‘factor endowment’, of a specified
long ran are noted, as well as the attempts at
number of units of labour in the simplest model,
empirical testing of the predictions derived from
and a ‘technology’, the productivity of this labour
them.
in terms of different goods, such as cloth and wine
in his example. Thus labour can move freely
Keywords between the production of cloth and wine in
Absolute advantage; Comparative advantage; England and in Portugal, but each labour force is
Com Laws; Distributive justice; Division of trapped within its own borders. Suppose that a unit
labour; Factor price equalization; Factor pro- of labour in Portugal can produce one unit of cloth
portions; Fixed factors; Free trade; Gains from or one unit of wine, while in England a unit of
trade; Flaberler, G.; Heckscher, E. F.; labour can produce four units of cloth or two units
Heckscher-Ohlin trade theory; Human capital; of wine. Thus the opportunity cost of a unit of wine
International trade; Intra-industry’ trade; Labour is one unit of cloth in Portugal while it is two units
mobility; Leontief, W.; Lemer, A. R; of cloth in England. On the assumption of
Monopolistic competition; Neoclassical general competitive markets and free trade, it follows that
equilibrium theory; Ohlin, B. G.; Product both goods will never be produced in both countries
differentiation; Returns to scale; Specialization; since wine in England and cloth in Portugal could
Specific factors; Stationary state; Stolper- always be undermined by a simple arbitrage
Samuelson theorem; Terms of trade; Time operation involving export of cloth from England
preference; Wages fund and import of wine from Portugal. Thus wine in
England or cloth in Portugal must contract until at
least one of these industries produces zero output. If
JEL Classifications both goods are consumed in positive amounts, the
FI ‘terms of trade’ in equilibrium must lie in the closed
interval between one and two units of cloth per unit
The modem economy, and the very world as we
of wine. Which of the two countries specializes
know it today, obviously depends fundamentally on
completely will depend upon the relative size of
specialization and the division of labour, between
each country (as measured by the labour force and
individuals, firms and nations. The principle of
its productivity in each industry) and upon the
comparative advantage, first clearly stated and
extent to which each of the two goods is favoured
proved by David Ricardo in 1817, is the
by the pattern of world demand. Thus Portugal is
fundamental analytical explanation of the source of
more
these enormous ‘gains from trade’. Though an
1894
likely to specialize the smaller she is compared with
England in the sense defined above and the more
world demand is skewed towards the consumption
of wine relative to the consumption of cloth.
The Gains from Trade
Viewed as a ‘positive’ theory, the principle of com-
parative advantage yields predictions about (a) the
direction of trade: that each country exports the
good in which it has the lower comparative oppor-
tunity cost ratio as defined by the technology in that
country, and about (b) the terms of trade: that it is
bounded above and below by these comparative
cost ratios. From a ‘normative’ standpoint the prin-
ciple implies that the citizens of each country
become ‘better off as a result of trade, with the
extent of the gains from trade depending upon the
degree to which the terms of trade exceed the
domestic comparative cost ratio. It is the ‘norma-
tive’ part of the doctrine that has always been the
more controversial, and it is therefore necessary to
evaluate it with the greatest care.
In Ricardo’s example the total labour force in
each country is presumably supplied by an aggre-
gate of different households, each having the same
relative productivity in the two sectors. Thus all
households in each country must become better off
as a result of trade if the terms of trade lie strictly in
between the domestic comparative cost ratios. The
import-competing sector in each country simply
switches over instantaneously and costlessly to
producing the export good (moving to the opposite
comer of its linear production- possibilities frontier,
in terms of the familiar geometry), obtaining the
desired level of the other good by imports, raising
utility in the process. When one country is
incompletely specialized, then all households in that
country remain at unchanged utility levels, all of the
gain from trade going to the individuals in the
‘small’ country. Thus we have a situation in which
everybody gains, in at least one country, while
nobody loses in either country, as a result of trade.
This very strong result depends upon Ricardo’s
assumption of perfect occupational mobility in
Comparative Advantage
each country. Suppose we take the opposite extreme
of completely specific labour in each sector, so that
each country produces a fixed combination of cloth
and wine, with no possibility of transformation. In
this case, labour in the import- competing sector in
each country must necessarily lose, as a result of
trade, while labour in each country’s export sector
must gain. It can be shown, however, that trade will
improve potential welfare in each country in the
Samuelson ( )
sense that the utility-possibility frontier with trade
will dominate the corresponding frontier without
trade, so that no one need be worse off, and at least
some one better off, if lump-sum taxes and transfers
are possible (Samuelson ).
International Factor Mobility and World
Welfare
Another very important normative issue is the
question of the relationship between the free- trade
equilibrium and world efficiency and welfare. In the
Ricardian model world welfare in general will not
be maximized by free trade alone. In the numerical
example considered here Ricardo stresses the fact
that England can still gain from trade even though
she has an absolute advantage in the production of
both goods, her productivity being greater in both
cloth and wine, though comparatively greater in
cloth. Suppose that labour in Portugal could
produce at English levels, if it moved to England;
that is, the English superiority is based on climatic
or other ‘environmental’ factors and not on
differences in aptitude or skill. Then, if labour was
free to move, and in the absence of ‘national’
sentiment, all production would be located in
England, and Portugal would cease to exist. The
former Portuguese labour would be better off than
under free trade, since their real wage in terms of
wine will now be two units instead of one. The
English labour would be worse off, if the terms of
trade were originally better than 0.5 wine per unit of
cloth, but it is easy to show that they could be
sufficiently compensated since the utility-possibility
frontier for the world economy as a whole is moved
out by the integration of the labour forces.
Comparative Advantage
The case when each country has an absolute
advantage in one good is more interesting. As is
easy to see, from Findlay ( ), this case will
involve a movement of labour to the country with
the higher real wage under free trade, increasing the
production of this country’s exportable and reducing
that of the lower-wage country under free trade. The
terms of trade turn against the higher-wage country
until eventually the real wage is equalized. The
terms of trade that achieve this equality of real
wages will be equal to the ratio of labour
productivities in each country’s export sector; that
is, the ‘double factoral’ terms of trade will be unity.
This solution of free trade combined with perfect
labour mobility will achieve not only efficiency for
the world economy as a whole but equity as well.
‘Unequal exchange’ in the sense of Emmanuel (
) would not exist, while liberal,
utilitarian and Rawlsian criteria of distributive
justice would be satisfied as well, as pointed out in
Findlay ( ). Despite all this, it still seems
utopian to expect a policy of ‘open borders’, in
either direction, for the contemporary world of
nation-states.
Extensions of the Basic Ricardian Model
The two-country, two-good Ricardian model was
extended to many goods and countries by a number
of subsequent writers, whose efforts are described in
detail by Haberler ( ) and Viner
( ). In the case of two countries and n goods
the concept of a ‘chain of comparative advantage’
has been put forward, with the goods listed in
descending order in terms of the relative efficiency
of the two countries in producing them. It is readily
shown that with a uniform wage in each country all
goods from 1 to some number j must be exported,
while all goods from (J + 1) to n must be imported.
The number j itself will depend upon the relative
sizes of the two countries and the composition of
world demand. Dombusch et al. ( ) generalize
this result to a continuum of
goods in an extremely elegant and powerfiil model
that has been widely used in subsequent literature.
An analogous chain concept applies to the case of
two goods and n countries, this time
1895
ranking the countries in terms of the ratio of their
productivities in the two goods, with country 1
having the greatest relative efficiency in cloth and
country n in wine. World demand and the sizes of
the labour forces will determine the ‘marginal’
country j, with countries 1 to j exporting cloth and
(/' + 1) to n exporting wine.
The simultaneous consideration of comparative
advantage with many goods and many countries
presents severe analytical difficulties. Graham (
) considered several elaborate
numerical examples, his work inspiring the Roch-
ester theorists McKenzie ( ) and Jones ( )
to apply the powerful tools of activity analysis to
this particular case of a linear general equilibrium
model. It is interesting to note in connection with
mathematical programming and activity analysis
that Kantorovich ( ) in his celebrated book on
planning for the Soviet economy worked out an
example of optimal specialization patterns for fac-
tories that corresponds exactly to the Ricardian
model of trade between countries.
The Three-Factor Ricardian Model
While most of the literature on the Ricardian trade
model has concentrated on the model of Chapter 7
of the Principles in which it appears that labour is
the sole scarce factor, his more extended model in
the Essay on Profits has been curiously neglected,
though the connections between trade, income
distribution and growth which that analysis
explores are quite fascinating. The formal structure
of the model was laid out very thoroughly in
Pasinetti ( ). The economy produces two
goods, com and manufactures, each of which has a
one-period lag between the input of labour and the
emergence of output. Labour thus has to be
supported by a ‘wage fund’, an initially given stock
that is accumulated over time by saving out of
profits. Com also requires land as an input, which is
in fixed supply and yields diminishing returns to
successive increments of labour. The wage-rate is
given exogenously in terms of com, and
manufactures are a luxury good consumed only by
the land-owning class, who obtain rents determined
by the marginal product of land.
1896
Profits are the difference between the marginal
product of labour and the given real wage, which is
equal to the marginal product ‘discounted’ by the
rate of interest, in this model equal to the rate of
profit, defined as the ratio of profits to the real wage
that has to be advanced a period before. Momentary
equilibrium determines the relative price of com
and manufactures, the rent per acre and the rate of
profit, as well as the output levels and allocation of
the labour force between sectors. The growth of the
system is at a rate equal to the product of the rate of
profit and the propensity to save of the capitalist
class. It is shown that the system approaches a
stationary state, with a monotonically falling rate of
profit and rising rents per acre.
The opportunity to import com more cheaply
from abroad will have significant distributional and
growth consequences. Just as Ricardo argued in his
case for the repeal of the Com Laws, cheaper
foreign com will reduce domestic rents and raise the
domestic rate of profit, and thus the rate of growth.
The approach to the stationary state is postponed,
though of course it cannot be ultimately averted,
while the growth consequences for the com exporter
are definitely adverse. The main doctrinal
significance of this wider Ricardian model,
however, is to reveal the extent to which the
subsequent ‘general equilibrium’ or ‘neoclassical’
approach to international trade is already present
within the Ricardian framework. For one thing, the
pattern of comparative advantage itself depends
upon the complex interaction of technology, factor
proportions and tastes. In his Chapter 7 case the
pattern of comparative advantage is exogenous,
simply given by the four fixed technical coefficients
indicating the productivity of labour in cloth and
wine in England and Portugal. The production-
possibility frontiers for each country are linear, and
comparative advantage is simply determined by the
relative magnitudes of the slopes. As demonstrated
in Findlay ( ), however, the Essay on Profits
model implies a concave production-possibilities
frontier at any moment, since there are diminishing
returns to labour in com even though the marginal
productivity of labour in manufactures is constant.
With two countries the pattern of
Comparative Advantage
comparative advantage will depend upon the slopes
of these curves at their autarky equilibria, which are
endogenous variables depending upon the sizes of
the ‘wage fund’ in relation to the supply of land and
the consumption pattern of landowners, as well as
the technology for the two goods.
As Burgstaller ( ) points out, however, the
steady-state solution of the model restores the linear
structure of the pattern of comparative advantage.
The zero profit rate in the steady state requires the
maiginal product of labour to be equal to the given
real wage, and this implies a fixed land-labour ratio
and hence output per unit of labour in com. We thus
once again have two fixed technical coefficients, so
that the slope of the linear production-possibilities
frontier is once again an exogenous indicator of
comparative advantage.
The ‘neo-Ricardian’ approach of Steedman (
, ) considers more general time-phased
structures of production. Technology alone deter-
mines negatively sloped wage-profit or factor- price
frontiers, any point on which generates a set of
relative product prices and hence a pattern of
comparative advantage relative to another such
economy.
Factor Proportions and the Heckscher-
Ohlin Model
While J.S. Mill, Marshall and Edgeworth all made
major contributions to trade theory, the concept of
comparative advantage did not undergo any evo-
lution in their work beyond the stage at which
Ricardo had left it. They essentially concentrated on
the determination of the terms of trade and on
various comparative static exercises. The interwar
years, however, brought fundamental advances,
stemming in particular from the work of the Swedes
Heckscher ( ) and Ohlin ( ). The
development of a diagrammatic apparatus to handle
general equilibrium interactions of tastes,
technology and factor endowments by Haberler (
), Leontief ( ), Lemer ( ) and others
culminated in the rigorous establishment of trade
theory and comparative advantage as a branch of
neoclassical general equilibrium theory.
Comparative Advantage
The essentials of this approach can be
expounded in terms of the familiar two-country,
two-good and two-factor model, on which see Jones
( ) for a detailed and lucid algebraic
exposition. The given factor supplies and constant
returns to scale technology define concave
production-possibility frontiers, on the assumption
that the goods differ in factor intensity. This
determines the ‘supply side’ of the model, which is
closed by the specification of consumer preferences.
Economies that have identical technology, factor
endowments and tastes will have the same autarky
equilibrium price-ratio and so will have no
incentive to engage in trade. Countries must
therefore differ with respect to at least one of these
characteristics for differences in comparative
advantage to emerge. With identical technology and
factor endowments, a country will have a
comparative advantage in the good its citizens
prefer less in comparison to the foreign country,
since then this good will be cheaper at home.
Similarly, if factor endowments and tastes are
identical, differences in comparative advantage will
be governed by relative technological efficiency;
that is, a country will have a comparative advantage
in the good in which its relative technological
efficiency is greater, just as in the Ricardian model.
These differences in technological efficiency could
be represented, for example, by the magnitude of
multiplicative constants in the production functions;
that is, ‘Hicks-neutral’ differences.
In keeping with the ideas of Heckscher and
Ohlin, however, it is differences in factor pro-
portions that have dominated the explanation of
comparative advantage in the neoclassical literature.
The Heckscher-Ohlin theorem, that each country
will export the commodity that uses its relatively
abundant factor most intensively, has been
rigorously established and the necessary
qualifications carefully specified, as in Jones ( ).
Among the more important of these is
the requirement that factor-intensity ‘reversals’ do
not take place; that is, that one good is always more
capital-intensive than the other at all wage- rental
ratios or at least within the relevant range defined
by the factor proportions of the trading countries.
1897
The Stolper-Samuelson Theorem
Associated with the Heckscher-Ohlin theorem is the
Stolper-Samuelson theorem ( ), that trade
benefits the abundant and harms the scarce factor
while protection does the opposite, and the cele-
brated factor price equalization theorem of Lemer (
, though written in 1932) and Samuelson
( , , ), which states that under certain
conditions free trade will lead to complete equali-
zation of factor rewards even though factors are not
mobile internationally. The normative significance
of this theorem is that free trade alone can achieve
world efficiency in production and resource alloca-
tion, unlike the case of the Ricardian model as
pointed out earlier. The requirements for the theo-
rem to hold, however, are very stringent, such as
that the number of tradable goods produced be
equal to the number of factors. It also requires
factor proportions to be sufficiently close to each
other in the trading partners so that the production
patterns are fairly similar. Thus it would be far-
fetched to expect the price of unskilled labour to be
equalized between Bangladesh and the United
States, for example.
The Specific-Factors Model
An important and popular variant of the factor
proportions approach is what Jones ( ) calls
the ‘specific factors’ and Samuelson ( ) the
Viner-Ricardo model. In this model each produc-
tion sector has its own unique fixed factor, while
labour is used in all sectors and is perfectly mobile
internally between them. Trade patterns still reflect
factor endowments but factor price equalization
does not hold in this model since the number of
factors is always one greater than the number of
goods. Gruen and Corden ( ) pre
sent an ingenious three-by-three extension of this
approach, in which one sector uses land and labour,
while the two others use capital and labour, thus
neatly integrating the ‘specific factors’ model with
the regular two-by-two Heckscher-Ohlin model.
Findlay ( , chs.
4 and 6) uses adaptations and extensions of the
Gruen-Corden specification to introduce human
1898 Comparative Advantage

capital formation and the concept of a natural question posed by Samuelson ( ) of whether
resource ‘frontier’ into the Heckscher-Ohlin trade equalizes not only the marginal product or
framework. rental of capital but the rate of interest itself.

Long-Run Extensions of the Factor Empirical Testing

Proportions Model
One limitation of the Heckscher-Ohlin model was
that the stock of ‘capital’, however conceived,
should be an endogenous variable determined by
the propensity to save or time preference of each
trading community, rather than being taken as
exogenously fixed. Oniki and Uzawa ( ) than her imports. Leontiers dramatic finding spurred
extended the model to a situation where the labour
force is growing in each country at an exogenous
rate and capital is accumulated in response to given
propensities to save in each country. One of the
goods is taken to be the ‘capital’ good, conceived of
as a malleable ‘putty-putty’ instrument. They
demonstrated that the system converges in the long
run to a particular capital-labour ratio for each
country, which will be higher for the country with
the larger saving propensity. In Findlay ( , US exports were considerably more skillintensive
), it is shown
that as the capital-labour ratio evolves the pattern of
comparative advantage for a given ‘small’ country
in an open trading world will also shift over time
towards more capital-intensive goods, thus
formalizing the concept of a ‘ladder of comparative
advantage’ that countries ascend in the process of
economic development. Thus comparative
advantage should not be conceived as given and
immutable, but evolving with capital accumulation
and technological change. Much of the loose talk
about ‘dynamic’ comparative advantage in the
development literature, however, is misconceived
since it attempts to change the pattern of production
by protection before the necessary changes in the
capacity to produce efficiently have taken place.
Other models which endogenize the capital stocks
of the trading countries are Stiglitz ( ) and Findlay
( ) which utilizes
a variable rate of time preference and an ‘Austrian’
point-input/point-output technology, which implies
a continuum of capital goods as represented by the
‘trees’ of different ages, and Findlay ( , ch.
2), which addresses the
Empirical testing of the positive side of the theory
of comparative advantage begins in a systematic
way only with the work of MacDougall ( ) on
the Ricardian theory and the celebrated article of
Leontief ( ) that uncovered the apparent par
adox that US exports were more labour-intensive
considerable further empirical research motivated
by the desire to find a satisfactory explanation. The
increasing scarcity of natural resources in the USA,
by causing capital to be substituted for it in import-
competing production, was stressed as an
explanation for the paradox by Vanek ( ).
The role of ‘human capital’ as an
explanation was pointed to by Kenen ( ) and a
number of empirical investigators, who found that
than her imports, even though physical capital-
intensity was only weakly correlated with exports
and imports. This pointed to the need to reinterpret
the simple Heckscher-Ohlin model in terms of
skilled and unskilled labour as the two factors,
rather than labour of uniform quality and physical
capital. Since the formation of skill through
education is an endogenous variable, a function of a
wage differential that is itself a function of trade, we
need a general equilibrium model that can
simultaneously handle both these aspects, as in
Findlay and Kierzkowski ( ) and
Findlay ( , ch. 4).
Many other extensions of the
Heckscher-Ohlin theory are surveyed in Jones and
Neary ( ) and Ethier ( ), while
Deardorf ( ) and Feenstra ( ) give very
incisive accounts of the attempts at empirical testing
of the theory of comparative advantage in its
different manifestations. Further important progress
in empirical testing of the Heckscher-Ohlin model
has been achieved by the work of Learner ( ),
Treflerf ), Harrigan ( ) and Davis
and Weinstein ( ).
Comparative Advantage
Increasing Returns
Finally, the crucial role of increasing returns to
scale in specialization and international trade has
only recently been rigorously investigated, since it
implies departures from perfect competition.
Krugman ( ) and Lancaster ( ) introduced
international trade into models of monopolistic
competition with differentiated products, showing
the possibility of gains from trade due to the
provision of greater variety of similar goods rather
than differences in comparative advantage, what is
referred to as ‘intra-industry’ trade. Helpman and
Krugman ( ) thoroughly examine and
extend our knowledge in this area, while Grossman
and Helpman ( ) expertly extend
the monopolistic competition approach to deal with
a number of issues involving endogenous
technological progress and growth in the world
economy.
See Also
► .)!'
► )b
► . ::
► U/.
Bibliography
Burgstaller, A. 1986. Unifying Ricardo’s theories of
growth and comparative advantage. E c o n o m i c a 53:
467-481.
Davis, D.R., and D.E. Weinstein. 2001. An account of
global factor trade. A m e r i c a n Economic
R e v i e w 91:
1423-1453.
Deardorf, A. 1984. Testing trade theories. In H a n d b o o k
o f i n t e r n a t i o n a l e c o n o m i c s , ed. R.W. Jones
and
RB. Kenen, vol. 1. Amsterdam: North-Holland.
Dombusch, R., S. Fischer, and FA. Samuelson. 1977.
Comparative advantage, trade and payments in a
Ricardian model with a continuum of goods.
American
E c o n o m i c R e v i e w 67: 823-839.
Emmanuel, A. 1972. U n e q u a l e x c h a n g e . New York:
1899
Ethier, W. 1984. Higher dimensional issues in trade theory. In
H a n d b o o k o f i n t e r n a t i o n a l e c o n o m i c s , ed.
R.W. Jones and P.B. Kenen, vol. 1. Amsterdam: North-
Holland.
Feenstra, R.C. 2004. A d v a n c e d international
t r a d e . Princeton: Princeton University Press.
Findlay, R. 1970. Factor proportions and comparative
advantage in the long run. J o u r n a l o f P o l i t i c a l
E c o n o m y 78: 27-34.
Findlay, R. 1974. Relative prices, growth and trade in a
simple Ricardian system. E c o n o m i c a 41: 1-13.
Findlay, R. 1978. An ‘Austrian’ model of international trade
and interest equalization. J o u r n a l o f P o l i t i c a l
E c o n o m y 86: 989-1007.
Findlay, R. 1982. International distributive justice. J o u r n a l
o f I n t e r n a t i o n a l E c o n o m i c s 13: 1-14.
Findlay, R. 1984. Growth and development in trade models.
In H a n d b o o k o f i n t e r n a t i o n a l e c o n o m i c s ,
ed. R.W. Jones and P.B. Kenen, vol. 1. Amsterdam:
North-Holland.
Findlay, R. 1995. F a c t o r p r o p o r t i o n s , t r a d e , a n d
g r o w t h . Cambridge, MA: MIT Press.
Findlay, R., and H. Kierzkowski. 1983. International trade
and human capital: A simple general equilibrium model.
J o u r n a l o f P o l i t i c a l E c o n o m y 91: 957-978.
Graham, F. 1948. T h e t h e o r y o f i n t e r n a t i o n a l
v a l u e s . Princeton: Princeton University Press.
Grossman, G.M., and E. Helpman. 1991. I n n o v a t i o n
and growth in the global economy.
Cambridge, MA: MIT Press.
Gruen, F., and W.M. Corden. 1970. A tariff that worsens the
terms of trade. In S t u d i e s i n i n t e r n a t i o n a l
e c o n o m i c s , ed. I.A. MacDougall and R. Snape.
Amsterdam: North-Holland.
Haberler, G. 1933. T h e t h e o r y o f i n t e r n a t i o n a l
t r a d e . Trans. A. Stonier and F. Benham. London: W.
Hodge, 1936; revised edn, 1937.
Harrigan, J. 1997. Technology, factor supplies and interna-
tional specialization: Estimating the neoclassical model.
A m e r i c a n E c o n o m i c R e v i e w 87: 475-494.
Heckscher, E. 1919. The effects of foreign trade on the
distribution of income. In E k o n o m i s k T t d s k r i f t .
English translation in R e a d i n g s i n t h e T h e o r y
o f I n t e r n a t i o n a l T r a d e , ed. H.S. Ellis and L.A.
Metzler. Philadelphia: Blakiston, 1949.
Helpman, E., and P. Krugman. 1985. M a r k e t s t r u c t u r e
a n d f o r e i g n t r a d e . Cambridge, MA: MIT Press.
Jones, R.W. 1956. Factor proportions and the Heckscher-
Ohlin theorem. R e v i e w o f E c o n o m i c S t u d i e s
24: 1-10.
Jones, R.W. 1961. Comparative advantage and the theory of
tariffs. R e v i e w o f E c o n o m i c S t u d i e s 28: 161-
175.
Jones, R.W. 1965. The structure of simple general equilib-
rium models. J o u r n a l o f P o l i t i c a l E c o n o m y
73: 557-572.
Jones, R.W. 1971. A three-factor model in theory, trade and
history. In T r a d e , b a l a n c e o f p a y m e n t s , a n d
g r o w t h , ed. J. Bhagwati et al. Amsterdam: North-
Holland.
1900
Jones, R.W., and P. Neary. 1984. Positive trade theory. In
H a n d b o o k o f i n t e r n a t i o n a l e c o n o m i c s , ed.
R.W. Jones and P.B. Kenen, vol. 1. Amsterdam: North-
Holland.
Kantorovich, L. 1965. T h e b e s t u s e o f e c o n o m i c
r e s o u r c e s . Cambridge, .VIA: Harvard University
Press.
Kenen, P.B. 1965. Nature, capital and trade. J o u r n a l o f
P o l i t i c a l E c o n o m y 73: 437 460; Erratum,
December, 658.
Kingman, P.R. 1979. Increasing returns, monopolistic
competition and international trade. J o u r n a l o f
I n t e r n a t i o n a l E c o n o m i c s 9: 469 479.
Lancaster, K.J. 1980. Intra-industry trade under perfect
monopolistic competition. Journal of
I n t e r n a t i o n a l E c o n o m i c s 10: 151-175.
Learner, E.P. 1984. Sources of international comparative
advantage: Theory and evidence. Cambridge, MA: MIT
Press.
Lemer, A.P. 1952. Factor prices and international trade.
E c o n o m i c a 19: 1-15.
Leontief, W.W. 1933. The use of indifference curves in the
analysis of foreign trade. Q u a r t e r l y J o u r n a l o f
E c o n o m i c s 47: 493-503.
Leontief, W.W. 1954. Domestic production and foreign trade:
The American capital position re-examined. E c o n o m i c
I n t e r n a t i o n a l e 7: 9-38.
Lemer, A.P. 1932. The diagrammatic representation of cost
conditions in international trade. E c o n o m i c a 12: 345-
356.
Lemer, A.P. 1952. Factor prices and international trade.
E c o n o m i c a 19: 1-15.
MacDougall, G.D.A. 1951. British and American exports.
E c o n o m i c J o u r n a l 6 1: 697 724.
McKenzie, L.W. 1954. Specialization and efficiency in world
production. R e v i e w o f E c o n o m i c S t u d i e s 21
(3): 165 180.
Ohlin, B. 1933. I n t e r - r e g i o n a l a n d i n t e r n a t i o n a l
t r a d e . Cambridge, MA: Harvard University Press.
Oniki, H., and H. Uzawa. 1965. Patterns of trade and
investment in a dynamic model of international trade.
R e v i e w o f E c o n o m i c S t u d i e s 32: 1 5 38.
Pasinetti, L. 1960. A mathematical formulation of the
Ricardian system. R e v i e w o f E c o n o m i c S t u d i e s
27: 78-98.
Ricardo, D. 1951. T h e w o r k s a n d c o r r e s p o n d e n c e
o f D a v i d R i c a r d o , ed. P. Sraffa. vols. 1 and 4.
Cambridge: Cambridge University Press.
Samuelson, PA. 1948. International trade and the equalization
of factor prices. E c o n o m i c J o u r n a l 58: 163 184.
Samuelson, PA. 1949. International factor price equalization
once again. E c o n o m i c J o u r n a l 59: 181 197.
Samuelson, PA. 1950. Evaluation of real national income.
Oxford Economic Papers 2: 1 29.
Samuelson, PA. 1953. Prices of factors and goods in general
equilibrium. R e v i e w o f E c o n o m i c S t u d i e s 21:
1-20.
Samuelson, PA. 1962. The gains from international trade once
again. E c o n o m i c J o u r n a l 72: 820 829.
Samuelson. P.A. 1965. Equalization by trade of the interest
rate along with the real wage. In T r a d e , g r o w t h
Comparative Statics
b a l a n c e o f p a y m e n t s , ed. R. Baldwin et al.
Chicago: Rand McNally.
Samuelson, P.A. 1971. Ohlin was right S w e d i s h J o u r n a l
o f E c o n o m i c s 73: 365-384.
Steedman, I. 1979a. T r a d e a m o n g s t g r o w i n g
e c o n o m i e s . Cambridge: Cambridge University Press.
Steedman, I., ed. 1979b. F u n d a m e n t a l i s s u e s i n
t r a d e t h e o r y . London: Macmillan.
Slight/, .1. 1970. Factor-price equalization in a dynamic
economy. J o u r n a l o f P o l i t i c a l E c o n o m y 78:
456—4SS.
Stolper, W.. and P.A. Samuelson. 1941. Protection and real
wages. R e v i e w o f E c o n o m i c S t u d i e s 9: 58-73.
Trefler, D. 1995. The case of missing trade and other
mysteries. A m e r i c a n E c o n o m i c R e v i e w 85:
1029-1046.
Vanek, .1. 1963. The natural resource content of U.- S. Foreign
trade 1870-1955. Cambridge. MA: MIT Press.
Vincr, .1. 1937. Studies in the theory of international trade. New
York: Harper.
Comparative Statics
John Nachbar
Abstract
Comparative statics in competitive general
equilibrium (GE) environments provide insight
into the operation of GE models and a means to
confront GE models with data. This article
focuses on a canonical comparative statics
prediction: in an exchange economy, aggregate
endowment changes are negatively related to
equilibrium price changes. In particular, an
increase in the aggregate endowment of a com-
modity lowers its equilibrium price.
Keywords
Aggregate excess demand; Asset pricing;
Comparative statics; Competitive general equi-
librium; Debreu-Mantel-Sonnenschein theorem;
Demand shocks; Endowment changes; General
equilibrium; Gross substimtes; Hicks, J.; Partial
equilibrium; Production economies; Rybcyznski
theorem; Stolper-Samuelson theorem; Supply
shocks; Weak axiom of revealed preference
Comparative Statics 1901

JEL Classifications Good 2

C6

Comparative statics in competitive general equi-

librium (GE) environments provide insight into the
operation of GE models and a means, at least in
principle, to confront GE models with data.
For concreteness, I focus most of this article on
what is arguably the canonical GE comparative
statics conjecture: in finite exchange economies
(that is, no production), equilibrium price changes
are negatively related to endowment changes. In
particular, if the endowment of good 1 increases
and the endowments of other goods remain the
same, then the price of good 1 falls. At the end of
this article, I briefly survey other GE comparative
statics results.
I break the analysis into three cases of increasing
complexity.
Case I
There is a single consumer and two commodities. In
an equilibrium of this trivial economy, the con-
sumer eats her own endowment. Equilibrium rela-
tive prices (which are well defined even though
there is no trade) are given by the slope of the
consumer’s indifference curve through her con-
sumption/endowment point, co*. Let E denote her
wealth expansion path through her initial endow-
ment; E is the set of points where the slope of her
indifference curve is the same as at to *.
If the new endowment, co lies below E, as in
Fig. , then the equilibrium price ratio Pi/pi falls.
Ifco lies above E, then p , /p2 rises. Ifco lies along
E, then p\/p2 remains unchanged.
The differential version of Fig. is given by Fig. .
The vector p, the tangent to E, is the derivative with
respect to nominal wealth of each good’s demand
(p is mnemonic for marginal propensity to consume
vector). To first order, the rule is that p\lp2 falls if
and only if Aco, the vector of endowment changes,
lies within 180° clockwise from p. The vector p
incorporates second order information from the
utility function and is, in particular, typically not
collinear with the utility gradient.
Comparative Statics, Fig. 1 Comparative statics with one
consumer and two commodities
Comparative Statics,
Fig. 2 First order
comparative statics with
one consumer and two
commodities
If the endowment of good 1 increases while the
endowment of good 2 remains unchanged, then Aco
lies along the positive good 1 axis. Figure implies
that, in this case, p\tp2 falls if and only if good 2 is
normal (p2 > 0); whether good 1 is normal or
inferior (or even a Giffen good) is irrelevant.
Case II
There is again one consumer but L commodities. If
Aco lies along the positive good 1 axis, then a
natural conjecture, to generalize the above obser-
vation for L = 2, is that p\/pi falls for each good £ >
1, provided each of these goods is normal. Hicks (
) showed that this conjecture is false
in general but that it is true if the gross substitute
property holds (GS; the matrix of partial derivatives
of excess demand with respect to price has positive
off-diagonal entries). GS holds automatically in the
one-consumer, L 2, case because, at equilibrium,
when L 2, GS is equivalent to the weak axiom of
revealed preference (WA).
1902
Matters are more complicated if two or more
endowments are shifting at the same time. For
multivariate endowment shocks, there appears to
be little one can say in general about changes in
the price ratio of any specific pair of commodities.
Instead, the conjecture is that Am is negatively
related to Ap, the vector of equilibrium price
changes. Formally,
Ap ■ Am < 0.
Call this the comparative statics inequality, CS
for short. Geometrically, CS says that the vectors
Ap and A g o lie at least 90° apart.
To interpret Ap as a change in relative prices,
prices must be normalized. Consider linear price
normalizations, in which all prices, in both the
original economy and the perturbed economy,
satisfy p ■ A = 1, where A is an L vector. If all the
coordinates of A, are positive, then a fall in the
normalized price of good 1 means that the ratio
Pi
P- I'^-i
falls, where p_\ and A_\ are subvectors
corresponding to all goods other than the first: the
price of good 1 falls relative to the value of a
commodity bundle consisting of AS: units of each
good £ > 1. Standard choices of A include X = (0, ...
, 0, 1) (use the last commodity as numeraire) and A,
m* (normalize prices so that GNP remains constant;
this is the Laspeyres normalization). Regardless of
how, or whether, actual prices are normalized, one
canre-normalize prices using whatever X one
chooses.
I can provide intuition for CS most easily by
continuing to use figures for a two-good economy.
Fix a normalizing vector X Since p ■ X = 1 for all p,
Ap ■ A 0. Therefore, Ap lies along the line that is at
right angles to X, labelled T-, in Fig. .
As drawn, Am lies within 180° clockwise from
p and hencep\/pi falls. Therefore, Ap, normalized by
A, lies on the upper left-hand branch of 7). As
illustrated, Am and Ap are more than 90° apart;
hence CS holds.
On the other hand, suppose that Am lies in the
cone spanned by X and p. Since Am again lies
Comparative Statics
Comparative Statics, Fig.
3 Condition CS with two
goods
within 180° clockwise from p, p\tpi again falls and
Ap again lies on the upper left-hand branch of Tx.
Now, however, AGO and Ap are less than 90° apart.
CS fails.
In general, in a one-consumer economy, for any
number of commodities and for any preferences, CS
fails whenever there is a gap between X and p and
Aot falls into this gap. Conversely, if X p (or, more
generally, if A is a scalar multiple of p) then CS
holds for any endowment change: Ap ■ Am < 0 with
Ap ■ Am = 0 if and only if Am is collinear with p
(which is the differential analog of Aot landing on
the wealth expansion path E in Fig. ). In one
consumer economies, X = /(is thus the unique (up to
scalar multiplication) linear price normalization for
which CS holds for all endowment changes.
If preferences are quasi-linear in good L, and
consumption is interior, then A = p implies A =
(0, ... , 0, 1); the last good is used as numeraire. If
the preferences are homothetic then p is a scalar
multiple of the reference endowment, <’>*, and so
one can set A m*. Typically, however, A p i s
different from price normalizations commonly used
in economics.
The X = p normalization, although nonstandard,
does have a sensible interpretation, provided p is
positive (all goods are weakly normal). If p is
positive, then a decrease in pi means that Pi/(P-i ■ p
i) falls: the price of good 1 falls relative to the value
of the consumer’s marginal consumption of all other
goods.
If Am lies along the positive good 1 axis and
goods 2, ..., L are normal, then a minor variation on
CS implies that p\Kp \ ■ p () falls, even if good 1 is
inferior. This is a Weaker conclusion then that of
Hicks ( ) but it has a weaker
hypothesis, since it does not assume the gross
substitute property.
Comparative Statics 1903

Case III which CS holds for all endowment changes. As the

endowment distribution changes, the price
There are I consumers and L commodities. The normalization may have to change.
generalization of CS is This illustrates an issue that has become a
central theme in the recent literature on GE com-
Ap • Acd < 0, parative statics. Given an arbitrary price normali-
zation, standard GE assumptions impose no
where Aco denotes the change in the aggregate
restrictions on the relationship between changes in
endowment. CS holds provided one uses an
equilibrium prices and changes in the aggregate
appropriate aggregate version of ji. Consider two
endowment (see Chiappori et al. ). This negative
alternatives. Each is a weighted sum of the indi-
result, a cousin of the Debreu-Mantel-
vidual marginal propensity to consume vectors, ji‘.
Sonnenschein theorem (DMS), has a loophole:
In the first, 7^, the weight on /(' is consumer Vs
standard GE assumptions do provide comparative
share in the change in the value of consumption,
statics restrictions if one works with micro-level
evaluated at the prices of the reference equilibrium.
information (for example, on the endowment dis-
In the second, JIAm, the weight on /(' is Vs share in
tribution) rather than exclusively with aggregates.
the change in the value of the endowment, again
In the CS results, micro-level data is used in the
evaluated at reference equilibrium prices.
price normalization. Note that CS requires micro
If one normalizes prices using X = JtAv, then
data even if one assumes that aggregate excess
inequality CS holds provided individual excess
demand satisfies WA.
demand satisfies the weak axiom (WA) at equilib-
Relative to the objectives laid out in the first
rium. If X = JiAco, then CS holds provided aggre-
paragraph of this article, the results on CS compar-
gate excess demand satisfies WA at equilibrium.
ative statics fare reasonably well in providing
See Nachbar ( ).
insight into the operation of GE models. The JiAo}
The hypothesis that aggregate excess demand
result is much the easier to interpret, since it is
satisfies WA is not implied by standard GE
computed with the use of endowment changes,
assumptions. One justification for nevertheless
which are exogenous. The result, on the other hand,
assuming WA is that it seems to be connected to the
extends easily to production economies. In contrast,
dynamic stability of the price adjustment process.
I do not know whether the JiA(0 result has a useful
WA holding at equilibrium is sufficient and almost
analog for production economies.
necessary for local asymptotic stability under the
The CS inequality fares less well as a tool for
Walrasian tatonnement, for example. Comparative
empirical work, because it requires a large amount
statics, by assuming that economies are at
of data just to estimate the normalization vector.
equilibrium, may therefore implicitly assume that
The necessity results imply that this difficulty is
aggregate excess demand satisfies WA. A second
intrinsic to CS.
justification for assuming that aggregate excess
demand satisfies WA is that this assumption, while
strong, is not implausible in exchange economies. Other Comparative Statics Results
For some sufficient conditions for WA, see Becker
( ), Hildenbrand Brown and Matzkin ( ), a path-breaking paper
( ), Grandmont ( ) and Quah ( ). that has heavily influenced subsequent work in this
In the one-consumer case, the X = /r price area, exploits the DMS loophole noted above to
normalization was necessary as well as sufficient. give testable restrictions linking equilibrium prices
There are analogous, but weaker, necessity results with individual endowments. For related work, see
for Jifc and/ZA(J). The important implication is that, Snyder ( ), Williams
because both Ji^ and JiAo} can vary with how the ( ), Kiibler ( ) and Chiappori et al.
endowment changes are distributed across con- ( ). Relative to CS, the Brown-Matzkin
sumers, there may be no price normalization for restrictions are easier to implement empirically
1904
because they do not require estimating normaliza-
tion vectors, but they are harder to interpret.
As already noted, CS-type reasoning can be
extended to production economies (see Quah ;
Nachbar ). CS-type reasoning can also be extended
to asset pricing environments (see Quah ).
For shocks to preferences rather than endow-
ments or technologies, the analog of CS is
Ap ■ Ax < 0 ,
where Ax is the change in equilibrium consumption.
Profit maximization implies that this inequality
holds for any price normalization. In this respect,
the analysis of demand shocks is trivial compared
with the analysis of supply shocks.
Interest in comparative statics has helped moti-
vate research on the uniqueness, regularity, and
stability of equilibria (see Kehoe ). Note that some
of the comparative statics results cited above (for
example, the Brown-Matzkin results and the /7,\v
CS result) do not assume uniqueness or stability.
Finally, perhaps the most famous comparative
statics results are the Stolper-Samuelson theorem
and its dual, the Rybcyznski theorem (for a recent
treatment, see Echenique and Manelli ). Stolper-
Samuelson links changes in factor prices with factor
intensities and changes in output prices.
Rybcyznski links changes in final goods production
with factor intensities and changes in factor
supplies. Although it is possible to embed these
results within a highly restricted GE model, they are
partial equilibrium in spirit; wealth effects play no
role.
See Also
Bibliography
Becker, G. 1962. Irrational behavior and economic theory.
J o u r n a l o f P o l i t i c a l E c o n o m y 70: 1-13.
Compensated Demand
Brown, D., and R. Matzkin. 1996. Testable restrictions on the
equilibrium manifold. E c o n o m e t r i c a 64:1249-1262.
Chiappori, P., I. Ekelund, F. Kiibler, and H. Polemarchakis.
2004. Testable implications of general equilibrium the-
ory: A differentiable approach. J o u r n a l of
M a t h e m a t i c a l E c o n o m i c s A O : 105-119.
Echenique, F., and A. Manelli. 2005. C o m p a r a t i v e
statics, English auctions, and the
S t o l p e r - S a m u e l s o n t h e o r e m . Mimeo. Tempe:
Arizona State University.
Grandmont, J. 1992. Transformations of the commodity
space, behavioral heterogeneity and the aggregation
problem. J o u r n a l o f E c o n o m i c T h e o r y 57: 1-
35.
Hicks, J. 1939. V a l u e a n d c a p i t a l . Oxford: Clarendon
Press.
Hildenbrand, W. 1983. On the law of demand.
E c o n o m e t r i c a 51: 997-1018.
Kehoe, T. 1987. Comparative statics. In T h e n e w
P a l g r a v e : A d i c t i o n a r y o f e c o n o m i c s , ed.
J. Eatwell, M. Milgate, andP. Newman, vol. 1.
Basingstoke: Palgrave.
Kiibler, F. 2003. Observable restrictions of general equi-
librium with financial markets. J o u r n a l of
E c o n o m i c T h e o r y 110: 137-153.
Nachbar, J. 2002. General equilibrium comparative statics.
E c o n o m e t r i c a 1 9 : 2065-2074.
Nachbar, J. 2004. General equilibrium comparative statics:
The discrete case with production. J o u r n a l o f
M a t h e m a t i c a l E c o n o m i c s 40: 153-163.
Quah, J. 1997. The law of demand when income is price
dependent. E c o n o m e t r i c a 65: 1421-1442.
Quah, J. 2003. Market demand and comparative statics when
goods are normal. J o u r n a l o f M a t h e m a t i c a l
E c o n o m i c s 39: 317-333.
Snyder, S. 1999. Testable restrictions of Pareto optimal public
good provision. J o u r n a l o f P u b l i c E c o n o m i c s
71:97-119.
Williams, S. 2002. Equations on the derivatives of an initial
Compensated Demand
Tatsuo Hatta
Hicks Compensation: Definition
When a consumer faces a price change under a
given nominal income, his utility (or real income)
level as well as his demand vector changes. Sup-
pose, however, that his income level is
Compensated Demand 1905

simultaneously changed as the price is changed so e(p,p) = mm{p'x\u(x) = p}.

as to keep his utility at the initial level. This
operation may be regarded as a compensation for
the price change, and we call the resulting demand
vector the compensated demand for the new price.
Thus the compensated demand is a function of
the price vector and the utility level, and we may
write it as
X
By definition, we obviously have
e p i p ,p ) ^ p ' h (p ,p ) . (6)
There is a less obvious, but extremely useful,
relationship between the compensated demand and
the expenditure functions:

x
= h(p,p), (1 ) ep{p,p)=h{p,p). (7)

where x and p are the consumption and price This identity usually referred to as the S h e p h a r d -
vectors, while p is the utility level. We call h the S a m u e l s o n L e m m a was obtained by Hicks
compensated (or Hicks) demand function. Formally, (1946, p. 331), Samuelson ( , p. 68,
it may be defined as the solution function of the -54, pp. 15-16), and Shephard ( ).
following minimization problem: To prove the Shephard-Samuelson Lemma, let x
be an expenditure-minimizing vector that yields p
minp'x subject to u(x) = p, (2) where u is at the price p , i.e.,

x*=h(p*,p). (8)
the utility function.
Define the gain function g by

Basic Properties g(p) = e(p,p)-p'x*. (9)

This and the definition of e imply that g(p) < 0.
Hicksian Demand Rules Also from ( ) and ( ), we have g(p*) = 0. Hence the
The Jacobian matrix of h with respect to p, denoted function g takes its minimum value of 0 when p p .
as hp, is nothing but the Hicks substitution matrix. It Therefore, the first and the second order
has well-known properties: minimization conditions yield

p'hp(p,p) = 0 (3) gp(p*) = 0 ( 10 )

y h P ( p , p ) y < 0 for all y . (4) and

hP(p,/j) = h'p(p,p). (5) y'gPPip*)y > o for a11

y ± o. (ii)
Condition ( ) is called the homogeneity condition, Equation immediately proves ( ).
since it shows that the function h is homogeneous of To demonstrate the usefulness of the Shephard-
degree zero with respect to p. Conditions ( ) and ( ) Samuelson Lemma, let us prove the Hicksian
are called the negative semi- definiteness and the Demand Rules from this Lemma. From ( ) we have
symmetry conditions, respectively. We will call
these three conditions the Hicksian Demand Rules. ePP(p,p) =hp(p,p).

This immediately yields ( ). In view of ( ) and ( ),

Shephard-Samuelson Lemma
this also proves ( ). On the other hand, ( ), ( ) and ( )
The minimized expenditure value of problem ( ) is a
yield
function ofp and p. This is called the expenditure
function. Formally, we define it by
tion by

v(jj,y) = u[m(p,y)\. (13)

1 = efl\p,v(p,y)]vy(p,y),
Problem ( ) may be regarded as the expenditure
(1 8)
minimization problem associated with ( ). Newman
( ) calls it the reflection or the mirror image of respectively. Multiplying by vy(p,y) on both sides of
maximization problem ( ). If we let p v(p,y) in ( ), and then applying ( ), ( ) and ( ), we
problem ( ), the resulting minimum expenditure obtain ( ).
must equal y and the compensated demand must be
equal to m(p,y) in Problem ( ). Thus we have Slutsky-Hicks Decomposition
Identities ( ) and ( ) also yield the
y = e\p,v(p,y)\ (14) Slutsky-Hicks decomposition,

and mp(p,y) = hp\p,v(p,y)\ - my(p,y)m(p,y).

(19)
m(p,y) = h\p,v(p,y)]. (15) Thus the slope of the ordinary demand function
equals the slope of the compensated demand func-
Roy's Identity
tion adjusted to the income effect.
These identities yield Roy’s identity, To prove ( ), differentiate ( ) with respect to
v
p and y to get
p(p,y) = -vy(p,y)m(p,y). (16)
mp(p,y) = hp\p,v(p,y)} + ty\p,v(p,y)}vp(p,y)
To see this, differentiate ( ) with respect to p (20)
and y to get
and
e
i\P’v(p,y)\vp(p,y) = - ep\p,v(p,y)}
f (17)
nb(p,y) =A^|p.v(p,y)]vy(p,y), (21)
and
respectively. The only difference between ( ) and ( )
is their income terms. Applying ( ) and ( ) to the last
term of ( ), we get ( ).

Slutsky Compensation

After a price change takes place, the Hicks com-

pensation keeps the consumer on the same utility
level as before the price change. As Mosak ( )
pointed out, however, Slutsky had a dif
ferent concept of compensating the loss of real
income. Slutsky considered a compensation that
‘makes possible the purchase of the same quantities
of all the goods that had formerly been bought’,
When a price change takes place, the Hicks-
compensated and the Slutsky- compensated demand
effects are generally different. When the price
change is infinitesimal, however, they become
equal, and this equality is called Mosak’s Equality.
Compensated Demand
Mosak’s equality has played an important role in
index number theory. The Laspeyres index, which
is widely adopted in practice, is based on the
Slutsky compensation, since it indicates the change
in income that would be needed in the current year
in order to buy the commodity bundle bought in the
base year. Under the Slutsky substitution effect, the
individual can be no worse off and is likely to be
better off since he is able to purchase at least the
bundle he had before the price change. Thus he is
‘overcompensated’ for the price change (see
Samuelson , pp. 4-5). The price index that truly
reflects the utility change should be based on the
Hicks compensation. Such an index is difficult to
compute because utility levels are not observable.
Mosak’s equality reveals, however, that for small
price changes the Laspeyres index is a good
approximation to the ‘ideal’ index.
Let us now formally state Mosak’s equality.
Define the function s by
s(p,x) = m(p,p'x).
( 22 )
The function s is the demand function with the fixed
endowment bundle x. We call sp \p, m (p, y)\ the
Slutsky substitution matrix. It represents the
variation in demand when the price change is
accompanied by an income compensation that
keeps the original consumption bundle m(p,y) on
the budget plane. Mosak’s Equality may now be
expressed as
sp\p,m(p,y)\ = hp\p,v{p,y)\. ( 23 )
To prove this, first define the function w by
w(p,x) = v(p,p'x). ( 24 )
The value of w(p,x) represents the maximized
utility level when the endowment bundle .r is given.
Differentiating this and applying Roy’s identity, we
get
wp\p,m(p,y)]=0. ( 25 )
Thus, if the utility maximizing bundle under a
certain price vector happens to be equal to the
1907
endowment bundle, the utility level is hardly
changed by a slight change of the price vector away
from the initial one. Equations ( ), ( )
and ( ), yield
s(p,x) = h\p,w{p,x)\.
Differentiating this with respect to p and noting ( ),
we immediately have ( ).
Historical Notes
Slutsky ( ) first established the homogeneity
and symmetry conditions on the substitution matrix.
Since he did not have the concept of utility-
maintaining compensation or the compensated
demand function, he derived these properties for the
Slutsky compensated substitution matrix sp \p. m(p,
y)] rather than for the Hicksian matrix hp [p. v(p,y)\.
During the 1930s, Hicks and Allen ( ) and Hicks (
, 1946) gave verbal
interpretations to the substitution matrix in terms of
the Hicksian compensation. But in their formal
derivation of its properties, they defined the sub-
stitution matrix to be the Slutsky substitution
matrix, as is clear from the following passage from
the Mathematical Appendix of Hicks (1946, p.
309):
.. .it follows from the equation that the substitution
term represents the effect on the demand for xs of a
change in the price of xr combined with such a
change in income as would enable the consumer, if he
chose, to buy the same quantities of all goods as
before, in spite of the change in Pr.
Thus they too did not explicitly state the func-
tion h, much less gave a name to it.
To the writer’s knowledge, Samuelson ( ) is
the first author who explicitly stated ( ) and derived
Hicksian rules directly from it, though he did not
give a name to it (see Samuelson , (43) on p. 103
and (99) onp. 114). Subsequently, Samuelson ( , p.
8, nl) gave a heuristic proof of the
symmetry and the negative-semidefiniteness condi-
tion of tiie Slutskycompensated substitution matrix
as envelope properties, i.e. in a spirit very much
similar to the one given above. In this path-breaking
proof, however, he relied upon the indirect utility
1908
junction rather than the expenditure junction, thus
without using the mirror image minimization prob-
lem or the compensated demand junction.
McKenzie ( ) and Karlin ( ) concept of consumers’ surplus, and Alonso’s ( )
explicitly
defined the function h, and derived its properties by
taking full advantage of the Shephard- Samuelson
Lemma. However, they had to use a global property
of the expenditure function in deriving the negative
semidefiniteness condition, which is a local
property. The proof of this condition employed
above is solely based on a local minimization
condition, and serves as the mirror image
counterpart of Samuelson’s ( ) proof for the
Slutsky substitution matrix. Diamond and McFad-
den (1973) attribute this proof method to Gorman
(see Gorman ). The above proof of Mosak’s
equality is due to Hatta and Willke ( ).
Silberberg ( ) considered the general max
imization problem with possibly many constraint
functions where both target and constraint functions
may be non-linear. Extending Samuelson’s ( )
proof method, he showed that the compen
sated solution function of that problem satisfies
generalized forms of the symmetry and non-
negative definite conditions, as long as the
constraint functions do not contain shift parameters.
Silberberg’s proof boils down to Gorman’s in the
standard expenditure minimization problem. Hatta (
) extended the concept of the compen
sated demand function to the case where the same
shift parameters may appear simultaneously in both
target and constraint functions in the general
problem. He showed that the properly compensated
solution function in that problem satisfies
generalized forms of symmetry and non-negative
definite conditions. He also established an envelope
theorem that contains both the Shephard-Samuelson
Lemma and Roy’s Identity as special cases. His
proof integrates Samuelson’s ( ) and
Gorman’s into one. The global char
acterization of the compensated demand function
by McKenzie and Karlin was extended into duality
theory, as surveyed by Diewert ( ).
In many branches of economics outside the
demand theory, the concept of the compensated
demand function was implicitly used without being
explicitly stated. Examples are Hotelling’s
Compensated Demand
( ) and Harbeiger’s ( ) analysis of the
excess burden of taxation, H i c k s ’ s ( ) compen
sating and equivalent variation that illuminate the
rent bidfunction, which keeps a consumer’s
utility constant regardless of the location he
chooses. A number of economists of the Chicago
School, including Friedman ( ), Bailey ( ),
and Becker ( ), used the concept of the Hicks
compensation in various welfare analyses. Each of
these authors gave a different name to the concept.
Friedman called its graph the Marshallian demand
curve contrary to the current usage of this term;
Bailey, the constant-real-income demand curve, and
Becker, the pure demand curve.
The explicit use of the compensated demand
junction gave rise to dramatically clearer restate-
ments and proofs of many existing theorems. Its
usefulness has reached far beyond that, however.
Since the early 1970s, this function has been used
for the analyses of the welfare impacts of para-
metic shifts in various general equilibrium models,
as stated by Ohyama ( ), Takayama
( ), Diamond and MacFadden ( ), Dixit
( ) and Hatta ( , ), and comprehen-
sively studied by Dixit and Norman ( ) and
Woodland ( ).
The history of the compensated demand func-
tion is curious. The properties of its derivatives
were known and the concept of Hicks compensation
used iu many fields of economics before the
function i tself was stated or named. Perhaps this is
because economists has an unconscious reluctance
in putting the elusive concept of the utility level as a
variable of a function. Once explicitly stated and
well understood, however, the compensated demand
function has found a powerful use in welfare
economics, precisely because it has the utility level,
rather than income, as an explicit variable.
See Also
Compensating Differentials
Bibliography
Alonso, W. 1964. L o c a t i o n and land use.
Cambridge, MA: Harvard University Press.
Bailey, M. J. 1954. The Marshallian demand curve.
J o u r n a l o f P o l i t i c a l E c o n o m y 62: 255-261.
Becker, G.S. 1971. E c o n o m i c t h e o r y . New York:
Knopf.
Diamond, P.A., and D.L. McFadden. 1974. Some uses of
expenditure function in public finance. J o u r n a l o f
P u b l i c E c o n o m i c s 3: 3-21.
Diewert, W.E. 1982. Duality approaches to microeconomic
theory. In H a n d b o o k of mathematical
e c o n o m i c s , vol. 2, ed. K.J. Arrow and M.D.
Intrihgator, 535-599. Amsterdam: North-Holland.
Dixit, A.K. 1975. Welfare effects oftax and price changes.
J o u r n a l o f P u b l i c E c o n o m i c s 4 : 103-123.
Dixit, A.K., and V. Norman. 1980. T h e o r y of
i n t e r n a t i o n a l t r a d e . Cambridge: Cambridge
University Press.
Friedman, M. 1949. The Marshallian demand curve. J o u r -
n a l o f P o l i t i c a l E c o n o m y 57: 463-495.
Gorman, W.M. 1976. Tricks with utility functions. In
E s s a y s i n e c o n o m i c a n a l y s i s , ed. M. Artis and
R. Nobay. Cambridge: Cambridge University Press.
Harberger, A.C. 1974. T a x a t i o n a n d w e l f a r e . Boston:
Little, Brown.
Hatta, T. 1977. A recommendation for a better tariff struc-
ture. E c o n o m e t r i c a 45: 1859-1869.
Hatta, T. 1980. Structure of the correspondence principle at
an extremum point. R e v i e w of Economic
S t u d i e s 47: 987-997.
Hatta, T., and R.J. Willke. 1982. Mosak’s equality and the
theory of duality. I n t e r n a t i o n a l E c o n o m i c
R e v i e w 2 2 : 361-364.
Hicks, J.R. 1939. V a l u e a n d c a p i t a l . London: Oxford
University Press.
Hicks, J.R. 1956. A r e v i s i o n o f d e m a n d t h e o r y .
London: Oxford University Press.
Hicks, J.R., and Allen, R.D.G. 1934. A reconsideration of the
theory of value, I, II. E c o n o m e t r i c a , N.S.I, 52-75,
196-219.
Hotelling, H. S. 193 8. The general welfare in relation to the
problems of taxation and of railway and utility rates.
E c o n o m e t r i c a 6: 242-269.
Karlin, S. 1959. M a t h e m a t i c a l m e t h o d s a n d
theory in games programming and
economics, vol. 1. Reading: Addison-Wesley
Publishing Company.
McKenzie, L. 1956. Demand theory without a utility index.
R e v i e w o f E c o n o m i c S t u d i e s 24: 185-189.
Mosak, J. 1942. On the interpretation of the fundamental
equation of value theory. In S t u d i e s in
mathematical economics and
e c o n o m e t r i c s , ed. O. Lange, 69-74. Chicago:
University of Chicago Press.
Newman, P.K. 1982. Mirrored pairs of optimization prob-
lems. E c o n o m i c a 49: 109-119.
Ohyama, M. 1974. Tariffs and the transfer problem. K e i o
E c o n o m i c S t u d i e s 11(1): 29.
1909
Samuelson, P.A. 1953. Consumption theorems in terms of
overcompensation rather than indifference comparison.
E c o n o m i c a 20(February): 1-9.
Samuelson, P.A. 1965. Using frill duality to show that
simultaneously additive direct and indirect utilities
implies unitary price elasticity of demand.
E c o n o m e t r i c a 33: 781-796.
Shephard, R. 1953, 1970. C o s t a n d p r o d u c t i o n
f u n c t i o n s . Princeton: Princeton University Press.
Silberberg, E. 1974. A revision of comparative statics
methodology in economics, or, how to do comparative
statics on the back of an envelope. J o u r n a l o f
E c o n o m i c T h e o r y 7: 159-172.
Slutsky, E. 1915. Sulla teoria del bilancio del consumatore.
G i o r n a l e d e g l i E c o n o m i s t i 51: 1-26. English
trans. in R e a d i n g s i n p r i c e t h e o r y , ed. G.J.
Stigler and K.E. Boulding, Chicago: Chicago University
Press, 1953.
Takayama, A. 1974. On the analytical framework of tariffs
and trade policy. In T r a d e s t a b i l i t y a n d
macroeconomics, essays in honor of
L l o y d A . M e t z l e r , ed. G. Horwich and P.A.
Samuelson, 153-178. New York: Academic.
Woodland, A.D. 1982. I n t e r n a t i o n a l t r a d e a n d
r e s o u r c e a l l o c a t i o n . Amsterdam: North-Holland.
Compensating Differentials
Matthew E. Kahn
Abstract
Compensating differentials represent a wage
premium for unpleasant aspects of a job. Jobs
differ along several dimensions. Some jobs offer
generous health insurance benefits. Others entail
long hours or may expose workers to physical
risks. Some are available only in polluted cities.
In equilibrium, labour markets accommodate
diversity by establishing wages that tend to
make different jobs relatively close substitutes at
the margin. Using hedonic wage regression
techniques, researchers have estimated the
equilibrium implicit market price that workers
pay, through lower wages, for working in a
more pleasant setting. This technique is widely
used by labour and environmental economists.
1910
Keywords
Compensating differentials; Hedonic wage
function; Labour economics; Rosen, S.; Super-
stars, economics of; Unobserved skill; Urban
economics; Urban environment and quality of
life; Wage heterogeneity, sources of; Wage
premium; Worker heterogeneity
JEL Classifications
J300
Compensating differentials represent a wage pre-
mium for unpleasant aspects of a job. Jobs differ
along a number of dimensions. Some jobs offer
generous health insurance benefits. Other jobs entail
long hours or may expose workers to physical risks.
Some jobs are only available in polluted cities. The
theory of compensating differentials is based on the
simple premise that there is ‘no free lunch’. In
market equilibrium, more unpleasant jobs will offer
a wage premium relative to other jobs. Similarly,
homes in nicer communities or high-quality-of-life
cities will sell for a premium. To quote Sherwin
Rosen ( , p. 2), ‘Markets
accommodate diversity by establishing prices that
tend to make different things relatively close sub-
stitutes at the margin. Adam Smith’s insight that
market prices tend to equalize their net advantages
is fundamental to these problems. If one good has
more desirable characteristics than another, the less
preferred variety must compensate for its
disadvantages by selling at a lower price.’
Defining Compensating Differentials
Jobs represent tied bundles of attributes. Suppose
that a worker gains utility from earning a wage and
from a job attribute. This attribute could represent
job safety, or total days of vacation, or health
insurance benefits. As shown in Fig. , there are two
jobs, A and B. Each job represents a different
bundle of a wage and a non-market job-specific
amenity level. The two jobs differ: job B is the
more pleasant of the two. If all workers have the
same utility function, then in
Compensating Differentials
Amenity
Compensating Differentials, Fig. 1
equilibrium this representative worker must be
indifferent between the two jobs. Thus, job A must
pay a higher wage than job B to compensate this
worker.
The econometrician can collect data on each job
type’s wage and amenities. In a more realistic
economy where there are many types of jobs that
differ with respect to the wage and their amenity
level, the representative worker’s indifference curve
would be sketched out. The slope of the
representative worker’s indifference curve repre-
sents the compensating differential of how much
lower a wage this worker would accept in return for
a small increase in the job amenity.
To see how worker heterogeneity affects the
interpretation about observed compensating dif-
ferentials, consider the simple extension where we
introduce two types of workers. These workers are
equally productive but differ with respect to then-
demand for working in the more pleasant job. In
Fig. , worker 1 values the job amenity more than
worker 2. In equilibrium, job A will pay a com-
pensating differential to attract workers to be will-
ing to work in this job. Worker 2 will choose to
work in job Awhile worker 1 will choose to work in
job B. Firm A will prefer to hire worker 2 rather
than worker 1 because worker 1 requires a larger
compensating differential for working in the more
unpleasant job. The profit maximizing firm seeks to
minimize its costs of production.
The econometrician will observe the equilibrium
wage paid to workers in job A and B. As shown in
Fig. , this equilibrium wage-amenity relationship
called the hedonic wage function
Compensating Differentials
Compensating Differentials, Fig. 2
does not represent either worker 1 ’s or worker 2’s
indifference curves. Instead, this hedonic wage
function represents the envelope of the minimum
wage that heterogeneous workers are willing to
accept to do a job. This simple example highlights
how introducing worker heterogeneity affects
inference from observed data (see Rosen ). Figure
focuses on just one dimension of worker
heterogeneity. The recent compensating differen-
tials literature has explored the consequences of
other dimensions of worker heterogeneity such as
unobserved skill (IQ, for example) and a worker’s
ability to self-protect against injury on the job
(Hwang, Reed and Hubbard ; Shogren and
Stamland ).
Labour Econometric Applications of
Compensating Differentials Theory
An enormous applied econometrics literature has
estimated various versions of hedonic wage func-
tions to recover estimates of the marginal valuation
of non-market job attributes. One major focus of
this research has been to estimate the value of life
by measuring how much of a wage premium the
marginal worker requires for working in a job with
a higher probability of death (Viscusi and Aldy ).
Other studies have used hedonic methods to
measure the compensating differential for mandated
government health insurance benefits (Gruber ).
1911
The standard approach utilizes a large microdata
set. The dependent variable in such a study is a full-
time worker’s wage in a specific occupation,
industry or city. For example, in Eq. the dependent
variable is the log of worker /’s wage in industry j in
year t. In an urban application, j would refer to a city
rather than an industry. The researcher will include a
large number of demographic controls, such as age,
ethnicity, or education, to ‘standardize’ the worker.
If one controls for these factors, the key variables of
interest are the Z’s in Eq. . In a labour economics
application, the Z vector may represent a set of job
specific attributes (length of day, job risk). In an
luban economics application, the Z vector may
represent attributes of the city where the job is
located (climate, pollution, crime).
Log (Wage,*) = y0 + + fi^j, + Uv (1)
Ordinary least squares regression estimates of 72
are used to construct measures of the compensating
differentials for job tasks and characteristics of
employment locations. Estimates of such
coefficients have been used to rank city quality of
life (see Gyourko and Tracy ) and represent the first
stage of the hedonic two-step for recovering demand
functions for non-market goods such as air quality
or climate (Rosen ; Ekeland et al. ).
If the population differs with respect to its tastes
for job attributes, then y2 can be used to construct a
worker’s budget constraint. For example, in a job-
safety regression if y2 equals minus $100 then this
means that a one-unit increase in job safety will cost
the worker an extra $100 in wages. The rational
worker facing this budget constraint will take this
trade-off into account when choosing the job that
maximizes her utility.
Hedonic estimates of compensating differentials
can also be used to bound worker preferences. To
return to Fig. , a lower bound on worker l’s
willingness to accept work in risky job A is the
equilibrium wage paid to worker 2. Since we know
that worker 1 chose the safe job and refused to work
in job A at the wage that worker 2 accepted, worker
2’s wage offer provides a lower bound (see Rosen ).
1912
The typical hedonic wage regression study
estimates Eq. using ordinary least squares. This
econometric approach will yield consistent
estimates of y2 if the unobserved determinants of
wages (that is, the error term) are uncorrelated with
the explanatory variables. What is the error term in
this hedonic pricing equation? While a researcher
might hope that it represents measurement error in
the dependent variable, it is more likely that the
error term represents unobserved attributes of the
worker and unobserved attributes of the
geographical area where the worker lives and
works.
Unfortunately for researchers, people selfselect
where to live and work. A researcher would like to
know what wage the same worker would earn in
every industry and in every city. In a cosmopolitan
city such as New York, superstars of all fields,
ranging from Don Trump in real estate to Derek
Jeter in baseball, have all chosen to work there. A
naive crosscity hedonic researcher would observe
these stars living in New York City earning high
wages relative to observationally identical people in
Tulsa, and would conclude, based on the wage
regression, that New York City’s quality of life
must be worse than Tulsa’s. Clearly, the problem
with this inference is the ‘apples to oranges’
comparison. New York City’s amenities are a
normal good. The high-skilled earn higher salaries
and are attracted to living and working in this city.
Conclusion
A job’s wage is not a sufficient statistic for its
quality. Coal miners are paid a relatively high wage
but the work is dangerous and unpleasant. A major
research agenda in labour economics investigates
how much people implicitly pay for non-market job
attributes. Credible estimates of wage compensating
differentials for living in less polluted cities or
working in risky industries would greatly aid policy
analysis that seeks to measure the benefits of
environmental and safety regulation.
Compensation Principle
See also
► A- Y I /
► Wage Inequality, Changes ir
Bibliography
Ekeland, I., J. Heckman, and L. Nesheim. 2004. Identification
and estimation of hedonic models. J o u r n a l o f
P o l i t i c a l E c o n o m y 112: S60-S109.
Gmber, J. 1994. The incidence of mandated maternity
benefits. A m e r i c a n E c o n o m i c R e v i e w 84: 622-
641.
Gyourko, J., and J. Tracy. 1991. The structure of local public
finance and the quality of life. J o u r n a l o f P o l i t -
i c a l E c o n o m y 91: 774-806.
Hwang, H.-S., R. Reed, and C. Hubbard. 1992. Compensating
wage differentials and unobserved productivity.
J o u r n a l o f P o l i t i c a l E c o n o m y 100: 835-858.
Rosen, S. 1974. Hedonic prices and implicit markets: Product
differentiation in pure competition. J o u r n a l o f
P o l i t i c a l E c o n o m y 82: 34-55.
Rosen, S. 2002. Markets and diversity. A m e r i c a n E c o -
n o m i c R e v i e w 92: 1-15.
Shogren, J., and T. Stamland. 2002. Skill and the value of life.
J o u r n a l o f P o l i t i c a l E c o n o m y 110: 1168-1173.
Viscusi, W., and J. Aldy. 2003. The value of statistical life:
A critical review of market estimates
throughout the world. Journal of Risk and Uncertainty 27:
5-76.
Compensation Principle
John S. Chipman
Abstract
The compensation principle holds that one of
two possible states constitutes an improvement
over the other if the gainers could compensate
the losers for their losses and still be at least as
well off as in the original state. The conflict
between potentiality and actuality - one situation
is judged better than another if everybody could
be made better off in the new situation even
though some in fact become worse off - ensures
that the compensation principle does not allow
for value-free policy decisions.
Compensation Principle
Keywords
Autarky; Baldwin envelope; Barone, E.;
Bergson-Samuelson social-welfare function;
Bickerdike, C. F.; Caimes, J. E.; Compensation
principle; Competitive equilibrium; Consumer
surplus; Cost-benefit analysis; Deadweight loss;
Debreu, G.; Dupuit, A.-J.-E. J.;
Exchange; Excise taxes; Free trade; Funda-
mental theorem of welfare economics; Gains
from trade; Game theory; General equilibrium;
Hicks, J. R.; Hotelling, H.; Ideal money;
Income-compensation function; Interpersonal
utility comparisons; Kaldor, N.; Kuznets, S.;
Lemer, A. R; Little, I. M. D.; Lump sum taxes;
Marshall, A.; Mishan, E. J.; Monopoly; New
welfare economics; Optimal tariffs; Pantaleoni,
M.; Pareto optimality; Partial equilibrium;
Pigou, A. C.; Proportional income tax; Revealed
preference; Ricardo, D.; Robbins, L. C.;
Samuelson, P. A.; Scitovsky indifference
surface; Scitovsky, T.; Taxation of income;
Transferable utility; Utility-possibility frontier;
Value judgements; Viner, J.
JEL Classifications
B4
The term ‘compensation principle’ refers to the
principle that, in comparing two alternative states in
which a given community of persons might find
itself, one of the states constitutes an improvement
over the other (in the weak sense including equiv-
alence) if it is possible for the gainers to compen-
sate the losers for their losses and still be at least as
well off as in the original state.
If the hypothetical compensation is actually car-
ried out, the principle reduces to the Pareto crite-
rion: all are at least as well off, in one state
compared to the other. There is no need to invoke
the compensation principle in such a case. On the
other hand, if the principle is used to compare two
unique alternative states in which a community
might find itself, neither of which is Pareto-
superior to the other, the principle seems quite
arbitrary unless interpreted in a broader context.
There is a sense in which one person might be said
to be basically healthier than another even
1913
though, at the particular moment, such a person
might have a cold and the other one not. The
compensation principle is usually used to make
comparisons in this sense; one state of the economy
is sounder, healthier, more robust, or has greater
productive potential, than another. What this
implies is that states under comparison are usually
not unique, singleton states but composite ones, or
sets of states. Formally, the objects being compared
are usually sets of commodity bundles that could be
made available to the aggregate of consumers,
described in the literature as ‘situations’ in contrast
to single ‘points’ in such sets (cf. Baldwin ).
Examples of comparisons in which the com-
pensation principle is typically used are those
between (a) a perfectly competitive system of
industrial organization and an imperfectly com-
petitive one; (b) free trade and no trade (or
restricted trade); (c) the state of an economy before
and after a war, or depression, or change in
productive techniques. Most but not all of these
types of comparisons are relevant to policy deci-
sions; and the policy decisions are usually not of an
ad hoc type (for which the compensation principle
would hardly be appropriate) but of a fundamental
nature concerning the underlying system of
industrial organization and trade.
Inasmuch as the principle can be applied with-
out the need to make interpersonal comparisons,
some of its more ardent proponents have maintained
that it is ‘value-free’. However, there can be no
doubt that it does require acceptance of some value
judgements, since the Pareto criterion itself
constitutes one - albeit a widely accepted one.
Another value judgement implicit in the principle as
it has usually been applied is that each individual is
the best judge of his or her own wellbeing; while
also quite widely accepted, this one is obviously
controversial, and in fact government policy
measures are often called for precisely in those
instances where it is clearly an untenable
assumption. But the most important and contro-
versial way in which value judgements enter into
the compensation principle is in the conflict
between potentiality and actuality: one situation is
judged better than another if everybody could be
made better off in the new situation even though
some in fact become worse off. This
1914 Compensation Principle

lacuna in the principle has led Little ( ) and , p. 272) and Marshall ( , p. 447; 1920,
Mishan ( p. 467) who used the concept of consumers’ surplus
) to formulations in which compen
to compare the losses of consumers (say from a
sation tests are combined with explicit distributional
value judgements, and Samuelson ( bridge toll or an excise tax) with the gains to the
,
government. The demonstration that the former
) into a full-fledged ethical system in which
exceed the latter, so that consumers cannot be
compensation is carried out to the extent that the
ethical norms dictate. compensated for their losses out of the government
revenues, provided a convincing case for the
In many applications the compensation principle
superiority of income tax to an excise tax (or for the
is difficult to formulate in a precise manner unless
one assumes absence of externalities in con- superiority of government subsidization of bridge
construction to its financing of it by tolls), and at
sumption, so it is usually formulated (but with some
the same time provided scientific prestige and great
notable exceptions - for example, Coase ) under the
intuitive appeal to a method that was able to reach
assumption that each person’s welfare depends only
on his or her own consumption of goods and such a definitive conclusion and furnish a measure
of the ‘deadweight loss’.
services. In most applications, the data available for
making comparisons are, almost inevitably, limited While Dupuit and Marshall used partial-
equilibrium analysis, Pareto (
to aggregative information on the actual state of the , p. 58) was
economy in each situation; much of the work in the first to introduce the concept into general-
equilibrium theory, in the course of an article
applying the principle therefore consists in using
devoted to proving the optimality of competitive
economic theory to make inferences from the actual
observations concerning underlying conditions in equilibrium. In the first part of this article
the economy. By its nature, the compensation (summarized by Sanger ), Pareto used as his
criterion of optimality the sum of individual
principle is limited in its application to comparing
alternative states (or sets of states) of a givenutilities; in the second part, however - acknowl-
edging the criticisms and suggestions of Pantaleoni
community of individuals; thus, it cannot be applied
and Barone (both admirers of Marshall, which
(at least not literally ) to historical comparisons of a
Pareto was not) - he reformulated the problem so as
country’s condition over time (since the population
has changed) or to comparisons of the living to sum not the utilities of different consumers but
conditions of different countries (since the the quantities they consume. His criterion of
optimality (1894, p. 60) was that it should be
populations are different). However, extensions of
impossible for one person to gain without another
the principle to cover such comparisons are possible
losing - ‘Pareto optimality’ - a criterion that had
provided suitable additional empirical assumptions
also been introduced by Marshall ( , pp. 449-50;
and value judgements are accepted; for example, if
all individuals are assumed to have identical 1920, pp. 470-1). A more
preferences, one could ask whether there exists arefined version of Pareto’s argument later appeared
in the Corns (Pareto -7, vol. 1, pp. 256-62; vol. 2, pp.
redistribution of income in each period (or country)
88-94).
such that each individual in the one situation would
be better off than each individual in the other. This The proposition formulated by Pareto ( )
anticipated what has now come to be known as the
would obviously entail additional value judgements
along with the additional empirical assumptions. ‘fundamental theorem of welfare economics’,
namely, that every competitive equilibrium is
Pareto optimal and, conversely, every Pareto opti-
Historical Development: From Dupuit to mum can be sustained by a competitive equilibrium.
Hotelling Pareto considered the problem faced by a socialist
state striving to attain an outcome in which it was
The compensation principle may be traced back to impossible for one person to gain without another
Dupuit ( , pp. 359-60; Arrow and Scitovsky losing. The Ministry of Justice would concern itself
with problems of income
Compensation Principle
distribution, and the Ministry of Production with
resource allocation and choice of production coef-
ficients. A weakness of Pareto’s argument was that
he assumed a price system already to be established
- perhaps our socialist state needs the prices of its
capitalist neighbours to guide it. Pareto further
assumed that each individual’s budget constraint
was adjusted by the addition of a parameter (a
lump-sum subsidy or tax) controlled by the
government. The government’s objective was to
maximize the sum of these parameters, which he
showed was equal to aggregate profit - the value of
commodities consumed less the value of factor
services supplied, equal to the value of firms’
output less the outlay on their factor inputs. If it
were possible to increase all the parameters, the
existing situation would not be Pareto optimal; if
their sum were a maximum, it would not be
possible to increase one of them without decreasing
another, and the outcome would be Pareto optimal.
Pareto showed that maximization of aggregate
profit at the given prices, subject to the resource-
allocation and production- function constraints,
would lead to cost- minimization and zero profits.
(For mathematical details of Pareto’s arguments see
Chipman , pp. 88-92). Pareto summarized this result
by stating (1896-7, vol. 2, p. 94):
Free competition of entrepreneurs yields the same
values for the production coefficients as would be
obtained by determining them by the condition that
commodity outputs should be chosen in such a way
that, for some appropriate distribution, maximum
ophelimity would be achieved for each individual in
society.
The last clause was Pareto’s unfortunately
awkward way of stating the criterion of Pareto
optimality.
Barone ( ), who had originally spurred
Pareto on to this line of argument, developed it
further himself. He noted that a competitive equi-
librium has the property that aggregate profit is at a
maximum at the equilibrium prices, hence, for any
feasible departure from this equilibrium, valuing
consumption and factor services at the equilibrium
prices, some individuals may gain and others will
lose, the losses outweighing the gains so that, even
if the gainers part with all their gains,
1915
the rest will still be worse off than originally.
(Barone used what is now known as the criterion of
revealed preference to make inferences concerning
preferences from data on prices and incomes). Such
a state was described by Pareto and Barone as
‘destruction of wealth’, and its measure by
aggregate income loss at the competitive-
equilibrium prices provided an alternative to the
deadweight loss considered by Dupuit and
Marshall. Barone ( ) also related
his arguments to those of Marshallian consumers’-
surplus analysis.
Lemer ( ) invoked the compensation prin
ciple in his proposed method for measuring
monopoly power, describing it as ‘a loss to the
consumer which is not balanced by any gain reaped
by the monopolist’. In this paper Lemer also
formulated, apparently independently, the concept
of Pareto optimality.
Hotelling ( ) made a noteworthy contribu
tion by providing an alternative demonstration of
the inferiority of excise taxes to income taxes, using
the compensation principle directly. He considered
a single individual consuming n commodities in
amounts qj and facing market prices pj. Prior to the
imposition of the excise taxes (or tolls), the
individual consumes a bundle q° at pricesp° and
income (or fixed component of income) m°, which
maximizes a utility function U(q) subject to the
budget constraint p° ■ q = m\ Subsequent to the
introduction of taxes, market (tax-inclusive) prices
and after-tax income are p1 and tn respectively, and
a bundle q is chosen which maximizes U(q) subject
to p1 ■ q = m . The government collects r=(/?1 —
p°)-c^ — (m1 m°)
in revenues. Since the government is assumed to
collect [p'j — p^j ■ q'j in taxes on commodity j.pj
must be identified with the production cost after the
tax (as well as with the market price = production
cost before the tax); this is a fairly restrictive
assumption, since it implies that the tax does not
affect production costs. (In this respect Hotelling’s
treatment is less general than Dupuit’s and Mar-
shall’s, involving infinite elasticities of supply). We
may denote the ad valorem excise-tax rate on
commodity j by tj = p)/p<j — L and a proportional
income-tax rate by to = 1 - ni1 /m° (negative
1916 Compensation Principle

taxes are interpreted as subsidies). The govern- charitable interpretation of Hotelling’s contribution
ment’s revenues are see Silberberg ).
Hotelling went on to assert that his proposition
n
r=
Yl lJph 'j + = 0, could be extended to many consumers (though no
jI details or proof were provided), and he proceeded to
examine the consumers’-surplus measure of loss \
assumed zero since the government distributes the
(pl -p°) ■ (q1 - q ) = \rIp]- (</' -</') (whereTisa
total proceeds of these excise taxes back to the
diagonal matrix of excise-tax rates tj). He also made
consumer (or taxes the consumer if these are neg-
some general observations ( , p. 267) that, to
ative). The consumer’s budget constraint after the
this day, constitute what is probably the best state-
imposition of the taxes is
ment to be found of the philosophy underlying the
compensation principle.
]C(i1 ij)phj o 'o)w°.
The Years of the New Welfare
These two equations together imply that q1 satisfies Economics
the budget constraintp° ■ q = m°, hence ql was in
the consumer’s original budget set. Therefore, In the cases to which the compensation principle
setting aside the ‘infinitely improbable ... was applied by Dupuit, Marshall, Lemer and
contingency’ that q and q1 lie on the same indif- Hotelling, compensation was made between the
ference surface, Hotelling concluded ( p. 252) that class of consumers on the one hand and a govern-
‘if a person must pay a certain sum of money in ment or a monopolist on the other. While Pareto
taxes, his satisfaction will be greater if the levy is and Barone had discussed compensation between
made directly on him as a fixed amount than if it is different classes of consumers (as had Hotelling in
made through a system of excise taxes which he can his general remarks) their work was unknown to
to some extent avoid by rearranging his production English-speaking economists until the publication
and consumption’. in 1935 of the English translation of Barone’s work.
Unfortunately Hotelling overlooked the fact that Even this seems not to have struck home, however,
if tj = I for all j then the government’s budget since Kaldor ( ) cited pas-
constraint implies p° ■ q1 = —m°to/t, whence /(, = sages from Harrod ( ) and
—t and q1 = q°. That is, a system of uniform ad Robbins ( ) to
valorem excise taxes is equivalent to a proportional the effect that, since movement towards free trade
income tax. This was pointed out by Frisch ( ) would affect different classes differently, no sci-
and accepted by Hotelling ( ). entific statement could be made concerning the
As Frisch made clear, what Hotelling really proved beneficial effect of free trade without making
was the non-optimality of a system of non- interpersonal comparisons of utility.
proportional excise taxes or subsidies when selling Kaldor ( ) proceeded to sketch an argu
prices are given. If these selling prices are equal to ment to the effect that removal of an import duty
marginal costs, Hotelling’s theorem shows that (using the classical example of repeal of the Com
market prices should be proportional to marginal Laws) would result in a situation in which the
costs. Since incomes are fixed in Hotelling’s losses incurred by the landlords could be compen-
formulation, income taxes may be regarded as sated by the gains (through lower import prices)
lump-sum taxes. If institutional consideration make obtained by the other consumers. Such an argument
excise taxes impossible for one commodity (say cannot be correct, however, since, as Kaldor ( )
leisure), then they must be zero for all commodities pointed out only a year later, it follows
and optimality requires that prices be equal to from Bickerdike’s theory of optimal tariffs that a
marginal costs. (For a less country can gain from the imposition of a suffi-
ciently small duty, and, as Graaff ( ) and
others later demonstrated, the compensation prin-
ciple can be used to show that, with suitable
Compensation Principle
compensation, all persons can gain. Unless the rate
of com duty was above the optimal tariff rate, the
opposite conclusion would follow to that indicated
by Kaldor ( ).
A previous attempt by Pareto ( ) to show
by means of the compensation principle that a tariff
would lead to ‘destruction of wealth’ was defective,
since he assumed trade to be balanced in domestic
prices and thus he failed to take account of the
improvement in the terms of trade and the beneficial
effect of the tariff revenues.
Other attempts prior to 1939 to make the case
for free trade suffered from vagueness both in
specifying the criterion of gain and in specifying the
alternative with which free trade was being
compared. Ricardo ( , p. 25) stated: ‘There are
two ways in which a country may be benefited by
trade - one by increase of the general rate of
profits ... the other by the abundance of commod-
ities, and by a fall in their exchangeable value, in
which the whole community participate’. According
to Caimes ( , p. 418), ‘the true
criterion of the gain on foreign trade [is] the degree
in which it cheapens commodities, and renders them
more abundant’. A hint of a compensation principle
is found in Viner ( pp. 533—4):
free trade ... necessarily makes a v a i l a b l e to the
community a s a w h o l e a greater physical real
income in the fomr of more of a l l commodities,
and ... the state ... can, by appropriate supplementary
legislation, make certain that removal of duties shall
result in more of e v e r y commodity for e v e i y
class of the community.
Like Kaldor’s statement, this is formally incorrect;
but it was sufficiently suggestive to stimulate
Samuelson ( ) into providing a formal proof
of a gains-from-trade theorem, albeit under very
restrictive assumptions.
Samuelson ( ) assumed that an open econ
omy had a locus <p(y, l) = 0 of efficient combina-
tions of outputs y and (variable) factor services /,
and asserted that vectors of prices p and factor
rentals w in competitive equilibrium would be such
that aggregate profit p ■ y - w ■ l is a maximum. This proves that for any free-trade equilibrium it is
This is the same as the proposition of Pareto ( ),
and Barone ( ) referred to above. Let
ting x denote the bundle of commodities
1917
consumed, under both (balanced) free trade and
autarky the budget equation p ■ x = p ■ y holds.
Letting superscripts 0 and 1 denote equilibrium
values under autarky and free trade respectively, it
follows that
pl-xl-wl-ll >pl-x ° - w 1 -1°.
Assuming all N individuals to be identical in their
preferences and ownership of factors, and dividing
this inequality through by the number of indi-
viduals, it states that each person chooses (x'/A, l1
IN) under free trade when (x°/A, PiIV) is available,
hence (if/?1 / p°) each person prefers (x'/A, l1 IN) to
(x°/A, PlN). Therefore free trade is Pareto- superior
to autarky.
Samuelson went on to assert ( , p. 204)
that, if the assumption of identical individuals is
dropped, then, although it could no longer be said
that each individual was better off under free trade,
‘it would always be possible for those who desired
trade to buy off those opposed to trade, with the
result that all could be made better off. This
argument went unchallenged until Olsen ( )
pointed out that, if compensation were
paid from gainers to losers, a new equilibrium price
constellation p1 would result, and the argument no
longer follows. For this reason Samuelson’s 1939
results has come to be known as the gains-from-
trade theorem for the ‘small- country case’, though
this interpretation was not suggested by Samuelson
at the time. But this description of Samuelson’s
result is inaccurate. Generalizing his argument we
can say that if (*!.<) are the allocations of (x, ll) to
individual i, where f , xti ri and fy!! , l\ = I', the given
the allocations (xj, l)) of (x1, /') under free trade one
can find Pareto-optimal allocations (x®, if) of (y\ P)
under autarky such that
1
• xj - w1 ■ l) >p' -x" H- • /f for
z = 1, 2, .. . ,A .
possible to find a weakly Pareto-inferior Pareto-
optimal autarky equilibrium. It does not prove the
obverse proposition that for any autarky
1918
equilibrium it is possible to find a weakly Pareto-
superior free-trade equilibrium. A general gains-
from-trade theorem was therefore yet to be
established, but Samuelson had provided an
important first step.
Hicks ( ) ushered in the ‘new welfare eco
nomics’ with a synthesis building on Hotelling (
) and Kaldor ( ) and based on the com
pensation principle, making it possible, according to
him, to make policy proposals in favour of
economic efficiency which were free of value
judgements. Hicks’s most original contribution
(Hicks ) was his attempt to apply the compensation
principle to data on a country’s real national income.
This was a natural thing to try to do, since Pigou’s (
) main work was
devoted to evaluating a country’s welfare by
national-income comparison, and it was largely
Pigou’s resort to interpersonal comparisons in order
to justify this that was the object of Robbins’s (
) criticism.
Hicks’s ( ) basic tool was the ‘revealed-
preference’ comparison which had been employed
by Barone ( ) and Hotelling
( ). If observations are available at times
0 and 1 of a country’s national income in period-1
prices, and it is recorded that p1 ■ y1 > p' ■ y° (where
p\ yl are vectors of prices and outputs at time t),
what can be inferred? In the first place, to make any
headway one must assume that the observed
situations are competitive equilibria. Let us define
an allocation of a commodity bundle x as an N x n
matrix X whose /th row, .r„ is the bundle of n
commodities allocated to individual i, and whose
row sum a{X) = , v, is equal to .r.
As between two bundles x?,xjconsumed by indi-
vidual i, let us define xjRpf to mean that x} is
preferred or indifferent to xf by individual i, where
Rt is a continuous, convex, monotonic total order,
with Pt denoting strict preference and /, indifference.
(This relation assumes the absence of externalities in
consumption). Finally, let XlRX° (resp. X1 PX°) mean
that X1 is weakly (resp. strictly) Pareto-superior to
X° (i.e. xjRjxf for all i, resp. xjRixf for all i and
xjRixf for some i). Then, from the real-income
comparisonp1 ■yl>pI ■ y°, Hicks noted that there
does not exist an allocation Xof y° that is weakly
Pareto-superior to the actual
Compensation Principle
allocation A'1 of v1. This follows from the same
argument that establishes the Pareto optimality of
the assumed competitive equilibrium in period 1.
The non-existence of an allocation X of y° such that
XRX1, where n(X') = y1, constituted for Hicks the
definition of an ‘increase in real social income’.
Kuznets ( ) pointed out by an example
that, in the case considered by Hicks, it could also
be true that there is no allocation X of y1 which is
weakly Pareto superior to the actual allocation X° of
y°. Accordingly he suggested that Hicks’s criterion
be supplemented by the condition that there should
exist an allocation X of y1 that is weakly Pareto
superior to the actual allocation X° of y°. But while
the latter criterion impliesp° ■ y1 > p° ■ y°, it is not
implied by it, so a national-income comparison
using current and base prices would still not yield
Kuznets’s criterion.
Kuznets’s criticism of Hicks was similar to the
objection raised by Scitovsky ( ) to the criterion
proposed by Kaldor ( ). According to
Scitovsky’s interpretation of Kaldor, an allocation
X1 of y1 is better than an allocation X° of y°, if there
exists a reallocation A ofy , which is Pareto superior
to X°. Scitovsky objected that this gave preference
to the status quo ante, and besides, he pointed out
that the criterion was internally inconsistent in the
sense that it allowed two such pairs (X, y ) to be
superior to each other. He therefore proposed that
Kaldor’s test be supplemented by the criterion that
there exist a reallocation X° ofy° that is Pareto
inferior to X1.
The literature on ‘compensation tests’ suffered
from ambiguity as to the domain of definition of the
relations and internal inconsistency of the relations.
It was pointed out by Gorman ( )
that the relations were intransitive. It was shown in
Chipman and Moore ( ) that the
Hicks-Kuznets and Scitovsky double criteria, as
well as the national-income comparisons in terms of
base- and current-year prices, could lead to cycles of
three competitive equilibria each superior to its
successor.
The definitive contribution to the subject of
national-income comparisons was that of Samuelson
( ) who introduced what Chipman and
Compensation Principle
Moore ( ) described as the ‘Kaldor-
Hicks-Samuelson (KHS) ordering’. The objects
under comparison in this approach are sets Y of
commodity bundles y, e.g. production-possibility
sets. Letting A(Y) denote the set of allocation
matrices X such that a(X) c K this ordering is
defined by
Y1 > RYr 0
<£> [VX° eA(r0)] [3X1 eA^1)]*1/?*0.
In words, Yl, is potentially superior to T° if, for all
allocations of commodity bundles in Y°, there exists
a (weakly) Pareto superior allocation of a
commodity bundle in Y]. This is a reflexive and
transitive relation; it also satisfies the condition that
y° C Y 1 implies Y 1 > R Y°. Samuelson also
introduced the important concept of a utility- pos-
sibility frontier, which is the relative boundary of a
utility-possibility set U(Y, R;f); this in turn is a set
of /V-tuples of individual utilities, u f(X), for some
X eA(Y), where/is an A'-tuple of positivevalued
utility functions representing R. If the sets Y are
‘disposable’ (that is, containing for every ye Y the
bundles / with 0 < y < y), and the Rj continuous and
monotonic, then the utility- possibility sets are also
disposable. If Y is nonempty, compact disposable,
and convex, and the Rj are continuous, monotonic,
and convex, then, provided the f are continuous and
concave, U(Y, R, f) is non-empty, compact, and
convex (cf. Chipman and Moore , p. 24). If the f are
only quasi-concave and not concave, U(Y,R,f) need
not be convex (cf. Kannai and Mantel ).
The KHS ordering among consumption-possibility
sets translates into set-inclusion of the corres-
ponding utility-possibility sets. Samuelson (1959, p.
10) gave an example of a case of crossing utility-
possibility frontiers in which X2 C_ A (Y2) was
Pareto superior to X1 C A(Y1) yet Y1 would be ranked
higher than Y2 in terms of some value judgement.
This established that the ‘compensation tests’ were
not ‘relatively wertfref.
Another approach was followed by Chipman
and Moore ( , ), who asked the follow
ing question: if competitive equilibria (X, y, p) are
observed satisfying pl ■ y1 > pl ■ y° and
1919
P° • y1 > P° • where y‘ e Yt for t = 0, 1, under what
conditions on preferences must this imply that Y1 >
RY°? For the case Y' = {/} they showed that the
preference relations R, must be identical and
homothetic. This is a global result; with positive
consumptions of all commodities the condition
could no doubt be weakened to the aggregation
criterion of Antonelli ( ),
Gorman ( ), and Nataf ( ), namely, that
consumer i’s demand for commodify j have the
form
x
ij = aij(P) + bj{p)nii
where mt is consumer i’s income.
Samuelson ( ) applied the compensation
principle in a striking way in his proposed alter-
native to the new welfare economics. He discovered
that, if a social- welfare function has the separable
form W[/(x)], then a social utility function fw(x) =
max {IL[/(x)] : X c A(x)\ has the property that it can
be achieved in a decentralized manner by means of
an income-distribution policy assigning individual
shares of aggregate income as functions of prices
and aggregate income. The first complete proof of
this result was presented in Chipman and Moore ( )
(see also Chipman and Moore ; Chipman ). The
main tool of analysis used was the concept of a
Scitovsky indifference surface (Scitovsky ) which is
defined as the boundary of the set Yj=i ^‘x‘ where
Rpj is the set of all commodity bundles preferred or
indifferent to xt by individual i. This set is
necessarily a subset of the set R^x of aggregate
bundles preferred or indifferent to x by the
Samuelson social ordering. In a competitive
equilibrium the aggregate consumption bundle
minimizes aggregate expenditure at the equilibrium
prices over both sets, hence the bundle x, minimizes
each individual’s expenditure over Rpc, (cf.
Koopmans , pp. 12-13).
Gains From Trade and Optimal Tariffs
The new tools developed by Scitovsky ( ) and
Samuelson ( , ) made possible a rigorous
1920
proof of a gains-from-trade theorem, as well as of
the proposition that a country could gain by a tariff.
Kemp ( ) noted that Samuelson’s 1939
theorem implied that for any point on the free- trade
utility-possibility frontier, the autarky utility-
possibility frontier must pass below it; he reasoned
that, as a result, for any point on the autarky utility-
possibility frontier, the free-trade utility-possibility
frontier must pass above it. If this argument can be
accepted, it follows that for every allocation^ C
A(Y°) where Y° is the autarkic production-
possibility set, there exists a (weakly) Pareto-
superior allocation A 1 £ A(Yh) where Y] is the free-
trade consumption-possibility set. Then free trade is
superior to autarky by Samuelson’s 1950 criterion.
The trouble with this argument, however, is that
it requires that one can define a free-trade utility-
possibility frontier (or consumption- possibility
frontier) with the strong topological property of
homeomorphism to the (N — 1)- dimensional unit
simplex (intuitively, absence of ‘holes’). That this
need not always be possible, was shown by Otani (
, p. 149), and indeed
admitted by Kemp and Wan ( , p. 513). It is
always possible if world prices are fixed, beyond
our country’s control. In that case the free-trade
consumption-possibility set Y1 is the budget set
enclosing the production-possibility set
Y° (cf. Samuelson , p. 821), and the gains- from-
trade theorem follows immediately from the
property I1 S
RY . In similar fashion the
famous ‘Baldwin envelope’ (Baldwin ) defines a
well-behaved consumption-possibility set
containing the production-possibility set, from
which one can prove the superiority of restricted
trade (with an optimal tariff) to autarky (cf.
Samuelson ).
For the general case in which a country can
influence world prices, a method was shown by
Kenen ( ). If all but 1 of the N individuals are
constrained to have the same level of satisfaction
under trade as achieved under autarky, a net
production-possibility set can be constructed which
indicates the amount available for the Mh person. It
remains only to show that the Mh person will gain
from a movement from autarky to
Compensation Principle
free trade. A similar approach was indicated by
Vanek ( ).
Grandmont and McFadden ( ) and
Chipman and Moore ( ) both used the concept
of an income-distribution policy to establish the
gains-from-trade theorem. In Chipman and Moore
this policy was chosen to be one that maximizes a
separable Bergson-Samuelson social-welfare
function. A standard argument is used to show that
social utility is at least as high under free trade as
under autarky. It remains to show that a function
W(u) can be chosen so that the corresponding
distribution policy ensures that an increase in social
utility implies an increase in each individual’s
utility. This is achieved by choice of W(u) = miiij
(iij — uf)/cj where ct > 0 and it1- is the level of
utility achieved by individual i under autarky.
General-Equilibrium Theory
The compensation principle is used in the proof of
the theorem that every competitive equilibrium is
Pareto-optimal (Arrow , pp. 516, 519; Koopmans ,
p. 49; Debreu , pp. 94-5), in the sense that arbitrary
allocations of feasible output bundles among
consumers are assumed possible, regardless of
resource-ownership constraints. A pair (A0, p°) is a
competitive equilibrium for the production-
possibility set Y if J^'RX for all X C A(Y) satisfying
Xp° < A°p° and y°p° < yp° for all y £ Y, where y° =
ff(A°) C_ Y. Pareto-optimalify means that one
cannot find an X £ A(Y) such that XPX°. The proof is
by contradiction: XPX° implies Xp° > A°p° (the
vector inequality being weak in all components and
strict
in at least one) hence taking column sums,
0 - 0 0
yp >yp ■
The converse theorem, that every Pareto opti-
mum can be sustained by a competitive equilibrium,
requires stronger assumptions which are awkward to
state (cf. Arrow , p. 518; Koopmans , p. 50; Debreu ,
p. 95). The basic idea of the proof (Koopmans , pp.
50-52; Debreu , p. 96) can be sketched in terms of
the concept of a Scitovsky ( )
indifference surface. If A0 is a Pareto-optimal
Compensation Principle
allocation for a closed, convex production-
possibility set K then the interior of the Scitovsky
set of X° can be written pkx°k + f°r
some k.
Defining the allocation X1 by x\pkxQk and x\Rt xf
for i ^ k we have X1 PX° hence X1 £ A(Y). Therefore
the interior of the Scitovsky set does not intersect Y,
and these convex sets can be separated by a
hyperplane defining the equilibrium prices. It is then
verified that at these prices the properties of a
competitive equilibrium are satisfied.
Debreu ( , p. 590) introduced an alternative
equilibrium concept according to which the condi-
tion that consumer preferences be maximized sub-
ject to their budget constraints was replaced by the
condition that consumer expenditures be minimized
subject to the constraints that the bundles considered
be at least as desirable as the equilibrium bundles.
(The second of the above theorems follows more
easily under this alternative definition). For a given
set of positive-valued utility functions representing
consumer preferences, Arrow and Hahn ( , p.
108) called this a ‘com
pensated equilibrium’. As a means of proving exis-
tence of the latter they studied the utility-possibility
frontier or ‘Pareto frontier’ ( , p. 96), and
obtained a new proof of the result of Chipman and
Moore ( ) that the set of Pareto-optimal
allocations X of Y (the ‘contract curve’) and the
utility-possibility frontier are topologically homeo-
morphic to the unit simplex of dimension one less
than the number of individuals. These results were
further developed by Moore ( ).
Cost-Benefit Analysis
Hicks ( , p. 112) made an interesting distinc
tion between two tasks of welfare economics: (1)
the study of (Pareto)-optimal organizations of the
economy and (2) the study of deviations from such
optima. More precisely, the first was concerned with
when there was a deviation and the second with the
size of the deviation. He also identified these two
tasks with general- and partial-equilibrium analysis
respectively, although there appears to be no
justification for
1921
this other than the historical accident that con-
sumers’ surplus developed as a partial- equilibrium
tool. He remarked that consumers’ surplus is not
needed for the first task, since lack of fulfilment of
the proportionality between marginal utilities and
marginal costs provides the needed information
immediately. For the second task, he was not
content with a ranking of the non- optimal states,
but with measuring the size of their deviations from
optimality, which of coiuse would provide such a
ranking. Thus, the staunch ordinalist in consumer
theory became an equally ardent cardinalist in
consumer theory.
Hicks’s concepts of compensating and equiva-
lent variation (Hicks ) may most conveniently be
defined in terms of the minimum- income or
income-compensation functions of McKenzie () and
Hurwicz and Uzawa
( ). Denoting the ith consumer’s demand
function by x,- = h^p, in,) (where x, and p are ;?-
vectors), and defining the indirect preference
relation R* by (p°,mk) R* (/;', m\) if and only if h,
(p°, )Rtht (px, m\), the income-compensa
tion function is defined by
pi(p;p°,m°) i n f {m , : (p,ml)R*(p0,m^) }.
Following Chipman and Moore ( ), the gen
eralized compensating variation in going from (p\
w?) to {p, nij) is defined as
Ci(p, mp,p°, »if) = mt - Pi (p;p°, »i?)
and the generalized equivalent variation by
Ei(p,mi;p0,m<}) = Pi(p0;p,nii) - mk.
These reduce to Hicks’s concepts when m,- = mf.
The compensating variation expresses for each
consumer the amount of money income he or she
would be willing to give up (or the negative of the
amount by which he or she would have to be
compensated), at the new prices, to make up for the
change in prices and income. One of the reasons for
the great appeal of the concept is that these are
amounts that can be added up over the set of
consumers. In Hicks’s words ( , p. 127):
1922 Compensation Principle

the general test for a particular reform being an improvement ordering between pairs of price-
i m p r o v e m e n t is that the gainers should gain
income pairs (p‘, A/) as follows. Let A* (p, M) be
sufficiently for them to be able to compensate the
the set of (n + /V)-tuplcs (p, m) such that 22^=1 m‘ =
losers and still remain gainers on balance. This test
M, and let R* be the relation such that (p°, m°)
would be carried out by striking the balance of the
Compensating Variations. R*(p\ m1) if and only if (p°,m9 )R* (p\ m\) for i =
Denoting by m' the vector of N incomes in state t, 1,2,..., N. Then we define the dual KHS relation
and by M' their sum, we can define a dual potential- >R* by

(p°,M°)>*. (p1,M1) ^[V(p',m') eA*(p\M1 )(3(p,m)eA*(p°,M0)] :

(p,m)R*(p',m').

Choosing price-income pairs (p°, m°) and (p1, ml) Boadway ( ) considered the relationship
satisfying this definition, since p,(p‘‘, p, mj) is an between the condition of positive summed com-
indirect utility function representing/?* for t 0 or 1, pensating variations and the fulfilment of com-
we have P;(p';p°,w;f) > Pi(p';p 1,»il-) for all pensation tests and came to the negative conclusion
individuals i, hence that the former was neither necessary nor sufficient
for satisfaction of the latter in general, but was
sufficient in the case of identical and homothetic
M° = Y^m<i preferences. Foster ( ) showed
1=1
that, if there are no price distortions (but not
otherwise), satisfaction of the compensation tests
i=1 2=1 implies satisfaction of the ‘cost-benefit criterion’
(positive summed compensating variations). This
so one obtains a multi-consumer analogue to the conclusion is in accord with the above inequalities.
compensating variation from the formula What about the Hicksian tenet that the size of
the compensating variation is important so that one
can compare two suboptimal states? This would
M ° =^2fj.i(p°;p1,m1i)^
2=1 require one to be able to conclude that, if the
compensating variation from state 0 to state 2 is
positive and greater than the compensating variation
M1 - =0.
2=1 from state 0 to state 1, then state 2 should be
superior to state 1 in terms of the dual KHS
Likewise for the equivalent variation, ordering. But this is not true even in the case of the
single consumer. It was shown in Chipman and
o=y>,(Pv.m”) Moore (1980) that the function C, (p, mt; p\ m°)
2=1 cannot be an indirect utihty function for unrestricted
domain (p, m,) > 0, and can be if m[ is held constant
-M° ^ ^pi(p°;p1,m!1)-M°. i=i
if and only if preferences are homothetic, and if p\
In the latter case the same indirect utihty is held constant if and only if preferences are
‘parallel’ with respect to commodity 1. If
functions are summed on both sides of the
preferences are identical and homothetic, since p, =
inequality sign; it is a case where Benthamites and
pis homogeneous of degree 1 in mi, I p,(p°;p, mi)
compensationists can find common ground.
so exact
aggregative analogues are obtained to both the
Compensation Principle 1923

compensating and equivalent variations. If the that the exchange in commodity 2 would be con-
equivalent variation, which is an indirect utility stant, but in commodity 1 ‘indeterminate’; see the
function, is used, restrictions on consumer prefer- reply by Berry , on behalf of Marshall, and Marshall
ences are not needed, and the problem of finding an 1891, p. 756; 1920, p. 845). The above form for the
adequate indicator of the size of the deviation from utility function has been rediscovered many time,
a given Pareto optimum is satisfactorily resolved. by Wilson ( ), Samuelson ( ),
and others; cf. Chipman and Moore ( ,
p. 115). Barone ( , p. 213n) gave the name
Game Theory ‘ideal money (numeraire)’ to a good with a constant
marginal utility (commodity 1 in the above). For the
One of the striking aspects of von Neumann and
case c, = c for all i, these ‘parallel’ preferences (cf.
Morgenstem’s theory of games ( ) was not
Boulding ) yield a special case of the family of
only its postulate of measurability of utility but also
aggregable Antonelli-Gor- man-Nataf demand
that of its transferability between players. Since this
functions referred to above.
was introduced as a positive rather than a normative
assumption, it has met with even greater resistance
of the part of economists than the hedonist calculus. Concluding Observations
Indeed, it was not until Debreu and Scarf ( )
showed how game the As Scitovsky ( ) pointed out, the compensa
ory could be liberated from this restriction with tion principle has been used in two quite different
their development of the concept of the core of an ways. Prior to Hicks ( ), it was used only to
economy that game theory began to be taken really compare efficient with inefficient states of a given
seriously by economists. The replacement of economy with a given technology or trading system.
transferability of utilities by transferability of Starting with Hicks ( ), its use was
commodities bears a striking resemblance to the extended to comparison of efficient states of an
replacement in welfare economics of the calculus of economy under different technologies. It has turned
utilities by the principle of compensation. out that, in order for national-income comparisons
In some branches of game theory the assump- to provide a correct indicator of potential- welfare
tion of transferable utility is still retained, but it has improvement, very strong conditions are required
been made somewhat more plausible, or at least concerning similarity of individual preferences:
interpretable, by means of the postulate that the locally, the Antonelli- Gorman-Nataf conditions,
utility functions of all individuals are linear in one and, globally, identical homothetic preferences. It is
distinguished commodity used for making side not even enough to assume that aggregate demand
payments (cf. Owen , p. 122). These utility can be generated by an aggregate preference
functions have the form relation - for example, that preferences are
homothetic and relative income- distribution
Ui{xi\, Xi2i —, ) CjXj\ V/ (x/2, —, XLn ).
constant (cf. Chipman and Moore ). Even in such
cases, strong value judgements (such as acceptance
This form of the utility function goes back to
of a particular Bergson-Samuelson social-welfare
Edgeworth ( , p. 237n) and even earlier
function) are required in order to draw welfare
(though in garbled form) to Auspitz and Lieben (
conclusions from national-income comparisons.
, p. 471). In Edgeworth it was used to illus
When attention is restricted to the efficient
trate the phenomenon of exchange when the mar-
operation of an economy with a given technology, it
ginal utility of one commodity serving as money
turns out that, in most cases of interest, the ranking
was held constant, in accordance with one possible
of consumption-possibility criterion sets according
interpretation of Marshall’s theory of consumers’
to the Kaldor-Hicks-Samuelson criterion follows
surplus. (In the case n = 2 he showed
from their ranking by set-inclusion.
1924
This does not mean, however, that the set- inclusion
is always obvious or easy to prove.
The KHS ordering of consumption-possibility
sets could be given simply a factual interpretation as
indicating the ‘productive potential’ of an economy.
But if it is given a normative interpretation then it
obviously involves a value judgement, since a more
efficient outcome, if it is not Pareto-superior, can
obviously be judged worse in terms of some social-
welfare function.
Samuelson’s ( ) model of the ‘good soci
ety’, elegant though it is, is too sweeping for most
economists to accept, and it begs the question of
how the social-welfare function will be chosen.
Little’s ( ) and Mishan’s ( ) attempts to
link plausible distributional value judgements with
compensation criteria have encountered
unresolvable logical difficulties (cf. Chipman and
Moore ). The hope that the compensation principle
would allow policy decisions to be made free of
value judgements has not been fulfilled.
Nevertheless, much has been learned about the
interrelationships among values, facts, and policies,
and it can certainly be said that the development of
the compensation principle has led to clearer
thinking about economic policy issues.
See Also
► Social Welfare Function
► Welfare Economics
Bibliography
Antonelli, G.B. 1886. S u l l a t e o r i a m a t e m a t i c a
d e l l a e c o n o m i a p o l i t i c a . Pisa: Folchetto.
English translation: On the mathematical theory of
political economy. In P r e f e r e n c e s , u t i l i t y , a n d
d e m a n d , ed. J.S. Chipman, L. Hurwicz, M.K. Richter,
and H.F. Sonnenschein. New York: Harcourt Brace
Jovanovich, 1971.
Arrow, K.J. 1952. An extension of the basic theorems of
classical welfare economics. In P r o c e e d i n g s o f t h e
second Berkeley symposium
mathematical statistics and probability,
ed. J. Neyrnan. Berkeley/Los Angeles: University of
California Press.
Arrow, K.J., and F.H. Hahn. 1971. G e n e r a l
c o m p e t i t i v e a n a l y s i s . San Francisco: Holden-
Compensation Principle
Arrow, K.J., and T. Scitovsky, eds. 1969. R e a d i n g s i n
w e l f a r e e c o n o m i c s . Homewood: Irwin.
Auspitz, R., and R. Lieben. 1889. U n t e r s u c h u n g e n
i i b e r d i e T h e o r i e d e s P r e i s e s . Leipzig:
Duncker & Humblot.
Baldwin, R.E. 1948. Equilibrium in international trade: A
diagrammatic analysis. Q u a r t e r l y J o u r n a l o f
E c o n o m i c s 62: 748-762.
Baldwin, R.E. 1954. A comparison of welfare criteria.
R e v i e w o f E c o n o m i c S t u d i e s 21: 154-161.
Barone, E. 1894. Sulla ‘consumers’ rent’. G i o r n a l e d e g l i
E c o n o m i s t i . Series 2, 9, September, 211-224.
Barone, E. 1908. II Ministerio della produzione nello stato
colletivista. G i o r n a l e d e g l i E c o n o m i s t i Series
2. 37: 267-293; 391-414. English translation: The ministry
of production in the collectivist state. In C o l l e c t i v i s t
e c o n o m i c p l a n n i n g , ed. F.A. Hayek. London:
Routledge & Regan Paul, 1935.
Berry, A. 1891. Alcune brevi parole sulla teoria del baratto di
A. Marshall. G i o r n a l e d e g l i E c o n o m i s t i .
Series 2, 2, June, 549-553.
Boadway, R.W. 1974. The welfare foundations of cost- benefit
analysis. E c o n o m i c J o u r n a l 84: 926-939. A reply,
E c o n o m i c J o u r n a l 86 (1976), 358-361.
Boulding, K.E. 1945. The concept of economic surplus.
A m e r i c a n E c o n o m i c R e v i e w 35: 851-869.
Caimes, J.E. 1874. Some leading principles of political
economy newly expounded. New York: Harper & Brothers.
Chipman, J.S. 1976. The paretian heritage. Revue europeenne
des sciences societies et Cahiers Vilfredo Pareto 14 (37):
65-171.
Chipman, J.S. 1982. Samuelson and welfare economics. In
Samuelson and neoclassical economics,
ed. G.R. Feiwel. Boston: Kluwer-Nijhoff Publishing.
Chipman, J.S., and J.C. Moore. 1971. The compensation
principle in welfare economics. In P a p e r s i n
q u a n t i t a t i v e e c o n o m i c s , ed. A.M. Zarley, vol. 2.
Lawrence: University Press of Kansas.
Chipman, J.S., and J.C. Moore. 1972. Social utility and the
gains from trade. J o u r n a l o f I n t e r n a t i o n a l
E c o n o m i c s 2: 157-172.
Chipman, J.S., and J.C. Moore. 1973. Aggregate demand, real
national income, and the compensation principle.
I n t e r n a t i o n a l E c o n o m i c R e v i e w 14: 152-181.
Chipman, J.S., and J.C. Moore. 1976a. Why an increase in
GNP need not imply an improvement in potential wel-
fare. K y k l o s 29 (3): 391-418.
Chipman, J.S., and J.C. Moore. 1976b. The scope of con-
sumer’s surplus arguments. In Evolution,
w e l f a r e , a n d t i m e i n e c o n o m i c s , ed. A.
Tang, F.M. Westfield, and J.S. Worley. Lexington:
Heath.
Chipman, J.S., and J.C. Moore. 1978. The new welfare
economics, 1939-1974. I n t e r n a t i o n a l E c o n o m i c
on
R e v i e w 19: 547-584.
Chipman, J.S., and J.C. Moore. 1979. On social welfare
functions and the aggregation of preferences. J o u r n a l
o f E c o n o m i c T h e o r y 21: 111-139.
Compensation Principle
Chipman, J.S., and J.C. Moore. 1980a. Real national income
with homothetic preferences and a fixed distribution of
income. E c o n o m e t r i c a 48: 401-422.
Chipman, J.S., and J.C. Moore. 1980b. Compensating
variation, consumer’s surplus, and welfare. A m e r i c a n
E c o n o m i c R e v i e w 70: 933-949.
Coase, R.H. 1960. The problem of social cost. J o u r n a l o f
L a w a n d E c o n o m i c s 3: 1-44.
de Graaff, J. V. 1949. On optimum tariff structures. R e v i e w
o f E c o n o m i c S t u d i e s 17: 47-59. Reprinted in
Arrow and Scitovsky (1969).
de Graaff, J.V. 1957. T h e o r e t i c a l welfare
e c o n o m i c s . Cambridge: Cambridge University Press.
Debreu, G. 1951. The coefficient of resource utilization.
E c o n o m e t r i c a 19: 273-292.
Debreu, G. 1954. Valuation equilibrium and Pareto optimum.
Proceedings of the National Academy of
S c i e n c e s 40: 588-592.
Debreu, G. 1959. T h e o r y o f v a l u e . New York: Wiley.
Debreu, G., and H. Scarf. 1963. A limit theorem on the core of
an economy. I n t e r n a t i o n a l E c o n o m i c R e v i e w
4: 235-246.
Dupuit, J. 1844. De la mesure de l’utilite des travaux publics.
Annales des Fonts et Chaussees,
Memoires et documents relatifs a l ’art
des constructions et au service de
I ’ i n g e n i e u r . Series 2, 2, 2e semestre, 332-375, PI. 75.
English translation: On the measurement of the utility of
public works, in Arrow and Scitovsky (1969).
Edgeworth, F.Y. 1891. Osservazioni sulla teoria matematica
delTeconomia politica con riguardo speciale ai principi di
economia di Alfredo Marshall. G i o r n a l e d e g l i
E c o n o m i s t I . Series 2, 2: 233-245. Ancora a proposito
della teoria del baratto. Giornale d e g l i E c o n o m i s t i .
Series 2, 2: 316-318. Abridged English translation: On the
determinateness of economic equilibrium, in F.Y.
Edgeworth, Papers relating t o p o l i t i c a l e c o n o m y ,
vol. 2. London: Macmillan., 1925.
Foster, E. 1976. The welfare foundations of cost-benefit
analysis - A comment. Economic Journal 86: 353-358.
Frisch, R. 1939. The Dupuit taxation theorem.
E c o n o m e t r i c a ! : 145-150. A further note on the
Dupuit taxation theorem. Econometrica 7: 156-157.
Gorman, W.M. 1953. Community preference fields.
E c o n o m e t r i c a 2 1: 63-80.
Gorman, W.M. 1955. The intransitivity of certain criteria
used in welfare economics. Oxford Economic Papers, N.S.
7: 25-35.
Grandmont, J.M., and D. McFadden. 1972. A technical note
on classical gains from trade. Journal of Internat i o n a l
E c o n o m i c s 2: 109-125.
Harrod, R.F. 1938. Scope and method of economics. Eco-
n o m i c J o u r n a l 48: 383^412.
Hicks, J.R. 1939. The foundations of welfare economics.
E c o n o m i c Journal 49: 696-712.
Hicks, J.R. 1940. The valuation of social income.
E c o n o m i c a , N.S. 7: 105-124.
1925
Hicks, J.R. 1941. The rehabilitation of consumers’ surplus.
R e v i e w o f E c o n o m i c S t u d i e s 8: 108-116.
Reprinted in Arrow and Scitovsky (1969).
Hicks, J.R. 1942. Consumers’ surplus and index-numbers.
Review of Economic Studies 9: 126-137.
Hicks, J.R. 1957. A r e v i s i o n o f d e m a n d t h e o r y .
Oxford: Clarendon Press.
Hotelling, H. 1938. The general welfare in relation to
problems of taxation and of railway and utility rates.
E c o n o m e t r i c a 6: 242-269. Reprinted in Arrow and
Scitovsky (1969).
Hotelling, H. 1939. The relation ofprices to marginal costs in
an optimum system. E c o n o m e t r i c a 7: 151-155. A
final note, E c o n o m e t r i c a 7: 158-159.
Hurwicz, L., and H. Uzawa. 1971. On the integrability of
demand functions. In P r e f e r e n c e s , u t i l i t y , a n d
d e m a n d , ed. J.S. Chipman, L. Hurwicz, M.K. Richter,
and H.F. Sonnenschein. New York: Harcourt Brace
Jovanovich.
Kaldor, N. 1939. Welfare propositions in economics and
interpersonal comparisons of utility. E c o n o m i c
J o u r n a l 49: 549-552. Reprinted in Arrow and
Scitovsky (1969).
Kaldor, N. 1940. A note on tariffs and the terms of trade.
E c o n o m i c a , N.S. 7: 377-380.
Kannai, Y., andR. Mantel. 1978. Non-convexifiable Pareto
sets. E c o n o m e t r i c a 46: 571-575.
Kemp, M.C. 1962. The gains from international trade.
Economic Journal 72: 803-819.
Kemp, M.C., and H.Y. Wan Jr. 1972. The gains from free
trade. I n t e r n a t i o n a l E c o n o m i c R e v i e w 13: 509-
522.
Kenen, P.B. 1957. On the geometry of welfare economics.
Quarterly Journal of Economics 71: 426—447. Koopmans,
T.C. 1957. Three essays on the state of economic science. New
York: McGraw-Hill.
Kuznets, S. 1948. On the valuation of social income -
Reflections on professor Hicks’ article. E c o n o m i c a ,
N.S. 15: 1-16, 116-131.
Lemer, A.P. 1934. The concept of monopoly and the mea-
surement of monopoly power. R e v i e w of
Economic
Studies 1: 157-175.
Little, I.M.D. 1950. A critique of welfare economics.
2nd ed. London: Oxford University Press. 1957.
Marshall, A. 1890. P r i n c i p l e s o f e c o n o m i c s .
London: Macmillan. 2nd ed, 1891; 8th ed, 1920.
McKenzie, L.W. 1957. Demand theory without a utility index.
Review of E c o n o m i c S t u d i e s 24: 185-189. Mishan, E.J.
1969. W e l f a r e e c o n o m i c s : A n a s s e s s m e n t .
Amsterdam: North-Holland.
Moore, J.C. 1975. The existence of‘compensated equilibrium’
and the structure of the Pareto efficiency frontier.
International E c o n o m i c R e v i e w 16: 267-300.
Nataf, A. 1953. Sur des questions d’agregation en
econometrie. P u b l i c a t i o n s d e I ' I n s t i t u t d e
S t a t i s t i q u e de l'Universite d e P a r i s 2 (4): 5-61.
Olsen, E. 1958. Udenrigshandelens gevinst [The gains of
international trade]. Nationalokonomisk
1926 Competing Risks Model

Otani. Y. 1972. Gains from trade revisited. J o u r n a l o f

I n t e r n a t i o n a l E c o n o m i c s 2: 127 156. Competing Risks Model
Owen, G. 1982. G a m e t h e o r y , 2nd ed. Orlando:
Academic Press.
Pareto, V. 1894. II massimo di utilita dato dalla libera
Gerard J. van den Berg
coneorrenza. G i o m a l e d e g l i E c o n o m i s t i Series
2, 9: 48 66.
Pareto, V. 1895. Teoria matematica del commercio
intemazionale. G i o m a l e d e g l i E c o n o m i s t i Series
2, 10: 476-498. Abstract
Pareto, V. 1896-7. C o i n s d ' e c o n o m i c p o l i t i q u e , A competing risks model is a model for multiple
vols. 2. Lausanne: F. Rouge.
durations that start at the same point in time for
Pigou, A.C. 1920. T h e e c o n o m i c s o f w e l f a r e , 4th
ed. London: Macmillan. 1932. a given subject, where the subject is observed
Ricardo, IF. 1815. . in essay on the influence of a low price of until the first duration is completed and one also
corn on the profits of stock. London: John Murray. In The observes which ofthe durations is completed
works and correspondence of David Ricardo, ed. P. Sraffa,
first. This article gives an overview ofthe main
vol. 4. Cambridge: Cambridge University Press, l‘)51.
Robbins, I.. 1938. Interpersonal comparisons of utility: A issues in the empirical econometric analysis of
comment. E c o n o m i c J o u r n a l 48: 635-641. competing risks models. The central problem is
Samuelson, PA. 1939. The gains from international trade. the non-identification of dependent competing
(Canadian Journal of Economics and Political Science 5: risks models. Models with regressors can
195 205.
Samuelson, PA. 1942. Constancy ofthe marginal utility of overcome this problem, but it is advisable to
income. In Studies in mathematical include additional data. Alternatively, effects of
economics and econometrics in memory interest can be bounded.
o f H e n r y S c h u l t z . ed. O. Lange, F. McIntyre, and
T.O. Yntema. Chicago: University of Chicago Press.
Samuelson, PA. 1947. Foundations of economic analysis. Keywords
Cambridge, MA: Harvard University Press. Bounds; Censoring; Competing risks models;
Samuelson, PA. 1950. Evaluation of real national income.
Copulas; Duration models; Hazard rates; Iden-
Oxford Economic Papers, N.S. 1: 1-29. Reprinted in
Arrow and Scitovsky (1969). tification; Latent durations; Marriage and
Samuelson, PA. 1956. Social indifference curves. Q u a r t e r l y divorce; Maximum likelihood; Mixed propor-
J o u r n a l o f E c o n o m i c s 70: 1-22. tional hazard models; Multiple-spell competing
Samuelson, PA. 1962. The gains from international trade once risks data; Multivariate duration models;
again. E c o n o m i c J o u r n a l 72: 820-829.
Sanger. C.P. 1895. Recent contributions to mathematical
Nonparametric kernel estimators; Regressors;
economics. E c o n o m i c J o u r n a l 5: 113-128. Roy models; Selection; Semi-parametric
Scitovsky, T. 1941. A note on welfare propositions in models; Stochastic dominance; Unemployment
economics. R e v i e w o f E c o n o m i c S t u d i e s 9: 77 durations; Unobserved heterogeneity; Weibull
88. Reprinted in Arrow and Scitovsky (1969).
Scitovsky, T. 1942. A reconsideration of the theory of tariffs.
distributions
R e v i e w o f E c o n o m i c S t u d i e s 9: 89 110.
Silberberg. E. 1980. Harold Hotelling and marginal cost
pricing. A m e r i c a n E c o n o m i c R e v i e w 70: 1054- JEL Classifications
1057. C13; C51
Vanek, .1. 1964. A rehabilitation of ‘well-behaved’ social
indifference curves. R e v i e w of Economic A competing risks model is a model for multiple
S t u d i e s 31: 87-89.
durations that start at the same point in time for a
Yiner. .1. 1937. Studies in the theory of international trade. New
York: Harper & Brothers. given subject, where the subject is observed until
von Neumann. .1. and Morgenstem, O. 1947. T h e o r y o f the first duration is completed and one also
g a m e s a n d e c o n o m i c b e h a v i o r . 2nd ed. observes which of the multiple durations is com-
Princeton: Princeton University Press.
pleted first.
Wilson. E.B. 1939. Pareto versus Marshall. Q u a r t e r l y
J o u r n a l o f E c o n o m i c s 53: 645-650. The teim 'competing risks' originates in the
interpretation that a subject faces different risks
Competing Risks Model
i of leaving the state it is in, each risk giving rise to
its own exit destination, which can also be denoted
by i. One may then define random variables Tt
describing the duration until risk i is materialized.
Only the smallest of all these durations Y := min, T,
and the corresponding actual exit destination, which
can be expressed as Z := argmin, 7], are observed.
The other durations are censored in the sense that
all is known is that then- realizations exceed Y.
Often those other durations are latent or
counterfactual, for example if T, denotes the time
until death due to cause i.
In economics, the most common application
concerns individual unemployment durations. One
may envisage two durations for each individual: one
until a transition into employment occurs, and one
until a transition into non-participation occurs. We
observe only one transition, namely, the one that
occurs first. Other applications include the duration
of treatments, where the exit destinations are relapse
and recovery, and the duration of marriage, where
one risk is divorce and the other is death of one of
the spouses. More generally, the duration until an
event of interest may be right-censored due to the
occurrence of another event, or due to the data
sampling design. The duration until the censoring is
then one of the variables T h
Sometimes one is interested only in the distri-
bution of Y. For example, an unemployment insur-
ance (UI) agency may be concerned only about the
expenses on UI and not in the exit destinations of
recipients. In such cases one may employ standard
statistical duration analysis for empirical inference
with register data on the duration of UI receipt.
However, in studies on individual behaviour one is
typically interested in one or more of the marginal
distributions of the 7). If these variables are known
to be independent, then again one may employ
standard duration analysis for each of the T t
separately, treating the other variables 7} ( j / i ) as
independent right-censoring variables. But often it
is not clear whether the 7) are independent. Indeed,
economic theory often predicts that they are
dependent, in particular if they can be affected by
the individual’s behaviour and
1927
individuals are heterogeneous. It may even be
sensible from the individual’s point of view to use
their privately observed exogenous exit rates into
destinations j as inputs for the optimal strategy
affecting the exit rate into destination i (i / j) (see,
for example, van den Berg ). Erroneously assuming
independence leads to incorrect inference, and in
fact the issue of whether the durations 7} are related
is often an important question in its own right.
Unfortunately, the joint distribution of all Tt is
not identified from the joint distribution of Y, Z, a
result that goes back to Cox ( ). In particular,
given any specific joint distribution, there is a joint
distribution with independent durations Tt that
generates the same distribution of the observable
variables Y, Z. In other words, without additional
structure, each dependent competing risks model is
observationally equivalent to an independent
competing risks model. The marginal distributions
in the latter can be very different from the hue
distributions.
Of course, some properties of the joint distri-
bution are identified. To describe these it is useful
to introduce the concept of the hazard rate of a
continuous duration variable, say W. Formally, the
hazard rate at time t is 9(t): = limj, j 0 Pr (W e [t, t +
dt))/dt. Informally, this is the rate at which the
duration W is completed at t given that it has not
been completed before t. The hazard rate is the
basic building block of duration analysis in social
sciences because it can be directly related to indi-
vidual behaviour at t. The data on Y, Z allow for
identification of the hazard rates of 7' at t given that
T > t. These are called the ‘crude’ hazard rates. If
the I) are independent, then these equal the ‘net’
hazard rates of the marginal distributions of the Tj.
We now turn to a number of approaches that
overcome the general non-identification result for
competing risks models. In econometrics, one is
typically interested in covariate or regressor effects.
The main approach has therefore been to specify
semi-parametric models that include observed
regressors X and unobserved heterogeneity terms V.
With a single risk, the most popular
1928
duration model is the mixed proportional hazard
(MPH) model, which specifies that 6 (t\ X x, V) = \J/
(t) exp (xfjyVfor some function Vis unobserved, and
the composition of the survivors changes selectively
as time proceeds, so identification from the
observable distributions of T\X is non-trivial.
However, it holds under the assumptions that XJL|
Vand var(X) > 0 and some regularity assumptions
(see van den Berg , for an overview of results).
With competing risks, the analogue of the MPH
model is the multivariate MPH (MMPH) model.
With two risks,
0 i( f | x , V)=\//1(t) exp(xJ/]1)Vi a nd
d2lt\x, V)=ij/2lt) exp{x'pi)V2.
where 7), T 2 \X, Pare assumed independent, so that
a dependence of the durations given X is modelled
by way of their unobserved determinants V\ and V2
being dependent. Many empirical studies have
estimated parametric versions of this model, using
maximum likelihood estimation.
The semi-parametric model has been shown to
be identified, under only slightly stronger condi-
tions than those for the MPH model (Abbring and
van den Berg ). Specifically, Var(X) > 0 is
strengthened to the condition that the vector X
includes two continuous variables with the
properties that (a) their joint support contains a non-
empty open set in R’, and (b) the vectors [il, [>2 of
the corresponding elements of Pi and p2 form a
matrix (jSj, P2) of full rank. Somewhat loosely, X
has two continuous variables that are not perfectly
collinear and that act differently on di and d 2 . Note
that, with such regressors, one can manipulate
cxp(x![>]) while keeping exp(x'/?2) constant. The
two terms exp(x'PP are identified from the
observable erode hazards at i 0 because at / 0 no
dynamic selection due to the
unobserved heterogeneity has taken place yet. Now
suppose one manipulates x in the way described
above. If 7\, T 2 \X are independent, then the
observable erode hazard rate of T2 at t > 0, given
that 71 > t, does not vary along. But, if 71; 72|A are
dependent, then this crude hazard rate does vary
along, for the following reason. First, changes in
explx'/^) affect the distribution of unobserved
heterogeneity V\ among
Competing Risks Model
the survivors at I, due to the well-known fact that
V i and X are dependent conditional on survival (i.e.
conditional on 7) > t > 0) even though they are
independent unconditionally. Second, if V\ and V)
are dependent, this affects the distribution of V)
among the survivors at t, which in turn affects the
observable crude hazard of T2 at t given that '/'i > t.
In sum, the variation in this crude hazard with
exp(x'j8i) for given exp(x'p2) is informative on the
dependence of the durations. An analogous
argument holds for the crude hazard rate
corresponding to cause 7=1.
Note that identification is not based on exclusion
restrictions of the sort encountered in instrumental
variable analysis, which require a regressor that
affects one endogenous variable but not the other.
Here, all explanatory variables are allowed to affect
both duration variables - they are just not allowed to
affect the duration distributions in the same way.
Identification with regressors was first established
by Heckman and Honore ( ), who considered a
somewhat larger class of models than the MMPH
model and accordingly imposed stronger conditions
on the support of X.
Although the MPH model is identified from
single-risk duration data where we observe a single
spell per subject, there is substantial evidence that
estimates are sensitive to misspecification of
functional forms of model elements (see van den
Berg , for an overview). This implies that estimates
of MMPH models using competing- risks data
should also be viewed with caution. It is advisable
to include additional data. For example, longitudinal
survey data on unemployment durations subject to
right-censoring can be augmented with register data
or retrospective data not subject to censoring (see
for example van den Berg et al. ). More in general,
one may resort to ‘multiple-spell competing risks’
data, meaning data with multiple observations of Y,
Z for each subject. For a given subject, such
observations can be viewed as multiple independent
draws from the subject-specific distribution of Y, Z,
on the assumption that the unobserved
heterogeneity terms V,, V2 are identical across the
spells of the subject. Here, a subject can denote a
single physical unit, like an individual, for which
we observe
Competing Risks Model
two spells in exactly the same state, or it can denote
a set of physical units for which we observe one
spell each. Multiple-spell data allow for
identification under less stringent conditions than
single-spell data. Abbring and van den Berg ( )
showed that such data identify models that
allow for full interactions between the elapsed
durations t and x in 6 t{t\ x, V), and, indeed, allow the
corresponding effects to differ between the first and
the second spell. The assumptions on the support of
X are similar to above. Fermanian ( ) developed
a nonparametric kernel estimator of the Heckman
and Honore ( ) model.
Another approach to deal with nonidentification
of dependent competing risks models is to
determine bounds on the sets of marginal and joint
distributions that are compatible with the observable
data. Peterson ( ) discrete-time or interval-censored competing risks
derived sharp bounds in terms of observable quan-
tities. They are often wide. In case of the marginal
distributions of two sub-populations distinguished
by a variable X, the bounds associated with the
different X may overlap, whether or not X
(monotonically) affects (one of) the marginal
distributions. With overlap, the causal effects of X
cannot even be signed.
Bond and Shaw ( ) combined bounds with
regressors. In the case of a single binary regressor,
the only substantive assumption made is that there
exist increasing functions g and h such that 7), T2\
X=0 equals g('I'\), h(T2)\X = 1 in distribution. In
words, the dependence structure is invariant to the
values of the regressors, so the latter affect only the
marginal distributions. Specifically, the copula (and
therefore Kendall’s x) of the joint distribution is
invariant to the value of X. The assumption is
satisfied by the aforementioned competing risks
models with regressors. Clearly, by itself the
assumption is insufficient for point identification.
The bounds concern the regressor effects on the
marginal distributions. If it is assumed that X affects
the marginal distributions of Tt in terms of first-
order stochastic dominance, the bounds are
sufficient to sign the effect of X on at least one of
the marginal distributions (so, in case of MMPH
models, also on at least one of the individual
marginal distributions conditional on V).
1929
We end this article by noting some connections
between competing risks models and other models.
First, they are related to switching regression
models, or Roy models. For example, if l',\X, Vin
the MMPH model have Weibull distributions, then
we can write log I) = x,j, + i = 1,2) (for example,
van den Berg et al. ), where we observe 7) iff J) < /}
(/ / /). Second, competing risks models are building
blocks of multivariate duration models, notably
models where one of the durations is always
observed (for example, 7) captures the moment of a
treatment and T2 is the observed duration outcome
of interest).
We have considered only continuous-time
duration variables 7) that have different realizations
with probability 1. Recently, semi-parametric and
nonparametric results have been derived for
models and models where different risks can be
realized simultaneously (see for example Bedford
and Meilijson ; van den Berg, van Lomwel and van
Ours ; Honore and Lleras-Muney ). The
biostatistical literature contains many studies in
which specific assumptions are made on the
dependence structure of the two durations Th
enabling inference on the marginal distributions
from data on Y, Z (see for example Moeschberger
and Klein , for a survey).
See Also
Bibliography
Abbring, J., and G. van den Berg. 2003. The identifiability of
the mixed proportional hazards competing risks model.
Journal of the Royal Statistical Society,
S e r i e s B 65: 701-710.
Bedford, T., and I. Meilijson. 1997. A characterization of
marginal distributions of (possibly dependent) lifetime
variables which right censor each other. A n n a l s o f
S t a t i s t i c s 25: 1622-1645.
Bond, S., and J. Shaw. 2006. Bounds on the covariate-time
transformation for competing-risks survival analysis.
L i f e t i m e D a t a A n a l y s i s 12: 285-303.
1930
Cox, D. 1959. The analysis of exponentially distributed life-
times with two types of failure. J o u r n a l o f t h e
R o y a l S t a t i s t i c a l S o c i e t y . S e r i e s B 21:
411-421.
Fermanian, J. 2003. Nonparametric estimation of competing
risks models with covariates. J o u r n a l
M u l t i v a r i a t e A n a l y s i s 85: 156-191.
Heckman, J., and B. Honore. 1989. The identifiability of the
competing risks model. B i o m e t r i k a 76: 325-330.
Honore, B., and A. Lleras-Muney. 2006. Bounds in competing
risks models and the war on cancer. E c o n o m e t r i c a
74: 1675-1698.
Moeschberger, M., and! Klein. 1995. Statistical methods for
dependent competing risks. L i f e t i m e
A n a l y s i s 1: 195 204.
Peterson, A. 1976. Bounds for a joint distribution function
with fixed sub-distribution functions: application to
competing risks. P r o c e e d i n g s o f t h e N a t i o n a l
A c a d e m y o f S c i e n c e s I S : 11-13.
van den Berg, G. 1990. Search behaviour, transitions to
nonparticipation and the duration of unemployment.
E c o n o m i c J o u r n a l 100: 842-865.
van ilcn Berg, G. 2001. Duration models: Specification,
identification, and multiple durations. In H a n d b o o k
o f e c o n o m e t r i c s , ed. J. Heckman and E. Learner,
Vol. 5. Amsterdam: North-Holland.
van den Berg, G., M. Lindeboom, and G. Ridder. 1994.
Attrition in longitudinal panel data, and the empirical
analysis of dynamic labour market behaviour. J o u r n a l
o f A p p l i e d E c o n o m e t r i c s 9: 421-435.
van den Berg, G., van Lomwel, A., and van Ours, J. 2003.
Nonparametric estimation of a dependent competing risks
model for unemployment durations. Discussion Paper No.
8l>x. Bonn: l/.A.
Competition
George J. Stigler
Abstract
Competition arises whenever two or more
parties strive for something that all cannot
obtain. The classical economists felt no need for
a very precise definition of competition because
they viewed monopoly as highly exceptional. In
the late 19th century competition became the
subject of intense analysis; the concept of
perfect competition emerged as the standard
model of economic theory and as first
approximation in the concrete studies of applied
microeconomics. The limitations of
Competition
the concept in dealing with conditions of per-
sistent and imperfectly predicted change will be
removed only when economics possesses a
developed theory of change.
of
Keywords
Bilateral monopoly; Clark, J.; Cliffe Leslie, T.;
Coalitions; Comparative statics; Competition;
Competitive equilibrium; Contracting; Cournot,
A.; Creative destruction; Darwin, C.; Dcmsctz,
Data H.; Edgeworth, F.; Entrepreneur- ship; Factor
price equalization theorem; Fisher, L; Industrial
organization; Innovation; .lenkin, F.; Jevons, W.;
Joint action; Kirzner, 1.; Knight, F.; Labour
markets; Laissez faire; Law of indifference; Law
of one price; Law of supply and demand; Long-
run equilibrium; Malthus, T.; Market price;
Markets; Marshall, A.; Mathematics and
economics; Mill, .1. S.; Monopoly; Oligopoly;
Pareto, V.; Perfect competition; Perfect
information; Perfect markets; Pigou,
A. ; Profit-maximixing behaviour; Rate of
return; Recontracting; Resource mobility;
Schumpeter, .1.; Second best; Senior, N.; Smith,
A.; Stigler, G. J.; Thornton, H.; Walras,
L. ; Workable competition
JEL Classifications
B0
Competition is a rivalry between individuals (or
groups or nations), and it arises whenever two or
more parties strive for something that all cannot
obtain. Competition is therefore at least as old as
man's history, and Darwin (who borrowed the
concept from economist Malthus) applied it to
species as economists had applied it to human
behaviour.
A concept that is applicable to two cobblers or a
thousand shipowners or to tribes and nations is
necessarily loosely drawn. When Adam Smith
launched economics as a comprehensive science in
1776, he followed this usage. He explained why a
reduced supply of a good led to a higher price: the
'competition [which] will immediately begin’
among buyers would bid up the price. Similarly if
the supply become larger, the price would sink
Competition
more, the greater ‘the competition of the sellers’
(Smith [ ] 1976, pp. 73—4). Here competition
was very much like a race: a race to obtain part of
reduced supplies or to dispose of a part of increased
supplies. Almost nothing except a number of buyers
and sellers was necessary for competition to
operate. And the greater the number of each, the
greater the vigour of competition:
If this capital [sufficient to trade in a town] is divided
between two different grocers, their competition will
tend to make both of them sell cheaper, than if it
were in the hands of one only; and if it were divided
among twenty, their competition would be just so
much the greater, and the chance of their combining
together, in order to raise the price, just so much the
less, (ibid., pp. 361-2)
With such a loose concept, there was little occasion
to speak of one market as being more or less
competitive than another, although this very pas-
sage presented the commonsense idea that larger
numbers of rivals increased the intensity of
competition.
The competition of grocers in a town pertained
to competition within a market or an industry. Smith
made much of the competition of different markets
or industries for resources, and he developed what
has always remained the main theorem on the
allocation of resources in an economy composed of
private, competing individuals or enterprises. The
argument may be stated: Each owner of a
productive resource will seek to employ it where it
will yield the largest return. As a result, under
competition each resource will be so distributed that
it yields the same rate of return in every use. For if a
resource were earning more in one use than another,
it would be possible for its return in the lower-
yielding use to be increased by reallocating it to the
higher-yielding use. And this theorem led to what
John Stuart Mill called the most frequently
encountered proposition in economics: ‘There
cannot be two prices in the same market’ (Mill ,
Book II, eh. IV, s. 3).
The competition of different markets or indus-
tries for the use of the same resources called
attention to some problems which are less important
within a single market such as the grocery trade in a
town. One must possess knowledge of the
investment opportunities in these different
1931
employments, and that knowledge is less commonly
possessed than knowledge within one market. It
often requires a good deal of time to disengage
resources from one field and install them elsewhere.
Both of these conditions were recognized by Smith,
who spoke of the difficulty of keeping secret the
existence of extraordinary profits, and of the long
mn sometimes required for the attainment of
equality of rates of return.
For the next three-quarters of a century the
prevailing treatment of competition followed the
practice of Smith. One can find occasional hints of a
more precise definition of competition, well
illustrated by Nassau W. Senior:
But though, under free competition, cost of produc-
tion is the regulator of price, its influence is subject
to much occasional interruption. Its operation can be
supposed to be perfect only if we suppose that there
are no disturbing causes, that capital and labour can
be at once transferred, and without loss, from one
employment to another, and that every producer has
frill information of the profit to be derived from
every mode of production. But it is obvious that these
suppositions have no resemblance to the truth. A
large portion of the capital essential to production
consists of buildings, machinery, and other
implements, the results of much time and labour,
and of little service for any except their existing
purposes... few capitalists can estimate, except upon
an average of some years, the amount of their own
profits, and still fewer can estimate those of their
neighbours. ( , p. 102)
Senior is hinting at a concept of perfect competi-
tion, but the hint is not pursued.
The classical economists felt no need for a
precise definition because they viewed monopoly as
highly exceptional: Harold Demsetz has counted
only one page in 90 devoted to monopoly in The
Wealth of Nations and only one in 500 in Mill’s
Principles of Political Economy. Indeed the word
‘monopoly’ was usually restricted to grants by the
sovereign of exclusive rights to manufacture, import
or sell a commodity; witness the entry in the Penny
Cyclopedia ( ):
It seems then that the word monopoly was never
used in English Law, except when there was a royal
grant authorizing some one or more persons only to
deal in or sell a certain commodity or article.
If a number of individuals were to unite for the
purpose of producing any particular article or com-
modity, and if they should succeed in selling such
1932
article very extensively, and almost solely, such
individuals in popular language would be said to
have a monopoly. Now, as these individuals have no
advantages given them by the law over other per-
sons, it is clear they can only sell more of their
commodity than other persons by producing the
commodity cheaper and better. (XV, p. 341)
The ability of rivals to seek out and compete away
supernormal profits, unless prevented by legal
obstacles, was bebeved to be the basic reason for
the pervasiveness of competition.
In the last third of the 19th century the concept
of competition became the subject of intense study.
The most popular reason given for this attention is
that the growth of large-scale enterprises, including
railroads, public utilities, and finally great
manufacturing enterprises, made obvious the fact
that a simple concept of competition no longer fit
the economy of an industrial nation such as
England.
A second source of misgiving with the broad
definition of competition is that it might not lead to
the uniformity of returns to a resource predicted by
the theory. The Irish economist Cliffe Leslie
repeatedly made this charge:
Economists have been accustomed to assume that
wages on the one hand and profits on the other are,
allowing for differences in skill and so forth, equal-
ized by competition, and that neither wages nor
profits can anywhere rise above ‘the average rate’,
without a consequent influx of labour or of capital
bringing things to a level. Had economists, however,
in place of reasoning from an assumption, examined
the facts connected with the rate of wages, they would
have found, from authentic statistics, the actual
differences so great, even in the same occupation,
that they are double in one place what they are in
another. Statistics of profits are not, indeed,
obtainable like statistics of wages; and the fact that
they are not so, that the actual profits are kept a
profound secret in some of the most prominent
trades, is itself enough to deprive the theory of equal
profits of its base. ( , pp. 158-9)
The easiest way to combat such criticisms was not
to confront them with data - that path was not
chosen for many years - but to define competition
in such a way as to ensure the desired results such
as uniformity of price. This definition of competition was especially
The complications possible with competition appropriate in Cournot’s system because,
were raised also on the theoretical side. William T.
Thornton, in his book On Labour ( ),
Competition
denied the fact that prices were determined by the
Taw of supply and demand’, particularly within
labour markets. He employed bizarre examples,
such as supply and demand curves which coincided
over a vertical range, to show that price could be
indeterminate or unresponsive to changes in supply
or demand. These objections naturally called forth
responses, from both J.S. Mill (Collected Works, V)
and Fleeming Jenkin, a famous engineer.
The most persuasive reason for the increasing
attention to the concepts of economics was the
gradual move of economic studies to the univer-
sities, which proceeded rapidly in the last decades of
the century. The expanding use of mathematics was
one major symptom of the development of the
formal and abstract theory of economics by Walras,
Pareto, Irving Fisher and others. That formalization
would scarcely be possible without a more precise
specification of the nature of competition, and the
precise specification of the nature of competition,
and the replies to Thornton’s criticisms were a
precursor to this literature.
The groundwork for the development of the
concept of perfect competition was laid by Augustin
Cournot in 1838 in his Mathematical Principles of
the Theory of Wealth. He made the first systematic
use of the differential calculus to study the
implications of profit-maximizing behaviour.
Starting with the definition, Profits = Revenue —
Costs, Cournot sought to maximize profits under
various market conditions. He faced the question:
How does revenue (say, pq) vary with output (qfi
The natural answer is to define competition as that
situation in which p does not vary with q - in which
the demand curve facing the firm is horizontal. This
is precisely what Cournot did:
The effects of competition have reached their limit,
when each of the partial productions D k [the output
of producer k ] is i n a p p r e c i a b l e , not only with
reference to the total production D = F ( p ) , but
also with reference to the derivative F ' ( p ) , so that
the partial production D k could be subtracted from
D without any appreciable variation resulting in the
price of the commodity. (Cournot [ ] 1927,
p. 90)
Competition
according to his theory of oligopoly, the excess of
price over marginal cost approached zero as the
number of like producers became large. The argu-
ment is as follows:
Let the revenue of the firm be q,p, and let n
identical firms have the same marginal costs, MC. By perfect knowledge Jevons meant only that each
Then the equation for maximum profits for one firm trader in a market knew the price bids of every other
would be
p + qfdp/dq) = MC.
The sum of n such equations would be np +
q(dp/dq) = nMC, for nqt q. This last equation may
be written, p = MC —p/nE,
where E is the elasticity of market demand (Cournot
, p. 84).
Cournot believed that this condition of compe-
tition was fulfilled ‘for a multitude of products, and,
among them, for the most important products’.
Cournot’s definition was enormously more
precise and elegant than Smith’s so far as the
treatment of numbers was concerned. A market
departed from unlimited competition to the extent
that prices exceeded the marginal cost of the firm,
and the difference approached zero as the number of
rivals approached infinity. This definition, however,
illuminated only the effect of number of rivals on
the power of individual firms to influence the
market price, on Cournot’s special assumption that
each rival believed that his output decisions did not
affect the output decisions of his rivals. It therefore
bore only on what we term market competition.
Cournot did not face the question of the role of
information possessed by traders, and this question
was taken up by William Stanley Jevons in 1871 in
his Theory of Political Economy. He characterized a
perfect market by two conditions:
(1.) A market, then, is theoretically perfect only when
all traders have perfect knowledge of the conditions
of supply and demand, and the consequent ratio of
exchange; ... (2.) ... there must be
1933
perfectly free competition, so that any one will
exchange with any one else upon the slightest
advantage appearing. There must be no conspiracies
for absorbing and holding supplies to produce
unnatural ratios of exchange. (Jevons , pp. 86, 87)
trader. The second condition ruled out any joint
actions by two or more traders, without his noticing
that with knowledge so perfect as to know the
behaviour of rivals, there might appear the very
conspiracies he ruled out. The two conditions
dictated that ‘there cannot be two prices for the
same kind of article’ in a perfect market, which he
called the Taw of indifference’.
The merging of the concepts of competition and
the market was unfortunate, for each deserved a full
and separate treatment. A market is an institution
for the consummation of transactions. It performs
this function efficiently when every buyer who will
pay more than the minimum realized price for any
class of commodities succeeds in buying the
commodity, and every seller who will sell for less
than the maximum realized price succeeds in selling
the commodity. A market performs these tasks more
efficiently if the commodities are well specified and
if buyers and sellers are fully informed of their
properties and prices. Also a complete, perfect
market allows buyers and seller to act on differing
expectations of future prices. A market may be
perfect and monopolistic or imperfect and
competitive. Jevons’s mixture of the two has been
widely imitated by successors, of course, so that
even today a market is commonly treated as a
concept subsidiary to competition.
Edgeworth was the first economist to attempt a
systematic and rigorous definition of perfect com-
petition. His exposition deserves the closest scrutiny
in spite of the fact that few economists of his time
or ours have attempted to disentangle and uncover
the theorems and conjectures of the Mathematical
Psychics ( ), probably the
most elusively written book of importance in the
history of economics. His exposition was the most
influential in the entire literature.
The conditions of perfect competition are stated
as follows:
1934
T h e f i e l d o f c o m p e t i t i o n with reference to a
contract, or contracts, under consideration consists of
all individuals who are willing and able to recontract
about the articles under consideration ...
There is free communication throughout a n o r -
m a l competitive field. You might suppose the con-
stituent individuals collected at a point, or connected
by telephones - an ideal supposition [1881], but
sufficiently approximate to existence or tendency for
the purposes of abstract science.
A p e r f e c t field of competition professes in
addition certain properties peculiarly favourable to
mathematical calculation; ... The conditions of a
p e r f e c t field are four; the first pair referable to the
heading m u l t i p l i c i t y or continuity, the second to
d i v i d e d n e s s or fluidity.
I. An individual is free to r e c o n t r a c t with any
out of an indefinite number, .. .
II. Any individual is free to c o n t r a c t (at the same
time) with an indefinite number;
... This condition combined with the first
appears to involve the indefinite divisibility
of each article of contract (if any X deal with
an indefinite number of Fs he must give each
an indefinitely small portion of .r); which
might be erected into a separate condition.
III. Any individual is free to r e c o n t r a c t with
another independently of, w i t h o u t t h e
c o n s e n t being required of, any third party,...
IV. Any individual is free to c o n t r a c t with
another independently of a third party; ...
The failure of the first [condition] involves the
failure of the second, but not vice versa; and the
third and fourth are similarly related (Edgeworth ,
pp. 17-19).
The essential elements of this formidable list of
conditions are two:
1. There are an indefinitely large number of inde-
pendent traders on each side of a market (the
Cournot condition).
2. Each trader can costlessly make tentative con-
tracts with everyone (hence the divisibility of
commodities) and alter these contracts
(recontract) so long as a more favourable con-
tract can be made. The result is perfect knowl-
edge (the Jevonian condition).
Edgeworth gave an intuitive argument for the
need for an indefinitely large number of traders on
both sides of a market. It proceeds as follows. Let
Competition
there be one seller and two buyers, and let the seller
gain all the benefits of the sale: each buyer is
charged the maximum price he would pay rather
than withdraw from the market. If now a second
seller appears, he will find it advantageous to offer
better terms to the two buyers: ‘It will in general be
possible for one of the [sellers] (without the consent
of the other), to recontract with the two [buyers], so
that for all those three parties the recontract is more
advantageous than the previously existing contract’
(ibid., p. 35). As the numbers of traders on each side
increase, the price approaches the competitive
equilibrium level where no individual trader can
influence it.
A defect in this argument is that it ignores the
fact that if the traders on one or both sides of the
market, be they 2, or 2000 or 2,000,000, join
together they can do better individually than by
competing. If traders on each side join, however,
there will be bilateral monopoly, not competition.
Edgeworth gives no reason why the combination of
traders tails to take place. Only in modem times has
the reason for independent behaviour by rivals been
established: the costs of reaching and enforcing
agreements on joint action increase with both the
number of rivals and the complexity of the trans-
actions. At a certain level - quite possibly with only
two traders under some conditions - the costs of
joint action exceed the gain to at least some of the
traders, and independent behaviour emeiges.
Edgeworth’s ‘conjecture’, as it is now often
called, that a unique, competitive price would
emerge when the number of traders became large,
has given rise to a modem literature vast in scope
and often highly advanced in its mathematical
techniques (for references, see Hildenbrand ). One
result in this literature is that in the case of a large
(infinite) number of traders, no coalition of a portion
of the traders can exclude traders outside the
coalition from trading at the price-taking
equilibrium.
Edgeworth’s introduction of the requirement
that the commodity or service that is traded be
highly divisible is a response to the following
problem:
Suppose a market, consisting of an equal number of
masters and servants, offering respectively wages
and service; subject to the condition that no man can
Competition 1935

serve two masters, no master employ more than one free movement of resources is even approached in
man; or suppose equilibrium already established the real economy. Nor is the assumption of perfect
between such parties to be disturbed by any sudden
influx of wealth into the hands of the masters. Then
mobility necessary to eliminate monopoly power in
there is no d e t e r m i n a t e , and very generally a market: in the Victorian age, the price of wheat of
u n i q u e , arrangement towards which the system Iowa was set in Liverpool even though
tends under the operation of, may we say, a law of transportation costs were substantial. The
Nature, and which would be predictable if we knew
beforehand the real requirements of each, or of the
assumption is usually necessary to attain strict
average, dealer; ... (Edgeworth , p. 46). equality in the price of a good at every point (the
law of one price ), although even this is not strictly
Consider the simple example: a thousand masters
true (as in the factor price equalization theorem).
will each employ a man at any wage below 100; a
Clark also demanded that the economy be stationary
thousand labourers will each work for any wage
for perfect competition, a condition we shall return
above 50. There will be a single wage rate: knowl-
to later.
edge and numbers are sufficient to lead a worker to
All the elements of a concept of perfect com-
seek a master paying more than the going rate or a
petition were in place by 1900, and this concept
master to seek out a worker receiving less than the
increasingly became the standard model of eco-
market rate. But any rate between 50 and 100 is a
nomic theory thereafter. The most influential
possible equilibrium. But if a single worker leaves
statement of the conditions for perfect competition
the market, the wage will rise to 100, and if a single
was made by Frank H. Knight in his doctoral
employer withdraws, the wage will fall to 50. This
dissertation, Risk, Uncertainty and Profit ( ).
ability of a single trader to affect the price arises
The conditions were stated in extreme form; for
because of the lumpiness of the article traded (here
example, ‘There must be perfect, continuous,
a worker’s labour for a given period). Once a
costless intercommunication between all individual
worker can work for two masters, the withdrawal of
members of the society’ (Knight , p. 78) -
one worker in a thousand will reduce the available
so Jones in Seattle would know the price of potatoes
hours of work per day to each employer by only
and be able costlessly to ship to Smith in Miami a
8/1000 hours or 4.8 minutes per day, with only
bushel of potatoes at every moment of time.
negligible influence upon the wage rate.
Of course these conditions are not necessary,
Alternatively, a distribution of wage offers and
but only sufficient, to achieve the competitive
demands would also eliminate the indeterminacy
equilibrium. For example, if even a considerable
and market power.
fraction of buyers knows that seller A is charging
Edgeworth’s analysis was limited to competition
more than B for a given commodity, their patronage
within a market, and it was left to John Bates Clark
may be quite enough to force A to reduce his price
to emphasize the need for mobility of resources if
to that of B. Nor are the various conditions
the return on each resource was to be equalized in
independent of one another: for example, if it is
every use.
.. .there is an ideal arrangement of the elements of very cheap for either a commodity or its buyers or
society, to which the force of competition, acting on sellers to move between two places, that will insure
individual men, would make the society conform. that the prices in the two places will be widely
The producing organism actually shapes itself about known.
his model, and at no time does it vary greatly from
it ... We must use assumptions boldly and advisedly,
Along with the development of the concept of
make labour and capital absolutely mobile, and competition as a standard component of the theory
letting competition work in ideal perfection. (Clark , of prices and the allocation of resources, it acquired
pp. 68, 71) a growing role as the criterion by which to judge the
Perfect and free mobility of resources is of efficiency of actual markets. Adam Smith had
course an even more extreme assumption than the already advanced the proposition that output was
other conditions required for perfect competition maximized in a private enterprise economy with
because there is less reason to believe that competition. If each owner of a
1936
resource maximized the return from his resources,
then (in the absence of ‘external’ effects of one
person’s actions on others) aggregate output would
be maximized. This theorem (labelled ‘on
maximum satisfaction’) was developed and qual-
ified by Leon Walras (1874), Alfred Marshall
(1980), Pareto (1895-6, 1907), Pigou (1912) and a
host of modem economists.
Competition is much too central a concept in
economics to remain unaffected when economists
change their interests or analytical methods. We
may illustrate this fact by the problem of economic
change.
In a regime of change, of growing population
and capital or innovations or new consumer
demands, the problem of defining competition is
much more difficult than it is for the stationary
economy. Unless the change is predictable with
precision, knowledge must necessarily be incom-
plete and errors and lags in adaptation to new
conditions can be large. For this reason, indeed, J.B.
Clark believed that perfect competition was
achievable only in the stationary economy.
Even short-run changes in market price raise the
question: is the change in price initiated by a
particular seller or buyer, and if so, is this trader not
facing a negatively sloping demand curve or a
positively sloping supply curve? The infinitely
elastic supply and demand curves of perfectly
competitive equilibrium seem inapplicable to
periods of changing market conditions. Some
economists nevertheless retain the condition that
individual traders cannot influence price by intro-
ducing a hypothetical auctioneer who announces
price changes.
A partial adaptation of the competitive concept
to change is made by making it a long-run equi-
librium concept. Even if resources are not costlessly
mobile and even if entrepreneurs do not have
perfect foresight, one can analyse the rate of
approach of returns on resources to equality. If an
industry experiences a once-for-all large change, it
could be in competitive equilibrium before and after
the change, and the equilibria could be studied by
competitive theory (comparative statics).
This adaptation did not satisfy Joseph
Schumpeter, who believed that incessant change
Competition
in products and production methods was the very
essence of competitive capitalism. He argued that
the displacing of one product or method by another,
a process which he called creative destruction, made
the concept of perfect competition irrelevant to
either positive analysis or welfare judgements. If the
monopoly that reduced output, compared to
competition, by 10 per cent in one year, increased
output by 100 per cent over the next two decades,
then monopoly might be preferred to stagnant
competition.
It is crucial to this argument that monopoly
provides large, though temporary, rewards to suc-
cessful innovators but competition does not:
But perfectly free entry into a n e w field may make
it impossible to enter it at all The introduction of new
methods of production and new commodities is
hardly conceivable with perfect - and perfectly
prompt - competition from the start. And this means
that the bulk of what we call economic progress is
incompatible with it. As a matter of fact, perfect
competition is and always has been temporarily
suspended whenever anything new is being
introduced - automatically or by measures devised
for that purpose - even in otherwise perfectly
competitive conditions. (Schumpeter , pp. 104-5)
Schumpeter relies on instantaneous rivalry to
eliminate the incentives to innovation under com-
petition, and the conclusion would not hold if
competition is defined in terms of long-run
equilibrium.
Nevertheless the issue is not disposed of so
easily. If change is continuous rather than sporadic,
long-run equilibria will never be frilly achieved.
Several economists have emphasized that
alterations in the concept of competition are called
for in periods of historical change. Kirzner has
emphasized the role of entrepreneiuial rivalry in
competition, whereas such rivalry is nonexistent in
a perfectly competitive equilibrium. Demsetz has
proposed a concept of laissez-faire competition, in
which freedom of resoiuces to move into any use is
the central element. Such realistic reversions to the
competitive concept of the classical economists
have not been systematically formalized into
theoretical models.
The concept of perfect competition, or indeed
any theoretically precise concept of competition,
Competition
will not be met by the actual condition of compe-
tition in any industry. John Maurice Clark made the
most influential effort to create a concept of
‘workable competition’ which would serve as a
working rule for public policies which seek to
preserve or increase competition.
Clark emphasized the fact that if one requisite of
perfect competition is absent, it may be desirable
that a second requisite also be unfulfilled. For
example, with instantaneous mobility but imperfect
knowledge, members of an occupation would keep
shifting back and forth between two cities, always
overshooting the amount of migration which would
equalize wage rates. This propensity to overshoot
equilibrium would be corrected with less mobility
of labour. This problem was later formalized as the
theory of the ‘second best’.
The essence of the concept of workable com-
petition was the belief that Tong- run curves, both
of cost and of demand, are much flatter than short-
run curves, and much flatter than the curves which
are commonly used in the diagrams of theorists’
(J.M. Clark , p. 460). This correct and sensible view
led to a proliferation of studies, usually in doctoral
dissertations, of individual industries, in which the
workableness of competition in each industry was
appraised. Unfortunately there were no objective
criteria to guide these judgements, and there was no
evidence that the studies were accepted by the
governmental agencies which administered
competitive policies.
The popularity of the concept of perfect com-
petition in theoretical economics is as great today as
it has ever been. The concept is equally popular as
first approximation in the more concrete studies of
markets and industries that comprise the field of
‘industrial organization’ (applied microeconomics).
The limitations of the concept in dealing with
conditions of persistent and imperfectly predicted
change will not be removed until economics
possesses a developed theory of change. Even
within a stationary economic setting the concept is
being deepened by mathematical economists (see
Mas-Colell ). Meanwhile the central elements of
competition - the freedom of traders to use their
resources where they will, and exchange them at
any price they wish - will
1937
continue to play a major role in the economics of an
enterprise economy.
See Also
Bibliography
Clark, J.B. 1899. T h e d i s t r i b u t i o n o f w e a l t h .
London: Macmillan.
Clark, J.M. 1940. Toward a concept of workable competition.
A m e r i c a n E c o n o m i c R e v i e w , June; reprinted
in R e a d i n g s i n t h e s o c i a l c o n t r o l o f
i n d u s t r y . Philadelphia: Blakiston, 1942.
Cliffe Leshe, T.E. 1888. E s s a y s in political
e c o n o m y a n d m o r a l p h i l o s o p h y . London:
Longmans Green.
Cournot, A. 1838. Researches into the mathematical principles
of the theory of wealth. Reprinted. New York: Macmillan,
1927.
Demsetz, H. 1982. Economic, legal and political dimensions of
competition. Amsterdam: North-Holland.
Edgeworth, F.Y. 1881. M a t h e m a t i c a l p s y c h i c s .
London: Kegan Paul, 1932.
Hildenbrand, W. 1974. C o r e a n d e q u i l i b r i a o f a
l a r g e e c o n o m y . Princeton: Princeton University
Press.
Jevons, W.S. 1871. T h e t h e o r y o f p o l i t i c a l
e c o n o m y . London: Macmillan.
Kirzner, I.M. 1973. Competition and
e n t r e p r e n e u r s h i p . Chicago: University of Chicago
Press.
Knight, F.H. 1921. R i s k , u n c e r t a i n t y a n d p r o f i t ,
Part 2. Boston: Houghton Mifflin Co.
Marshall, A. 1890. P r i n c i p l e s o f e c o n o m i c s .
London: Macmillan.
Mas-Colell, A., ed. 1982. Noncooperative approaches to the
theory of perfect competition. New York: Academic Press.
McNulty, P.J. 1967. A note on the history of perfect com-
petition. J o u r n a l o f P o l i t i c a l E c o n o m y 75:
395-399.
Mill, J.S. 1848. P r i n c i p l e s o f p o l i t i c a l e c o n o m y .
In C o l l e c t e d w o r k s , ed. J.M. Robson. Toronto:
University of Toronto Press, 1965.
Nutter, G.W. 1951. The extent of enterprise monopoly in the
United States, 1899-1939. Chicago: University of Chicago
Press.
Penny Cyclopedia ofthe Societyfor the Diffusion of Useful
Knowledge. 1839.
Schumpeter, J.A. 1942. C a p i t a l i s m , s o c i a l i s m a n d
d e m o c r a c y . New York: Harper & Bros.
Senior, N.W. 1836. P o l i t i c a l e c o n o m y . London: W.
Clowes.
Shepherd, W.G. 1982. Causes of increased competition in the
1938
Smith, A. 1776. T h e w e a l t h o f n a t i o n s . Glasgow ed.
Oxford: Oxford University Press. 1976.
Stigler, G.J. 1957. Perfect competition, historically con-
templated. J o u r n a l o f P o l i t i c a l E c o n o m y 65:
1-17.
Stigler, G.J., and R. Sherwin. 1985. The extent of the market.
J o u r n a l o f L a w a n d E c o n o m i c s 28: 555-585.
Thornton, W.T. 1869. O n l a b o u r . London: Macmillan.
Competition and Efficiency
John C. Panzar
The association between economic efficiency and
competition goes back at least as far as Adam
Smith’s ‘invisible hand’ metaphor. Indeed, a goodly
portion of the vast body of subsequent work in
value theory has dealt with the normative issues
arising from the workings of the competitive
economy. Thus any short essay on the topic must be
somewhat idiosyncratic, focusing upon the points
which are of greatest interest to the author.
Therefore, I shall limit my attention to the
properties of the (static, partial equilibrium)
economic model of perfect competition and how its
use has recently been extended to add to our
understanding of a larger range of real world
markets. I must leave to others the tasks of sorting
out the importance of competition in, for example,
the Schumpeterian process of ‘creative destruction’,
the aggregation and transmission of society’s stock
of information, or the evolutionary progress of
technological advance. Fortunately for my
purposes, the historical development of the
competitive model has been thoroughly analysed by
Stigler ( ). The formulation of
the model, as we know it today, was completed in
the work of Knight ( ). It is interesting to
note that the last refinement to be added was the
free mobility of resources across industries: i.e. the
entry and exist of firms. In his insightful concluding
section, Stigler points out that competition can
flourish within a market without this last ingredient.
(Consider an agricultural market with Ricardian
rents.) He suggested that the term ‘market
competition’ be used to describe such
Competition and Efficiency
situations, and that the term ‘industrial competition’
be applied when mobility across industries is
present. The work that I shall discuss deals with the
converse possibility: perfectly contestable markets,
situations in which competition may not necessarily
exist within a particular market, but firms (and
resources) are assumed to be perfectly mobile
across industries.
The role of entry and exit in assuring the equal-
ization of returns across markets is not logically
limited those cases in which it is technologically
feasible for the market to be populated by a large
number of firms, each capable of achieving an
efficient scale of operation. It may be expected that
the lure of profits might serve to make relevant
certain aspects of competitive theory even under
conditions of ‘natural monopoly’. The most striking
practical illustration of this point was the recent
deregulation of airlines in the United States. This
took place, in part, because the free mobility of
resources (aircraft) across markets led policy
makers to believe that satisfactory economic
performance could be achieved without the
stultifying effects of economic regulation. This,
despite the fact that most city-pair airline markets
are natural monopolies and none can be expected to
support the large numbers of firms required by the
perfectly competitive model. Thus the need to
extend at least part of the competitive paradigm to
incorporate such cases had become apparent.
In a classic article, Demsetz ( ) set forth
one way to break the commonly perceived link
between monopoly provision of certain increasing
returns services and monopoly conduct on the part
of the firm providing the service at any point in
time. By pointing out that the impossibility of
competition within the market need not preclude
effective competition for the market, Demsetz
raised a fundamental challenge to the conventional
wisdom that the only effective ways to deal with a
technological natural monopoly were through
economic regulation or public enterprise.
Demsetz chose to elaborate this idea in the
context of a franchise bidding scheme, in which the
franchise was to be awarded, not to the firm
offering the greatest lump sum payment to the
municipal coffers, but to the firm offering to
Competition and Efficiency
serve the market at the lowest price. Subsequent
authors have criticized this as a policy proposal,
focusing on the problems raised by considerations
of sunk costs and incomplete contracts, from which
Demsetz explicitly sought to abstract. However,
there is another sense in which the franchise
bidding example may have been an unfortunate
expository choice. Because it introduced a new
institution between the firm and the market - the
franchise auctioneer - this illustration may have
obscured the link between the analysis of
competition for the market and the earlier notion of
the role of free entry and exit in ensuring effective
industrial competition. Furthermore, Demsetz’s
simple bidding scheme cannot handle the realistic
cases in which the monopolist produces two or
more technologically related services.
The theory of contestable markets developed by
Baumol et al. ( ) is most usefully viewed
as an attempt to extend the neoclassical (partial
equilibrium ) theory of long-run competitive equi-
librium to the case of increasing returns to scale. In strictures of Definition can be interpreted to mean
so doing, they developed a model which achieved
the Demsetz solution to the monopoly problem as
the result of a market equilibrium process. This
extension was accomplished by emphasizing the
role played by potential entry in characterizing the
role defining properties of long- run competitive
equilibrium. To see this reinterpretation most
clearly, the following definitions are necessary:
Definition 1 A Feasible Industry Configuration making potential entrants ‘less optimistic’ in the
(FIC) is a collection of firms, i = 1, . . . , m output following sense:
vectors for each, y1, ..., and a market price vector p
such that each firm earns non-negative profits and Definition 3 A Sustainable Industry Configuration
the total quantity supplied equals the quantity (SIC) e ise any FIC
demanded; i.e. py‘ — C(y‘) >0, for all i = l, ... ,m
and J2y‘ = d(p) - where C is the (multiproduct)
minimum cost function and D the market demand
function.
Feasibility surely reflects the minimal condi-
tions one would expect to prevail in long-run
industry equilibrium in a private enterprise econ-
omy: All firms must earn non-negative profits and
the total quantity supplied by firms equals the
amount demanded by consumers at the market
1939
price. While feasibility requires financial viability,
it does not preclude the positive profits which may
attract entry. Therefore, the neoclassical notion of
long-run competitive equilibrium must encompass
some additional restrictions. More specifically,
Definition 2 A long-run competitive equilibrium is
any FIC which also has the property that py — c (y)
< 0 for ally.
While this characterization of long-run com-
petitive equilibrium may be unfamiliar, it is equiv-
alent to the standard notion of price taking firms
earning zero economic profits by equating marginal
cost to price. (To see this, note that since profits are
nonpositive for all output levels, the fact that py1
C(y‘) > 0 means that output level y1 maximizes the
rth firm’s profits. This, in turn, implies that MC(yl)
= p if firm i is producing.)
Characterizing competitive equilibrium via
Definitions and has the advantage of focusing
attention on the role played by potential entry. The
that the firms in an industry in long-mn competitive
equilibrium act as if they were policed by potential
entrants prepared to enter the market in pursuit of
any profit opportunity calculated at current market
prices. While this lack of attention to the possibility
of retaliatory price responses by rivals reflects the
noncooperative spirit of the competitive paradigm,
it ignores the response of consumers to a change in
the market price. Therefore it is useful to consider
e
which also satisfies the condition
that p y - C(y ) < 0 for all pe < p andye < D(pe).
Thus firms in a SIC behave as if the market
were policed by potential entrants that calculate the
profitability of entry under the assumption that
incumbent firms’ prices remained unchanged, but
that do take account of the reality that consumers
can be induced to purchase a larger quantity only at
a lower price. Put another way, an FIC is also an
SIC when no potential
1940
entrant can anticipate earning a positive profit by
quoting a price at or below that prevailing in the
market and serving all or a part of the resulting
demand. The following semantic clarification
completes the characterization of a contestable
market:
Definition 4 A perfectly contestable market is one
in which perfectly free entry and exit ensure that the
only possible long-ran equilibria are SICs.
An immediate implication of Definitions , , ,
and is:
Proposition 1 Any long-run competitive equilib-
rium is an SIC, but not conversely. Thus all
perfectly competitive markets are perfectly con-
testable, but not all perfectly contestable markets
are perfectly competitive.
The proof follows from the fact that the condi-
tions which an FIC must satisfy in order to be a
long-run competitive equilibrium are stronger than
those required of an SIC. Thus a long-run
competitive equilibrium is, by construction, an SIC.
To see that the converse is not true, consider the
case in which the average costs of production fall
throughout the relevant range; i.e. at least as far as
the intersection of the average cost curve and the
market demand curve. The point of intersection, the
Demsetz outcome, characterizes a sustainable
industry configuration, since profits are non-
positive and no point on or below the demand curve
can yield non-negative profits at a lower price.
However this outcome is clearly not a long-run
competitive equilibrium because price is equal to
average cost which, by hypothesis, is strictly greater
than marginal cost. The above demonstration points
out the fact that the concept of contestable markets
can be applied beyond the large-numbers case of
perfect competition. However it also raises
questions about the efficiency properties of such
markets. The fact that equilibrium may iiwolve a
price greater than marginal cost means that the First
Best optimality properties of the competitive model
need no longer apply. What efficiency properties,
then, can be associated with equilibria in
contestable markets? In the case of single product
markets it is intuitively clear (and straightforward to
prove) that,
Competition and Efficiency
when they exist, sustainable industry configurations
are solutions to the Second Best optimization
problem: maximize welfare (as measured, for
example, by the sum of producers’ and consumers’
surpluses) subject to the constraint that firms earn
non-negative profits. Clearly, when increasing
returns to scale render marginal cost pricing
unprofitable, the best that can be done, in the
absence of lump sum transfers and discriminatory
or non-linear pricing, is to set price equal to average
cost.
However, even this level of performance can no
longer be guaranteeed once one moves to the
realistic realm of multiple products. For example, a
monopolist producing two or more products can, in
general, find an infinite number of price com-
binations which will yield it exactly zero economic
profits. Some of these prices and resulting market
demand quantities may represent SICs. Call this set
P. If the underlying cost and demand functions are
sufficiently well-behaved, there will exist a unique
constrained welfare maximizing price vector p*.
The most desirable efficiency result would be for
the set P to consist of the single element p*.
Unfortunately, it is easy to construct examples in
which P does not contain p*, as well as cases in
which P is empty. What efficiency properties does
this generalized process of industrial competition
possess when extended beyond the realm of perfect
competition? The results that pertain generally he
entirely on the cost side.
Proposition 2 In any SIC, the industry’s output is
divided among the firms in a way that minimizes
total industry costs.
The proof is by contradiction. Consider an ini-
tial SIC composed of m firms producing output
vectors, y1, . . . , ym, at market prices p. Suppose,
contrary to hypothesis, that there exists an alterna-
tive group of Affirms with output vectors, z1,..., zk,
that could produce the current industry output at a
lower total cost That is JUz' = D(p) = JA/, but Y,fV)
< £,C(y'). Then the new group, in total, would earn
positive economic profits at the initial price p. This
is true because, by hypothesis, total revenues would
be equal, but total costs would be lower for the
alternative group, while the initial
Competition and Selection
group of firms must have been earning non- References
negative profits. Therefore at least one firm, say
firm j, in the alternative group would anticipate Baumol, W., J. Panzar, and R. Willig. 1982. C o n t e s t a b l e
earning strictly positive profits at the price vectorp.
But then there exists an entry plan // = p < p and ye Demsetz, H. 1968. Why regulate utilities? J o u r n a l o f
= zJ < D(pe) such that peye - C(ye) > 0 , which
contradicts the hypothesis that the initial group of
firms constituted a SIC.
Additional efficiency results for contestable
markets are presented in chapter 11 of Baumol,
Panzar and Willig. Here, I shall mention specifi-
cally a class of results which are relevant only in the
multiproduct context. One implication of the fact
that equilibrium in a contestable market presents no Competition and Selection
profit opportunities for potential entrants is that no
subset of services of a multiproduct enterprise can Sidney G. Winter
generate revenues in excess of the cost of providing
them alone. Thus, equilibrium in perfectly
contestable markets cannot involve one gorup of
services subsidizing another. Whether or not this
property is a desirable efficiency result is unclear.
Consider, for example, a situation in which a
monopoly firm uses common facilities to produce
two services, one of which has a very elastic
demand curve while that of the other is very
inelastic. Maximizing total surplus subject to a
break-even constraint leads to the well- known
inverse elasticity rule: the markup of price over
marginal cost is greater for services whose demand
is least elastic. However, this pricing policy may
easily lead to revenues from the inelastic service in
excess of the cost of providing it alone. Such an
outcome would not be an SIC and could not persist
in a perfectly contestable market. Thus while the
mobility of firms and resources can, even without
the presence of market competition, ensure
productive efficiency, it cannot in general guarantee
that an optimal relationship of output prices in a
multiproduct industry will prevail outside the
perfectly competitive realm.
See Also
►I
1941
markets and the theory of industry
s t r u c t u r e . New York: Harcourt Brace Jovanovich.
L a w a n d E c o n o m i c s 11: 55-65.
Knight, F. 1921. R i s k , u n c e r t a i n t y , a n d p r o f i t .
Chicago: University of Chicago Press.
Stigler, G. 1957. Perfect competition, historically contem-
plated. J o u r n a l o f P o l i t i c a l E c o n o m y 65: 1-
17.
Abstract
The claim that a business firm must maximize
profit if it is to survive serves as an informal
statement of the common conclusion of a class
of theorems characterizing explicit models of
economic selection processes. Such models, by
making explicit the strong assumptions needed
to generate this sort of result, are the basis for a
critique of standard economic theory which
relies on competitive equilibrium. Models of
Schumpeterian competition, emphasizing the
centrality of innovation, plainly provide a much
better description of the world we live in than do
models of static equilibrium.
Keywords
Adjustment, dynamic vs static; Alchian, A.;
Behavioural change; Comparative statics;
Competition and selection; Competitive equi-
librium; Constant returns to scale; Decreasing
returns to scale; Diminishing returns; Entrepre-
neurial rents; Entrepreneurship; Evolutionary
economics; Fixed factors; Friedman, M.;
Increasing returns to scale; Innovation; Latent
productivity; Market power; Mimicry theorems;
Natural selection; Neoclassical growth theory;
Patents; Present value; Profit maximization;
Research and development; Rules of
1942
behaviour; Satisficing; Schumpeter, J.;
Schumpeterian competition; Selection equilib-
rium; Simon, H.; Static equilibrium; Survival of
the fittest; Technological opportunity; Winter, S.
G.
JEL Classifications
BO
Under competitive conditions, a business firm must
maximize profit if it is to survive - or so it is often
claimed. This purported analogue of biological
natural selection has had substantial influence in
economic thinking, and the proposition remains
influential today. In general, its role has been to
serve as an informal auxiliary defence, or cratch, for
standard theoretical approaches based on
optimization and equilibrium. It appeared explicitly
in this role in a provocative passage in Milton
Friedman’s famous essay on methodology
(Friedman , eh. 1), and it seems that many
economists are familiar with it in this context only.
There is, however, an alternative role that the
proposition can and does play. It serves as an
informal statement of the common conclusion of a
class of theorems characterizing explicit models of
economic selection processes. A model in this class
posits, first, a range of possible behaviours for the
firm. This range must obviously extend beyond the
realm of profit maximization if the conclusion of
the argument is to be non-trivial, and it must include
behaviour that is appropriately termed ‘profit
maximizing’ if the conclusion is to be logically
attainable at all. The model must also characterize a
particular dynamic process that in some way
captures the general idea that profitable firms tend
to survive and grow, while unprofitable ones tend to
decline and fail. A stationary position of such a
process is a ‘selection equilibrium’.
Models of this type occupy an important but
non-central position in evolutionary economic
theory (Nelson and Winter ). They establish that the
equilibria of standard competitive theory can indeed
be ‘mimicked’ (in several different senses) by the
equilibria of selection models. More importantly, by
making explicit the strong assumptions that
apparently are required to
Competition and Selection
generate this sort of result, they are the basis for a
critique of its generality and an appraisal of the
strength of the cratch on which standard theory
leans. They also provide a helpful entry-way to the
much broader class of evolutionary models in which
mimicry results fail to hold. This entry-way has the
convenient feature that the return path to standard
theory is well marked; the sense in which
evolutionary theory subsumes portions of standard
theory becomes clear.
The concept of competition need not, of course,
be considered only in the context of perfectly
competitive equilibrium. In a broader sense of the
term, any nontrivial selection model in which the
‘fit’ prosper and the ‘unfit’ do not is a model of a
‘competitive’ process. The process need not have a
static equilibrium, or any equilibrium, and it may
easily lead to results that are clearly non-
competitive by the standards of industrial
organization economics.
The remainder of this essay first considers in
more detail the theoretical links between selection
processes and competitive equilibrium outcomes. It
then examines a more interesting and less well-
explored area that involves selection and, in a broad
sense, competition; Schumpeterian competition.
Competitive Equilibrium as a
Selection Outcome
The intention here is to describe the heuristic basis
of existing examples of this type of theorem, or,
alternatively, to describe the basic recipe from
which an obviously large class of broadly similar
results could be produced. There may be other basic
recipes, as yet unknown. There certainly are ways to
ignore individual instructions of the recipe and yet
preserve the result, though at the cost of delicately
contrived adjustments in other assumptions.
(To avoid confusion, it should be noted at the
outset that the word ‘equilibrium’ is used in two
different senses in this discussion, the ‘no incentives
to change behaviour’ sense employed in economic
theory and the ‘stationary position of a dynamic
process’ sense that is common outside of
Competition and Selection 1943

economics. The point of the discussion is, in positive profit that is potentially achievable at
fact, to relate these two equihbrium ideas in a selection equihbrium scale.
particular way.)
In standard theory, expansion in response to
(1) Constant returns to scale must prevail in the profitability may be seen as an aspect of the firm’s
specific sense that the supply and demand profit-seeking on the assumption that it regards
functions of an individual firm at any particular prices as unaffected by its capacity decisions. In
time are expressible as the scale (or ‘capacity’ ) turn, this ordinarily requires that the firm in
of that firm at that time multiplied by functions question be but one of an indeterminately large
depending on prices, but not directly on scale number of firms that all have access to the same
or time. Increasing returns to scale must be technological and organizational possibilities.
excluded for familiar reasons. Decreasing While the assumption that firms have identical
returns must be excluded because they will in production sets and behavioural rules is common
general give rise to equilibrium ‘entrepreneurial and appears inoffensive in orthodox theorizing, it is
rents’ which could be partially dissipated by very much at odds with evolutionary theory. The
departures from maximization without orthodox view comes down to the assertion that all
threatening the survival of the firm. Thus, for productive knowledge is freely available to one and
example, the U-shaped long ran average cost all - perhaps it is all in the public library. By
curve of textbook competitive theory does not contrast, evolutionary theory emphasizes the role of
provide a context in which selection necessarily firms as highly individualized repositories of
mimics standard theory if competitive productive knowledge, not all of which is
equilibrium would require some firms to be on articulable. From the evolutionary perspective, the
the upward sloping portion of the curve. fact that mimicry theorems rely on assumptions of
unimpaired access to a public knowledge pool is by
(2) Firms must increase scale when profitable and itself sufficient to make it clear that the selection
decrease scale (or go out of business entirely) argument can provide only a weak and shaky crutch
when unprofitable. Alternatively, profitability for standard competitive theory.
of a particular firm must lead to entry by
perfect imitators of that firm’s actions. In the (3) A firm that is breaking even with a positive
absence of such assumptions, it is plain that output at prevailing prices must not alter its
there w i l l in general be equilibria with non- behaviour; a potential entrant that would only
zero profit levels, which under assumption (1) break even at prevailing prices must not enter.
cannot mimic the competitive result. While the This assumption is needed to assure that the
‘decline or fail’ assumption is a plausible competitive equilibrium position is in fact a
reflection of long-run breakeven constraints stationary position of the selection process.
characteristic of actual capitalist institutions, no
such realistic force attaches to the requirement Models of natural selection in biology do not
that profitability lead to expansion. If firms do typically involve this sort of assumption, but neither
not pursue profits in the long-run sense of do they conclude that only the fittest genotypes
expanding in response to positive profitability, survive - the biological analogue of the proposition
stationary positions may involve positive discussed here. Rather, they show how constant
profits. Such stationary positions fail to mimic gene frequencies come to prevail as the selection
competitive equilibria for that reason alone forces that tend to eliminate diversity come into
(given constant returns), but they also introduce balance with mutation forces that constantly renew
once again the possibility that the short-run it. A strictly analogous treatment of economic
behavioural responses of surviving firms may selection would be much more appealing than the
dissipate some of the sort of result discussed here.
1944
It would admit that occasional disruptions may
arise from random behavioural change, or from
over-optimistic entrants. Thus, potentially at least,
it could better serve the purpose of establishing the
point that the results of standard competitive theory
are in some sense robust with respect to its
behavioural assumptions. Unfortunately, standard
theory offers no clue as to what this sense might be.
It is plain that the adjustment processes of the
system are centrally involved, and there is no
behaviourally plausible theory of adjustment that is
the dynamic counterpart in the disciplinary par-
adigm of static competitive equilibrium theory.
Within the limits defined by the requirement for
a strictly static competitive outcome, the most
plausible approach combines the idea of charac-
terizing the firms in the selection process by their
‘rules of behaviour’ - an idea advanced in a seminal
paper by Armen Alchian ( ) - with Her-
bert Simon’s idea of satisficing ( ). In the
simplest version, each firm simply adheres
unswervingly to its own deterministic behavioural
rule (or ‘routine’, in the language of Nelson and
Winter ). Such a rule subsumes or implies the
firm’s supply and demand functions, and given the
conditions set forth in (1) and (2) above, a constant
environment evokes a constant response. Satisficing
may be introduced as a complication of this picture
by an assumption that a firm that sustains losses
over a period of time will search for a better
behavioural rule; this adds behavioural plausibility
to the adjustment process but does not introduce the
possibility that random rule change might disrupt
an otherwise stationary competitive equilibrium
position.
(4) The final requirement can be succinctly but
inadequately stated as ‘some firms must actu-
ally be profit maximizers’. Although this for-
mulation does adequately cover some simple
cases, it does not suggest the depth and subtlety
of the issues involved.
Two points deserve particular emphasis here.
The first is the distinction between profit maxi-
mizing rules of behaviour (functions) and profit
maximizing actions. In general, a selection equi-
librium that mimics a particular competitive
Competition and Selection
equilibrium must clearly be one in which some
firms take actions that are profit maximizing in that
competitive equilibrium, and in this sense are profit
maximizers. But this observation does not imply
that the survivors in the selection equilibrium
possess maximizing rules, and in general it is not
necessary that survivors be maximizers in this
stronger sense. (Proof: Consider a competitive
equilibrium with constant returns to scale. Restrict
the firms’ supply and demand functions to be
constant up to a scale factor at the values taken in
the given equilibrium. Embed this static equilibrium
in a dynamic adjustment system in which firms’
scales of output respond to profitability in
accordance with assumption (2). Then the given
competitive equilibrium becomes a selection
equilibrium - since the only techniques in use make
zero profit - but the firms are not profit maximizers
in the stronger sense.)
The second point extends the first. The notion
of profit maximizing behavioural rules itself rests
on the conceptual foundation of a production set or
function that is regarded as a given. In evolutionary
theory, however, it is the rules themselves that are
regarded as data and as logically antecedent to the
values (actions) they yield in particular
environments. Thus, in this context, a problem
arises in interpreting the basic idea of a selection
equilibrium mimicking a standard competitive one:
there is no obvious set of ‘possibilities’ to which
one should have reference.
The most helpful approach here emphasizes
internal consistency. Assumptions about the struc-
ture of what is ‘possible’ can be invoked without
the additional assumption that there is a given set of
possibilities - for example, additivity and divis-
ibility may be assumed without implying that the
set of techniques to which these axioms apply is a
given datum of the system. Such an approach
provides a basis for discussing whether a particular
selection equilibrium is legitimately interpretable
as a competitive equilibrium given the other
assumptions in force. Along this path one can
explore a rich variety of selection equilibrium
situations that may be thought of as competitive
equilibria. Precisely because the variety is so rich,
to know only that an outcome is interpretable in this
fashion is to know very little about it.
Competition and Selection
In the light of formal analysis of selection
models of the sort described above, how strong is
the crutch that selection provides to standard
theory? For many analytical purposes, it is a crucial
weakness that the crutch relates only to equilibrium
actions and not to behavioural rules; it is from the
knowledge that the rules are maximizing that the
results of comparative statics derive. A selection
system disturbed by a parameter change from a
‘mimicking’ equilibrium does not necessarily go to
a new ‘mimicking’ equilibrium, let alone to one that
is consistent, in standard theoretical terms, with the
information revealed in the original equilibrium.
More fundamentally, selection considerations
cannot compensate for the inadequacies of standard
theory that arise from the basic assumption that
production possibilities are given data of the
system.
Schumpeterian Competition
In two great works and in many other writings,
Joseph Schumpeter proclaimed the central impor-
tance of innovative activity in the development of
capitalism. His early book, The Theory of Economic
Development, focused on the role and contribution
of the individual entrepreneur. From today’s
perspective the work remains enormously insightful
and provocative but may seem dated; the image of
the late 19th-century captains of industry lurks
implicitly in the abstract account of the
entrepreneur. The late work, Capitalism, Socialism
and Democracy, is likewise insightful, provocative
and a bit anachronistic. In this case, the
anachronism derives from the predictions of a
future in which the innovative process is bureau-
cratized, the role of the individual entrepreneur is
frilly usurped by large organizations, and the
sociopolitical foundations of capitalism are thereby
undercut. Present reality does not correspond
closely to Schumpeter’s predictions, and it seems
increasingly clear that he greatly underestimated the
seriousness of the incentive problems that arise
within large organizations, whether capitalist
corporations or socialist states.
Substantial literatures have accumulated around
a number of specific issues, hypotheses
1945
and predictions put forward in Schumpeter’s var-
ious writings. Regardless of the verdicts ultimately
rendered on particular points, everyday observation
repeatedly confirms the appropriateness of his
emphasis on the centrality of innovation in
contemporary capitalism. It confirms, likewise, the
inappropriateness of the continuing tendency of the
economics discipline to sequester topics related to
technological change in subsectors of various
specialized fields, remote from the theoretical core.
The purpose of the present discussion is to
assess the relationships of selection and competition
from a Schumpeterian viewpoint, that is, to extend
the discussion above by considering what difference
it makes if firms are engaged in inventing,
discovering and exploring new ways of doing
things. Plainly, one difference it makes is that
‘competition’ must now be understood in the broad
sense that admits a number of additional
dimensions to the competitive process, along with
price-guided output determination. In particular,
costly efforts to innovate, to imitate the innovations
of others, and to appropriate the gains from
innovation are added to the firm’s competitive
repertoire.
Selection now operates at two related levels.
The organizational routines governing the use made
of existing products and processes in every firm
interact through the market place, and the market
distributes rewards and punishments to the
contenders. These same rewards and punishments
are also entries on the market’s scorecard for the
higher level routines from which new products and
processes derive - routines involving, for example,
expenditure levels on innovative and imitative
R&D efforts. Over the longer term, selection forces
favour the firms that achieve a favourable balance
between the rents captured from successive rounds
of innovation and the costs of the R&D efforts that
yield these innovations.
In formal models constructed along these lines,
it is easy to see how various extreme cases turn out.
One class of cases formalizes the cautionary tale
told by Schumpeter ( , p. 105), in which
competition that is ‘perfect - and perfectly prompt’
makes the innovative role non-viable.
1946
Sufficiently high costs of innovation and low costs
of imitation (including costs of surmounting any
institutional barriers such as patents) will lead to the
eventual suppression of all firms that continue to
attempt innovation, and the system will settle into a
static equilibrium. (The character of this
equilibrium may, however, depend on initial
conditions and on random events along the evo-
lutionary path; the production set ultimately arrived
at is an endogenous feature of the process.) One can
also construct model examples to illustrate the
cautionary message ‘innovate or die’, the principal
requirement being simply a reversal of the cost
conditions stated above.
With the exception of some extreme or highly
simplified cases, models of Schumpeterian com-
petition describe complex stochastic processes that
are not easily explored with analytical methods. Of
course, the activity of writing down a specific
formal model is often informative by itself in the
sense that it illuminates basic conceptual issues and
poses key questions about how complex features of
economic reality can usefully be approximated by a
model. Some additional insight can then be obtained
using simulation methods to explore specific cases
(Nelson and Winter , Part V; Winter ). One of the
most significant benefits from simulation is the
occasional discovery of mechanisms at work that
are retrospectively ‘obvious’ and general features of
the model.
The discussion that follows pulls together a
number of these different sorts of insights, empha-
sizing in particular some issues that do not arise in
the related theoretical literature that explores var-
ious Schumpeterian themes using neoclassical
techniques (For the most part these neoclassical
studies explore stylized situations involving a single
possible innovation, and thus do not address issues
relating to the cumulative consequences of dynamic
Schumpeterian competition. See Kamien and
Schwartz ( ) and Dasgupta
( ) for references and perspectives on this
literature.)
A fundamental constituent of any dynamic
model of Schumpeterian competition is a model of
technological opportunity. Such a model establishes
the linkage between the resources that
Competition and Selection
model firms apply to innovative effort and their
innovative achievements. The long mn behaviour of
the model as a whole depends critically on the
answers provided for a set of key questions relating
to technological opportunity. Does the individual
firm face diminishing returns in innovative
achievement as it applies additional resources over
a short period of time? If so, from what ‘ fixed
factors’ does the diminishing returns effect arise,
and to what extent are these factors subject to
change over time either by the firm’s own efforts or
by other mechanisms? Are selection forces to be
studied in a context in which technological
opportunity presents more or less the ‘same
problem’ for R&D policy over an extended period,
or is the evolutionary sorting out of different
policies for the firm a process that proceeds
concurrently with historical change in the criteria
that govern the sorting?
Technological opportunity is said to be constant
if R&D activity amounts to a search of an
unchanging set of possibilities - in effect, there is a
meta-production set or meta-production function
that describes what is ultimately possible.
Increasing technological opportunity means that
possibilities are being expanded over time by causal
factors exogenous to the R&D efforts in question -
implying that, given a level of technological
achievement and a level of R&D effort, the effort
will be more productive of innovative results if
applied later. With constant technological
opportunity, returns to R&D effort must eventually
be decreasing, approaching zero near the boundary
of the fixed set of possibilities.
It is all too obvious that it may be very difficult
to develop an empirical basis for modelling tech-
nological opportunity in an applied analysis of a
particular firm, industry or national economy. There
is no easy escape from the conundrum that observed
innovative performance reflects both opportunity
and endogenously determined effort, not to mention
the fact that neither performance nor effort is itself
easily measured or the even more basic question of
whether analysis of the past can illuminate the
future. These difficulties in operationalizing the
concept of technological opportunity do not,
unfortunately, in any way diminish its critical role
in Schumpeterian competition.
Competition and Selection
The evolutionary analysis of Schumpeterian
competition has not, thus far, produced any coun-
terpart for the sorts of mimicry theorems that can be
proved for static equilibria. That is, there is no
model in which it can be shown that selection
forces, alone or in conjunction with adaptive
behavioural rules, drive the system asymptotically
to a path on which surviving firms might be said to
have solved the remaining portion of the dynamic
optimization problem with which the model situ-
ation confronts them - except in the cases where the
asymptotic situation is a static equilibrium with zero
R&D. The list of identified obstacles to a non-trivial
positive result is sufficiently long, and the obstacles
are sufficiently formidable, so as to constitute
something akin to an impossibility theorem. It
seems extremely unlikely that a positive result can
be established within the confines of an
evolutionary approach - that is, without endowing
the model firms with a great deal of correct
information about the structure of the total system
in which they are embedded.
The most formidable obstacle of all derives from
the direct clash between the future-oriented
character of a dynamic optimization and the fact
that selection and adaptation processes reflect the
experience of the past. If firms cannot ‘see’ the path
that technological opportunity will follow in the
future, if their decisions can only reflect past
experience and inferences drawn therefrom, then in
general they cannot position themselves optimally
for the future. They might conceivably do so if the
development of technological opportunity were
simple enough to validate simple inference
schemes. Such simplicity does not seem
descriptively plausible; who is to say that it is
implausible that in a particular case technological
opportunity might be constant, or exponentially
increasing, or following a logistic, or some sto-
chastic variant of any of these? And without some
restriction on the structural possibilities, how are
model firms to make inferences to guide their R&D
policies?
This obstacle is not featured prominently in the
simulations reported by Nelson and Winter, which
are largely confined to very tame and stylized
technological regimes in which opportunity is
summarized by a single exponentially increasing
1947
variable, called ‘latent productivity’. Such an
environment, reminiscent in some ways of neo-
classical growth theory, seems at first glance to be a
promising one for the derivation of a balanced
growth outcome in which actual and latent pro-
ductivity are rising at the same rate, the problem
facing the firms is in a sense constant, and selection
and adaptation might bring surviving firm R&D
policies to optimal values.
In fact, such a result remains remote even under
the very strong assumption just described. Demand
conditions for the product of the industry (or the
economy) affect the long ran dynamics, and in this
area also assumptions must be delicately contrived
to avoid excluding a balanced growth outcome. For
example, consider an industry model with constant
demand in which demand is (plausibly) less than
unit elastic at low prices. Then, cost reduction
continued indefinitely would drive sales revenue to
zero. Zero sales revenue will not cover the cost of
continuing advance. What is involved here is a
reflection of the basic economics of information;
costs of discovery are independent of the size of the
realm application, and on the assumption stated the
economic significance of that realm is dwindling to
nothing. The implication is that demand conditions
may check progress even if technological
opportunity is continually expanding. Indeed, this
may well be the pattern that is typically realistic for
any narrowly defined sector.
This difficulty too can be dispatched by an
appropriately chosen assumption. Beyond it he
some further problems. A model that acknowledges
the partially stochastic nature of innovative success
will display gradually increasing concentration
(Phillips ), unless some opposing tendency is
present. A good candidate for an opposing tendency
is the actual exercise of market power that has been
acquired by chance (Nelson and Winter , eh. 13).
But this market power can, presumably, also shelter
various departures from present value
maximization, including departures from
dynamically optimal R&D policy.
To reiterate, the quest for mimicry theorems in
the context of Schumpeterian competition seems
foredoomed to failure. Since models of
1948
Schumpeterian competition plainly provide a much
better description of the world we live in than do
models of static equilibrium, the overall conclusion
with regard to the strength of the selection crutch is
distinctly more negative than the conclusion for
static models alone. Assumptions that firms
maximize profit or present value will have to stand
on their own, at least until somebody invents a
better crutch for them. In the meantime, it will
continue to be the case that predictions based on
these assumptions are sometimes sound and
sometimes silly, and standard theory does not offer
a means of discriminating between the cases. More
direct attention should be paid to the mechanisms of
selection, adaptation and learning, which among
them probably account for as much sense as
economists have actually observed in economic
reality, and also leave room for a lot of readily
observable nonsense.
Bibliography
Alchian, A.A. 1950. Uncertainty, evolution and economic
theory. Journal of Political Economy 58: 211-221.
Dasgupta, P. 1985. The theory of technological competition. In
New developments in the analysis of
m a r k e t s t r u c t u r e , ed. J. Stiglitz and G.F.
Mathewson. Cambridge. ALA.: MIT Press.
Friedman. M. 1953. E s s a y s i n p o s i t i v e e c o n o m i c s .
Chicago: University of Chicago Press.
Kamien. M.. andN. Schwartz. 1981. M a r k e t s t r u c t u r e
a n d i n n o v a t i o n . Cambridge: Cambridge University
Press.
Nelson. R.. and S. Winter. 1982. A n e v o l u t i o n a r y
t h e o r y o f e c o n o m i c c h a n g e . Cambridge. ALA:
The Belknap Press of Harvard University Press.
Phillips. A. 1971. Technology and market structure: A Study of
the Aircraft Industry. Lexington: D.C. Heath.
Schumpeter. J.A. 1912. T h e t h e o r y o f e c o n o m i c
d e v e l o p m e n t . Trans. Redvers Opie. Cambridge.
AIA: Harvard University Press, 1934.
Schumpeter, J.A. 1950. C a p i t a l i s m , s o c i a l i s m a n d
d e m o c r a c y , , 3rd ed. New York: Harper.
Simon, H. 1955. A behavioral model of rational choice.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 69: 99-118.
Winter, S.G. 1964. Economic ‘natural selection’ and the
theory of the firm. Y a l e E c o n o m i c E s s a y s 4(1):
225 272.
Winter, S.G. 1971. Satisficing, selection and the innovating
remnant. Q u a r t e r l y J o u r n a l o f E c o n o m i c s
85: 237 261.
Winter, S.G. 1984. Schumpeterian competition in alternative
technological regimes. J o u r n a l o f E c o n o m i c
B e h a v i o u r a n d O r g a n i z a t i o n 5(3-4): 287-320.
Competition in International Trade
Competition in International Trade
D. K. Stout
In international markets the equivalent of the owned
assets and the skills of individuals and firms in
internal markets is the difference in relative national
resource endowments. In elementary theory, the
effects of international competition are defined by
comparison with an initial state of no trade. The
typical question addressed is ‘What will happen if
trade is opened up?’ On this view, international
competition ‘causes’ trade because of the
differences there would be, in the absence of trade,
in the relative costs ofproduction of pairs of goods
in two countries. And the effect of international
competition, and the ensuing trade, is to have ironed
out some of these differences and to have increased
in the process the aggregate equilibrium output of
each good.
In each country, initially, the cheaper good is
the one whose production technology calls for a lot
of the input that is relatively plentiful there. When
trade is opened up, the producers of the dearer
commodity in each country face competition from
the country better suited to produce it. Through
international competition, differences in the relative
scarcity (and cost) of labour, land and capital and
consequent differences in the ratio of the price of,
say, food to the price of, say, machinery are
lessened. With the growth of multinational
companies (MNCs), choices are made between
competing by offering products (i.e. by exporting)
and by licensing domestic producers overseas or
setting up subsidiary companies to produce and
market on the spot. These choices depend upon the
importance of economies of scale in production;
transport costs; the costs of transferring capital,
technology and management; and upon the
restrictions imposed by governments upon the free
movement of goods: that is, tariffs, import quotas
and other barriers.
If all industries were subject to diminishing
returns and perfect competition ruled, with homo-
geneous products, differences in the factor
Competition in International Trade
endowments of two countries starting to compete
would lead to specialization up to the point where,
at the margin, the relative costs of each pair of
goods were the same in each country. In fact, with
increasing returns to scale, equal relative marginal
costs are no guarantee of equilibrium. Competitive
advantage may accrue to an industry in the country
where home demand expands and is protected for a
time, so that relative costs fall as output increases.
This so-called ‘infant industry’ case is still the most
striking illustration of the inadequacy of simple
static theorems about the effects of international
competition upon trade flows.
The account of trade flows due to Heckscher and
Ohlin (namely that it depends on differing input
requirements for different goods and differing input
endowments in different countries) remains the
most important observation about international
competitive advantage.
This model of specialization through trade can
be modified to cope quite well with such market
imperfections as transport and other transaction
costs, limited information, a degree of imperfect
competition in factor markets and limited trade
barriers. It becomes seriously strained however
when one tries to use it to explain the results of
international competition in the world of many
traders and many commodities. The number of
separate ‘factors of production’ has to be increased
to equal the number of traded commodities and
defined to suit the observed trade flows.
Competitive advantage became something
‘revealed’, rather than predictable by taking a
census of resources. Cheap US skilled labour might
be redefined as capital; but since physical capital
was also cheap in the US the paradox which
Leontief observed - that the US tended to import
capital-intensive goods and export labour- intensive
- could not be resolved this way.
On another view, out of which grew an alter-
native and more robust model of international
competition, high technology was regarded as
complementary to (skilled) labour explaining the
bias of US trade. But a theory which relied first on
separating factors of production but then required
them to be lumped together is not a good theory.
Furthermore, competitive advantages were found
1949
to be continuously changing in ways that could not
be predicted by counting heads or hectares.
In the 1950s and 1960s, when industrial inter-
national competition became much more open and
many countries were newly industrializing, trade
both ways in different variants of the same products
grew prodigiously. It has become clear that
international competition drives intraindustry trade
in imperfectly competitive differentiated markets;
that in very many markets today’s net exporter is
tomorrow’s importer; and that international
competition involves the international transfer of
capital and knowhow as well as of goods.
Something is needed then to supplement the
explanation of broad directions of trade advantage
in terms of fixed differences in factor endowments
and given technology. There has arisen a less
determinate, more detailed account of disequilib-
rium change in trade flows under conditions of
imperfect competition. This account makes it clear
how international competition leads to two-way
flows of trade within the same industry. Where
there are economies of scale in each of eight plants,
four in each country, producing four variants of one
product, a likely outcome is that four of the eight
will survive, perhaps two in each country, and in
each country the loyal customers of each type will
import or buy at home accordingly. Net trade flows
might be zero but gross trade might be half of total
consumption. Or two countries, one rich and one
poor, might have developed at home (through
domestic competition) lower relative costs in the
‘superior’ variant in the richer country and in the
‘down-market’ variant in the poorer country. As
trade becomes easier, with cheaper transport and
communications or lower tariff barriers, the poor
country exports its variant to the poorer consumers
in the richer country and vice versa. As incomes
grow in both countries, demand in both will shift
more and more towards the rich country’s variant,
and the poorer country loses overall market share.
Where the pressures of competition are leading
to successive shifts in state-of-the-art technology in
a discontinuous way, competitive advantage may
move from the technological leader nation to the
fast followers as they acquire this
1950
technology; and back to the leader again with the
next major breakthrough, as has been observed in
industries as different as radio receivers and
textiles.
These models of changing comparative advan-
tage owe much to Joseph Schumpeter. They focus
upon the temporary monopoly advantages that
accrue to the innovator, extended sometimes by
patent protection and economies of scale.
As the technology gap is closed by imitators
competing away the monopoly profits, the tech-
nological leader nation, like Schumpeter’s indi-
vidual innovator, jumps to new dry steppingstones
in today’s ‘sunrise’ industries. The technology gap
is moved but kept open.
Whenever this evolution involves new products,
it becomes closely linked to a now well- established
view of the development of product markets:
namely, that products move through a life cycle of
youthful growth as comparative luxuries; into
explosive growth as they are mass-produced and
become available to poorer consumers; into
maturity, stagnation and sometimes death as they
are superseded by a more efficient or desirable way
of satisfying the same underlying needs. As
products move through their life cycles,
international competitive advantage will tend to
shift from developed technological leader
economies to newer or poorer industrializing
economies, particularly those with large domestic
markets allowing homegrown scale economies.
Technology gap and product life-cycle theories
do not explain why some economies lead and others
follow. The underlying point is that industrial
leadership creates a kind of human and physical
capital infrastructure, usually reflected in vigorous
R&D. This intellectual property is partly external to
the individual competitors in the leader economies
and acts as a catalyst for new processes and
products. There is a virtuous circle at work so long
as resources can move freely from yesterday’s
staples to tomorrow’s winners. When structural
adaptation is slow, international competition can
remove a historical leadership position from an
economy whose high relative wages then prevent it
from competing later in the cycle with newly-
industrializing fast-followers. The UK is
Competition in International Trade
an economy, open to fierce international competi-
tion, which has lost its place in this way and is
rapidly de-industrializing.
The typical market in international trade now-
adays is one in which there is a small number of
firms each with a strong home base but competing
in each other’s markets and in thud markets. Price
cuts are typically quickly matched, so oligopolists
compete (and try to secure their segment of the
market) by differentiating their products. Cars,
clothes and computers are therefore both imported
and exported as international competition drives
individual producers to specialize.
Much of the most recent work on international
competition examines how trade develops under
conditions of imperfect competition. Intraindustry
trade flows are affected by increasing returns to
scale in individual product lines, by product
differentiation, by competition in non- price terms
(design, reliability', durability, variety, packaging,
servicing, distribution, delivery speed, advertising)
and by the scope for vertically integrated
corporations to choose the separate locations of the
different stages of production of the final product.
With intra-industry trade specialization, the
threat of re-entry into each other’s specialisms, as
well as the possibilities of new entiy, limit
monopoly power. Other things being equal, entry is
easier where world demand is growing fast and
when the background technology is changing fast.
Trade under conditions of imperfect competition
tends to lead to increasing ‘narrow market’ spe-
cialization but decreasing ‘wide market’ monopoly
power, together with lower costs and equal or
greater breadth of consumer choice.
Two final international competition issues are
important: the occasionally perverse effects of
increased price competitiveness through currency
depreciation; and the effect upon trade flows of the
competitive options open to MNCs.
Suppose, through successive devaluations, one
country’s relative unit labour costs fall. (Relative
unit labour costs are a better measure of price
competitiveness than relative export prices because
the latter catch only actual transactions, not lost
orders.) The benefit to its trade in manufactures may
be short-lived. Orders wi 11 come in at
Competition Policy 1951

the most price-sensitive end of each of its industrial Caves, R.E., and R.W. Jones. 1981. W o r l d t r a d e a n d
markets. Since this is usually the less sophisticated p a y m e n t s : A n i n t r o d u c t i o n , 3rd ed. Boston:
Little Brown.
end, the nation’s industries tend to become less Cornwall, J. 1977. M o d e m c a p i t a l i s m , i t s g r o w t h
competitive in non-price terms: to be impelled to a n d t r a n s f o r m a t i o n . London: Martin Robertson,
specialize in those product versions whose demand ch. 10.
grows least fast as incomes grow. In the longer run, Gmbel, H., and P. Lloyd. 1975. Intra-industry trade: The theory
and measurement of international trade in differentiated
scale economies (and overall market share) may be products. New York: Wiley.
lost. Paradoxically, devaluation may have its best Houthakker, H.S., and S.P. Magee. 1969. Income and price
chance of increasing trade competitiveness when elasticities in world trade. T h e R e v i e w of
domestic competition is limited, so that home E c o n o m i c s a n d S t a t i s t i c s 51(2): 111-125.
Kanamori, H. 1968. Economic growth and exports. In
currency prices can be raised in price-inelastic Economic growth: The Japanese
markets, and the higher margins be used to e x p e r i e n c e s i n c e t h e M e i j i E r a , ed. L. Klein
overcome non-price disadvantages in world and K. Onkawa. Homewood: Richard Irwin.
markets. Leontief, W.W. 1934. Domestic production and foreign trade.
E c o n o m i c a I n t e m a z i o n a l e 7: 3-32. Reprinted in
Large firms can move technology, capital and American Economic Association, R e a d i n g s i n
management skills, as well as goods, and change i n t e r n a t i o n a l e c o n o m i c s , Homewood: Richard
domestic factor supplies by training. International Irwin, 1968.
competition drives firms to look for the lowest cost Linder, S. 1961. A n e s s a y o n t r a d e a n d
t r a n s f o r m a t i o n . Stockholm: Almqvist & Wiksell.
way of creating value-added. It is often cheaper for
Schumpeter, J.A. 1942. C a p i t a l i s m , s o c i a l i s m a n d
a MNC to transfer key inputs than to ship final d e m o c r a c y . New York: Harper.
products. Exporting, licensing, local assembly or Stout, D.K. 1977. International price competitiveness, non-price
packaging, joint ventures with local producers, and factors and export performance. London: National
Economic Development Office.
complete local production and marketing are
Thirlwall, A.P. 1980. Balance of payments theory and the UK
alternative modes of international competition. experience. London: Macmillan.
Each mode has private benefits and costs. The Venables, A.J. 1985. International trade, trade and industrial
choice between them is affected by the expected policy and imperfect competition: A survey. Centre for
Economic Policy Research Discussion Paper No. 74, Oct
reactions of competitors and by the policies of the
Vernon, R. (ed.) 1970. T h e t e c h n o l o g y f a c t o r i n
consumer governments. Protection, and its threat, international trade. National Bureau of
and dramatic improvements in communications in Economic Research Conference Series No. 22. New York.
recent years have led to a large increase in Wells, L. 1972. International trade: The product life cycle
approach. In T h e p r o d u c t l i f e c y c l e a n d
international competition between local subsidiaries
i n t e r n a t i o n a l t r a d e , ed. L. Wells. Cambridge,
of MNCs and an increase in the proportion of trade
in intermediate goods within vertically integrated
firms. The overall volume of trade is nowadays a
poor measure of the importance or strength of
Competition Policy
international competition.
Alan Hughes

See Also

► International Trade
The content and direction of competition policy is
inevitably conditioned by the domestic structure
References and international relations of the economy in which
Caves, R.E. 1971. International corporations: The industrial the policy is applied. Since my discussion, although
economics of foreign investment. E c o n o m i c a 38: theoretical, is essentially concrete, I will direct my
1-27. analysis to the case of competition policy in the
UK. However, the argument may readily be
generalized to other economies.
1952 Competition Policy

Economic Efficiency the demands for goods and services of consumers at

home, as well as sell enough of its products abroad
In the postwar period there have been considerableto pay for the nation’s import requirements subject
changes in the industrial structure and performance
to the constraint that this objective is fulfilled at
of the UK economy. In general, these changes have socially acceptable levels of output, employment
been associated with a decline in international and the exchange rate (Singh ).
competitiveness across broad areas ofIt will be assumed that the role of competition
manufacturing industry, and a more concentrated policy within this framework must be to regulate
structure of output in the hands of fewer, major, economic behaviour to assist in the achievement of
domestic producers in many manufacturing and the efficiency objective. In more operational terms
non-manufacturing sectors. The increased exposure we shall interpret this to mean that competitive
of the UK to international competition following behaviour should be so regulated as to ensure the
entry to the EEC, and the more general progressiveproduction, currently and in the future, of
liberalization of international trade in the worldinternationally tradeable output of a sufficiently
economy, have been associated, in this country, high quality and at a sufficiently low cost and price
with the development of substantial structural to compete effectively with international suppliers
unemployment, surplus capacity, and an inability to
at home and abroad, along with least- cost
match, by exports, the increased penetration of production of those non-tradeable elements of
domestic markets by foreign producers. domestic consumption. This means that the
The policy response to these changes has competitive process must so far as is possible
involved a constant interplay between macroeco- allocate sufficient inputs, including investment, to
nomic demand management on the one hand, and those uses necessary to achieve those objectives,
interventionist, cooperative and competitive supply
and ensure operation at lowest possible cost. To the
side strategies on the other. As one industrial policy
extent that certain forms of competitive behaviour
weapon, on the supply side, the regulation of are not compatible with the achievement of the
competition has evolved gradually in the postwar appropriate levels and allocations of investment,
years in response to the accumulation of evidence employment and output, and hinder the
about the effects of particular types of market achievement of the overall efficiency objective,
behaviour upon the amount, quality and price of then competition policy must be sufficiently flexible
output supplied, and upon the responsiveness of to accommodate non-competitive behaviour shown
supply to changing domestic and international to be necessary for this.
demand patterns. This evolution has occurred I am therefore abstracting throughout from
against a background of other supply side industrial
arguments in favour of a competition policy based
policies, in particular planning and cooperative on the notion of competitive behaviour as a ‘good
initiatives, and policies towards price and profitthing’ in itself, irrespective of its implications in
margin controls. Any discussion of the appropriateterms of an overall economic efficiency objective.
current, and future, stance of competition policy Such arguments may be considered by some to be
must therefore be carried out in the context of of overriding importance. It must, however, be
specific assumptions about the objectives and formrecognized that they are based on political,
of industrial policy; and about the form of overall
philosophical or moral grounds, and are not in
economic strategy, in particular about policies themselves susceptible to positive economic
towards the balance of payments and international analysis.
trade.
In an open economy we may define the role of
industrial policy to be the maintenance of efficient Competitive and Competition Policy
production of output. By efficiency in this sense we
mean that the economy must be able to meet In order to distinguish competitive and non-
competitive behaviour, and to consider the
Competition Policy
link between market behaviour and some notion of
economic efficiency we obviously need to define a
concept of competition. The best- developed notion
of competition in formal economic analysis is that
inherent in the state of affairs known as perfect
competition. It is easy to show that, under certain
very restrictive conditions, perfect competition will
lead to economic efficiency in the traditional Pareto
sense in which resources are so allocated that no
one can be made better off without someone else
being made worse off. Unfortunately, both the
notion of perfect competition and the related Pareto
efficiency criterion are too restrictive to serve as
useful foundations for the analysis of competition
policy. The state of affairs embodied in perfect
competition assumes away nearly all those aspects
of business behaviour with which competition
policy is concerned (such as rivalry in terms of
price setting, innovation in products and processes,
and advertising) and ignores the static welfare gains
to be had by improving the internal organization of
enterprises, in favour of an analysis concentrating
exclusively on the gains to be had by reallocating
resources between perfectly internally efficient
producers. Moreover, even within its own restricted
ambit, it yields few helpful decision rules for an
imperfect world to adopt since it is unfortunately
the case that if the conditions necessary for perfect
competition are unavoidably absent in any single
market then nothing can be said in general about the
correct behaviour to be followed in the rest of the
economy. In particular, it does not follow, for
example, that the perfectly competitive partial
equilibrium solution of setting prices equal to
marginal cost, so that normal profits are earned in
all sectors but the affected one, is the next best
solution to setting prices equal to marginal cost
everywhere. This ‘problem of the second best’
means that appropriate pricing rules can only be
derived after a piecemeal approach to individual
cases in which the input-output links with the
distorted sectors are examined in order to gauge in
which direction, and to what extent, price should
diverge from marginal cost. The Pareto efficiency
criterion is, of course, limited because of its
inability to deal with those economic changes which
involve alterations in the
1953
distribution of income. It is inevitable that the
achievement of the kind of efficiency we have
described above as the objective of UK economic
policy will involve changes in the distribution of
income. It not least between different industrial
nations, and between sectors of the domestic
economy. For all these reasons we have preferred to
adopt another approach to competition to use in
relation to our efficiency criterion set out earlier.
This approach analyses competition, not as a
state of affairs, but as a dynamic process linking
structural change with market behaviour. Compe-
tition is taken to mean that range of activities aimed
at meeting the objectives of one producer at the
expense of others, and is thus defined in the
business sense of the word, as a process involving
rivalry between producers. Competitive rivalry
takes both the form of contests within existing
markets, and the form of potential entry into new
areas when prospective returns appear relatively
attractive. It includes rivalry in terms of price, but
also in terms of altered or improved techniques of
production or products, and in terms of the provi-
sion of information to consumers about products.
All these forms of rivalry have consequences for the
level and rate of growth of the technical efficiency
of production and standards of consumption, for the
allocation of resources between industries, and for
the evolution of the structures of markets
themselves.
There are a number of theories from which we
may choose, which involve notions of competition
as a dynamic process. The best known are those of
Marx ( -94), Schumpeter ( ),
Downie ( ) and Clark ( ). This entry is
based on the characterization of the competitive
process by Downie. This is because he provides in a
relatively simple manner, an analysis which is both
more precise for our purposes than Clark, and
avoids the complications imposed by the richer,
more cosmic, implications of the Marxian and
Schumpeterian analyses of capitalism. Moreover,
the Downie model includes the main beneficial
effects usually claimed to follow from competitive
rivalry, in terms of cost efficiency, resource
allocation, and technical progress, whilst
emphasizing the possible structural changes flowing
from competition which may change its
1954
nature and effects. We can then, following Downie,
sketch an outline of a simple competitive process in
a given market and consider its efficiency
implications.
In any market we may expect that firms with the
lowest cost structure (inherited from the past), will
for any given market price have the highest profits
to finance expansion. In their own pricing policy,
firms are expected to set a price which promises to
attract customers and provide sufficient retained
profits to finance the capacity necessary to serve
them in a balanced manner. In this way, within
markets, relatively low costs will be associated with
relatively fast growth, and a competitive transfer of
market shares from the less to the more efficient
will occur. Moreover, this transfer mechanism may
have a feedback effect on efficiency in the sense
that the previously less efficient will be threatened
with an ever-diminishing share of the market unless
they can improve their cost position, for instance by
introducing more recent innovations in production
technique.
The transfer and innovation mechanisms within
markets thus have beneficial effects, both upon
allocative efficiency, by transferring output and
resources to the currently most efficient producers,
and upon technical progressiveness, by encouraging
the introduction of the least-cost techniques
available. The outcome of this competitive process,
however, in terms of the changing allocation of
output between different firms over time, and the
impact of this upon competitive behaviour itself,
will depend ultimately on the answer to a number of
empirical questions about the relationship between
past and future efficiency, and company size and
performance. If past success is repeated, and
resulting gains in relative size offer efficiency
advantages in themselves, through economies of
scale or enhanced innovative ability, then particular
markets may come to be dominated by ever fewer,
ever larger, films as a result of the transfer
mechanism, whilst the innovative mechanism,
acting as a spur for past losers to improve
performance, may not be powerful enough to offset
this tendency. Past failure may raise the desire, but
inhibit the ability', to recover. Innovation in process
or product may be expensive, and low profits, or
relatively small
Competition Policy
scale, may limit the power of losers to respond
actively enough to prevent the transfer mechanism
from leading to the concentration of ever more
output in fewer hands. Thus, in the absence of
revival the end product of the competitive process
may be domination of the market by the single
producer whose efficiency in the past has
outstripped all rivals.
However, before this stage is reached, the pri-
vate gains from collusion between remaining pro-
ducers may become apparent, since it can offer both
lower costs of competition and a higher market
price. The costs of adding to the stock of goodwill
accumulated by past advertising can be reduced,
and the necessity to indulge in competitive process
and product innovation is modified. Moreover,
collusion can offer the leading firms higher margins
by concerted pricing policies, at least in the short
run, and possibly in the long run too, if the scales of
production and advertising created by the
competitive process itself form effective barriers to
entry. Thus the competitive process seems to end
inevitably in dominant-firm monopolistic or
oligopolistic structures, and in the suppression of
the process itself. This reduces the pressure to
increase the level of efficiency through time and, in
the case of collusion, to reduce the dispersion of
efficiency. Market price is higher, and to some
extent output is lower than it might otherwise be.
High-cost plants and firms can remain in production
at the higher market price and the incentive to
improve products and production methods falls.
Thus, both levels and rates of growth of efficiency
in any market may be worsened by the cessation of
the competitive process in structural conditions
which may themselves limit the chances of
substantial new entry.
In terns of this analysis the objective of com-
petition policy would appear to be the identification
and regulation of those structures, restrictions on
entry and kinds of behaviour which arrest the
competitive process, and produce the efficiency
losses described above. This clearly poses problems
of establishing links between structure and
behaviour, of specifying behaviour inimical to the
competitive process and of establishing the exis-
tence of markets where the competitive process has
atrophied, and where it appears that efficiency
Competition, Austrian
gains are possible. In the case of an open economy
such as the United Kingdom the relevant markets,
and standards of comparison for efficiency, are
international in character, and we must recognize
that in many industries the degree of concentration
of purely domestic producers will be of little use in
assessing the fierceness of the transfer mechanism,
because of the importance of foreign competition.
However, more fundamental problems have also
to be recognized. It may be that a competitive
process of the kind outlined above may not be the
only, or the most efficient, means of achieving the
same ends, either before or after the stage is reached
at which the possibilities of market dominance or
collusion arise. Thus, we must consider planned or
cooperative solutions to the problem of raising
efficiency. Moreover, there is no guarantee that the
end result of the competitive process itself, in terms
of the levels and rates of change of operating
efficiency of producers, and their distribution
between activities, will, for any individual country,
necessarily meet the overall efficiency objective
outlined earlier, in terms of output, inflation,
employment and international trade performance
targets. There is then, for both the above reasons,
the possibility of conflict, in industrial policy
methods, between the promotion of arm’s-length
competitive behaviour through competition policy,
and the possible reduction in such behaviour
encouraged by other aspects of government policy
in pursuit, for instance, of industrial strategy. The
possibility of conflict is most obvious in relation to
restrictive trade practices, but is also present in the
creation or encouragement of domestic dominant-
firm positions through government-sponsored
mergers or rationalization schemes (NEDO ).
These conclusions run in some cases contrary to
the majority of recent academic argument in this
area, where increasing emphasis has been paid to
the claimed welfare losses due to merger, increased
monopoly power, and restrictive trade practices.
However, in the context of current UK industrial
performance and within the current overall
economic policy framework a detailed scrutiny of
the current evidence for the UK yields no
foundation for a generally more aggressive
1955
approach towards large-firm dominance than the
current legislation already adopts. This is not to
deny that competition policy is necessary, but to
assert that an appropriate competition policy must
be designed in the light of existing economic
conditions; of the evidence of the effects of various
forms of market structure and behaviour; and of the
overall objectives of economic policy.
See Also
► mti-ti ust Policy
► Mark Structure
Bibliography
Clark, J.M. 1961. Competition as a dynamic process.
Washington, DC: Brookings Institution.
Downie, J. 1958. T h e c o m p e t i t i v e p r o c e s s . London:
Duckworth.
Marx, K. 1867-94. C a p i t a l London: Lawrence & Wishart,
1970-72.
NEDO. 1978. C o m p e t i t i o n p o l i c y . London: HMSO/
National Economic Development Office.
Schumpeter, J. A. 1942. C a p i t a l i s m , s o c i a l i s m a n d
d e m o c r a c y . London: George Allen & Unwin, 1968.
Singh, A. 1977. The UK industry and the world economy: A
case of de-industrialisation? C a m b r i d g e J o u r n a l
o f E c o n o m i c s 1(2): 113-36.
Competition, Austrian
Paul J. McNulty
Keywords
Allocative efficiency; Austrian economics;
Competition, Austrian; Creative destruction;
Entrepreneurship; Imperfect competition;
Innovation; Market power; Mises, L.E. von;
Monopolistic competition; Monopoly; Perfect
competition; Schumpeter, J.A
JEL Classifications
BO
1956
The essence of Austrian economics is its emphasis
on the ongoing economic process as opposed to the
equilibrium analysis of neoclassical theory,
Austrian concepts of competition reflect this
emphasis. Indeed, one of the central challenges by
Austrians to the neoclassical model, and a common
denominator of virtually all Austrian economics, is
the rejection of the concept of perfect competition.
In this respect, a number of economists who cannot
be considered Austrian in all aspects of their work,
share, nonetheless, the Austrian emphasis on actual
market activities and processes - for example,
Joseph Schumpeter ( ), J.M. Clark ( ),
Fritz Machlup ( ) of the capitalist process.
and others.
When the concept of competition entered eco-
nomics at the hands of Adam Smith and his pre-
decessors, it was not clearly defined, but it generally
meant entry by firms into profitable industries (or
exit from unprofitable ones) and the raising or
lowering of price by existing firms according to
market conditions. There was little recognition, and
virtually no analysis, of entrepreneurship as it might
be reflected in these and other forms of competition,
but there was a recognition that business firms do in
most situations have some control over market
prices, with the degree of control varying inversely
with the number of firms in the industry. These
basic ideas, expanded and supplemented, are
generally compatible with most modem Austrian
analysis.
What is objectionable to Austrian economists is
the neoclassical concept of perfect competition,
developed during the 19th and early 20th centuries.
The development began with Cournot ( ),
whose concern it was to specify as rigor
ously as possible the effects of competition, after the
process of competition had reached its limits. His
conceptualization of this situation was a market
structure in which the output of any one firm could
be subtracted from total industry output with no
discernible effect on price. Later contributions by
Jevons, Edgeworth, J.B. Clark and Frank Knight led
to the model of perfect competition as we know it
today (Stigler ; McNulty )•
The trouble with the concept from the Austrian
point of view, as Hayek has emphasized, is that it
Competition, Austrian
describes an equilibrium situation but says nothing
about the competitive process which led to that
equilibrium. Indeed, it robs the firm of all business
activities which might reasonably be associated
with the verb ‘to compete’ (von Hayek ). Thus,
firms in the perfectly competitive model do not
raise or lower prices, differentiate their products,
advertise, try to change their cost structures relative
to their competitors, or do any of the other things
done by business firms in a dynamic economic
system. This was precisely the reason why
Schumpeter insisted on the irrelevance of the
concept of perfect competition to an understanding
For Schumpeter, any realistic analysis of com-
petition would require a shift in analytical focus
from the question of how the economy allocates
resources efficiently to that of how it creates and
destroys them. The entrepreneur, a neglected figure
in classical and neoclassical economics, is the
central figure in the Schumpeterian analytical
framework. The entrepreneur plays a
disequilibrating role in the market process by
interrupting the ‘circular flow’ of economic life,
that is, the ongoing production of existing goods
and services under existing technologies and
methods of production and organization. He does
this by innovating - that is, by introducing the new
product, the new market, the new technology, the
new source of raw materials and other factor inputs,
the new type of industrial organization, and so on.
The result is a concept of competition grounded in
cost and quality advantages which Schumpeter felt
is much more important than the price competition
of traditional theory and is the basis of the ‘creative
destruction’ of the capitalist economic process. It
produces an internal efficiency within the business
firm, the importance of which for economic welfare
is far greater, Schumpeter argued, than the
allocative efficiency of traditional economic theory
(Schumpeter ).
His emphasis on the advantages of the firm’s
internal efficiency led Schumpeter to a greater
tolerance for large-scale business organizations,
even for those enjoying some degree of monopoly
power, than was typical of many more traditional
theorists of his time. This is a not uncommon
Competition, Austrian
characteristic of Austrian economics. Hayek, for
example, makes the distinction between entrenched
monopoly, with its probable higher- than-necessary
costs, and a monopoly based on superior efficiency
which does relatively little harm since in all
probability it will disappear, or be forced to adjust
to market conditions, as soon as another firm
becomes more efficient in providing the same or a
similar good or service (von Hayek ). And that is
precisely Schumpeter’s point. The ground under
even large-scale enterprise is constantly shaking as
a result of the competitive threat from the new firm,
the new management, or the new idea.
Schumpeter’s competitive analysis was less a
defence of monopoly power than of certain business
activities which were judged to be monopolistic
only from the comparative standpoint of the model
of perfect competition. He insisted that the quality
of a firm’s entrepreneurship was of far greater
significance than its mere size.
The leading contemporary Austrian theorist of
competition is Israel Kirzner ( ). Kirzner’s
approach draws on the analysis of market processes
and the concept of ‘human action’ developed earlier
by Ludwig von Mises. For von Mises,
entrepreneurship is human action in the market
which successfully directs the flow of resources
toward the fulfillment of consumer wants (von
Mises ). Kirzner’s more fully developed theory of
competition is based on the idea that the means -
end nexus of economic life is not given but is itself
subject to creative human action. This creative role
Kirzner defines as entrepreneurship, and it is
essentially the ability to detect new but desired
human wants, as well as new resources, techniques,
or other ways through which to satisfy them.
Whether he discovers new wants or new means of
satisfying old ones, the Kirznerian entrepreneur is
the one who sees and exploits what others fail to
notice - the profit opportunities inherent in any
situation in which the prices of factor inputs fall
short of the price of the final product.
There is a difference between Kirzner’s theory
of entrepreneurship and that of Schumpeter.
Schumpeter’s entrepreneur is a disequilibrating
force in the economic system; he initiates
1957
economic change. Kirzner’s entrepreneur plays an
equilibrating role; the changes he brings about are
responses to the mistaken decisions and missed
opportunities he detects in the market. Unlike
Schumpeter’s entrepreneur, he is not so much the
creator of his own opportunities as a responder to
the hitherto unnoticed opportunities that already
exist in the market. Thus, in the competitive market
process, the Schumpeterian and Kirznerian
entrepreneurs may complement each other - the one
creating change, the other responding to it.
Austrian dissatisfaction with the perfectly com-
petitive model extends to the theories of imperfect
and monopolistic competition. Hayek’s and
Kirzner’s criticisms are the same as of perfect com-
petition, namely, that the analysis is limited to an
equilibrium situation in which the underlying data
are assumed to be adjusted to each other, whereas
the relevant problem is the process through which
adjustment occurs. Schumpeter criticized monop-
olistic competition for its continued acceptance of
an unvarying economic structure and forms of
industrial organization. Nonetheless, the incorpo-
ration into economic theory of quality competition
and sales efforts, complementing the traditional and
limited focus on price competition, as well as the
efforts on the part of some industrial organization
specialists and institutional economists to analyse
and explain actual market processes, are
developments that are generally within the Austrian
tradition.
See Also
► Austrian 5
► :o
► E cati \ c D.
Bibliography
C l a r k , J . M . 1 9 6 1 . Competition as a dynamic process.
Washington, DC: Brookings Institution.
C o u r n o t , A . A . 1 8 3 8 . Recherches sur les principes
mathematiques de la theorie des richesses. P a r i s :
Hachette.
Kirzner, I. 1973. Competition and entrepreneurship. Chicago:
University of Chicago Press.
1958 Competition, Classical

Machlup, F. 1942. Competition, pliopoly, and profit. Pts. competition (notably the conditions which ensure
I-II. E c o n o m i c a , N.S. 9, Pt. I, 1-23, Pt. II, 153-73. price-taking) are often read back, illegitimately, into
McNulty, P. 1967. A note on the history of perfect compe-
tition. J o u r n a l o f P o l i t i c a l E c o n o m y 75: 395-399.
classical discussions of competition.
Schumpeter, J. 1942. C a p i t a l i s m , s o c i a l i s m , a n d The mechanisms which determine the economic
d e m o c r a c y . New York: Harper & Row. 1962. behaviour of industrial capitalism are not self-
Stigler, G. 1957. Perfect competition, historically contem- evident. As a form of economy in which production
plated. J o u r n a l o f P o l i t i c a l E c o n o m y 65: 1-17.
von Hayek, F.A. 1948. The meaning of competition. In
and distribution proceed by means of a generalized
I n d i v i d u a l i s m a n d e c o n o m i c o r d e r , ed. F.A. process of exchange (in particular by the sale and
Hayek. London: Routledge. purchase of labour), it possesses no obvious direct
von Mises, L. 1949. H u m a n a c t i o n . New Haven: Yale mechanisms of economic and social coordination.
University Press.
Yet, in so far as these operations constitute a
system, they must be endowed with some degree of
regularity, the causal foundations of which may be
revealed by analysis. The first steps in economic
Competition, Classical investigation which accompanied the beginnings of
industrial capitalism consisted of a variety of
John Eatwell attempts to identify such regularities, often by
means of detailed description and enumeration, as
in the works of Sir William Petty, and hence to
establish the dominant causes underlying the
behaviour of markets. But what was required was
Keywords
not simply the description and classification which
Cantillon, R.; Capital accumulation; Classical
precedes analysis, but abstraction, the
theory of value and distribution; Competition,
transcendence of political arithmetic (Smith , p.
classical; Intrinsic value; Market price; Marx, K.
501).
H.; Natural price; Neoclassical economics;
The culmination of the search for a coherent
Perfect competition; Perfect liberty; Petty, W.;
abstract characterization of markets, and hence the
Quesnay, F.; Rate of profit; Ricardo, D.; Smith,
foundation of modem economic analysis, is to be
A.; Turgot, A. R. J.
found in Chapter 7 of Book I of Adam Smith’s
Wealth of Nations - ‘Of the Natural and Market
Price of Commodities’. In this chapter Smith pre-
JEL Classifications
BO sented the first satisfactory formulation of the
regularity inherent in price formation. The idea,
Only through the principle of competition has partially developed earlier by Cantillon, and by
political economy any pretension to the character of Turgot in his discussion of the circulation of money,
a science. So far as rents, profits, wages, prices, are was that
There is in every society ... an ordinary or average
determined by competition, laws may be assigned
rate of both wages and profits ... When the price of
for them. Assume competition to be their exclusive any commodity is neither more nor less than what is
regulator, principles of broad generality and sufficient to pay the rent of land, the wages of labour,
scientific precision may be laid down, according to and the profits of stock employed ... according to
their natural rates, the commodity is then sold for
which they will be regulated. (Mill , p. 242)
what may be called its natural price.
In all versions of economic theory ‘competi-
tion’, variously defined, is a central organizing and that
concept. Yet the relationship between different The natural price ... is, as it were, the central price,
to which the prices of all commodities are continually
definitions of competition and differences in the gravitating. Different accidents may sometimes keep
theory of value have not been fully appreciated. In them suspended a good deal above it, and sometimes
particular, the characteristics of ‘perfect’ force them down somewhat below
Competition, Classical 1959

it. But whatever may be the obstacles which hinder It is obvious that the annual products which can be
them from settling in this center of repose and derived from capitals invested in these different
continuance, they are continually tending towards it. employments are mutually limited by one another,
(Smith 177< , p. 65) and that all are relative to the existing rate of interest
on money. (Turgot , p. 70)
Thus the natural price encapsulates the persistent
element in economic behaviour. And that persis- However, Turgot neither related the determination
tence derives from the ubiquitous force of compe- of the rate of profit to production in general - he
tition: or, as Smith put it, the condition of ‘perfect accepted the Physiocratic idea that the incomes of
liberty’ in which ‘the whole of the advantages and the industrial and commercial classes were ‘paid’
disadvantages of the different employments of by agriculture - nor developed the conceptual
labour and stock must ... be either perfectly equal or framework which linked the formation of prices and
continually tending to equality’ (p. Ill), for the of the rate of profit to the overall organization of the
natural price is ‘the price of free competition’ (p. economy. These were to be Smith’s achievements:
68). If ... the quantity brought to market should at any
The relationship between competition and the time fall short of the effectual demand, some of the
establishment of what Petty called ‘intrinsic value’ component parts of its price must rise above their
natural rate. If it is rent, the interest of all other
had been discussed in the works of Petty,
landlords will naturally prompt them to prepare
Boisguillebert, Cantillon and Harris as the outcome more land for the raising of this commodity; if it is
of rival bargaining in price formation, competition wages or profit, the interest of all other labourers
being the greater when the number of bargainers and dealers will soon prompt them to employ more
labour and stock in preparing and bringing it to
was such that none has a direct influence on price.
market. The quantity brought thither will soon be
Quesnay expressed the formation of competitive sufficient to supply the effectual demand. All the
prices as being ‘independent of mens’ will... far different parts of its price will soon sink to their
from being an arbitrary value or a value which is natural rate, and the whole price to its natural price.
(Smith , p. 65)
established by agreement between the contracting
parties’ (inMeek ,p. 90), but he did not relate the So in a competitive market there will be a tendency
organization of production to the formation of for the actual prices (or ‘market prices’ as Smith
prices in competitive markets. Consideration of that called them) to be relatively high when the quantity
relationship required the development of a general brought to market is less than the effectual demand
conception of the role of capital, and with it the (the quantity that would be bought at the natural
notion of a general rate of profit formed by the price) and relatively low when the quantity brought
competitive disposition of capital between alterna- to market exceeds the effectual demand. This
tive investments (Vaggi ). working of competition was known as the ‘Law of
A significant step in this direction was made by Supply and Demand’. The working of competition
Turgot, who both conceived of the process of which constitutes the ‘Law’ do not identify the
production as part of the circulation of money: phenomena which determine natural prices. The
We see ... how the cultivation of land, manufactures ‘Law’ of supply and demand should not be
of all kinds, and all branches of commerce depend confused with supply and demand ‘theory’, that is,
upon a mass of capitals, or movable accumulated the neoclassical theory of price determination which
wealth, which, having been first advanced by the was to be developed one hundred years later. Nor
entrepreneurs in each of these different classes of
work, must return to them every year with a regular should Smith’s discussion of the tendencies of
profit ... It is this continual advance and return of concrete market prices be confused with supply and
capitals which constitutes w h a t o u g h t t o b e demand function, which are loci of equilibrium
called the circulation of money. prices.
(Turgot , p. 148)
Adam Smith’s conception of ‘perfect liberty’
and saw that the structure of investments would consists of the mobility of labour and stock between
tend to be that which yielded a uniform rate of different uses - the mobility that is
profit:
1960 Competition, Classical

necessary for the establishment of ‘an ordinary or monopolised, either by an individual, or by a

average rate both of wages and profits’ and hence company ... their price has no necessary connexion
for the gravitation of market prices toward natural with their natural value’ (p. 385). His analysis of
prices. Smith identifies four reasons why market value and distribution is accordingly confined to
prices may deviate ‘for a long time together’ above ‘By far the greatest part of those goods which are
natural price, creating differentials in the rate of the object of desire ... such commodities only as can
profit, all of which involve restriction of mobility: be increased in quantity by the exertion of human
labour, and on the production of which competition
(a) Extra demand can be ‘concealed’, though operates without restraint’ (p. 12).
‘secrets of this kind ... can seldom be long For Marx competition is synonymous with the
kept’; generalization of capitalist relations of production.
(b) Secret technical advantages; Competition is thus related to the rise to dominance
(c) ‘A monopoly granted either to an individual or of the capitalist mode of production.
a trading company’; While free competition has dissolved the barriers of
(d) ‘Exclusive privileges of corporation, statutes of earlier relations and modes of production, it is nec-
apprenticeship, and all those laws which essary to observe first of all that the things which
were a barrier to it were the inherent limits of earlier
restrain, in particular employments, the com- modes of production, within which they spontane-
petition to a smaller number than might oth- ously developed and moved. These limits became
erwise go into them’. barriers only after the forces of production and the
relations of intercourse had developed sufficiently to
For Smith there is some similarity in the forces enable capital as such to emerge as the dominant
principle of production. The limits which it tore
acting on wages and profits which derives from his
down were barriers to its motion, its development
conceiving of the capitalist as personally involved and realization. It is by no means the case that it
in the prosecution of a particular trade or business. thereby suspended all limits, nor all barriers, but
So the rate of profit, like the rate of wages, may be rather only the limits not corresponding to it ... Free
competition is the real development of capital. (Marx
differentiated between sectors by ‘the agreeableness 73, pp. 649-50)
of disagreeableness of the business’, even though
‘the average and ordinary rates of profit in the And as capitalism itself develops so does
competition:
different employments of stock should be more
On the one hand ... [capital] creates means by which
nearly upon a level than the pecuniary wages ofthe to overcome obstacles that spring from the nature of
different sorts of labour’ ( p. 124). Landlords, production itself, and on the other hand, with the
capitalists and workers are all active agents of development of the mode of production peculiar to
itself, it eliminates all the legal and extra-economic
mobility. In Ricardo’s discussion the emphasis impediments to its freedom of movement in the
shifted towards the distinctive role of capital: different spheres of production. Above all it
It is, then, the desire, which every capitalist has, of overturns all the legal or traditional barriers that
diverting his funds from a less to a more profitable would prevent it from buying this or that kind of
employment, that prevents the market price of com- labour-power as it sees fit, or from appropriating
modities from continuing for any length of time this or that kind of labour. (Marx 167, p. 1013)
either much above, or much below their naturalThe concentration of capital (increasing unit size of
price. (Ricardo , p. 91)
firms) and, in particular, the centralization of capital
Ricardo used the term ‘monopoly price’ to refer (cohesion of existing capitals) destroys and
to commodities ‘the value of which is determined recreates competition. Competition is one of the
by their scarcity alone’, such as paintings, rare most powerful ‘levers of centralization’, and
books and rare wines ( pp. 249-51) which have The centralization of capitals, or the process of their
attraction, becomes more intense in proportion as
‘acquired a fanciful value’, and he argued that for the specifically capitalist mode of production
‘Commodities which are develops along with accumulation. In its turn
Competition, Classical 1961

centralization becomes one of the greatest levers of Theory proceeds by the extraction from reality
its development. (Marx , p. 778n) of those forces which are believed to be dominant
Like Smith and Ricardo, Marx, relates the and persistent, and the formation of those elements
development of competition to the establishment of into a formal system, the solution of which is to
the general rate of profit: determine the magnitude or state of the variables
under consideration. It is obvious that the solution
What competition, first in a single sphere, achieves is
a single market value and market price derived from
will not, except by a fluke, correspond to the actual
the individual values of commodities. And it is magnitudes of the variables ruling at any one time,
competition of capitals in various spheres, which first for these will be the outcome not solely of the
brings out the price of production equalising the elements grouped under the heading ‘dominant and
rates of profit in the different spheres. The latter
persistent’, but also of the myriad of other forces
process requires a higher stage of capitalist produc-
tion than the previous one. (Marx , p. 180) excluded from the analysis as transitory, peculiar or
specific (lacking general significance) which may,
It is in his conception of the circuit of capital
at any moment, exert a more or less powerful effect.
that Marx best portrays capitalist competition. The
Nonetheless, the practice of analysis embodies the
image is one of capital as a homogeneous mass of
assumption that the forces comprising the theory
value (money) seeking its maximum return. Profits
are dominant, and that the determined magnitudes
are created by embodying capital in the process of
will, on average, tend to be established. In any
production, the commodity outputs of which must
satisfactory analytical scheme these magnitudes
be realized, that is, returned to the homogeneous
must be centres of gravitation, capturing the
money form to be reinvested. Competition is thus
essential character of the phenomena under
characteristic of the capitalist mode of
consideration.
accumulation; mobility and restructuring are two
The importance of Smith’s use of competition is
aspects of the same phenomenon.
now apparent. Theory cannot exist in a vacuum.
Marx’s general conception of capital as a system
Simply labelling forces dominant is not enough.
corroborates Quesnay’s notion of an economy oper-
These forces must operate through a process which
ating ‘independent of men’s will’. This does not
establishes their dominance and through which the
mean that there may not be circumstances in which
‘law-governed’ nature of the system is manifest.
individual capitals exercise some control in
That process is competition, which both enforces
particular markets - indeed, such limitations may be
and expresses the attempt of individual capitals to
necessary for the accumulation process to proceed
maximize profits. Thus important aspects of the
in certain lines. Capital removes only those barriers
behaviour of a capitalist market economy may be
which limit its accumulation. The market control
captured at a sufficient level of generality to permit
exercised in some lines of modem industry is not
the formulation of general causal statements, that is,
necessarily a limitation but may be a prerequisite of
to permit analysis. Without this step, which
production on an extended scale. Aggregate capital
constitutes the establishment of what was called
flows discipline the actions of individual capitals,
above the method of analysis, it would have been
and hence endow the system with the regularity
impossible to develop any general form of
manifest in the perpetual tendency, successfully
economic theory.
contradicted and recreated, towards a general rate of
The classical theory of value and distribution
profit and associated prices.
may be shown to provide a logically coherent
Competition not only establishes the object of
explanation of the determination of the general rate
analysis, natural prices and the general rate of
of profit and hence of natural prices (prices of
profit, but makes meaningful analysis possible,
production) taking as data (see Sraffa ):
since it allows the operations of the capitalist
economy to be characterized in a manner which (a) The size and composition of social output;
permits theoretical statements of general validity to (b) The technique in use; and
be made about them. (c) The real wage.
1962
The classical achievement is thus composed of
two independent elements: (a) the characterization
ofthe object ofthe theory of value; and (b) the
provision of a theory for the determination of that
object. Underlying the former is the concept of
gravitation imposed by competition, and underlying
the latter the concept of gravitation inherent in
theoretical abstraction. Any alternative system must
not simply provide a different theory but also
achieve a similar congruence with the traditional
method.
The development in the final quarter of the 19th
century of what was to become known as the neo-
classical theory of value and distribution was an
attempt to provide an alternative to a classical the-
ory embroiled in the logical difficulties inherent in
the labour theory of value and sullied by unsavoury
associations with radicalism and Marxism. But
despite the dramatic change in theory that was to be
heralded by the works of Jevons, Menger and
Walras, the method of analysis which characterized
the object the theory was to explain stayed funda-
mentally the same; the new theory was an alterna-
tive explanation of the same phenomena. Marshall
labelled natural prices ‘long-run normal prices’, and
declared that, as far as his discussion of value was
concerned ‘the present volume is chiefly
concerned ... with the normal relations of wages,
profits, prices etc., for rather long periods’ ( , p.
315).
The same continuity of method may be found in the
work of Walras ( -7, pp. 224, 380), Jevons
( , pp. 86, 135-6), Bohm-Bawerk ( , Macmillan.
p. 380) and Wicksell ( , p. 97).
Nonetheless, the structure of neoclassical theory
is such that a different notion of competition is
required. The classical emphasis on mobility must
be supplemented by a precise definition of the
relationships presumed to exist between individual
agents. The fundamental concept of ‘perfect’
competition, for example, encompasses the idea that
the influence of each individual participant in the
economy is ‘negligible’, which in turn leads to the
idea of an economy with infinitely many par-
ticipants (Aumann ). Such formulations are entirely
absent from the classical conception of competition,
since the classical theory is not constructed around
individual constrained utility maximization.
Competition: Marxian Conceptions
See Also
Bibliography
Aumann, R.J. 1964. Markets with a continuum of traders.
E c o n o m e t r i c a 32: 39-50.
von Bohm-Bawerk, E. 1899. C a p i t a l a n d i n t e r e s t .
Vol. 2.
South Holland: Libertarian Press. 195l>.
Jevons, W.S. 1871. Theory of political economy.
Harmondsworth: Penguin. 1970.
Marshall, A. 1920. P r i n c i p l e s o f e c o n o m i c s . 8th ed.
London: Macmillan.
Marx,K. 1867. C a p i t a l . Vol. 1. Harmondsworth: Penguin.
1976. Marx, K. 1894. C a p i t a l . Vol. 3. New York:
International Publishers. 1967.
Marx, K. 1973. G r u n d r i s s e . Harmondsworth: Penguin.
Meek, R.L. 1956. S t u d i e s i n t h e l a b o u r t h e o r y o f
value.
London: Lawrence & Wishart.
Meek, R.L. 1962. T h e e c o n o m i c s o f p h y s i o c r a c y .
London: Allen & Unwin.
Meek, R.L. 1973. I n t r o d u c t i o n t o t u r g o t (1973).
Mill, J.S. 1848. P r i n c i p l e s o f p o l i t i c a l e c o n o m y .
London: Parker.
Ricardo, D. 1817. In P r i n c i p l e s o f p o l i t i c a l
e c o n o m y a n d t a x a t i o n , ed. P. Sraffa. Cambridge:
Cambridge University Press. 1951.
Smith, A. 1776. An inquiry into the nature and causes of the
wealth of nations. London'. Methuen. 1961.
Sraffa, P. 1960. P r o d u c t i o n o f c o m m o d i t i e s b y
m e a n s o f c o m m o d i t i e s . Cambridge: Cambridge
University Press. Turgot, A.J.R. 1973. In T u r g o t o n
p r o g r e s s , s o c i o l o g y a n d e c o n o m i c s , ed. R.L.
Meek. Cambridge: Cambridge University Press.
Vaggi, G. 1987. The economics of franqois quesnay. London:
Walras, M.E.L. 1874-7. E l e m e n t s o f p u r e
e c o n o m i c s . Homewood: Irwin. 1954.
Wicksell, K. 1934. L e c t u r e s o n p o l i t i c a l e c o n o m y .
Competition: Marxian Conceptions
Willi Semmler
In the works of the classical economists such as
Smith ( ) and Ricardo ( ), competition
was identified as a central concept in economic
theory. Free competition was regarded as the
Competition: Marxian Conceptions
organizing and equilibrating force in an exchange
society, bringing about natural prices as centres of
gravity for market prices through capital flows from
areas with low rates of returns to areas with high
rates. Yet compared with the theory of perfect
competition, classical free competition was defined
more in terms of economic behaviour than of
market structure (Stigler ; McNulty ; Eatwell ).
Marx’s concept of competition, rooted in the
classical theory of free competition, also refers to
the behavioural activities of the capitalist firm.
Marx, however, more than the classics, cast serious
doubts on the stability properties of the competitive
process, and he conceptualized competition as inter-
firm dynamics carried out through reorganization of
the firm and technical change. In this it somewhat
resembles the modem theory of oligopolistic rivalry
(Friedman ) and Schumpeter’s notion of
competition as a process of ‘creative destruction’ (
).
The Dynamics of Competition in Marx
The marxian concept of competition though already
adumbrated in his early writings (
/8) is systematically developed in his later
work ( /3, , 1893, 1894). It is derived
from his theory of the behaviour of the capitalist
firm (Kuruma ). The driving force for economic
change and growth is the goal of the capitalist firm
to grow and to expand (‘the selfexpansion of
capital’). From the inter-firm dynamics results
economic evolution, accumulation and growth, but
also the downfall of old firms and the centralization
of capital (Marx ch. 25) by which the competition
and rivalry become fiercer. Firms are not conceived
as powerless economic agents adjusting passively
toward parametrically given techniques, prices and
quantities but as actively seeking the reorganization
of production and market activities in the context of
rivals’ possible reactions. Firms also are not seen as
price takers but rather as pricesetting firms with
their market shares adjusting through the reaction of
the rivals or as quantitysetting firms ( , ch. 10) with
prices and
1963
profits determined through market interactions.
Price differentiation even for homogeneous prod-
ucts is assumed to exist under disequilibrium con-
ditions ( , ch. 10). Monopoly firms are
considered exceptional cases as ‘temporary
monopolies’ ( , p. 178) when the
demand exceeds supply for a considerable period or
as ‘natural monopolies’ ( , p. 861)
when there is ownership of land or natural resources
( , pp. 178, 861).
In production activities, the reorganization of
the firm and technical change are seen as the main
weapons of competition (Marx , pp. 623). The goal
of the firm is to capture a transient surplus profit
and to transform it into long-run growth potentials,
leading to disequilibria and imbalances through
irreversible technical change and innovation, taking
place not in time- continuous form but in discrete
steps. Moreover, competition through technical
change results not in the existence of one optimal
technique but in the coexistence of multiple
techniques, and the weighted average technique is -
excluding some exceptional cases such as
decreasing returns to scale and rent - considered the
regulating technique determining the long-run
normal price ( , ch. 10).
Contrary to those forces generating disequilibria
and imbalances through inter-firm dynamics
competition is also conceived as a balancing force.
Capital as a homogeneous fund (money capital)
seeks its maximum return by flowing between
sectors (‘competition between industries’, , ch.
10). Free mobility of
labour and capital, no artificial or natural barriers
for its entry or exit and sufficiently widespread
knowledge of fields of investment are considered
preconditions for the free flow of funds. In Marx (
, ch. 10) as in Smith ( , ch. 7), and
Ricardo ( , ch. 4) a dynamical process is con
ceived in which capital funds flow into industries
with high rates of return away from industries with
low rates of return. Thus the relative output
proportions in industries will change, creating
imbalances of supply and demand. These, in turn,
cause relative market prices and profit rates to
change, tending to establish for the economic
system long-run prices of production as centres of
1964 Competition: Marxian Conceptions

gravity for market prices. Yet the stability proper- theorists regarding (i) the causes of the monopo-
ties of such a dynamic process were not demon- lization; (ii) the determination of the mark-ups and
strated rigorously. The arguments were put forward the rates of return; (iii) the different role of inverse
intuitively by analogy with Newton’s theory of the demand function and quantity reactions in their
planetary system that profit rates fluctuate or theory of price setting firms; (iv) the impact of the
oscillate within a bounded interval and actual pricesrise of oligopolies and firms size on technical
gravitate around their long-run production prices. change; and (v) the impact of large oligopoly firms
Differentials of profit rates between industries and on the stability of the economic system (increasing
firms were expected to exist for a shorter or longer stability or instability with stagnation tendencies).
period due to disequilibrium dynamics and due to Yet in spite of these differences, post- marxian
speed and ease of adjustment varying from one theory is influenced by the theory of imperfect
industry to another ( , p. 208). Though Marx competition arising in the 1930s, and competition is
anticipated thus identified more with market structure than with
possible institutional and structural changes, due torival behaviour. Moreover, the theory of mark-up
‘concentration and centralization of capital’, he didpricing was built more on a partial equilibrium view
not, however, assume that inter-firm and inter- and thus not well- founded in an interdependent
industry competition would become less severe with economic system. Though in the writings of some
the evolution of capitalism. of the post- Marxian scholars the existence of large
oligopoly firms does not preclude rivalry and
competition (in particular concerning technical
Post-Marxian Theory change, see Baran and Sweezy , eh. 3), post-
marxian writers seemed to have considered the
In the post-marxian theory since Hilferding ( ) the
theory of imperfect competition a more adequate
elimination of competition and the
framework for their analysis of advanced
delay and disruption of the formation of a
capitalism.
general profit rate through monopolization
became the main theoretical concern. Three
causes are posited as reasons for monopolization:
Recent Discussions
industrial concentration, increasing constraints
for the mobility of capital (in particular due to In recent discussions there is a certain revival of the
high proportion of fixed capital in total capital concepts of competition of the classics and Marx, in
outlay), and collusion (cooperative behaviour particular concerning the role of competition for (1)
and cartels). In this view these three causes industrial and corporate price and profit
result in monopoly prices and the persistence of determination; (2) technical change and innovation;
differential profit rates between industries and and (3) the formation of a general rate of profit.
size classes of firms (Sweezy ). For those theories,
the large firms are conceived as economic units 1. In new contributions attempts have been made to
endowed with discretionary price setting power elaborate a theory of mark-up pricing for large
determining their own environment (Kalecki , corporations in the context of a dynamic theory
; Sweezy of competition and long-run prices of
; Baran and Sweezy ; Eichner ). Here the production. In this context the economic
ideas of mark-up pricing, target rate of return behaviour of large corporations is explained
pricing and entry-preventing pricing have more in terms of change of the production
replaced the classical and marxian theory of pro- processes and the organization of the firm and
duction prices (natural prices, prices of less in terms of a change of market structures
production). (Clifton , ; Semmler ) as was
Given this general trend in post-marxian theory
there are, however, many differences among
Competition: Marxian Conceptions
attempted by post-marxian theory. According to
this new view, mark-up and target rate of return
pricing have their origin not in new market
structures but in the rise of a new type of firm:
the multi-plant and multi-product corporations
and their new financial management techniques.
Though there is, as the theory of mark-up
pricing predicts, sufficient empirical evidence of current and future cash flow and capital cost of
differential profit rates among industries and
size classes of firms - depending, however, also
on the time- period and the measure chosen for
the rate of profit - it has not been sufficiently
demonstrated that these differentials stem from
imperfect market structures or from a
disequilibrium dynamics. In addition, the
empirically observed mark-up, target rate of
return, and entry-preventing pricing, originally
developed by large corporations in the 1920s,
can be made consistent with a concept of long
period prices of production. Since, however,
large corporations are no longer single product
firms, it is more appropriate to apply the theory
of joint production to the economic behaviour
of large corporations (Semmler ). On this basis
fruitful attempts have been made to analyse the 3. In post-marxian theories the competitive for-
dynamics of competition, mark-ups, and rates of
return on the basis of an interdependent system
of prices and outputs.
2. The theory of technical change in marxian
economics has recently been given a firmer
foundation in the theory of competition (Okishio
; Shaikh ; Roemer ). In this discussion, the
marxian statement (Marx , ch. 12; , ch. 15) that
under
competitive pressure individual firms will
implement technical change and innovations and
capture a transient surplus profit, but that the
diffusion of techniques will entail a falling
general profit rate, was debated anew with the
tools of mathematical economics. The Okishio
theorem seemed to invalidate this statement,
since it implies that the capitalist firm in com-
petition will always choose a cost minimizing
technique that raises the individual as well as the
general profit rate (Okishio ). This
1965
Okishio result was disputed by Shaikh ( )
and extended by Roemer ( ). The latter
extended the Okishio result to a production price
model including fixed capital. In the debate,
however, it became clear that the Okishio result
holds only under the conditions of perfect
competition with perfect information about the
an innovation where rivals’ reactions either do
not occur or can be foreseen (Semmler ). In the
context of the dynamics of competition as
conceived by Marx - and also in the
Schumpeterian tradition - due to unforeseen
rivals’ reactions certainty concerning future
technology and markets cannot be expected
when firms choose or are forced to choose a
technique through competition. Thus the theory
of perfect competition does not seem to be
applicable as a framework in this context. But
choice of techniques with market and
technological uncertainties due to unforeseen
rivals’ reactions is by its nature difficult to
model appropriately and thus more precise
results are not yet available.
mation of a general profit rate was either taken
for granted or completely disputed (as in the
tradition since Hilferding). Recently, however, it
became clear that if it cannot be established
theoretically how profit rate differentials are
dynamically equalized through the forces of
competition then the concept of prices of pro-
duction would become empirically irrelevant. In
order to solve this problem, many scholars have
begun to formalize Marx’s conceptualization of
competition by means of dynamical systems
with price and quantity changes over time.
Nikaido ( ) presented results on the
dynamic equalization of profit rates and the
stability properties of prices of production
showing that in general they are not even locally
stable. In subsequent discussion, however, it was
shown by Dumenil and Levy ( ), Steedman (
), and Flaschel and
Semmler ( ) that better results may be
obtained if the dynamics of competition are
formalized as indicated above. For an n-sector
1966
model, a dynamics which includes changes not References
only in prices but also in production levels can
Baran, P., and P. Sweezy. 1966. M o n o p o l y
be formulated as follows:
Xj = di [rip, x)t - r(p, x ) ] (1 )
and
Pi = h [Dip, x); = Sip, x);] (2)
where D(p,x) is the average or expected demand
for an industry i for a growing economic system,
S(p,x)j the industry’s supply, r(p, x)j the
industry’s profit rate, r(p,x) the average profit
rate and px, x, the time rate of change of prices
and outputs. This dynamical process of
competition refers to capital flows across
industries according to differential profit rates
(and changes in respective production levels) as
well as to changes in prices due to imbalances in
supply and demand. The results obtained in
recently published articles on this process range
from the demonstration of complete instability
of the dynamic equalization of profit rates
(Nikaido ) to the demonstration that prices of
production are at least locally stable (Dumenil
and Levy ). It can also be demonstrated (by
utilizing a proper Lyapunov function) that
prices, outputs, and profit rates are fluctuating or
oscillating within boundaries (Flaschel and
Semmler ). Most of these attempts, however,
refer only to a circulating capital model when
the inter-industry competitive process is
analysed, and the demonstrated results depend
on the type of formalization, the reaction
coefficients as well as on additional stabilizing
forces (such as substitution in capitalist
consumption, rate of change of inventories or
rate of change of profit rate differentials).
Models of inter-industry competition including
fixed capital, returns to scale, or multiple
techniques are still rare.
See Also
Competition: Marxian Conceptions
c a p i t a l . New
York: Monthly Review Press.
Clifton, J.A. 1977. Competition and the evolution of the
capitalist mode of production. C a m b r i d g e J o u r n a l
o f E c o n o m i c s 1(2): 137-151.
Clifton, J.A. 1983. Administered prices in the context of
capitalist development. Contributions to
P o l i t i c a l E c o n o m y 2: 23-38.
Dumenil, G., and Levy, D. 1984. The dynamics of competition:
A restoration of the classical analysis. CEPREMAP, No.
8416. Paris.
Eatwell, J. 1982. Competition. In C l a s s i c a l a n d
Marxian political economy: Essays in
m e m o r y o f R . M e e k , ed. I. Bradley and M.
Howard. London: Macmillan.
Eichner, A.S. 1976. The Megacorp and oligopoly:
Microfoundations of macrodynamics. Cambridge:
Cambridge University Press.
Flaschel, P, and Semmler, W. 1987. C l a s s i c a l a n d n e o -
classical competitive adjustment
p r o c e s s e s . The Manchester School of Economic and
Social Studies.
Friedman, J. 1982. Oligopoly theory. In H a n d b o o k o f
m a t h e m a t i c a l e c o n o m i c s , vol. II, ed. K.J. Arrow
and M.D. Intriligator. North-Holland: Amsterdam.
Hilferding, R. 1910. D a s F i n a n z k a p i t a l . Trans, as
Finance Capital. London: Routledge & Regan Paul, 1981.
Kalecki, M. 1938. D i s t r i b u t i o n of national
i n c o m e . Reprinted in M. Kalecki, Selected essays on
the dynamics of the capitalist economy. Cambridge:
Cambridge University Press, 1971.
Kalecki, M. 1943. C o s t a n d p r i c e s . Reprinted in M.
Kalecki, Selected essays (1971).
Kuruma, S. 1973. M a r x L e x i k o n z u r P o l i t i s c h e n
O e k o n o m i e , Konkurrenz, vol. 1. Berlin: Oberbaum.
Marx, K. 1847. L o h n a r b e i t u n d K a p i t a l . Trans, as
Wage labor and capital. New York: International
Publishers, 1933.
Marx, K. 1857/8. G r u n d r i s s e d e r K r i t i k d e r
P o l i t i s c h e n O k o n o m i e . Trans, as Grundrisse.
New York: Random House, 1973.
Marx, K. 1861/3. T h e o r i e n i i b e r d e n M e h r w e r t .
Trans, as Theories of surplus value, 3 vols. Moscow:
Progress Publishers, 1963.
Marx, K. 1867, 1893, 1894. D a s K a p i t a l , K r i t i k d e r
P o l i t i s c h e n O k o n o m i e . 3 vols. Trans, as Capital,
a critique of political economy. New York: International
Publishers, 1967.
McNulty, P.J. 1968. Economic theory and the meaning of
competition. Quarterly Journal of
E c o n o m i c s 82: 639-656.
Nikaido, H. 1983. Marx on competition. Z e i t s c h r i f t f u r
N a t i o n a l o k o n o m i e 43(4): 337-362.
Okishio, N. 1961. Technical changes and the rate of profit.
K o b e U n i v e r s i t y E c o n o m i c R e v i e w 7: 85-99.
Ricardo, D. 1817. Principles of political economy and taxation.
In W o r k s a n d c o r r e s p o n d e n c e , ed. P. Srafia,
vol. 1. Cambridge: Cambridge University Press, 1951.
Competitive Market Processes
Roemer, J.E. 1979. Continuing controversy on the falling rate
of profit: Fixed capital and other issues. C a m b r i d g e
J o u r n a l o f E c o n o m i c s 3(4): 379-398.
Schumpeter, J. 1943. C a p i t a l i s m , s o c i a l i s m a n d
d e m o c r a c y . London: George Allen & Unwin.
Semmler, W. 1984a. C o m p e t i t i o n , m o n o p o l y a n d
d i f f e r e n t i a l p r o f i t r a t e s . New York: Columbia
University Press.
Semmler, W. 1984b. Marx and Schumpeter on competition,
transient surplus profit and technical change.
E c o n o m i e A p p l i q u e e 37(3-4): 419-455.
Shaikh, A. 1978. Political economy and capitalism: Notes on
Dobb’s theory or crisis. C a m b r i d g e J o u r n a l o f
E c o n o m i c s 2(2): 233-251.
Smith, A. 1776. An inquiry into the nature and causes of the
wealth of nations, ed. E. Cannan. London: Methuen, 1961.
Steedman, I. 1984. Natural prices, differential profit rates,
and the classical competitive process. The
Manchester School of Economic and
S o c i a l S t u d i e s 52(2): 123-140.
Stigler, G.J. 1957. Perfect competition, historically con-
templated. J o u r n a l o f P o l i t i c a l E c o n o m y
65(1): 1-17.
Sweezy, P.M. 1939. Demand under conditions of oligopoly.
J o u r n a l o f P o l i t i c a l E c o n o m y 47: 568-573.
Sweezy, P. M. 1942. T h e t h e o r y o f c a p i t a l i s t
d e v e l o p m e n t . New York: Monthly Review Press,
1968.
Competitive Market Processes
J. A. Clifton
1. Do fully competitive price signals from intense
rivalry in the market justify the moral sentiment
of laissez-faire? On grounds of distributive
justice among risk-takers, the answer has
generally been ‘yes’ throughout the history of
economic analysis.
In considerations of optimum economic
efficiency, however, the answer seems to have
become more difficult over the course of devel-
opment. The cheapening of commodities
witnessed by the classical economists is the most
virtuous example of efficient competitive market
processes in which the distribution of returns
tends to be equalized. Against the theoretical
standard of perfect competition, non-price forms
of competition came to be viewed in the 1930s
as second-class
1967
virtues - imperfectly or monopolistically com-
petitive practices.
Questions of intervention on grounds of
allocative inefficiency have continued to hinge
on the existence of classical monopoly profits.
By the 1970s, the weight of empirical evidence
and the acknowledged fact of intensified global
competition served to eliminate the credibility of
the market concentration doctrine derived from
perfect and imperfect competition (see Demsetz ,
).
At a more fundamental level, Joan Robinson
was persuaded to abandon imperfect competition
in favour of trying to more fully develop a
classical line of analysis only partially worked
out by the classical economists and largely
ignored ever since (see Clifton )•
Yet, after a decade of deregulation and the
strongest sentimentality to let the free market
reign, evidence of static and dynamic ineffi-
ciencies in industry is accumulating (Business
Week ). Has the sheer intensity of competition in
the rate of economic change and the pace of
economic life become so severe as to hinder
economic efficiency even under the strongest
possible tendencies to equalized returns in the
market? Competition, however complex and full
of discontinuities, is still evident as a systematic
and general force in the empirically observed
fact that accounting rates of return across firms
and industries tend toward uniformity over time.
This dynamic tendency is stronger among
larger than smaller firms and is stronger today
than a century ago (see Singh and Whittington
chapter 6; Brazen , , ,
pp. 239-40). But it is not explained by the
neoclassical theory of perfect competition, which
requires atomism of independent agents imder
static premises of maximization. It is not
explained by imperfect or monopolistic com-
petition for stable positions of some degree of
monopoly power are less and less in evidence all
the time. Yet this dynamic tendency is not
associated with any optimum or unique state of
industrial efficiency, as under perfect competi-
tion. Finally, the intensity of competitive
1968
rivalry that leads to this tendency cannot be
measured by neoclassical standards - the number
of firms in a market. It exists primarily under
market conditions of concentrated oligopoly.
It seems pointless to try to reconstitute the
general theory of competitive value by still more
a priori game theorizing which only adds to the
false perception of indeterminacy and lack of
systematic generality in ‘price behaviour’ under
contemporary market conditions. A recent
alternative has been to apply game theory to
perfect competition (see Mas-Colell ). What used
to be a static state of affairs distinguished by the
absence of any and all rivalry is now a non-
cooperative equilibrium, independent of the
number of agents, that may entail dynamic
strategies of M periods contingent on past
histories.
This re-introduces the long forgotten classi-
cal principle that interdependent, dynamic
rivalries are what lead to the tendency toward
uniformity in returns across the price system. A
possible virtue of the approach is that not all
games need have positive sum outcomes, so the
question of competitive rivalry and economic
efficiency is left open, not closed as in the pure
neoclassical doctrine of perfect competition.
With all the intellectual baggage imposed by
perfect and imperfect competition, however, is it
not preferable to start fresh by examining and
explaining in classical price-theoretic terms that
systematic empirical tendency toward uniformity
in returns? The first point is that the institutional
conditions for free capital mobility in the
industrial context of fixed capital have developed
gradually and progressively over the course of
economic development during the past two
hundred years. They are to be found in the first
instance not in the atomistic enterprise but in the
evolution of the organizational structure and
competitive strategies of today’s representative
firm, the industrially and geographically diver-
sified, publicly held corporation. Top manage-
ment in industry has increasingly assumed the
role once reserved for bankers in day-to-day
Competitive Market Processes
affairs, moving capital from areas of lower to
areas of higher returns.
When finance is committed to industry as
fixed capital, it is at once immobilized for its
economic life. It does not have the character of
putty which enables it to be moulded for any use
promising today’s highest return in the market.
The greatest barriers to capital mobility existed
for the single factory enterprise which typified
organizational structures in the United States in
the 1840s. Railroad firms created the first degree
of capital mobility directly within the enterprise
by pioneering the coordinated, multi-unit
organization. From the 1890s on, a degree of
capital mobility across industries was added by
integrated manufacturing companies and by
mass retailers. Truly diversified industrial
corporations began appearing in the 1920s and
by the 1930s, mass retailers were national in
scope (see Chandler , for a definitive history).
Beyond these structural elements in the
development of capital mobility in the firm, the
number of competitive strategies available to it
from economies of large-scale organization and
the intensity of the search for competitive
advantage available from large budgets and
staffs have also increased. Product innovations
from a permanent R&D staff, advertising
campaigns, takeovers and divestitures, together
with price and credit competition give the firm
added flexibility in responding to changes in
market conditions and in initiating them.
Free capital mobility is not synonymous with
the ability of atomistic firms or individual agents
to move freely throughout the economy,
whatever utopian analogies with a system of
perfect liberty and individual freedom that may
conjure up. What matters is the freedom of
capital, however organized, so to move. As
theoretical constructs, perfect and imperfect
competition left a vision of capitalist develop-
ment that is at complete odds with the actual
historical development of conditions of free
capital mobility. In this view, which is also
espoused by many non-neoclassical economists,
barriers to free capital mobility have
Competitive Market Processes
grown with the evolution of large corporations,
and the system has become less competitive, not
more competitive.
Even beyond considerations of corporate
organization and strategy, free capital mobility
is nowhere more folly developed in history than
in the institutions of today’s capital markets.
Ever more integrated on a world scale, ever
more innovative in the range of ‘products’ and
services offered, the large firms which dominate
these markets have such powerful and all-
encompassing information networks as to
approximate the economic assumption of perfect
information in the short run, if not rational
expectations in the long run.
The acceptance of market processes in ever
more spheres of human existence beyond basic
needs is a third sense in which free capital
mobility is more highly developed today. Scale
economies in automobile production are not
barriers to entry into new fields of endeavour
like the child care industry. Finally, with the
growth of labour-intensive services as a
proportion of the economy, more businesses
take on the characteristics of merchant capital
once again in history since even learned human
capital is more malleable than fixed stock.
2. Beyond considerations of free capital mobility
in explaining the uniformity in returns are other
key issues that fall outside the scope of perfect
and imperfect competition, whether or not
amalgamated with game theory.
If today’s oligopolistic firms are the slaves of
the market as never before in history, in what
sense are they ‘pure price takers’? Such corpo-
rations are entirely unable to dictate their ex post
rate of return in the market, whatever their ex
ante pricing behaviour. It is with the ex post rate
of return that the theory of competitive price is
concerned, and that will be determined by many
forms and intensities of competitive behaviour
in the market, of which a suggested mark-up ex
ante is only one. Partial equilibrium mark-up
theories have never comprehended the
difference between ex ante and ex post and err
in believing ex ante pricing discretion implies
some degree of ex-post monopoly power.
1969
The very interdependence in decisionmaking
between oligopolistic firms is what causes that
ebb and flow of business and profits across
firms, industries and markets so as to render the
ex post rate of return fully competitive and
beyond the control of the individual firm.
Unfortunately, game theory was used for
decades to deny the generality of contemporary
competitive behaviour rather than to explain its
most systematic feature in the convergence of
accounting rates of return over the long run.
A virtue of the neoclassical theory of perfect
competition was to provide a readily quantifiable
means of measuring the intensity of competition
- by the number of firms in a market. In
consideration of non-price forms of competition,
this precision in economic theory became lost,
appearing in lieu of theory as an industrial
organization ‘paradigm’ of market structure and
conduct and performance. Can quantitative
precision be resurrected in a general theory of
competitive value for the modem age?
Observation tells us that the intensity of
rivalry in contemporary markets can be mea-
sured by the frequency and voracity of changes
in market conditions - the sum total of strategic
moves and countermoves made by firms in that
market per unit of time. The common denom-
inator among all types of competition is to what
measured degree does the action move business
and profits from one sphere to another or one
firm to another.
There is a clear analogy to perfect competi-
tion that can be made here. Were oligopolists in
a market limited to the type of action an atom-
istic firm entering that market could take, pure
price taking behaviour would emerge as the
frequency of such strategic moves and coun-
termoves increased without limit. Price for a
homogeneous good would be bid down to its
normal competitive minimum not by unlimited
entry by one small firm after another, but by an
unlimited number of atomistic-like strategic
moves by the competing oligopolists.
Game theory to date appears to have over-
looked the primacy of numbers of actions in
1970
the marketplace over numbers of actors in
resurrecting the general theory of competitive
value, on measuring the intensity of competition
as the frequency of strategic moves and
countermoves in the first instance.
Of course, large firms are not restricted to
atomistic competition. Cut-rate ‘two percent’
financing by General Motors Corporation in
August 1986 was a competitive move that had
the potential to draw a great amount of business
and profits away from other firms. For that
reason, this voracious move was imitated
quickly by Ford and Chrysler.
Competition in the personal computer market
has been intense not only because of voracious
price breaks from time to time, but because the
frequency of changes in market conditions has
been enormous from real and cosmetic
innovations in hardware and software. The
frequency of competition among the commercial
television networks in changing the time slots of
programmes has at times approached the
irrational from the consumer’s standpoint.
3. When market processes are intensely compet-
itive in the frequency, voracity or complexity of
strategic moves and countermoves applied, what
will be the nature of decisionmaking by the
individual firm? Does active rivalry in the
market necessarily mean ‘maximizing behav-
iour’, optimally efficient performance from
decision-making at the margin?
One strong clue to the answer is the rejection
of the marginal method and the assumption of
constant returns to scale in recent classical
general equilibrium models of competitive price
determination (see Sraffa ). If maximizing
behaviour underlies the classical approach, it
certainly is not of pure neoclassical vintage, for
decision-making at the margin requires marginal
units which, according to Sraffa, are ‘nowhere to
be found’ in the pure classical theory of compet-
itive value.
Nor is any notion of maximization or optimal
efficiency to be found in the statement of
technology or ‘production function’ of the pure
classical system. The technology is not
Competitive Market Processes
specified by input-output coefficients, which
imply minimum input per unit of output. Only
viability conditions for each industry at a given
scale of output are listed. Viability is not the
same thing as optimum efficiency in the use of a
technique of production, whether under con-
ditions of simple reproduction or the production
of a surplus.
The entirely unsophisticated requirements for
specifying technology in the classical
determination of competitive value is an
advantage, because it formally leaves open the
question of whether fully competitive price
behaviour in ongoing market processes is always
efficient.
The empirically observed tendency of
accounting rates of return to converge in the long
run seems more assuredly decisionmaking by
oligopolists where the intensity of competitive
behaviour is asymmetric around a normal or
average rate of profit. Whether from creditor or
stockholder admonition, team pride or the threat
of takeover, firms whose performance is below
the normal rate of return are under stronger
pressure to improve profitability than those
whose performance is above the norm (see Cyert
and March ).
Further, the attributes of intensely competi-
tive market processes cause decision-making by
the firm facing such discontinuities and
complexities in its external environment to be
the kind of ‘bounded rationality’ highlighted in
the administrative theories of decision-making
for different reasons related to the internal
characteristics of large organizations (see
Simon ).
The paradox of how ‘maximum’ effort or
greater and greater rivalry directed through
market processes can result in sub-optimal out-
comes is precisely the question the business
world, especially in America, seems to be asking
itself today (see President’s Commission ).
While associated with even stronger and faster
movements to capture new markets or eliminate
excess profitability than less intensely
competitive behaviour in bygone eras,
classifying it as ‘maximizing behaviour’ or
‘satisficing’ can only lead to confusion. The
Competitive Market Processes
Competitive Market
Processes,
Fig. 1 Efficiency and
Inefficiency in Competitive
Resource Allocation *For
neoclassical theory, this
axis measures the number
of competitors, where the
limit, represented by the a fully equalized
On
vertical dashed line, is the o
familiar large numbers case o
of atomistic or perfect d
competition. For classical
theory, the axis measures
the frequency and
magnitude of changes in
market conditions. There is
no limit to the intensity of
such competitive behaviour
former implies efficiency where no such impli-
cation is warranted a priori, while the latter
implies an absence of highly energetic behaviour
from constantly striving, an implication at
complete odds with the facts. A more neutral
term like ‘competitive behaviour’ seems
preferable.
4. If frilly competitive price signals can exist under
different degrees of industrial efficiency, then
the moral sentiment of laissez-faire is not so
readily justified in a competitive free enterprise
system. Welfare economics must focus on
competition as both virtue and vice, rather than
competition as virtue and monopoly as vice, as
in the past fifty years. Consider Fig. , which
relates the intensity of competition to the degree
of economic efficiency. In modem economic
doctrine there are three unambiguous situations:
pure monopoly
(point y), perfect competition (point 7i) and the
long mn shutdown point beyond which a firm
cannot cover its total costs (point 0).
In the context of a single industry, ruinous
competition is rightly viewed as self-correcting
by market forces. Therefore, the entire scope of
economic investigation is believed to have been
between 0 and (5. The curve CT TI expresses the
sentiment that the more competition the better
for efficiency as measured by the rate
1971
neoclassical range of
- range of --------^ - sub-optimal
competitive —
behaviour competitive
behaviour with
returns
\4>
________________i______________________________V- -
p '
Intensity of Competition*
of return. The curve y n expresses the propo-
sition that the more competition, the lower the
degree of monopoly, and the stronger the ten-
dency toward uniformity in returns around a
normal rate of profit p. All inefficiency is due to
the absence of competition in sufficient degree,
and may be measured as social welfare losses
like the area y a n.
The principle justification for laissez-faire
through history has been that ‘competition
without limit’ must always enhance the general
welfare by improving static or dynamic effi-
ciencies, as expressed in the positive slope of
cni. Competition in effect can never become so
intense, or of a character or complexity, that it
pushes a market or an economic system beyond
point n in the long run. In neoclassical theory,
this is expressed as an increase in the number of
firms without limit tending to produce a state of
perfect competition.
Yet once we admit that ruinous competition
has existed in history, is there no range of sub-
optimal competitive behaviour between n and
<p? Competition that is sufficiently intense to
bid away all excess profits, but too intense to
maximize efficiency and the general welfare?
Fully competitive market processes that lead to
sub-optimal outcomes - zero sum or even neg-
ative sum games?
1972
If and only if such business practices are
isolated in one or a few markets will they be self-
correcting by the market. If they are, or have
become, systemic throughout the economy, there
is no reason to believe they will be self-
regulating in the market in a way which leads to
movement from a position like 5 to the unique
point of optimum efficiency associated with
equality of returns, n.
I submit that today’s general competitive
equilibrium in resource allocation lies at a point
like 5 and that the free enterprise system in an
atmosphere of laissez-faire is experiencing social
welfare losses of the form n a S, not of the form
n /> y from monopolistic distortions.
There is no distortion in price signals asso-
ciated with contemporary social welfare losses.
They exist in a climate of intensely competitive
market processes where the tendency toward
equality of returns is stronger, not weaker. The
real issue is becoming whether all this incessant
change still represents a Schumpeterian process
of creative destruction or an inefficient process
of ‘destructive creations'.
Free capital mobility has become so highly
developed in financial markets and top man-
agement behaviour in corporations that it has led
to the virtual collapse of the long period in
setting aspiration levels for the rate of return on
real capital formation in industry. This increase
in the intensity of competition is generating an
ongoing bias against efficiency-enhancing forms
of strategic corporate behaviour in favour of
stop-gap or crisis management forms of
competition such as ‘asset juggling’, which does
not affect the quality of products or the
efficiency with which they are produced,
distributed and sold.
The rate of change in and complexity' of
market conditions to which the film must
respond strategically has accelerated, not only in
product and input markets, but also in economic
policy variables here and abroad. The intensity'
of these competitive pressures is leading to the
creation of corporate cultures that are very risk
averse, and to decision-making of strictly
bounded rationality' that, however
Competitive Market Processes
energetic, can hardly be called ‘maximizing
behaviour’.
The growing inability to protect positions of
differential rent or supra-normal profits for a
period necessary to sustain some of the most
productive forms of risk-taking entrepreneurial
behaviour is caused by the very intensity of
competition in contemporary market processes.
The crowding out of these Schumpeterian forms
of dynamically efficient market processes is a
third social welfare loss that exists in today’s
laissez-faire atmosphere.
The capitalization of finance on pure finance
rather than real asset creation has become
almost an epidemic of market processes that are
of dubious value to the general welfare and that,
moreover, increase the cost of capital for
productive uses. For example, the increase in
takeover divestiture type activities is associated
with the creation of a distinct market for
corporate control which simply changes the
distribution of ownership and or control of
existing productive assets.
5. All seem to be agreed that competition has
become more intense in recent decades and
especially in recent years. I continue to main-
tain, as well, that there has been a secular
increase in the intensity and complexity of
competition over the course of capitalist devel-
opment and that the free enterprise system
continues to develop fundamentally along the
lines of ever greater capital mobility.
But it is also my contention that over the
course of capitalist development and especially
evident in recent years in America, the intensity
of competition has become so great as to hinder
industrial efficiency. Change for the sake of
change rather than for economic and social
progress. Competition, that engine of prosperity
that has propelled us forward for two centuries,
now seems to be of a character that it is holding
us back.
This suggests a very different role for economic
doctrine and public policy than either laissez-faire
or the regulation of monopolistically competitive
practices. It implies that intervention in the market
which reduces the intensity or scope of certain
Computation of General Equilibria 1973

fully competitive practices will not inexorably lead

to protected positions of monopoly or associated Computation of General Equilibria
inefficiencies. Intervention may in all probability
enhance economic growth or improve statical Herbert E. Scarf
resource allocation while fully maintaining that
attribute of distributive justice among risk-takers,
insofar as the equality of returns is concerned, that
is the hallmark of capitalism and freedom. Abstract
The Walrasian model of economic equilibrium
See Also is a generalization to the entire economy of the
basic notion that prices move to levels that
► Av equilibrate supply and demand. Although the
► Competition, Austrian model avoids some factors of economic signif-
► Competition, Classical icance, it is extremely useful in helping us
► A :tition L . ■ Conceptions evaluate the effects of changes in economic
policy or the economic environment. A
References moderately realistic model designed to illustrate
a significant economic issue typically involves a
Brozen, Y. 1970. The antitrust task force deconcentration large system of highly nonlinear equations and
recommendations. Journal of Law and inequalities. Existence of a solution is
E c o n o m i c s 13(2): 279-292.
Brozen, Y. 1971. The persistence of ‘high rates of return ’ in
demonstrated by non-constructive fixed point
high stable concentration industries. J o u r n a l o f L a w theorems. The explicit numerical solution of
and Economics 14(2): 501-512. such a model requires sophisticated
Brozen, Y. 1982. Concentration, m e r g e r s a n d p u b l i c computational techniques.
policy. New York: Macmillan.
Business Week. 1986. The hollow corporation. 3 Mar.
Chandler Jr., A.D. 1977. The v i s i b l e h a n d . Cambridge, Keywords
MA: Harvard University Press.
Arrow-Debreu model; Barone, E.; Brouwer’s
Clifton, J.A. 1977. Competition and the evolution of the
capitalist mode of production. C a m b r i d g e J o u r n a l fixed-point theorem; Cobb-Douglas function;
o f Economics 1(2): 137-151. Computation of general equilibria; General
Cyert, R., and J. March. 1956. Organizational factors and the equilibrium; Harberger, A.; Johansen, L.;
theory of oligopoly. Q u a r t e r l y J o u r n a l o f
Kakutani’s fixed-point theorem; Kuhn-Tucker
E c o n o m ics 70: 44-64.
Demsetz, H. 1973. The m a r k e t c o n c e n t r a t i o n Theorem; Lange, O.; Non-convexity; Spemer’s
d o c t r i n e . American Enterprise Institute for Public lemma; Technical coefficients of production;
Policy Research, Hoover Institution on War, Revolution Uncertainty; Walras’s Law; Walrasian model
and
Peace.
Demsetz, H. 1982. Economic, l e g a l , a n d p o l i t i c a l
d i m e n sions of competition. New York: North-Holland. Mas-
Colell, A. 1980. Noncooperative approaches to the theory of JEL Classifications
perfect competition: Presentation. J o u r n a l o f Economic C68
Theoty 22(2): 121-135.
President’s Commission on Industrial Competitiveness. 1985.
Global competition: T h e n e w r e a l i t y , vol. 1.
The general equilibrium model, as elaborated by
Washington, DC: U.S. Government Printing Office. Simon, H. Walras and his successors, is one of the most
1945. Administrative b e h a v i o r . New York: Macmillan and comprehensive and ambitious formulations in the
Free Press. current body of economic theory. The basic
Singh, A., and G. Whittington. 1968. G r o w t h ,
p r o f i t a b i l i t y and valuation. Cambridge: Cambridge
ingredients with which the Walrasian model is
University Press. Sraffa, P. 1960. Production of constructed are remarkably spare: a specification of
c o m m o d i t i e s b y m e a n s o f commodities. Cambridge: the asset ownership and preferences for goods and
Cambridge University Press. services of the consuming units in the
1974
economy, and a description of the current state of
productive knowledge possessed by each of the
firms engaged in manufacturing or in the provision
of services. The model then yields a complete
determination of the course of prices and interest
rates over time, levels of output and the choice of
techniques by each firm, and the distribution of
income and patterns of saving for each consumer.
The Walrasian model is essentially a generali-
zation, to the entire economy and to all markets
simultaneously, of the ancient and elementary
notion that prices move to levels which equilibrate
supply and demand. No intellectual construction of
this scope, designed to address basic questions in a
subject as complex and elusive as economics, can
be described as simply true of false - in the sense in
which these terms are used in mathematics or
perhaps in the physical sciences. The assertions of
economic theory are not susceptible to crisp and
immediate experimental verification. Moreover, the
Walrasian model disregards obvious aspects of
human motivation which are of the greatest
economic significance and which cannot be
addressed in the language of our subject: economic
theory is mute about our affective lives, about our
opposing needs for community and individual
assertion, and about the non-pecuniary determinants
of entrepreneurial energy.
There are, in addition, aspects of economic
reality which are capable of being described in the
framework of the Walrasian model but which must
be assumed away in order for the model to yield a
determinate outcome. Uncertainty' about the future
is an ever-present fact of economic life, and yet the
complete set of markets for contingent commodities
required by the Arrow-Debreu treatment of
uncertainty is not available in practice. Economies
of scale in production are a central feature in the
rise of the large manufacturing entities which
dominate modem economic activity; their
incorporation into the Walrasian model requires the
introduction of non-convex production possibility
sets for which the competitive equilibrium will
typically fail to exist.
In spite of its many shortcomings, the Walrasian
model - if used with tact and circumspection - is an
important conceptual framework for evaluating the
consequences of
Computation of General Equilibria
changes in economic policy or in the environment
in which the economy finds itself. The effects of a
major shock to the economy ofthe United States -
such as the four-fold increase in the price of
imported oil which occurred in late 1973 - can be
studied by contrasting equilibrium prices, real
wages and the choice of productive techniques both
before and after the event in question. Generations
of economists have used the Walrasian model to
analyse the terms of trade, the impact of customs
unions, changes in tariffs and a variety of other
issues in the theory of International Trade. And
much of the literature in the field of Public Finance
is based on the assumption that the competitive
model is an adequate description of economic
reality.
In these discussions the analysis is frequently
conducted in terms of simple geometrical diagrams
whose use places a severe restriction on the number
of consumers, commodities and productive sectors
that can be considered. This is in contrast to formal
mathematical treatments of the Walrasian model,
which peimit an extraordinary generality' in the
elaboration of the model at the expense of
immediate geometrical visualization. Unfortunately,
however, it is only under the most severe
assumptions that mathematical analysis will be
capable of providing unambiguous answers
concerning the direction and magnitude of the
changes in significant economic variables, when the
system is perturbed in a substantial fashion. In order
for a comparative analysis to be carried out in a
multi-sector framework it is necessary to employ
computational techniques for the explicit numerical
solution of the highly nonlinear system of equations
and inequalities which represent the general
Walrasian model.
The Use of Fixed-Point Theorems in
Equilibrium Analysis
One ofthe triumphs of mathematical reasoning in
economic theory has been the demonstration of the
existence of a solution for the general equilibrium
model of an economy, under relatively mild
assumptions on the preferences of consumers and
the nature of production possibility sets (see
Computation of General Equilibria
Debreu ). The arguments for the existence of
equilibrium prices inevitably make use of
Brouwer’s fixed-point theorem, or one of its many
variants, and any effective numerical procedure for
the computation of equilibrium prices must
therefore be capable of computing the fixed points
whose existence is asserted by this mathematical
statement.
Brouwer’s fixed-point theorem, enunciated by
the distinguished Dutch mathematician L.E.J.
Brouwer in 1912, is the generalization to higher
dimensions of the elementary observation that a
continuous function of a single variable which has
two distinct signs at the two endpoints of the unit
interval, must vanish at some intermediary point. In
Brouwer’s Theorem the unit interval is replaced by
an arbitrary closed, bounded convex set S in R", and
the continuous function is replaced by a continuous
mapping of the set S into itself: x—y g(x) Brouwer’s
Theorem then asserts the existence of at least one
point .r which is mapped into itself under the
mapping; that is, apoint .r for which .r g(x).To see
how this conclusion is used in solving the existence
problem let us begin by specifying, in mathematical
form, the basic ingredients of the Walrasian model
(Fig. )•
The typical consumer is assumed to have a
preference order for, say, the non- negative com- 1. ftp) is homogeneous of degree zero.
modity bundles .r = {xhx2,.. .,x„) in R"-, the pref- 2. ftp) is continuous in the interior of the positive
erence ordering is described either by a specific
utility function u(x hx2,.. ,,xn) or by means of an 3. ftp) satisfies the Walras LawP'ftp) 0.
1975
abstract representation of preferences. The con-
sumer will also possess, prior to production and
trade, a vector of initial assets w=(w/, w2,.. .,w„).
When a non- negative price vector p= (pi,p2,.. p„) is
announced the consumer’s income will be I =p ■ w
and his demands will be obtained by maximizing
preferences subject to the budget constraint p ■ x <
p ■ w. If the preferences satisfy sufficient regularity
assumptions, the consumer’s demand functions x(p)
will be single-valued functions ofp, continuous
(except possibly when some ofthe individual prices
are zero), homogeneous of degree zero and will
satisfy the budget constraint p-x(p)=p-w(Fig. ).
The market demands are obtained by aggregat-
ing over individual demand functions and, as such,
will inherit the properties described above. The
market excess demand functions, which I shall
denote by ftp), arise by subtracting the supply of
assets owned by all consumers from the demand
functions themselves. It is these functions which are
required for a complete specification of the
consumer side of the economy in the general
equilibrium model: they may be obtained either by
the aggregation of individual demand functions - as
we have just described - or they may be directly
estimated from econometric data. The following
properties will hold, either as a logical conclusion or
by assumption:
orthant.
 '-1 0 .. . 0 ^1, N+1 A\,K
-1 ^2, N+1 AI,K
A= 0 0

_0 0 -1 &N, N+1 ^N,K _

simultaneously and at arbitrary non-negative levels
x = (X],x2,. . -,xk); the net production plan is theny =
Ax (Fig. ).
With this formulation, a competitive equilibrium
is defined by a non-negative vector of prices p =
(pi,p2 >• • -Pi,) ar|d a non-negative vector of activity
levels x = (xhx2,.. .xk) satisfying the following
conditions:

fp) = Ax,
2. pA < 0.

The first condition states that supply and demand

are equal in all markets, and the second that there
are not opportunities for positive profits when the
profitability of each activity is evaluated at the
equilibrium prices. Taken in conjunction with the
Walras’s Law, these conditions imply that those
activities which are used at a positive level in the
equilibrium solution make a profit of zero.
Given the assumption of continuous and single-
valued excess demand functions and the description
of the production possibility set by means of an
activity analysis model, the following rather direct
application of Brouwer’s Theorem is sufficient to
demonstrate the existence of a equilibrium solution.
Under weaker assumptions on the model, variants
Each column of A describes a particular productive such as Kakutani’s Fixed- Point Theorem may be
process, with inputs represented by nonnegative required.
entries and outputs by positive entries. The
activities are assumed capable of being used

Computation of General
Equilibria, Fig. 3
Computation of General Equilibria
Let prices be normalized so as to lie on the unit
simplex S = {p = (pltp2,-- 5>i=l}-
The set of prices p for which pA<() is termed the
dual cone of the production possibility set generated
by the activity matrix A. Its intersection with the
unit simplex is a convex polyhedron C consisting of
those normalized prices which yield a profit less
than or equal to zero for all activities.
We construct a continuous mapping of S into
itself as follows: for each p in .S' consider the point
P +ftp)\ a point which is generally not on the unit
simplex itself. We then define g(p) - the image of p
under the mapping - to be that point in C which is
closest, in the sense of Euclidean distance, to P
+
.f{p)' is ^en an elementary application of the Kuhn-
Tucker Theorem to show that a fixed point of this
mapping is, indeed, an equilibrium price vector.
The Equilibrium Model as a Tool for
Policy Evaluation
Brouwer’s original proof of his theorem was not
only difficult mathematically, but it was decidedly
non-constructive; it offered no method for effec-
tively computing a fixed point of the mapping.
Brouwer did, in fact, reject his own argument during
the later ‘intuitionist’ phase of his career, in which
he proclaimed the acceptability of only those
mathematical conclusions obtained by constructive
procedures. In spite of the many simplifications in
the proof of Brouwer’s Theorem offered during the
subsequent half-century, it was not until the mid-
1960s that constructive methods for approximating
fixed points of a continuous mapping finally made
their appearance on the scene (Scarf ) - aided by the
development of the modem electronic computer and
by the rapid methodological advances in the
discipline of operations research.
In the early decades of this century, the question
of the explicit numerical solution of the general
equilibrium model was an active topic of discussion
- not by numerical analysts - but rather by
economists concerned with the techniques of
An Elementary Algorithm
economic planning in a socialist economy. The issue Barone’s negative conclusion is certainly valid if
was raised in the remarkable paper published
the full production possibility set, including all of
1977
by Enrico Barone in 1908, entitled ‘The Ministry of
Production in a Socialist Economy’. Barone, and
subsequently Oskar Lange ( ), accepted the
Walrasian model - with suitable transfers of income
- as an adequate description of ideal economic
activity in an economy in which the means of
production were collectively owned. In the absence
of markets, prices, levels of output and the choice of
productive techniques were to be obtained by an
explicit numerical solution of file Walrasian system.
A key feature of Barone’s analysis was the concept
of the ‘technical coefficients of production’ - the
input/output coefficients associated with those
activities in use at equilibrium. Barone’s contention
was that the equilibrium could be found - by an
extremely laborious calculation which might indeed
claim a significant share of the national product -
only if die correct activities were known in advance.
For Barone, rational economic calculation in a
socialist economy was defeated by the many oppor-
tunities for substitution in production: the particular
activities in use at equilibrium would be impossible
to determine by a prior computation. It is instructive
to quote Barone on this point.
The determination of the coefficients economically
most advantageous can only be done in an e x p e r i -
m e n t a l way: and not on a s m a l l s c a l e , as
could be done in a laboratory; but with experiments
on a v e r y l a r g e s c a l e , because often the
advantage of the variation has its origin precisely in
a new and greater dimension of the undertaking.
Experiments may be successfid in the sense that they
may lead to a lower cost combination of factors; or
they may be unsuccessfid, in which case the
particular organization may not be copied and
repeated and others will be preferred, which
e x p e r i m e n t a l l y have given a better result.
The Ministry of Production could not do without
these experiments for the determination of the
e c o n o m i c a l l y most advantageous technical
coefficients if it would realize the condition of the
minimum cost of production which is e s s e n t i a l
for the attainment of the maximum collective
welfare.
It is on this account that the equations of the
equilibrium with the maximum collective welfare are
not soluble a p r i o r i , on paper.
1978
the possibilities for substitution in production, is not
known to the central planner. In this event,
numerical calculation is impossible, and Lange’s
suggestion, made some 20 years later, may be
appropriate: the problem can be turned on its head
and the market, itself, can be used as a mechanism
of discovery as well as a giant analogue computer.
But if the production possibility set can be
explicitly constructed, substitution - in and of itself
- does not seem to me to be a severe impediment to
numerical computation.
At the present moment, some 20 years after the
introduction and continued refinement of fixed-
point computational techniques, I have in my
possession a small floppy disk with a computer
program which will routinely solve - on a personal
computer - for equilibrium prices and activity levels
in a Walrasian model in which the number of
variables is on the order of 100. (The authors of the
program suggest that examples with 300 variables
can be accommodated on a mainframe computer. )
Substantial possibilities of substitution, if known in
advance, offer no difficulty to the successful
functioning of this algorithm. In my opinion, the
modem restatement of Barone’s problem is rather
that even 300 variables are extremely small in
number in contrast to the millions of prices and
activity levels implicit in his account. The
computer, while expanding our capabilities
immeasurably, has taught us a severe lesson about
the role of mathematical reasoning in economic
practice and forced us to shift our point of view
dramatically from that held by our predecessors.
We realize that our preoccupations are not with
universal laws which describe economic
phenomena with full and complete generality, but
rather with intellectual formulations which are an
imperfect representation of a complex and elusive
reality. The application of general equilibrium
theory to economic planning, and more generally to
the evaluation of the consequences of changes in
economic policy, must be based on highly
aggregated models whose conclusions are at best
tentative guides to action.
An exercise in comparative statics is begun by
constmcting a general equilibrium model whose
solution reflects the economic situation existing
prior to the proposed policy change. The number
Computation of General Equilibria
of parameters required to describe demand func-
tions, initial endowments and the production pos-
sibility set is considerable, and in practice the
constraint of reproducing the current equilibrium
must be augumented by a variety of additional
statistical estimates in order to specify the model.
The limitations of data in the form required by the
Walrasian model inevitably make this estimation
procedure less than fully satisfactory.
The second step in the exercise is to calculate
the solution after the proposed policy changes are
explicitly introduced into the model. In some cases
the policy variables being studied can be directly
incorporated as parameters in the equations whose
solution yields the equilibrium values; if the
changes are small, their effects on the solution may
be obtained by differentiating these equations and
solving the resulting linear system for the
corresponding changes in the equilibrium values
themselves. This approach was adopted by Leif
Johansen ( ) and by Arnold
Harberger ( ) in his study of the incidence of a
tax on corporate profits. The use of this method in
policy analysis continues in Norway, and it forms
the basis of the amibitious programme carried out
by Peter Dixon and his collaborators in Australia (
). If, on the other hand, the policy changes
are large, the equilibrium position may be shifted
substantially, and its determination may require the
use of more sophisticated computational methods.
Fixed-point algorithms can be divided into two
major classes: those based on the elements of
differential topology, surveyed by Smale ( ),
and those which are combinatorial in nature. The
most elementary of the combinatorial algorithms for
approximating a fixed point of a continuous
mapping of the unitsimplex S = {(x = (xj,x2,.. xn)\xi
> = 1} begins by dividing the sim
plex into a large number of small subsimplices as
illustrated in Fig. . In our notation the simplex is of
dimension n 1 and has faces of dimension n 2, .. .,1.
It is a requirement of the subdivision that the
intersection of any two of the subsimplices is either
empty or a full lower dimensional face of both of
them.
Each vertex of the subdivision will have asso-
ciated with it an integer label selected from the set
Computation of General Equilibria
Computation of General
Equilibria, Fig. 4
(1, 2,..n). When the method is applied to the
determination of a fixed point of a particular map-
ping, the labels associated with a vertex will depend
on the mapping evaluated at that point. For the
moment, however, the association will be arbitrary
aside from the requirement that a vertex on the
boundary of the simplex will have a label i only if
the /th coordinate of that vertex is positive.
The remarkable combinatorial lemma demon-
strated by Emanuel Spemer ( ) in his doctoral
thesis is that at least one subsimplex must have all
of its vertices differently labelled. Assuming this
result to be correct, let us consider a mapping of the
simplex in which the image of the vector ,r = Oi,..
.,xn) is fix) = [/}(*),..., /,(*)]. The requirement that
the image be on the simplex implies that f(x) > 0 and
that Yfi{x) = 1. It follows that for every vertex of the
subdivision v, unless v is a fixed point of the
mapping, there will be a least one index i for which
fi(y) vi <0. If we select such an index to be the label
associated with the vertex v, then the assumptions of
Spemer’s Lemma are clearly satisfied, and the
conclusion asserts the existence of a simplex whose
vertices are distinctly labelled.
If the simplicial subdivision is very fine, the
vertices of this sub-simplex are all close together; at
each vertex a different coordinate is decreasing
under the mapping, and by continuity every point in
the small subsimplex wi 11 have the property that
1979
3
each coordinate is not increasing very much under
the mapping. Since the sum of the coordinate
changes is by definition zero, the image of any point
in the completely labelled subsimplex will be close
to itself, and such a point will therefore serve as an
approximate fixed point of the mapping. A formal
proof of Brouwer’s Theorem requires us to
construct a sequence of finer and finer subdivisions,
to find, for each subdivision, a completely labelled
simplex, and to select a convergent sequence of
these simplices tending to a fixed point of the
mapping.
Spemer’s Lemma may be applied to the equi-
librium problem directly. For simplicity, consider
the model of exchange in which the market excess
demand functions are given by g(p), with p on the
unit price simplex. As before, we subdivide the
simplex and associate an integer label from the set
(1,..., n) with each vertex v of the subdivision,
according to the following rule: the label i is to be
selected from the set of those indices of which gi(p)
< 0. It is an elementary consequence of Walras’s
Law that a selection can be made which is
consistent with the assumptions of Spemer’s
Lemma, and there will therefore be a subsimplex all
of whose vertices bear distinct labels. By virtue of
the particular labelling rule, any point in such a
completely labelled simplex will be an approximate
equilibrium price vector in the sense that all excess
demands, at this price, will be either negative or, if
positive, very small.
1980
Spemer’s original proof of his combinatorial
lemma was not constructive; it was based on an
inductive argument which required a complete
enumeration of all completely labelled simpli- ces
for a series of lower dimensional problems. In order
to develop an effective numerical algorithm for the
determination of such a simplex let us begin by
embedding the unit simplex, and its subsimplices, in
a larger simplex T, as in Fig. . The larger simplex is
subdivided by joining its n new vertices to those
vertices of the original subdivision lying on the
boundary of the unit simplex. The assumptions of
Spemer’s Lemma permit the new vertices to be
given distinct labels from the set (1,..., n), in such a
way that no additional completely labelled simplices
are generated. For concreteness, let the new vertex
receiving the label i be denoted by v'.
We begin our search for a completely labelled
simplex by considering the simplex with vertices
v2,..., v" and one additional vertex, say v*. If v*
Computation of General Equilibria
has the label 1, this simplex is completely labelled
and our search terminates; otherwise we move to an
adjacent simplex by removing the vertex whose
label agrees with that of v and replacing it with that
unique other vertex yielding a simplex in the
subdivision. As the process continues, we are, at
each step, at a simplex whose vertices bear the
labels 2 , . . w i t h a single one of these labels
appearing on a pair of vertices. Precisely two n 2
dimensional faces have a complete set of labels
2,..,,n. The simplex has been entered through one of
these faces; the algorithm proceeds by exiting
through the other such face.
The argument first introduced by Lemke ( ) in
his study of two person non- zero sum
games was carried over by Scarf ( ) to show
that the above algorithm never returns to a simplex
previously visited and never requires a move
outside of T. Since the number of simplices is
finite, the algorithm must terminate, and termina-
tion can only occur when a completely labelled
simplex is reached.
Computation of General Equilibria
Improvements in the Algorithm
The algorithm can easily be programmed for a
computer, and it provides the most elementary
numerical procedure for approximating fixed points
of a continuous mapping and equilibrium prices for
the Walrasian model. Since its introduction in 1967,
the algorithm, in this particular form, has been
applied to a great number of examples of moderate
size, and it performs sufficiently well in practice to
conclude that the numerical determination of
equilibrium prices is a feasible undertaking. The
algorithm does, however, have some obvious
drawbacks which must be overcome to make it
available for problems of significant size. For
example, the information which yields the labelling
of the vertices, and therefore the path taken by the
algorithm, is simply the index of a coordinate which
happens to be decreasing when the mapping is
evaluated at the vertex. More recent algorithms
make use of the full set of coordinates of the image
of the vertex instead of a single summary statistic.
Second, this primitive algorithm is always ini-
tiated at the boundary of the simplex. If the
approximation is not sufficiently good, the grid size
must be refined, and a recalculation, which makes
no use of previous information, must be performed.
It is of the greatest importance to be able to initiate
the algorithm at an arbitrary interior point of the
simplex selected as our best a priori estimate of the
answer.
The following geometrical setting (Eaves and
Scarf ) for the elementary algorithm suggests the
form these improvements can take. Let us construct
a piecewise linear mapping, h(x), of T into itself as
follows: for each vertex v in the subdivision let
h( v) v , where i is the label associated with v. We
then complete the mapping by requiring h to be
linear in each simplex of the subdivision. The
mapping is clearly continuous on Tand maps every
boundary point of T into itself. Moreover, every
subsimplex in the subdivision whose vertices are
not completely labelled is mapped, by /?, into the
boundary of T. If none of the simplices were
completely labelled, this construction would yield a
most improbable conclusion: a continuous mapping
of T into itself which is the identity on the boundary
and which maps the entire simplex into
1981
the boundary. That such a mapping cannot exist is
known as the Non-Retraction Theorem, an assertion
which is, in fact, equivalent to Brouwer’s Theorem.
The impossibility of such a mapping reinforces our
conclusion that a completely labelled simplex does
exist.
Select a point c interior to one of the boundary
faces of T and consider the set of points which map
into c; that is, the set of x for which h(x) c. As Fig.
indicates, this set contains a piecewise linear path
beginning at the point c, and trans- versing precisely
those simplices encountered in our elementary
algorithm. There are however, other parts of the set
{x|/i(x)=c}: closed loops which do not touch the
boundary of T and other piecewise linear paths
connecting a pair of completely labelled simplices.
Stated somewhat informally, the general conclusion,
of which this is an example, is that the inverse
image of a particular point, under a piecewise linear
mapping from an n dimensional set to an n— 1
dimensional set, consists of a finite union of interior
loops, and paths which join two boundary points
(see Milnor , for the differentiable version).
To see how this observation can be used, con-
sider the product of the unit simplex S and the
closed unit interval [0, 1]; that is, the set of points
(x, 1) with x in S and 0 < / < 1, as in Fig. .Extend
the mapping from the unit simplex to this large set
by defining F(x,f)=(l—t) f(x)+tx* , with x* a
preselected point on the simplex, taken to be an
estimate ofthe true fixed point. The set ofpoints for
which F(x,t) x 0 is, by our general conclusion, a
finite union of paths and loops. Precisely one of
these paths intersects the upper boundary of the
enlarged set. If the path is followed, its other end-
point must lie in the face / 0 and yield a fixed point
of the original mapping.
The path leading to the fixed point can be
followed on the computer in several ways. We can,
for example, introduce a simplicial decomposition
of the set S x [0, 1] and approximate />' by a
piecewise linear mapping agreeing with F on the
vertices of the subdivision. Following the path then
involves the same type of calculation we have
become accustomed to in carrying out linear
programming pivot steps. There are a great many
variations in the mode of simplicial subdivision
1982 Computation of General Equilibria

Computation of General
Equilibria, Fig. 7

leading to substantial improvements in the effi- mapping so that differentiation of F(x,t)—x = 0

ciency of our original fixed-point algorithm (Eaves yields a set of differential equations for the path
; Merrill ; van der Laan and joining x to the fixed point on l = 0. This leads to a
Talman ). variant of Newton’s method which is global in the
An alternative procedure, adopted by Kellogg et sense that it need not be initiated in the vicinity of
al. ( ) and Smale ( ), is to impose suf the correct answer. But, whichever of these
ficient regularity conditions on the underlying alternatives we select, the numerical
Computation of General Equilibria
difficulties in computing equilibrium prices can be Bibliography
overcome for all problems of reasonable size.
Applied General Equilibrium Analysis
During the last 15 years, the field of Applied
General Equilibrium Analysis has grown consid-
erably; instead of the few tentative examples illus-
trating our ability to solve general equilibrium
problems, we have seen the construction of a large
number of models of substantial size designed to
illuminate specific policy issues. The number of
books and papers which have appeared in the field
is far too large for a complete enumeration in this
essay, and I shall mention only a few public ations
which may be consulted to obtain an indication of
the diversity of this activity. The paper by Shoven
and Whalley ( ) in the Jour
nal of Economic Literature is a survey of applied
general equilibrium models in the fields of taxation
and international trade constructed by these authors
and their colleagues. The volume by Adelman and
Robinson ( ) is concerned with
the application of general equilibrium analysis to
problems of economic development. Whalley ( ) has
written on trade liberalization, and
Ballard et al. ( ) on the evaluation of tax
policy. Jorgenson (Hudson and Jorgenson ) and
Manne ( ) have made extensive applica
tions of this methodology to energy policy, and
Ginsburg and Waelbroeck ( ) provide a
refreshing discussion of alternative computational
procedures applied to a model of international trade
involving over 200 commodities. The volume
edited by Scarf and Shoven (1985) contains a
collection of papers presented at one of an annual
series of workshops in which both applied and
theoretical topics of interest to researchers in the
field of Applied General Equilibrium Analysis are
discussed.
See Also
►V on of s:.: . 3'
j. lew Developments)
► General Equilibrium
1983
Adelman, I., and S. Robinson. 1978. Income distribution policy
in developing countries: A case study of Korea. Stanford:
Stanford University Press.
Ballard, C.L., D. Fullerton, J.B. Shoven, and J. Whalley. 1985.
A general equilibrium model for tax
p o l i c y e v a l u a t i o n . Chicago: University of Chicago
Press.
Barone, E. 1908. II Ministerio della Produzione nello stato
colletivista. G i o r n a l e d e g l i E c o n o m i s t i e
R e v i s t a d i S t a t i s t i c a . Trans, as The Ministry of
Production in the Collectivist State. In C o l l e c t i v i s t
e c o n o m i c p l a n n i n g , ed. F.A. ITayek. London: G.
Routledge & Sons, 1935.
Brouwer, L.E.J. 1912. UberAbbildungen von Mannigfal-
tigkeiten. M a t h e m a t i s c h e A n n a l e n 71: 97-115.
Debreu, G. 1982. Existence of competitive equilibrium. In
H a n d b o o k o f m a t h e m a t i c a l e c o n o m i c s , ed.
K J. Arrow and M. Intriligator. Amsterdam: North-
Elolland.
Dixon, P.B., B.R. Parmenter, J. Sutton, and D.P. Vincent.
1982. O R A N I : A m u l t i s e c t o r a l m o d e l o f t h e
A u s t r a l i a n e c o n o m y . Amsterdam: North-Holland.
Eaves, B.C. 1972. Homotopies for the computation of fixed
points. M a t h e m a t i c a l P r o g r a m m i n g 3: 1-22.
Eaves, B.C., and H. Scarf. 1976. The solution of systems of
piecewise linear equations. M a t h e m a t i c s of
O p e r a t i o n s R e s e a r c h 1: 1-27.
Ginsburg, V.A., and J.L. Waelbroeck. 1981. A c t i v i t y ’
analysis and general equilibrium
m o d e l l i n g . Amsterdam: North-Holland.
Harberger, A. 1962. The incidence of the corporation income
tax. J o u r n a l o f P o l i t i c a l E c o n o m i c s 70: 215-
240.
Hudson, E.A., and D.W. Jorgenson. 1974. US energy policy
and economicgrowth. Bell Journal of
E c o n o m i c s a n d M a n a g e m e n t S c i e n c e 5:
461-514.
Johansen, L. 1960. A m u l t i - s e c t o r a l s t u d y o f
e c o n o m i c g r o w t h . Amsterdam: North-Holland.
Kellogg, R.B., T.Y. Li, and J. Yorke. 1976. A constructive
proof of the Brouwer fixed point theorem and compu-
tational results. S L A M J o u r n a l o f N u m e r i c a l
A n a l y s i s 13: 473-483.
Kuhn, H.W. 1968. Simplicial approximation of fixed points.
Proceedings of the National Academy of
S c i e n c e s 61: 1238-1242.
Lange, O. 1936. On the economic theory of socialism.
R e v i e w o f E c o n o m i c S t u d i e s A (53-71): 123-
142.
Lemke, C.E. 1965. Bimatrix equilibrium points and math-
ematical programming. M a n a g e m e n t S c i e n c e 11:
681-689.
Manne, A.S. 1976. ETA: A model of energy technology
assessment. B e l l J o u r n a l o f E c o n o m i c s a n d
M a n a g e m e n t S c i e n c e 7: 379-2406.
Merrill, O.H. 1971. Applications and extensions of an
algorithm that computers fixed points of certain non-
empty convex upper semicontinuous point to set map-
1984
Scarf, H.E. 1967. The approximation of fixed points of a
continuous mapping. SIAM Journal of Applied Mathe-
matics 15: 1328-1343.
Scarf, H.E., with the collaboration of T. Hansen. 1973. The
computation of economic equilibria. London/New Haven:
Yale University Press.
Scarf, H., and J.B. Shoven, eds. 1984. Applied general
equilibrium analysis. Cambridge: Cambridge University
Press.
Shoven, J.B., and J. Whalley. 1972. A general equilibrium
calculation of the effects of differential taxation of income
from capital in the U.S. Journal of Public Economy 1:
281-321.
Shoven, J.B., and J. Whalley. 1984. Applied general-
equilibrium models of taxation and international trade.
Journal of Economic Literature 22: 1007-1051.
Smale, S. 1976. A convergent process of price adjustment and
global Newton methods. Journal of Mathematical
Economics 3: 107-120.
Smale, S. 1981. Global analysis and economics. \n Handbook
of mathematical economics, ed. K.J. Arrow and M.
Intriligator, vol. I. Amsterdam: North-Holland.
Spemer, E. 1928. Neur Beweis fur die Invarianz der
Dimensionszahl und des Gebietes. Abhandlungen an den
mathematischen Seminar der Universitdt Hamburg 6:
265-272.
van der Laan, G., and A.J.J. Talman. 1979. A restart algo-
rithm for computing fixed points without an extra
dimension. Mathematical Programming 17: 74-84.
W h a l l e y , J . 1 9 8 5 . Trade liberalization among major world
trading areas. C a m b r i d g e , M A : M I T P r e s s .
Computation of General Equilibria (New
Developments)
Felix Kubler * I
Abstract
In this article, I review two recent
developments in the theory of computation
of general equilibria. First, following Brown
et al. ( ) exponentially in 1/e. Under restrictive assumptions
several
papers have developed globally convergent
algorithms for of
I review some thethe
computation
developmentsof general
in that
area Second, new developments in
computational algebraic geometry lead to
algorithms to compute effectively all
equilibria of systems of polynomial
equations. I point out some applications of
these algorithms to general equilibrium
Computation of General Equilibria (New Developments)
Keywords
Computation of general equilibria; Grobner
bases; Homotopy algorithms; Incomplete asset
markets; Kuhn-Tucker conditions; Multiple
equilibria; Newton-Kantarovich conditions;
Real business cycles; Semi-algebraic
economies; Smale’s alpha method; Tarski-
Seidenberg th; Uncertainty
JEL Classifications
D5
Introduction
After Scarf ( ) showed that there exist glob
ally convergent (and effectively applicable) algo-
rithms to compute economic equilibria, there is now
a class of computable applied models which are
routinely used to evaluate the economic con-
sequences of different taxes and tariff structures
(see, for example, Shoven and Whalley ). Research
on efficient algorithms for the computation of
general equilibria in these models largely took place
outside of economics.
A large literature in numerical analysis has
developed algorithms that are much faster than
Scarf’s original method and that can be used for
large-scale applications. Efficient iterative schemes,
mostly based on global Newton methods, now
allow applied researchers to solve for competitive
equilibria in models with hundreds of commodities
and agents (see, for example, Ferris and Pang ).
Recently, there has been substantial research in
theoretical computer science on the development of
polynomial time algorithms for the computation of
general equilibria. For most existing methods, the
number of operations needed to approximate
equilibria within a fixed precision s grows
on preferences, in models without production,
researchers have developed algorithms to
approximate equilibria ‘in polynomial time’, that is,
the running time of the algorithm increases
polynomially in the input parameters and in the
precision with which equilibria are
Computation of General Equilibria (New Developments) 1985

computed. Codenotti et al. ( ) give an over There are H agents with endowments eh e M++1^
view on recent developments along this line. and utility functions uh : 11 /
* ' —>
In this article I will not discuss any of these M. It is assumed throughout this article that utility
practical aspects of the solution of large-scale functions are smooth in the sense of Debreu ( ) -
models. I will instead focus on the following two that is, utility is C2, strictly increasing,
unrelated developments in the computation of strictly quasi-concave, exhibits non-zero Gaussian
general equilibria in economics. curvature and indifference curves do not cut the
axes.
1. The computation of equilibria in models with There are / assets available for trade. In each
time, uncertainty and missing asset markets. state s, asset j pays a bundle of commodities a,{s)
2. The computation of all equilibria and the rela- C_ M£. It is without loss of generality to assume that
tionship between exact and approximate equi- the LS x /matrix
libria in the standard Arrow-Debreu model.
/ai(l) •••
Models with Asset Markets \ai(S) .. . aj(S) J

Due to their essential static nature, standard com-
putable general equilibrium models suffer from an
oversimplified treatment of uncertainty. Agents
either solve a static problem or have myopic expec-
tations, and the model can therefore not explicitly
incorporate investment and saving decisions. The
general equilibrium model with incomplete asset
markets (GEI model) provides a basic framework
with several agents and several commodities to
incorporate uncertainty and financial markets. See,
for example, Magill and Qunizii ( ) for
an overview of the literature. The computation of
equilibria in these models is challenging because in
some specifications equilibria fail to exist while in
others they are often numerically unstable.
Kehoe and Prescott ( ) argue that real busi
ness cycle models provide an alternative way to
extend computable general equilibrium to models
with time and uncertainty. There is now a large
literature on the computation of equilibria in
dynamic stochastic economies. This is reviewed
elsewhere in this dictionary; see approximate
solutions to dynamic models (linear methods); see
also Judd ( ).
In the standard GEI model there are two time
periods (Kubler and Schmedders , show how the
problem of computation of equilibria in multiperiod
finance models can be essentially reduced to the
two period case) and S possible states of the world
in the second period. There are L perishable
commodities available for trade at each state.
has frill rank /. Allowing assets to pay in different
commodities is crucial when one wants to extend
the model to several time periods and long-lived
securities.
In the following, it will be useful to write
commodity prices as
P=(P(0),P(1)........pISDeV""1-1
and the S x /asset payoff matrix (as a function of
spot prices p(\)... p(S)), R(p), as
(p{\)-ax{\) ... p(l)-aj(l)\
m: •• :
\p{S)-ax{S) . .. p(S) ■ aj(S) J
In part of the discussion we assume an exogenous
short-sale constraint, that is, there is a number 0 < K
< oo such that the two-norm of an agent’s portfolio
must always be less than or equal to K. One can
then write an agent’s aggregate excess demand
function as the solution of his maximization
problem in the GEI economy.
(zh{p),(ph{p)) = arg max life1' + z) s.t.p
■z
=
0(p(l) • z(l)>... ,p(S) ■
z(S)) T = R(P) •‘PIMI < - K
1986 Computation of General Equilibria (New Developments)

A GEI equilibrium is a collection of prices, port- z “( p) = aigmaxu1(e + z) s. t .p ■ z = 0,

Z
folios and a consumption allocation such that
markets clear and each agent maximizes her utility, and aggregate demand as z(p) = zu(p)+ YL 2zh(p).
i.e. equilibrium prices p are characterized by Note that p is a GEI equilibrium (given that K = oo)
=a
Ei/(p) if and only if z(p) = 0. An e-equilibrium is
In a slight idealization (see also the more precise characterized by ||z(p)|| < e.
definition in the next section), we assume that the Define the expenditure of the unconstrained
maximization problem can be solved exactly and agent y' as
we define an e-equilibrium as a price p such that
yu=(p(l)-Y‘1{p),...,p(S)-4(p)). D e f i n e a n

iE^(p)ii<e-
hI e x t e n d e d pa yo f f ma t r i x R* (p) by
R*(p) = [Rip), f ip)]
A General Algorithm
and let R* i{p) be R*(p) with the f’th column
Although generally R(p) will have full rank ./, there deleted. For the constrained agents h = 2, ...,
will be so-called ‘bad prices’ at which the rank of //define
R(p) drops. When there are no short sale
constraints, that is, K oo, this leads to a discon- = a rg ma x uh{eh + z) s . t p ■ z = 0
tinuity of excess demand. Scarf’s algorithm fails: no
matter how fine the simplicial subdivision, if the {1y{\)-z{\),...,p{S)-z{S))T =R*_i(p)-p.
algorithm terminates at somc/J, one cannot
necessarily infer a bound on |z(/i)| and hence cannot Now consider a family of homotopies, indexed by i
find an e-equilibrium.
Homotopy continuation methods (see Garcia (H \
zii
(p)+tYyip,R-iip))
and Zangwill ; Eaves ) turn out to be ideally suited Hi(p,t,6) = h=2
for this numerical problem. In order to solve a R*(p)6
\e-e-i
system of equations/(.r) = 0,/: X—> Y, the basic
idea underlying homotopy methods is to find a Jo prove existence of a homotopy path, Brown et al.
smooth map H: X x [0, 1] —> Y with ( ) show that LJ/1H~1 contains a smooth
path connecting the starting point to a solution at t=
H{x, 1) = /(,v) and H(x, 0) = g(x), 1.
While generically in endowments a homotopy
where g : A * Y has a known unique zero. The map path turns out to exist, the algorithm is hardly
H is called a smooth homotopy. In using homotopy applicable in medium-sized problems, since the
methods it is crucial to set up the function, H, to number of homotopies one has to consider can
ensure that there is a smooth path that connects become quite large. An alternative is to focus on
(rs,0) with y(rs) = 0 to some (I, 1) with f{x) = 1. models with K < oo (or alternatively models with
Brown et al. ( ) develop a homotopy algo transaction costs) or to consider algorithms which
rithm which can be shown to be globally conver- might fail in a small class of problems but which
gent in that it finds an e-equilibrium for any e > 0 in are generally more efficient.
a finite number of steps. Following the so-called
Cass-trick, it is usefiil to introduce an unconstrained
agent, that is, to define the first agent maximization
problem as
Computation of General Equilibria (New Developments)
Short-Sale Constraints
In the presence of short-sale constraints, the excess
demand fhnction is continuous and equilibrium
existence can be proven with Brouwer’s theorem.
Therefore, one could presumably use a version of
Scarf’s algorithm to compute equilibria in this case.
However, while there are no new mathematical
problems to be solved, the fact that the rank of the
asset-payoff matrix can still collapse in equilibrium
poses difficult numerical problems. Simple Newton
method-based algorithms often do not work (see
Kubler and Schmedders ) unless one has a starting
point very close to the actual solution. It turns out
that, just as in the problem without short-sale
constraints, homotopy continuation methods can
provide a basis for reliable algorithms.
Schmedders ( ) develops a homotopy
algorithm which can be used to solve models with a
large number of heterogeneous households and
goods. The basic idea of his algorithm is to modify
the agents’ problem by introducing a homotopy
parameter t C [0, 1] as follows.
(zh(p,t),iph(p,t)) =a r g m a x u(eh + z)
z£R L ( . S + 1),ip£Vj
-(i-O^IMI2 s.t.p-z = o(p(i)-z(i),...,p(s)- z(S))
=R(p) •‘FIM <K.
Under the assumptions on utilities this is still a
convex problem and the first order Kuhn-Tucker
conditions are necessary and sufficient.
Schmedders provides various examples that show
that even for K = oc his algorithm, although not
guaranteed to converge, performs well in practice.
For K < oo, the Kuhn-Tucker inequalities can be
converted into a system of equalities via a change
of variables (see Garcia and Zangwill , eh. 4).
Kubler ( ), Herings and
Schmedders ( ) and others subsequently
used this idea to solve models with transaction
costs, trading constraints and other market
imperfections.
Of course, it is an important practical problem how
to trace out a homotopy path numerically.
1987
See Watson ( ) for a theoretical algorithm.
For a practical description of numerical homotopy
path-following methods see Schmedders ( ).
Equilibria in Semi-algebraic Economies
While it is clear that sufficient assumptions for the
global uniqueness of competitive equilibria are too
restrictive to be applicable to models used in
practice, it remains an open problem how serious a
challenge the non-uniqueness of competitive
equilibrium poses to applied equilibrium modelling.
In the presence ofmultiple equilibria, comparative
statics exercises become meaningless. Furthermore,
even when for a given specification of the economy
equilibrium is globally unique, as Richter and
Wong ( ) point out,
the possibility of multiple equilibria for close-by
economies implies that it is generally impossible to
compute prices and allocations that are close- by
exact equilibrium prices and allocations (as
opposed to computing prices at which aggregate
excess demand is close to zero). In this section I
argue that one can solve these problems by focusing
on so-called ‘semi-algebraic’ economies.
While the arguments are also applicable to the
GEI model, for simplicity, consider a standard
Arrow-Debreu exchange economy, (uh,eh)^=l. There
are H agents trading L commodities. Each agent h
has individual endowments eh G and ‘smooth
preferences’ characterized by an utility function uh :
IR/' > R.
A Walrasian equilibrium is a collection of con-
sumption vectors and prices p c AL 1
such that
x* e arg max uh(x) s.t.p ■ x < p ■ eh (1)
X) (** - eh) = 0. (2)
h=1
An approximate (e-) equilibrium consists of an
allocation an prices such that
1988
ll"*(**) m a x uh{x) s . t .p ■ x < p ■ eh <£ (3)
£(■* < £. (4)
Given any £ > 0, Scarfs algorithm (as well as the
more efficient algorithms used in practice) finds a
p, xh which constitute an £-equilibrium.
This leaves open two important theoretical
questions.
1. Can one relate the approximate equilibrium
prices and allocations, to exact equilibria, that
is, given a computed £-equilibrium (p, (x/r)),
does there exist a Walrasian equilibrium p, (x‘)
with ||(p, (x1')) — (p, (x;'))|| small? Can one find
good bounds on this distance which tend to zero
as £ —> 0?
2. Given an economy (uh, A)^ ] with N Walrasian
equilibria (/A (xh)")^=l and any 8 > 0, is it
possible to approximate all N equilibria, that is,
to find N £-equilibria (p'\ (xh)")^=1 with ||(pn,
(xh)n) - (pn, (xh)n)\\ < 8, for all n = 1, ..., N1
Clearly, the second problem is strictly more
difficult to tackle than the first. Richter and Wong (
) show that for general economies even the
answer to the first question is negative. In order to
obtain positive answers to both qsts, one needs to
restrict possible preferences. One approach is to
assume that better sets are semialgebraic sets. 1 will Therefore, the problem of finding Walrasian equi-
make the slightly more useful assumption that libria reduces to finding the real roots of polynomial
marginal utilities are semi-algebraic functions.
Semi-algebraic Economies
We assume that for each /?, Dxit'(x) is a semi- Walrasian equilibria to finding roots of a polyno-
algebraic function, that is, its graph {(x,y) e W2^ : y mial system of equations, one can then answer the
= Dxiih(x)} is a finite union and intersection of sets two qsts above affirmatively.
of the form
21
{(x,y) e R : g(x,y) > 0} or
2£
{(x,y)eR :f(x,y) = 0}
for polynomials with real coefficients,/and g.
Computation of General Equilibria (New Developments)
For practical purposes, the focus on semi-
algebraic preferences is quite general. First note that
Afriat’s theorem implies that a finite set of
observations on an individual’s choices that can be
rationalized by any utility function can also be
rationalized by semi-algebraic preferences (in fact,
Afriaf s construction is piece-wise linear).
Furthermore, note that the constant elasticity of
substitution utility function which is often used in
applied work is semi-algebraic if the elasticities of
substitution are rational numbers.
It follows from the Tarski-Seidenberg theorem
that for semi-algebraic economies the answers to
both qsts above are positive, since the relevant
statements can be written as first order sentences
(see Basu et al. ). Flowever, algorithmic quantifier
elimination which needs to be used to answer
general qsts in this framework is so computationally
inefficient that for practical purposes this does not
help towards solving the above qsts for interesting
specifications of economies.
Nevertheless, given a semi-algebraic economy it
is possible to find a system of polynomial equations
./(x) = 0,/: -1 ^ Rw+iy«. - is
and finitely many inequalities g'(x) > 0, g' : M//(L+1)+
£ - i _> R" i = i, ..n < oo such that p, (xh) is a
Walrasian equilibrium for the economy (uh, eh) if
and only if there exist X1' e R—, h = 1,.. .,11 such
that for some i = 1,..., N,
f(p,(xl\Ah))=0,gi(p,(xh,lh))>0.
systems of equations and verifying polynomial
inequalities (see Kubler and Schmedders ).
Having reduced the problem of finding
Question 1: Smale's Alpha Method
Smale’s alpha method provides a simple sufficient
conditions for approximate zeros to be close to
exact zeros and can be viewed as an extension of
the Newton-Kantarovich conditions. The following
results are from Blum et al. ( , ch. 8).
Computation of General Equilibria (New Developments) 1989

Let D c R" be open and let/: D > R" be analytic.

computationally convenient since modem computer
algebra systems perform exact computations over
For z e D, define/'7'/2) to be the &’th derivative off at
z. This is a multi-linear operator which maps ^- the field Q. Economic parameters are typically real
tuples of vectors in D into R". Define the norm ofnumbers, and equations characterizing equilibria he
an operator A to be in M[x], The algorithms to compute all solutions to
polynomial systems always compute all solutions in
an algebraically closed field, in this case C[x].
Mil = sup Given a polynomial system of equations/: CM >
xjiO M
C there is now a variety of algorithm to
Suppose that the Jacobian of / at z, /' \z) is approximate numerically all complex and real zeros
1

invertible and define of/. Sturmfels’s monograph ( ) provides

an excellent overview. In this survey I briefly
(f(i)(Z))-yw(Z) mention two possible approaches, homotopy con-
y(z) = sup A:! t-i tinuation methods and solution methods based on
k>2
Grobner bases.
At the writing of this article, both approaches
and are too inefficient to be applicable to large eco-
nomic models, but they can be used for models with
P ( z ) = ll(/(1)(z)) /(z)ll- four or five households and four or five
commodities. To find all equilibria for a given
Theorem 1 Given a z <E D, suppose the ball of economy, homotopy methods seem slightly more
radius ^1 — /y (z) around z is contained in D and efficient, while Grobner bases allow for statements
that about entire classes of economies.
/?(z)y(z) < 0.157.

All Solution Homotopies

Then there exists a: e D with
/(f) = 0 and ||z - z|| < 2/3(z).
While the theorem applies to any locally ana-
lytic function, the bound y(z) can in general only be
obtained if the system is in fact polynomial. For this
case, the bound can be computed fairly easily.
Given an e-equilibrium the result gives an imme-
diate bound on the distance between the approxi-
mation and an exact Walrasian equilibrium, hence
answering Question 1 above.
Question 2: Polynomial System Solving
In the following, I denote the collection of all
polynomials in the variable xb x2, ..., x„ with
coefficients in a field K by K [xb ..., x„]. The for this
survey relevant examples of K are the field of
rational numbers Q, the field of real numbers R, and
the field of complex numbers C. Polynomials over
the field of rational numbers are
Solving polynomial systems numerically means
computing approximations to all isolated solutions.
Homotopy continuation methods can provide paths
to all approximate solutions. There are well-known
bounds on the maximal number of complex
solutions of a polynomial system. The basic idea is
to start at a generic polynomial system g(x) whose
number of roots is at least as large as the maximal
number of solutions tof(x) 0 and
whose roots are all known. Then one needs to trace
out all paths (in complex space) of the homotopy
H(x, t ) = tg(x) + (1 — /)/(x), which do not diverge
to infinity. Smale’s alpha method can be applied
along the path to ensure that the approximate
solutions are close to real exact solutions (see Blum
et al. ). It can be shown that all solutions to fix) = 0
can be found in this manner.
Sommese and Wampler ( ) provide a
detailed overview. Applications of these methods in
economics have so far been largely restricted to
game theory, but the method is also applicable to
Walrasian equilibria.
1990
Grobner Basis
For given polynomials f\,..fk in 0[x] the set
/= = <fv ■■■,/*)
is called the ideal generated by/i, It turns
out that under conditions which can often be shown
to hold in practice, the so-called ‘reduced Grobner
basis’ of this ideal, /, in the lexicographic term
order has the shape
# = { X \ - ?i(-v„),.v2 - q 2{ x „ ) , . . . - Garcia, C., and W. Zangwill. 1981. P a t h w a y s t o
qn_fx„),r(x„)}
where r is a polynomial of degree d and the q, are
polynomials of degree d — 1.
This basis can be computed exactly, using
Buchberger’s algorithm (recently, much more
efficient versions of the basic algorithm have been
developed; see for example Faugere ). The number
of real solutions to the original system then equals
the number of real solutions of the univariate
polynomial r(.) which can be determined exactly by
Sturm’s method (see Sturmfels , for details). The
roots of/-(.) can be approximated numerically with
standard methods and the remaining solution to the
original system is linear in these roots.
Kubler and Schmedders ( ) use the method
to test for uniqueness of equilibria in semi-
algebraic classes of economies.
See Also
Bibliography
Basu, S., R. Pollack, and M.-F. Roy. 2003. A l g o r i t h m s i n
r e a l a l g e b r a i c g e o m e t r y . New York: Springer.
Blum, L., F. Cucker, M. Shub, and S. Smale. 1998. C o m -
p l e x i t y a n d r e a l c o m p u t a t i o n . New York:
Springer.
Computation of General Equilibria (New Developments)
Brown, D.J., P.M. DeMarzo, and B.C. Eaves. 1996. Com-
puting equilibria when asset markets are incomplete.
E c o n o m e t r i c a 64: 1-27.
Codenotti, B., S. Pemmaraju, and K Varadarajan. 2004.
Algorithms column: The computation of market equi-
libria. A C M S I G A C T N e w s 35(4): 23-37.
Debreu, G. 1972. Smooth preferences. E c o n o m e t r i c a 40:
603-615.
Eaves, B.C. 1972. Homotopies for the computation of fixed
points. M a t h e m a t i c a l P r o g r a m m i n g 3: 1-22.
Faugere, J.C. 1999. A new efficient algorithm for computing
Grobner bases (f4). J o u r n a l o f P u r e a n d
A p p l i e d A l g e b r a 139: 61-88.
Ferris, M.C., and J.S. Pang. 1997. Engineering and economic
applications of complementarity problems. S I A M
R e v i e w 39: 669-713.
s o l u t i o n s , fixed p o i n t s , a n d e q u i l i b r i a .
Englewood Cliffs: Prentice Hall.
Herings, P.J.J., and K Schmedders. 2006. Computing
equilibria in finance economies with incomplete markets
and transaction costs. E c o n o m i c T h e o r y 27: 493-
512.
Judd, K. 1998. N u m e r i c a l m e t h o d s i n e c o n o m i c s .
Cambridge: MTT Press.
Kehoe, T.J., and E.C. Prescott. 1995. Introduction to the
symposium, the discipline of applied general equilibrium.
E c o n o m i c T h e o r y 6: 1-11.
Kubler, F. 2001. Computable general equilibrium with
financial markets. E c o n o m i c T h e o r y 18: 73-96.
Kubler, F., and K. Schmedders. 2000. Computing equilibria
in stochastic finance economies. C o m p u t a t i o n a l
E c o n o m i c s 15: 145-172.
Kubler, F., and Schmedders, K. 2006. Uniqueness of equi-
libria in semi-algebraic economies. Discussion paper,
Northwestern University.
Magill, M.J.P., and M. Qunizii. 1996. T h e o r y o f
i n c o m p l e t e m a r k e t s . Cambridge: MIT Press.
Richter, M.K., and K.-C. Wong. 1999. Non-computability of
competitive equilibrium. E c o n o m i c T h e o r y 14: 1-
27.
Scarf, H. 1967. On the computation of equilibrium prices. In
Ten economic studies in the tradition of
i r v i n g f i s h e r , ed. W.J. Fellner. New York: Wiley.
Schmedders, K. 1998. Computing equilibria in the general
equilibrium model with incomplete asset markets.
Journal of Economic Dynamics and
C o n t r o l 2 2 : 1375-1403.
Schmedders, K. 2004. Homotopy path-following with
easyhomotopy: Solving nonlinear equations for
economic models. Working paper, Northwestern
University.
Shoven, J.B., and J. Whalley. 1992. A p p l y i n g g e n e r a l
e q u i l i b r i u m . Cambridge: Cambridge University
Press.
Sommese, A.J., and C.W. Wampler. 2005. The numerical
solution of systems ofpolynomials arising in engineering
and science. Singapore: World Scientific Press.
Sturmfels, B. 2002. S o l v i n g s y s t e m s o f p o l y n o m i a l
e q u a t i o n s , CBMS Regional Conference Series in
Computational Methods in Econometrics
Mathematics No. 97. Providence: American Mathematical
Society.
Watson, L.T. 1979. A globally convergent algorithm for
computing fixed points of C2 maps. A p p l i e d M a t h e -
m a t i c s a n d C o m p u t a t i o n 5: 297-311.
Computational Methods
in Econometrics
Vassilis A. Hajivassiliou
Abstract
The computational properties of an econometric
method are fundamental determinants of its
importance and practical usefulness, in con-
junction with the method’s statistical properties.
Computational methods in econometrics are
advanced through successfully combining ideas
and methods in econometric theory, computer
science, numerical analysis, and applied
mathematics. The leading classes of computa-
tional methods particularly useful for econo-
metrics are matrix computation, numerical
optimization, sorting, numerical approximation
and integration, and computer simulation. A
computational approach that holds considerable
promise for econometrics is parallel com-
putation, either on a single computer with
multiple processors, or on separate computers
networked in an intranet or over the internet.
Keywords
Bayesian inference; Bootstrap; Classical infer-
ence; Computational methods; Generalized least
squares; Generalized method of moments;
Importance sampling simulation; Jackknife;
Least absolute deviations; Maximum likelihood;
Numerical integration; Optimal control;
Ordinary least squares; Random effects models;
Simulation-based estimation; Stone, J. R. N;
Markov chain Monte Carlo methods; Parallel
computation
JEL Classification
C15
1991
Introduction
In evaluating the importance and usefulness of
particular econometric methods, it is customary to
focus on the set of statistical properties that a
method possesses - for example, unbiasedness,
consistency, efficiency, asymptotic normality, and
so on. It is crucial to stress, however, that
meaningful comparisons cannot be completed
without paying attention also to a method’s com-
putational properties. Indeed the practical value of
an econometric method can be assessed only by
examining the inevitable interplay between the two
classes of properties, since a method with excellent
statistical properties may be computationally
infeasible and vice versa. Computational methods in
econometrics are evolving over time to reflect the
current technological boundaries as defined by
available computer hardware and software
capabilities at a particular period, and hence are
inextricably linked with determining what the state
of the art is in econometric methodology.
To give a brief illustration, roughly from the late
1950s until the early 1960s we had the ‘Stone Age’
of econometrics, when the most sophisticated
computational instrument was the slide rale, which
used two rulers on a logarithmic scale, one sliding
into the other, to execute approximate multiplication
and division. In this Stone Age, suitably named in
honour of Sir Richard Stone, winner of the 1984
Nobel Prize in Economics, the brightest Ph.D.
students at the University of Cambridge were toiling
for days and days in back rooms using slide rales to
calculate ordinary linear regressions, a task which
nowadays can be achieved in a split second on
modem personal computers.
The classic linear regression problem serves to
illustrate the crucial interaction between statistical
and computational considerations in comparing
competing econometric methods. Given data of size
S, with observations on a dependent variable
denoted by S x 1 vector y and corresponding
observations on k explanatory factors denoted by S x
k matrix X (k < X), the linear plane fitting exercise is
defined by Gauss’s minimum quadratic distance
problem:
1992
/ ? = ar g m in (y — Xb)'(y — Xb)
h
S (1)
x=arg min ^(ys - xJsb)2
5=1
where x's is the sth row of matrix X and h is a k / 1
vector of real numbers defining the regression plane
Xb. Under the assumption that X has hill column
rank k, the solution to this ordinary least squares
minimization problem is the I inear-in-v expression
[i (X'X) X!y. which only requires the matrix
operations of multiplication and inversion. Suppose,
however, that Gauss had chosen instead as his
measure of distance the stun of absolute value of the
deviations, and defined instead:
s
) = vg minV|ys-A>| (2)
b In the following section we discuss in turn the
s =1
The vector fl that solves the second minimi-
zation is known as the least absolute deviations
(LAD ) estimator and has no closed-form matrix
expression. In fact, calculation of p requires highly
nonlinear operations for which computationally
efficient algorithms were developed only in the
1970s. To give a concrete example, consider the
intercept-only linear regression model where X is
the S x 1 vector of ones. Then the single P
coefficient that solves ( ) is the sample mean of y,
while ft that solves ( ) is the sample median of y.
The latter is orders of magnitude more difficult to
compute than the former since it involves sorting y
and finding the value in the middle, while the
former simply adds all elements of y and divides by
the sample size. Clearly, it could be quite
misleading if p and p where compared solely in
terms of statistical properties without any
consideration of their substantially different
computational requirements.
A second example in a similar vein is the
following parametric estimation problem. Suppose a
sample of size S is observed on a single variabley. It
is believed that each observation^ is
Computational Methods in Econometrics
drawn independently from the same uniform dis-
tribution on the interval [0,c] where the lower value
of the support is the single unknown parameter that
needs to be estimated, while c is known. Two
parametric estimation methods with particularly
attractive statistical properties are the generalized
method of moments (GMM) and the method of
maximum likelihood (MLE). Indeed, for relatively
large sample sizes these two methods are
comparably attractive in terms of statistical
properties, while they differ drastically in terms of
computational requirements: the GMM solution is
6gmm = f ys — c, thus requiring only the simple
calculation of the sample mean y, while the MLE
involves the highly nonlinear operation of finding
the minimum of the data vector y, 6mie =
min(y1,... ,ys).
leading classes of methods that are of particular
importance in modem econometrics, while section “
” introduces the con
cept of parallel processing and describes its current
value and future promise in aiding dramatically
econometric computation.
Computational Methods
Important for Econometrics
The advancement of computational methods for
econometrics relies on understanding the interplay
between the disciplines of econometric theory,
computer science, numerical analysis, and applied
mathematics. In the five subsections below we
discuss the leading classes of computational
methods that have proven of great value to modem
econometrics.
Matrix Computation and Specialized
Languages
To start with the fundamental econometric frame-
work of linear regression, the sine qua non of
econometric computation is the ability to program
and perform efficiently matrix operations. To this
end, specialized matrix computer languages have
( 1 y y2 ... /-i\
y i y
y2 y ••• r
1 y
W~l y2 y i /

Other matrix algebra methods especially important

in econometrics are the Cholesky factorization (see
Golub ) of a positive definite matrix A into the
product A = R'R where R is an upper- triangular
matrix, and the singular value decomposition that
allows the calculation of pseudoinverse of any
matrix B which may be non-square, and if square,
not positive definite (see Belsley ).
It is important to note that on occasion a brilliant
theoretical development can simplify enormously
the computational burden of econometric methods
that, though possessing attractive statistical
properties, were thought to be infeasible with
existing computation technology in the absence of
the theoretical development. A case in point is the
GLS/MLE estimator for the one-factor random
effects model proposed by Balestra and Nerlove (
), which is of great importance in the anal
ysis of linear panel data models. The standard
formulation gives rise to the GLS formula
1994 Computational Methods in Econometrics

requiring the inversion of an equi-correlated var- problems in computational economics, for example,
iance covariance matrix Q of dimension S x S, the problem of optimal control.) Algorithms for
where S is of the order of the product of the number optimizing functions of many variables are a key
of available observations in the cross- section component in the collection of tools for
dimension times the number available in the time econometric computation. The suitability of a cer-
dimension. For modem panel data-sets, this can tain algorithm to a specific optimization econo-
exceed 300,000, thus making the calculation of Q 1 metric problem depends on the following
infeasible even on today’s supercomputers, let alone classification:
with the slide rales available in 1966. Fuller and
Battese ( ), however,1. Algorithms that require the calculation of first
showed that the equi-correlated nature of the one- and possibly second derivatives Versus algo-
factor random effects model made calculation of the rithms that do not. Clearly, if the function to be
GLS estimator equivalent to an OLS problem, optimized is not twice continuously differen-
where the dependent variable y and the regressors X tiable (as is the case with LAD) or even dis-
are simple linear combinations of the original data continuous (as is the case with the maximum
score estimator for the semiparametric analysis
v„, „rllY,..., xklt and its time ^averages yu,x\it, ... ,xku
defined by of the binary response model - see Manski
{ylt

and y,, =yit — /f,, and analogously for the regressor ), algorithms that require differentiability
variables. This realization allowed the calculation of will not be suitable. The leading example of an
the GLS estimator without the need for inverting the algorithm not relying on derivatives is the non-
usually problematically large Q matrix. linear simplex method of Nelder and Meade (
Another important case where a theoretical ).
development in methodology led to a dramatic 2. Local Versus global algorithms. Optimization
lowering of the computational burden and hence algorithms of the first type (for example, Gauss-
allowed the calculation of models that would oth- Newton, Newton-Raphson, and Bemdt et al. (
erwise have had to wait perhaps for decades for )) search for an optimum in the vicinity
sufficient advancements in computer technology is of the starting values fed into the algorithm.
the simulation-based inference for Limited This strategy may not necessarily lead to a
Dependent Variable models, associated with the global optimum over the full set of parameter
name of Daniel McFadden ( ). See section space. This is of particular importance if the
“ ” below, McFadden, Daniel function to be optimized has multiple local
and simulation-based estimation. optima, where typically the estimator with the
desirable statistical properties corresponds to
Optimization locating the overall optimum of the function. In
Many econometric estimators with attractive sta- such cases, global optimization algorithms (for
tistical properties require the optimization of a example, simulated annealing and genetic
(generally ) nonlinear function of the form: optimization algorithm) should be employed
instead.
q=aig m ax F(6; data) ( 3)
9 Special methods are necessary for constrained
optimization, where a function must be maximized
over a vector of unknown parameters 0 of dimen- or minimized subject to a set of equality or
sion p, typically considerably larger than 1. Exam- inequality constraints. These problems, in general
ples are: the method of maximum likelihood, considerably more demanding than unconstrained
minimum-distance (OLS, LAD, GMM), and other optimization, can be handled through three main
extremum estimators. (The need to optimize alternative approaches: interior, exterior and re-
functions numerically is also important for certain parameterization methods.
Comprehensive reviews of optimization
methods in econometrics can be found in Goldfeld
Computational Methods in Econometrics
and Quandt ( ), Quandt ( ), and Dennis
and Schnabel ( ). These studies also discuss
the related issue of the numerical approximation of
derivatives and illustrate the fundamental link in
terms of computation between optimization and the
problem of solving linear and nonlinear equations.
For similar methods used in economics, see
numerical optimization methods in economics and
nonlinear programming.
Sorting
Of special importance for computing the class of
estimators known as robust or semiparametric
methods is the ability to sort data rapidly and
computationally efficiently. Such a need arises in
the calculation of order statistics, for example, the
sample median and sample minimum required by
the first two estimation examples given above. The
leading sorting algorithms, bubble-, heap- and
quick-sort, have fundamentally different properties
in terms of computation speed and memory
requirements, in general depending on how close to
being sorted the original data series happens to be.
For a practical review of the leading sorting
algorithms, see Press et al. ( , ch. 8).
Numerical Approximation and Integration
Numerical approximation is necessary for any
mathematical function that does not have a closed
form solution, for example, exponential, natural
logarithm and error functions. See Abramowitz and
Stegun ( ) for an exhaustive study of
mathematical functions and their efficient approx-
imation. Judd ( ) focuses on numerical
approximation methods particularly useful in eco-
nomics and econometrics.
Numerical integration, also known as numerical
quadrature, is a related approximation problem that
is cmcial to modem econometrics. There are two
key fields of econometrics where integrals without a
closed form must be evaluated numerically. The
first is Bayesian inference where moments of
posterior densities need to be evaluated, which take
the form of high-dimensional integrals. See, inter
aha, Zellner et al. ( ).
The second main class is classical inference in
limited dependent variable (LDV) models; for
example, Hajivassiliou and Ruud ( ). See
1995
Geweke ( ) for an exhaustive review of
numerical integration methods in computational
economics and econometrics, and Davis and
Rabinowitz ( ) for earlier results.
It is important to highlight a cmcial difference
between the numerical integration problems in
Bayesian inference and those in classical inference
for LDV models, which makes various integration-
by-simulation algorithms be useful to one field and
not the other: in the Bayesian case, typically a
single or a few high-dimensional integrals have to
be evaluated accurately. In contrast, in the classical
LDV inference case, quite frequently hundreds of
thousands of such integrals need to be
approximated.
Computer Simulation
The need for efficient generation of pseudorandom
numbers with good statistical properties on a
computer appears very routinely in econometrics.
Leading examples include:
• Statistical methods based on resampling, pri
marily the ‘jackknife’ and the ‘bootstrap’, as
introduced by Efron ( ). These methods
have proven of special value in improving the
small sample properties of certain econometric
estimators and test procedures, for example in
reducing estimation bias. They are also used to
approximate the small sample variance of esti-
mators for which no closed form expressions can
be derived.
• Evaluation of econometric estimators through
Monte Carlo experiments, where hypothetical
data-sets with certain characteristics are simu-
lated repeatedly and the econometric estimators
under study are calculated for each set. This
allows the calculation of empirical (simulated)
properties of the estimators, either to compare to
theoretical mathematical calculations or because
the latter are intractable.
• Calculation of frequency probabilities of pos-
sible outcomes in large-scale decision trees, for
which the outcome probabilities are impossible
to characterize theoretically.
• Sensitivity analyses and what-if studies, where
an econometric model is ‘ran’ on a computer
under different scenarios of policy measures.
1996 Computational Methods in Econometrics

• Simulation-based Bayesian and classical infer- in the case of simulation-based estimators. See
ence, where integrals are approximated through Nagumey ( ) for a discussion of parallel com
computer simulation (known as Monte Carlo putation in econometrics.
integration). Particularly important methods in An alternative approach for parallel computation
this context are the following: frequency that does not involve a single computer with many
simulation; importance sampling; and Markov processors has been developed recently and offers
chain Monte Carlo methods (the leading considerable promise for computational
exponents being Gibbs resampling and the econometrics. Through the use of specialized
Metropolis/Hastings algorithm). A related class computer languages, many separate computers are
of methods, known as variance-reduction harnessed together over an organization’s intranet
simulation techniques, includes control variates or even over the internet, and an econometric
and antithetics. See Geweke ( ) and computation task is distributed across them. The
Hajivassiliou et al. ( ) for reviews. See also benefits of this approach depend critically on the
simulation-based estimation. relative burden of the overhead of communicating
across the individual computers when organizing
the splitting of the tasks and then collecting and
Parallel Computation processing the separate partial results. Such
distributed parallel computation has the exciting
Parallel processing, where a computation task is potential of affording formidable super-computing
broken up and distributed across different com- powers to econometric researchers with only
puters, is a technique that can afford huge savings modest computer hardware.
in terms of total time required for solving partic-
ularly difficult econometric problems. For example,
the simulation-based estimators mentioned in the
previous section exhibit the potential of significant
See Also
computational benefits by calculating them on
computers with massively parallel architectures, ► Longitude: Data Analysis
because the necessary calculations can be organized ► addi ?om 1537)
in essentially an independent pattern. An example ► i Ton-' v:: Programming
of such a computer is the Connection Machine CM- ► i f o i m : .
5 at the National Center for Supercomputing
Applications in Illinois with 1024 identical ► o . - rs in Econometrics
processors in a multiple- instruction/multiple-data ► Simulation-Based Estimation
(MIMDI) configuration. The benefits of such a
parallel architecture on the problem of solving an
Bibliography
econometric optimization classical estimator not
involving simulation can also be substantial, since Abramowitz, M., and I. Stegun. 1964. H a n d b o o k o f
such estimators involve the evaluation of m a t h e m a t i c a l f u n c t i o n s . Washington, DC:
contributions to the criterion (for example, National Bureau of Standards.
likelihood) function in the case of independently Balestra, P., and M. Nerlove. 1966. Pooling cross-section and
time-series data in the estimation of a dynamic model.
and identically distributed (i.i.d.) observations. E c o n o m e t r i c a 34: 585-612.
Since typical applications in modem applied Belsley, D. 1974. Estimation of system of simultaneous
econometrics using cross-sectional and longitudinal equations and computational specifications of GREM-
data sets involve several thousands of i.i.d. LIN. A n n a l s o f E c o n o m i c a n d S o c i a l
M e a s u r e m e n t 3: 551-614.
observations, the potential benefits of parallel Bemdt, E.K., B.H. Hall, R.E. Hall, and J.A. Hausman. 1974.
calculations of such estimators should be obvious. Estimation and inference in nonlinear structural models.
The benefits of a massively parallel computer Annals of Economic and Social
architecture become even more pronounced M e a s u r e m e n t 3: 653-666.
Computer Industry 1997

Davis, P.J., and P. Rabinovitz. 1984. M e t h o d s o f Zellner, A., L. Bauwens, and H. VanDijk. 1988. Bayesian
n u m e r i c a l i n t e g r a t i o n . New York: Academic. specification analysis and estimation of simultaneous
Dennis, J.E., and R.B. Schnabel. 1984. U n c o n s t r a i n e d equation models using Monte Carlo methods. J o u r n a l
optimization and nonlinear equations. o f E c o n o m e t r i c s 38: 73-90.
Englewood Cliffs: Prentice-Hall.
Drud, A. 1977. An optimization code for nonlinear econo-
metric models based on sparse matrix techniques and
reduced grades. A n n a l s o f E c o n o m i c a n d
S o c i a l M e a s u r e m e n t 6: 563-580. Computer Industry
Efron, B. 1982. T h e j a c k k n i f e , t h e b o o t s t r a p ,
a n d o t h e r r e s a m p l i n g p l a n s , CBMS-NSF
Shane Greenstein
monographs No. 38. Philadelphia: SIAM.
Fuller, W.A., and G.E. Battese. 1973. Transformations for
estimation of linear models with nested-error structure.
Journal of the American Statistical
A s s o c i a t i o n 68: 626-632.
Geweke, J. 1988. Antithetic acceleration of Monte Carlo Abstract
integration in Bayesian inference. J o u r n a l o f Commercial computing has grown to include an
E c o n o m e t r i c s 38: 73-90. extraordinary range of economic undertakings.
Geweke, J. 1996. Monte Carlo simulation and numerical
integration. In H a n d b o o k o f c o m p u t a t i o n a l
In any given era, computing markets are
e c o n o m i c s , vol. 1, ed. H. Amman, D. Hendrik, and J. organized around platforms - a cluster of tech-
Rust. Amsterdam: North-Holland. nically standardized components that buyers use
Goldfeld, S., and R. Quandt. 1972. N o n l i n e a r m e t h o d s together to make the wide range of applications.
i n e c o n o m e t r i c s . Amsterdam: North-Holland.
There has been an increasing secular trend in the
Golub, G.H. 1969. Matrix decompositions and statistical
calculations. In S t a t i s t i c a l c o m p u t a t i o n , ed. number of firms that possess the necessary
R.C. Milton and J.A. Milder. New York: Academic. technical knowledge and commercial
Hajivassiliou, V.A., and PA. Ruud. 1994. Classical estimation capabilities to bring to market some component
methods using simulation. In H a n d b o o k of
or service. While general improvements in
e c o n o m e t r i c s , vol. 4, ed. R. Engle and D. McFadden.
Amsterdam: North-Holland. technical capabilities are readily apparent, it is
Hajivassiliou, V.A., D.L. McFadden, and P.A. Ruud. quite difficult to calculate the productivity
1996. Simulation of multivariate normal rectangle improvements arising from increased investment
probabilities and derivatives: Theoretical and compu-
in and use of computing.
tational results. J o u r n a l o f E c o n o m e t r i c s 72(1,
2): 85-134.
Judd, K. 1996. Approximation, perturbation, and projection Keywords
methods in economic analysis. In H a n d b o o k o f
c o m p u t a t i o n a l e c o n o m i c s , vol. 1, ed. H.
Computer industry; Economic growth; Infor-
Amman, D. Hendrik, and J. Rust Amsterdam: North- mation technology; Innovation; Moore’s law
Holland.
Manski, C. 1975. Maximum score estimation of the stochastic
utility model of choice. J o u r n a l o f E c o n o m e t r i c s
JEL Classifications
3: 205-228.
McFadden, D. 1989. A method of simulated moments for
L63
estimation of multinomial discrete response models.
E c o n o m e t r i c a 57: 995-1026. The commercial computing industry accounts for a
Nagumey, A. 1996. Parallel computation, l a H a n d b o o k large fraction of economic activity. From its
o f c o m p u t a t i o n a l e c o n o m i c s , vol. 1, ed. H. military and research origins in the late 1940s, it
Amman, D. Hendrik, and J. Rust Amsterdam: North-
Holland.
spread into the commercial realm and has since
Nelder, J.A., and R. Meade. 1965. A simplex method for grown to include an extraordinary range of eco-
function minimization. C o m p u t e r J o u r n a l 7: 308- nomic undertakings. Many economists believe this
313. expansion of applications for computing has been a
Press, W.H., B.P. Flannery, S.A. Teukolsky, and W.T.
Vetterling. 2001. Numerical recipes in
driver of economic growth.
Fortran 77: T h e art of scientific Computing aids the automated tracking of
c o m p u t i n g . Cambridge: Cambridge University Press. transactions, a function that finds use, for
Quandt, R. 1983. Computational problems and methods. In
1998
example, in automating billing, managing the
pricing of inventories of airline seating, and
restocking retail outlets in a geographically dis-
persed organization. It also facilitates the coordi-
nation of information-intensive tasks, such as the
dispatching of time-sensitive deliveries or emer-
gency services. Computing also enables perfor-
mance of advanced mathematical calculations,
useful in such diverse activities as calculating
interest on loans and generating estimates of
underground geological deposits. Computer- aided
precision also improves the efficiency of processes
such as manufacturing metal shapes or the
automation of communication switches, to name
just two.
In any given era, computing markets arc orga-
nized around platforms - a cluster of technically
standardized components that buyers use together to
make the aforementioned wide range of appli-
cations. Such platforms involve long-lived assets,
both components sold in markets (that is, hardware
and some software) and components made by
buyers (that is, training and most software).
Important computing platforms historically include
the UNIVAC, the IBM 360 and its descen- dents,
the Wang minicomputers, IBM AS/400, DEC VAX,
Sun SPARC, Intel Windows PC, Unix/Linux, and,
after the mid-1990s, TCP/IP- based client-server
platforms linked together.
Vendors tend to sell groups of compatible
products under umbrella strategies aimed at the
users of particular platforms. In the earliest eras of
computing markets, the leading films integrated all
facets of computing and offered a supply of goods
and services from a centralized source. In later eras,
the largest and most popular platforms historically
included many different computing,
communications and peripheral equipment films,
software tool developers, application software
writers, consultants, system integrators, distributors,
user groups, news publications and service
providers.
Until the early 1990s, most market segments
were distinguished by the size of the tasks to be
undertaken and by the technical sophistication of
the typical user. Mainframes, minicomputers,
workstations, and personal computers, in decreasing
order, constituted different size-based market
Computer Industry
segments. Trained engineers or programmers made
up the technical user base, while the commercial
market was geared more towards administrators,
secretaries and office assistants.
The most popular platform in the late 1980s and
1990s differed from the prominent platforms of
earlier years. The personal computer (PC) began in
the mid- 1970s as an object of curiosity among
technically skilled hobbyists, but became a common
office tool after the entry of IBM's design. Unlike
prior computing platforms, this one has diffused
into both home and business use. From the
beginning, this platform involved thousands of large
and small software developers, third-party
peripheral equipment and card developers, and a
few major players. In more recent experience,
control over the standard has completely passed
from IBM to Microsoft and Intel. Microsoft
produces the Windows operating system and Intel
produces the most commonly used microprocessor.
For this reason the platform is often called Wintcl.
The networking and internet revolution in the
late 1990s is responsible for blurring once- familiar
distinctions. These new technologies have made it
feasible to build client-server systems within large
enterprises and across ownership boundaries. It
employs internet-based computing systems
networked across potentially vast geographic
distances, supporting the emergence of a ‘network
of networks’.
Despite frequent and sometimes dramatic tech-
nical improvements in specific areas of technology,
many features of the most common platforms in use
tend to persist or change very slowly. Many durable
components make up platforms. And, though they
lose their market value as they become obsolete in
comparison with frontier products, they do not as
quickly lose their ability' to provide a flow of
services to users. Consequently, new technology
tends to be most success- fill when new components
enhance and preserve the value of previous
investments, a factor that creates demand for
‘backward compatible’ upgrades or improvements.
It also creates a demand for support and service
activities to reduce the costs of making the
transition from old to new.
Computer Industry
Control over changes to design and other aspects
of technical standards shapes the backward
compatibility for key components. Control of these
decisions is coincident with platform leadership -
determining the rate and direction of change in
technical features of components around which
other firms build their businesses. In each platform,
it is very rare to observe more than a small number
of firms acquiring leadership positions. Since such
positions have been historically associated with
high firm profitability, firms compete fiercely for
market dominance in component categories where
standards are essential. Not surprisingly,
competitive behavioiu affiliated with obtaining and
retaining market leadership does occasionally
receive attention from antitrust authorities.
Though innovative change in computing began
well prior to the invention of the integrated circuit,
in popular discussion advances in computing have
become almost synonymous with advances in
microprocessors. This is due to an observation by
Gordon Moore, who co-founded and became
chairman at Intel. In 1965 he foresaw a doubling of
circuits per chip every two years. This prediction
about the rate of technical advance later became
known as ‘Moore’s law’. In fact, microprocessors
and DRAMS have been doubling in capability
every 18 months since the mid-1970s.
Moore’s prediction pertained narrowly to inte-
grated circuits. However, a similar pattern of
improvement - though with variation in the rate -
characterizes other electronic components that go
into producing a computer or that are
complementary with computing in many standard
uses. This holds for disk drives, display screens,
routing equipment, and data-transmission capacity,
to name a few. Such widespread innovation creates
opportunities for new entry and rearrangements in
the conditions of supply.
Accordingly, there has been an increasing sec-
ular trend in the number of firms that possess the
necessary technical knowledge and commercial
capabilities to bring to market some component or
service of value to computing users. This factor
alone explains the increasing complexity of supply
chains for the supply of most computing
1999
hardware and software products. It is also coinci-
dent with their increasing geographic reach. In
addition, as in other manufacturing processes, the
increasing use of sophisticated information
technology helps coordinate design and production
involving firms from many countries and
continents.
While the spawning of new information tech-
nology businesses in North America has tended to
be concentrated in a small number of locations,
such as the Boston area and Santa Clara Valley
(popularly known as Silicon Valley), every other
facet of the supply chain for computing involves
firms headquartered and operating in a much wider
set of locations. In North America, these range from
Seattle, Austin, Los Angeles, the greater New York
area, Denver-Boulder, Washington DC, the North
Carolina Research Triangle, Chicago, and virtually
all major cities in the United States. The supply
chain for many complementary components has
also been associated with many firms in Western
Europe and as well as in India, Israel, South Korea,
Singapore, Taiwan and China. Even more
widespread are computing service firms, which
follow business and home users dispersed across
the globe.
Despite this geographic dispersion since the
1950s, US companies have retained leadership in
generating new platforms and commercializing
frontier technologies in forms that most users find
valuable. Part of this results from the persistence of
platform leadership for a time within a segment. In
addition, US firms have historically been ascendant
whenever platform leadership has changed.
However, this pattern seems likely to change in the
21st century, as non-US firms already have found
leadership positions in producing components of
many platforms and in related areas of electronics,
such as consumer electronics, communication
equipment and specialized software.
While general improvements in technical capa-
bilities are readily apparent, it is quite difficult to
calculate the productivity improvements arising
from increased investment in and use of computing.
There is no question that existing computing
activities have become less expensive, while new
capabilities have been achieved. This has allowed
2000 Computer Science and Game Theory

economic actors to attain previously unobtainable Brensnahan, T., and F. Malerba. 1999. Industrial dynamics
outcomes. This shift in economic possibilities has and the evolution of firm’s and nations’ competitive
capabilities in the world computer industry. In S o u r c e s
generated a restructuring of organizational routines, o f i n d u s t r i a l l e a d e r s h i p , ed. D. Mowery and R.
market relationships, and other activities associated Nelson. Cambridge, UK: Cambridge University Press.
with the flow of goods, which inevitably improves Brynjolfsson, E., and H. Lorin. 2000. Beyond computation:
Information technology, organizational transformation
the economy’s ability to transform inputs into
and business performance. J o u r n a l o f E c o n o m i c
consumer welfare. P e r s p e c t i v e s 14(4): 23-48.
Yet altering the business use of computing can Dedrick, J., and K. Kraemer. 2005. The impacts of IT on firm
be slow. It often demands large adjustment costs and industry structure: The personal computer industry.
C a l i f o r n i a M a n a g e m e n t R e v i e w 47(3): 122-
and gradual learning about which organizational
142.
processes can best employ advances in computing. Flamm, K. 2003. The new economy in historical perspective:
It can involve a reallocation of decision rights and Evolution of digital electronics technology. In N e w
discretion inside a large organization, especially e c o n o m y h a n d b o o k , ed. D. Jones. San Diego:
Academic Press/Elsevier.
when business units alter a wide array of Greenstein, S. (ed.). 2006. T h e i n d u s t r i a l e c o n o m i c s
intermediate routine processes (such as billing, o f c o m p u t i n g . Northampton: Edward Elgar.
account monitoring, and inventory management) or Jorgenson, D., and C. Wessner (eds.). 2005.
the coordination of services (such as the delivery of Deconstructing the computer: Report of a symposium.
Washington, DC: National Academies Press.
data for decision support). Moreover, the largest McKinsey Global Institute. 2001. U.S. productivity growth,
changes come from altering many complementary 1995-2000: Understanding the contribution of information
activities that respond to new and unanticipated technology relative to other factors. Washington, DC:
McKinsey and Co.
opportunities, setting off new waves of invention.
Each wave’s productivity effect is interwoven with
others.
Along with these improvements the boundaries
of the ‘computing market’ have changed. A Computer Science and Game Theory
hardware-based definition for the computing market
was barely adequate in the 1960s and is no longer Joseph Y. Halpern
adequate for economic analysis. However, there is
no consensus about what alternative framing will be
appropriate for understanding value creation,
supplier behaviour, and user adoption in computing Abstract
in the 21st century. Work at the intersection of computer science and
game theory is briefly surveyed, with a focus on
the work in computer science. In particular, the
See Also following topics are considered: various roles of
computational complexity in game theory,
► Diffusion of Technology including modelling bounded rationality, its role
► General urpose Technologies in mechanism design, and the problem of
► -iformation Ic .. ..gy I Wc: computing Nash equilibria; the price of anarchy,
1:1 that is, the cost of using decentralizing solution
to a problem; and interactions between
► Internet, Economics of the distributed computing and game theory.
► Technical Change

Bibliography
Bresnahan, T., and S. Greenstein. 1999. Technological
competition and the structure of the computer industry.
J o u r n a l o f I n d u s t r i a l E c o n o m i c s 47(1): 1-40.
Keywords
Algorithmic knowledge; Algorithmic mecha-
nism design; Bayesian networks; Bounded
rationality; Byzantine agreement; Cheap talk;
Computer Science and Game Theory 2001

Coalitions; Combinatorial auctions; Complexity solution to a problem. In section “

theory; Computational complexity; Computer ”, I consider interac-
science and game theory; Distributed tions between distributed computing and game
computing; Efficient representation of games; theory. In section “ ”,
Game theory; Gibbard-Satterthwaite th; I consider the problem of implementing mediators,
Implementing mediators; Interactive episte- which has been studied extensively in both
mology; ^-resilient equilibrium; (&,t)-robust computer science and game theory. I conclude in
equilibrium; Learning; Mechanism design; section “ ” with a discussion of a few
Markov networks; Nash equilibrium; Price of other topics of interest.
anarchy; Prisoner’s Dilemma; Regret; Strategic
voting; Tit for tat; Voting
Complexity Considerations

JEL Classifications
The influence of computer science in game theory
CO; C6 has perhaps been most strongly felt through com-
plexity theory. I consider some of the strands of this
research here. There are a numerous basic texts on
complexity theory that the reader can consult for
Introduction more background on notions like NP-completeness
and finite automata, including Hopcroft and Ullman
There has been a remarkable increase in work ( ) and Papadimitriou
at the interface of computer science and ( )•
game theory in the past decade. Game theory
forms a significant component of some major Bounded Rationality
computer science conferences (see, for example, One way of capturing bounded rationality is in
Kearns and Reiter ; Sandholm and Yokoo ); leading terms of agents who have limited computational
computer scientists are often invited to speak at power. In economics, this line of research goes back
major game theory conferences, such as the World to the work of Neyman ( ) and Rubin-
Congress on Game Theory 2000 and 2004. In this stein ( ), who focused
article I survey some of the main themes of work in on finitely repeated
the area, with a focus on the work in computer Prisoner’s Dilemma. In n-round finitely repeated
2 1
science. Given the length constraints, make no Prisoner’s Dilemma, there are 2 " strategies (since
attempt at being comprehensive, especially since a strategy is a function from histories to {cooperate,
other surveys are also available, including Halpem ( defect}, and there are clearly 2" 1
), Linial ( ), Papadimitriou histories of length < n). Finding a best response to a
( ), and a comprehensive survey book particular move can thus potentially be difficult.
(Nisanetal. ). Clearly people do not find best responses by doing
The survey is organized as follows. I look at the extensive computation. Rather, they typically rely
various roles of computational complexity in game on simple heuristics, such as ‘tit for tat’ (Axelrod ).
theory in section “ Such heuristics can often be captured by finite
”, including its use in modelling bounded automata; both Neyman and Rubinstein thus focus
rationality, its role in mechanism design, and the on finite automata playing repeated Prisoner’s
problem of computing Nash equilibria. In section ”, Dilemma. Two computer scientists, Papadimitriou
I consider a game- theoretic problem that originated and Yannakakis ( ), showed
in the computer science literature, but should be of that if both players in an n-round Prisoner’s
interest to the game theory community: computing Dilemma are finite automata with at least 2" 1
the price of anarchy, that is, the cost of using a states, then the only equilibrium is the one where
decentralizing they defect in every round. This result says that a
finite automaton with exponentially many states
2002
can compute best responses in Prisoner’s Dilemma.
We can then model bounded rationality by
restricting the number of states of the automaton.
Neyman ( ) showed, roughly speaking, that if
the two players in «-round Prisoner’s Dilemma are
modelled by finite automata with a number of states
in the interval \nyk, //'] for some k, then collaboration
can be approximated in equilibrium; more precisely,
if the payoff for (cooperate, cooperate) is (3, 3)
there is an equilibrium in the repeated game where
the average payoff per round is greater than 3 j. for
each player. Papadimitriou and Yannakakis ( )
sharpen
this result by showing that if at least one of the
players has fewer than 2c‘n states, where
= -p-j-pj—y, then for sufficiently large n, there
is an equilibrium where each player’s average pay-
off per round is greater than 3 — e. Thus, compu-
tational limitations can lead to cooperation in
Prisoner’s Dilemma.
There have been a number of other attempts to
use complexity-theoretic ideas from computer sci-
ence to model bounded rationality (see Rubinstein ,
for some exs). However, it seems that there is much
more work to be done here.
Computing Nash Equilibrium
Nash ( ) showed every finite game has a Nash
equilibrium in mixed strategies. But how hard is it
to actually find that equilibrium? On the positive
side, there are well known algorithms for computing
Nash equilibrium, going back to the classic Lemke-
Howson ( ) algorithm, with a spate of
recent improvements (see, for example, Govindan
and Wilson ; Blum et al. ; Porter et al. ). Moreover,
for certain classes of games (for example,
symmetric games,
Papadimitriou and Roughgarden ), there are known
to be polynomial-time algorithms. On the negative
side, many qsts about Nash equilibrium are known
to be NP-hard. For example, Gilboa and Zemel (
) showed that, for a game pre
sented in normal form, deciding whether there
exists a Nash equilibrium where each player gets a
payoff of at least r is NP-complete. Interestingly,
Gilboa and Zemel also show that computing
Computer Science and Game Theory
whether there exists a correlated equilibrium
(Aumann ) where each player gets a payoff of at
least r is computable in polynomial time. In general,
qsts regarding correlated equilibrium seem easier
than the analogous qsts for Nash equilibrium; see
Papadimitriou ( ) and
Papadimitriou and Roughgarden ( ) for fur-
ther examples. Chu and Halpem ( ) prove
similar NP-completeness results if the game is
represented in extensive form, even if all players
have the same payoffs (a situation that arises
frequently in computer science applications, where
we can view the players as agents of some designer,
and take the payoffs to be the designer’s payoffs).
Conitzer and Sandholm ( ) give a
compendium of hardness results for various qsts
regarding Nash equilibria.
Nevertheless, there is a sense in which it seems
that the problem of finding a Nash equilibrium is
easier than typical NP-complete problems, because
every game is guaranteed to have a Nash
equilibrium. By way of contrast, for a typical NP-
complete problem like prptal satisfiability, whether
or not a prptal formula is satisfiable is not known.
Using this observation, it can be shown that if
finding a Nash equilibrium is NP-complete, then NP
= coNP. Recent work has in a sense completely
characterized the complexity of finding a Nash
equilibrium in normal-form games: it is a PPAD-
complete problem (Chen and Deng ; Daskalis et
al. ). PPAD stands for ‘polynomial parity argument
(directed case)’; see Papadimitriou ( ) for a formal
definition and
examples of other PPAD problems. It is believed
that PPAD-complete problems are not solvable in
polynomial time, but are simpler than NP-complete
problems, although this remains an open problem.
See Papadimitriou ( ) for an
overview of this work.
Algorithmic Mechanism Design
The problem of mechanism design is to design a
game such that the agents playing the game, moti-
vated only by self-interest, achieve the designer’s
goals. This problem has much in common with the
standard computer science problem of designing
protocols that satisfy certain specifications (for
example, designing a distributed protocol that
Computer Science and Game Theory
achieves Byzantine agreement; see section “
”). Work on
mechanism design has traditionally ignored com-
putational concerns. But Kfir-Dahav et al. ( ) involved here.
show that, even in simple settings, optimizing social
welfare is NP-hard, so that perhaps the most
common approach to designing mechanisms,
applying the Vickrey-Groves-Clarke (VCG)
procedure (Clarke ; Groves ; Vickrey ), is not going
to work in large systems. We might hope that, even
if we cannot compute an optimal mechanism, we
might be able to compute a reasonable
approximation to it. However, as Nisan and Ronen (
, )
show, in general, replacing a VCG mechanism by
an approximation does not preserve truthfulness.
That is, even though truthfully revealing one’s type
is an optimal strategy in a VCG mechanism, it may
no longer be optimal in an approximation.
Following Nisan and Ronen’s work, there has been
a spate of papers either describing computationally
tractable mechanisms or showing that no
computationally tractable mechanism exists for a
number of problems, ranging from task allocation
(Archer and Tardos ; Nisan and Ronen ) to cost-
sharing for multicast trees (Feigenbaum et al. )
(where the problem
is to share the cost of sending, for example, a movie
over a network among the agents who actually want
the movie) to finding low-cost paths between nodes
in a network (Archer and Tardos ).
The problem that has attracted perhaps the most
attention is combinatorial auctions, where bidders
can bid on bundles of items. This becomes of
particular interest in situations where the value to a
bidder of a bundle of goods cannot be determined
by simply summing the value of each good in
isolation. To take a simple example, the value of a
pair of shoes is much higher than that of the
individual shoes; perhaps more interestingly, an
owner of radio stations may value having a licence
in two adjacent cities more than the sum of the
individual licences. Combinatorial auctions are of
great interest in a variety of settings including
spectrum auctions, airport time slots (that is, take-
off and landing slots), and industrial procurement.
There are many complexity-theoretic issues
2003
related to combinatorial auctions. For a detailed
discussion and references see Cramton et al. ( ); I
briefly discuss a few of the issues
Suppose that there are n items being auctioned.
Simply for a bidder to communicate her bids to the
auctioneer can take, in general, exponential time,
since there are 2" bundles. In many cases, we can
identify a bid on a bundle with the bidder’s
valuation of the bundle. Thus, we can try to care-
fully design a bidding language in which a bidder
can communicate her valuations succinctly. Simple
information-theoretic arguments can be used to
show that, for every bidding language, there will be
valuations that will require length at least 2" to
express in that language. Thus, the best we can
hope for is to design a language that can represent
the ‘interesting’ bids succinctly. See Nisan ( ) for
an overview of various bidding
languages and their expressive power.
Given bids from each of the bidders in a com-
binatorial auction, the auctioneer would like to then
determine the winners. More precisely, the
auctioneer would like to allocate the m items in an
auction so as to maximize his revenue. This prob-
lem, called the winner determination problem, is
NP-complete in general, even in relatively simple
classes of combinatorial auctions with only two
bidders making rather restricted bids. Moreover, it
is not even polynomial-time approximable, in the
sense that there is no constant d and polynomial-
time algorithm such that the algorithm produces an
allocation that gives revenue that is at least 1/d of
optimal. On the other hand, there are algorithms
that provably find a good solution, seem to work
well in practice, and, if they seem to be taking too
long, can be terminated early, usually with a good
feasible solution in hand. See Lehmann et al. ( ),
for an overview of the results in
this area.
In most mechanism design problems, compu-
tational complexity is seen as the enemy. There is
one class of problems in which it may be a friend:
voting. One problem with voting mechanisms is
that of manipulation by voters. That is, voters may
be tempted to vote strategically rather than ranking
the candidates according to their true preferences, in
the hope that the final outcome will be
2004
more favourable. This situation arises frequently in
practice; in the 2000 US presidential election,
American voters who preferred Nader to Gore to
Bush were encouraged to vote for Gore, rather than
‘wasting’ a vote on Nader. The classic Gibbard-
Satterthwaite theorem (Gibbard ; Satterthwaite )
shows that, if there are at least three alternatives,
then in any nondictatorial voting scheme (that is,
one where it is not the case that one particular voter
dictates the final outcome, irrespective of how the
others vote), there are preferences under which an
agent is better off voting strategically. The hope is
that, by constructing the voting mechanism
appropriately, it may be computationally intractable
to find a manipulation that will be beneficial. While
finding manipulations for the plurality protocol (the
candidate with the most votes wins) is easy, there
are well-known voting protocols for which manipu-
lation is hard in the presence of three or more
candidates. See Conitzer et al. ( ) for a sum
mary of results and further pointers to the literature.
Communication Complexity
Most mechanisms in the economics literature are
designed so that agents truthfully reveal their
preferences. However, in some settings, revealing
one’s full preferences can require a prohibitive
amount of communication. For example, in a
combinatorial auction of m items, revealing one’s
full preferences may require revealing what one
would be willing to pay for each of the 2"' 1 worst social welfare (assuming that the social
possible bundles of items. Even if in 30,
this requires revealing more than one billion num-
bers. This leads to an obvious qst: how much
communication is required by various mechanisms?
Formal work on this question in the economics
community goes back to Hurwicz (
and Mount and Reiter ( ); their definitions
focused on the dimension of the message space.
Independently (and later), there was active work in
computer science on communication complexity, the
number of bits of communication needed for a set
of n agents to compute the value of a function / :
xni=l 0, —> X , where each agent i knows 9j C 0,
(Think of 6t as representing agent fs type.) Recently
there has been an
Computer Science and Game Theory
explosion of work, leading to a better understanding
of the communication complexity for many
important economic allocation problems; see Segal
( ) for an overview. Two important
themes in this work are understanding the role of
price- based market mechanisms in solving allo-
cation problems with minimal communication, and
designing mechanisms that provide agents with
incentives to communicate truthfully while having
low communication requirements.
The Price of Anarchy
In a computer system, there are situations where we
may have a choice between a centralized and a
decentralized solution to a problem. By ‘central-
ized’ here, I mean that each agent in the system is
told exactly what to do and must do so; in the
decentralized solution, each agent tries to optimize
his own selfish interests. Of course, centralization
comes at a cost. For one thing, there is a problem of
enforcement. For another, centralized solutions tend
to be more vulnerable to failure. On the other hand,
a centralized solution may be more socially
beneficial. How much more beneficial can it be?
Koutsoupias and Papadimitriou ( ) formal
ized this question by considering the ratio of the
social welfare of the centralized solution to the
social welfare of the Nash equilibrium with the
welfare function is always positive). They called
this ratio the price of anarchy; and proved a number
of results regarding the price of anarchy for a
scheduling problem on parallel machines. Since the
original paper, the price of anarchy has been studied
) in many settings, including traffic routing
(Roughgarden and Tardos ), facility location games
(for example, where is the best place to put a
factory) (Vetta ), and spectrum sharing (how should
channels in a WiFi network be assigned)
(Halldorsson et al. ).
To give a sense of the results, consider the
traffic-routing context of Roughgarden and Tardos (
). Suppose that the travel time on a
road increases in a known way with the congestion
on the road. The goal is to minimize the
Computer Science and Game Theory
average travel time for all drivers. Given a road
network and a given traffic load, a centralized
solution would tell each driver which road to take.
For example, there could be a rule that cars with
odd-numbered licence plates take road 1, while
those with even-numbered plates take road 2, to
minimize congestion on either road. Roughgarden
and Tardos show that the price of anarchy is
unbounded if the travel time can be a nonlinear
function of the congestion. On the other hand, if it is
linear, they show that the price of anarchy is at most
4/3.
The price of anarchy is but one way of com-
puting the ‘cost’ of using a Nash equilibrium.
Others have been considered in the computer sci-
ence literature. For example, Tennenholtz ( ) ), and has been studied in detail since
compares the safety level of a game - the optimal
amount that an agent can guarantee himself, inde-
pendent of what the other agents do - to what the
agent gets in a Nash equilibrium, and shows, for
interesting classes of games, including load-
balancing games and first-price auctions, that the
ratio between the safety level and the Nash equi-
librium is bounded. For example, in the case of
first- price auctions, it is bounded by the constant e.
Game Theory and
Distributed Computing
Distributed computing and game theory are inter-
ested in much the same problems: dealing with
systems where there are many agents, facing
uncertainty and having possibly different goals. In
practice, however, there has been a significant
difference in emphasis between the two areas. In
distributed computing, the focus has been on
problems such as fault tolerance, asynchrony,
scalability, and proving correctness of algorithms;
in game theory, the focus has been on strategic
concerns. I discuss here some issues of common
interest. Most of the discussion in the remainder of
this section is taken from Halpem ( ).
To understand the relevance of fault tolerance
and asynchrony, consider the Byzantine agreement
problem, a paradigmatic problem in the distributed
systems literature. In this problem, there
2005
are assumed to be n soldiers, up to t of which may
be faulty (the t stands for traitor)', n and / are
assumed to be common knowledge. Each soldier
starts with an initial preference, to either attack or
retreat. (More precisely, there are two types of
nonfaulty agents - those that prefer to attack, and
those that prefer to retreat.) We want a protocol that
guarantees that (1) all nonfaulty soldiers reach the
same decision, and (2) if all the soldiers are
nonfaulty and their initial preferences are identical,
then the final decision agrees with their initial
preferences. (The condition simply prevents the
obvious trivial solutions, where the soldiers attack
no matter what, or retreat no matter what.)
The problem was introduced by Pease et al. (
then; Chor and Dwork ( ), Fischer ( ), and
Linial ( ) provide overviews. Whether the
Byzantine agreement problem is solvable depends
in part on what types of failures are considered, on
whether the system is synchronous or asynchro-
nous, and on the ratio of n to t. Roughly speaking, a
system is synchronous if there is a global clock and
agents move in lockstep; a ‘step’ in the system
corresponds to a tick of the clock. In an asynchro-
nous system, there is no global clock. The agents in
the system can mn at arbitrary rates relative to each
other. One step for agent 1 can correspond to an
arbitrary number of steps for agent 2 and vice versa.
Synchrony is an implicit assumption in essentially
all games. Although it is certainly possible to model
games where player 2 has no idea how many moves
player 1 has taken when player 2 is called upon to
move, it is not typical to focus on the effects of
synchrony (and its lack) in games. On the other
hand, in distributed systems, it is typically a major
focus.
Suppose for now that we restrict to crash fail-
ures, where a faulty agent behaves according to the
protocol, except that it might crash at some point,
after which it sends no messages. In the round in
which an agent fails, the agent may send only a
subset of the messages that it is supposed to send
according to its protocol. Further suppose that the
system is synchronous. In this case, the following
rather simple protocol achieves Byzantine
agreement:
2006
• In the first round, each agent tells every other
agent its initial preference.
• For rounds 2 to t + 1, each agent tells every
other agent everything it has heard in the pre-
vious round. Thus, for example, in round 3,
agent 1 may tell agent 2 that it heard from agent
3 that its initial preference was to attack, and
that it (agent 3) heard from agent 2 that its initial
preference was to attack, and it heard from agent
4 that its initial preferences was to retreat, and
so on. This means that messages get
exponentially long, but it is not difficult to
represent this information in a compact way so
that the total communication is polynomial in n,
the number of agents.
• At the end of round t+ 1, if an agent has heard
from any other agent (including itself) that its
initial preference was to attack, it decides to
attack; otherwise, it decides to retreat.
Why is this correct? Clearly, if all agents are
correct and want to retreat (resp., attack), then the
final decision will be to retreat (resp., attack), since
that is the only preference that agents hear about
(recall that for now we are considering only crash
failures). It remains to show that if some agents
prefer to attack and others to retreat, then all the
nonfaulty agents reach the same final decision. So
suppose that i and j are nonfaulty and i decides to
attack. That means that i heard that some agent’s
initial preference was to attack. If it heard this first
at some round t' <t + 1, then i will forward this
message to j, who will receive it and thus also
attack. On the other hand, suppose that i heard it
first at round t + 1 in a message from . Thus, this
message must be of the form % said at round t
that... that 12 said at round 2 that /'/ said at round 1
that its initial preference was to attack. ’ Moreover,
the agents i\v ..., /, + 1 must all be distinct. Indeed, it
is easy to see that ik must crash in round k before
sending its message to i (but after sending its
message to /'/, ,), for
k 1............/, for otherwise i must have gotten the
message from ik, contradicting the assumption that i
first heard at round t + 1 that some agent’s initial
preference was to attack. Since at most t agents can
crash, it follows that /, ,, the agent that sent the
message to i, is not faulty, and thus
Computer Science and Game Theory
sends the message to j. Thus, j also decides to attack.
A symmetric argument shows that if j decides to
attack, then so does i.
It should be clear that the correctness of this
protocol depends on both the assumptions made:
crash failures and synchrony. Suppose instead that
Byzantine failures are allowed, so that faulty agents
can deviate in arbitrary ways from the protocol; they
may Tie’, send deceiving messages, and collude to
fool the nonfaulty agents in the most malicious
ways. In this case, the protocol will not work at all.
In fact, it is known that agreement can be reached in
the presence of Byzantine failures iff t < n/3, that is,
iff fewer than a third of the agents can be faulty
(Pease et al. ). The effect of asynchrony is even
more devastating: in an asynchronous system, it is
impossible to reach agreement using a deterministic
protocol even if t = 1 (so that there is at most one
failure) and only crash failures are allowed (Fischer
et al. ). The problem in the asynchronous setting is
that if none of the agents have heard from, say,
agent 1, they have no way of knowing whether agent
1 is faulty or just slow. Interestingly, there are
randomized algorithms (that is, behavioural
strategies) that achieve agreement with arbitrarily
high probability in an asynchronous setting (Ben-
Or ; Rabin ).
Byzantine agreement can be viewed as a game
where, at each step, an agent can either send a
message or decide to attack or retreat. It is essen-
tially a game between two teams, the nonfaulty
agents and the faulty agents, whose composition is
unknown (at least by the correct agents). To model it
as a game in the more traditional sense, we could
imagine that the nonfaulty agents are playing against
a new player, the ‘adversary’. One of the
adversary’s moves is that of ‘corrupting’ an agent:
changing its type from ‘nonfaulty’ to ‘faulty.’ Once
an agent is corrupted, what the adversary can do
depends on the failure type being considered. In the
case of crash failures, the adversary can decide
which of a corrupted agent’s messages will be
delivered in the round in which the agent is
corrupted; however, it cannot modify the messages
themselves. In the case of Byzantine failures, the
adversary essentially gets to make the moves for
agents that have been
Computer Science and Game Theory
corrapted; in particular, it can send arbitrary
messages.
Why has the distributed systems literature not
considered strategic behaviour in this game? Crash
failures are used to model hardware and software
failures; Byzantine failures are used to model
random behaviour on the part of a system (for
example, messages getting garbled in transit),
software errors, and malicious adversaries (for
example, hackers). With crash failures, it does not
make sense to view the adversary’s behaviour as
strategic, since the adversary is not really viewed as
having strategic interests. While it would certainly
make sense, at least in principle, to consider the
probability of failure (that is, the probability that
the adversary corrupts an agent), this approach has
by and large been avoided in the literature because
it has proved difficult to characterize the probability
distribution of failures over time. Computer
components can perhaps be characterized as failing
according to an exponential distribution (see
Babaoglu , for an analysis of Byzantine agreement
in such a setting), but crash failures can be caused
by things other than component failures (faulty
software, for ex); these can be extremely difficult to
characterize probabilistically. The problems are
even worse when it comes to modelling random
Byzantine behaviour.
With malicious Byzantine behaviour, it may
well be reasonable to impute strategic behaviour to
agents (or to an adversary controlling them).
However, it is often difficult to characterize the
payoffs of a malicious agent. The goals of the
agents may vary from that of simply trying to delay
a decision to that of causing disagreement. It is not
clear what the appropriate payoffs should be for
attaining these goals. Thus, the distributed systems
literature has chosen to focus instead on algorithms
that are guaranteed to satisfy the specification
without making assumptions about the adversary’s
payoffs (or nature’s probabilities, in the case of
crash failures).
Recently, there has been some work on adding
strategic concerns to standard problems in distrib-
uted computing; see, for example, Alvisi et al. ( )
and Halpem and Teague ( ).
Moving in the other direction, there has also been
some work on adding concerns of fault
2007
tolerance and asynchrony to standard problems in
game theory; see, for example, Eliaz ( ),
Monderer and Tennenholtz ( , ) and the
definitions in the next section. This seems to be an
area that is ripe for further developments. One such
development is the subject of the next section.
Implementing Mediators
The question of whether a problem in a multiagent
system that can be solved with a trusted mediator
can be solved by just the agents in the system,
without the mediator, has attracted a great deal of
attention in both computer science (particularly in
the cryptography community) and game theory. In
cryptography, the focus on the problem has been on
secure multiparty computation. Here it is assumed
that each agent i has some private information xt.
Fix functions//, ..., /n. The goal is to have agent i
learn f(xu ..., x,j without learning anything about Xj
for j / i beyond what is revealed by the value off
(xj ..., x„). With a trusted mediator, this is trivial:
each agent i just gives the mediator its private value
x, the mediator then sends each agent i the value
f(xi, ..., x„). Work on multiparty computation
(Goldreich et al. ; Shamir et al. ; Yao ) provides
conditions under which this can be done. In game
theory, the focus has been on whether an
equilibrium in a game with a mediator can be
implemented using what is called cheap talk - that
is, just by players communicating among
themselves (cf. Barany ; Ben-Porath ; Forges ;
Gerardi ; Heller ; Urbano and Vila ). As suggested
in the previous section, the focus in the computer
science literature has been in doing multiparty
computation in the presence of possibly malicious
adversaries, who do everything they can to subvert
the computation, while in the game theory literature
the focus has been on strategic agents. In recent
work, Abraham et al. ( ) and Abraham
et al. ( ) considered deviations by both ratio
nal players, who have preferences and try to max-
imize them, and players who can viewed as
malicious, although it is perhaps better to think
2008 Computer Science and Game Theory

of them as rational players whose utilities are not like Kazaa or Gnutella, it would seem that no
known by the other players or mechanism designer. rational agent should share files. Whether or not
I briefly sketch their results here; the following you can get a file depends only on whether other
discussion is taken from Abraham et al. ( ). people share files. Moreover, there are disincentives
The idea of tolerating deviations by coalitions of for sharing (the possibility of lawsuits, use of
players goes back to Aumann ( ); more bandwidth, and so on). Nevertheless, people do
recent refinements have been considered by Moreno share files. However, studies of the Gnutella net-
and Wooders ( ). Aumann’s defini work have shown almost 70 per cent of users share
tion is essentially the following. no files and nearly 50 per cent of responses are
from the top one per cent of sharing hosts (Adar and
Definition 1 is a k-resilient ’ equilibrium if, for Huberman ).
all sets C of players with | C | < k, it is not the One reason that people might not respond as we
case that there exists a strategy x=* ■ such that expect is that they have utilities that are different
ufx^ c, <7^—c) > for all i G C . from those we expect. Alternatively, the players
As usual, the strategy (x c, a - c) is the one may be irrational, or (if moves are made using a
where each player i e C plays x, and each player i computer) they may be playing using a faulty
C plays c r A s the prime notation suggests, this is computer and thus not able to make the move they
not quite the definition we want to work with. The would like, or they may not understand how to get
trouble with this definition is that it suggests that the computer to make the move they would like.
coalition members cannot communicate with each Whatever the reason, it seems important to design
other during the game. Perhaps surprisingly, strategies that tolerate such unanticipated
allowing communication can prevent certain equi- behaviours, so that the payoffs of the users with
libria (see Abraham et al. , for an ex). Since we ‘standard’ utilities do not get affected by the
should expect coalition members to communicate, nonstandard players using different strategies. This
the following definition seems to capture a more can be viewed as a way of adding fault tolerance to
reasonable notion of resilient equilibrium. Let the equilibrium notions.
cheap-talk extension of a game T be, roughly
speaking, the game where players are allowed to Definition 3 A joint strategy cr > is t-immune if,
communicate among themselves in addition to for all T C N with |T| < t, all joint strategies T=>, and
performing the actions of F and the payoffs are just all i £ T, we have u,{a x - i ) > w,(cr=').
as in F. The notion of /-immunity and F-resilienee
address different concerns. For t immunity, we
Definition 2 is a k-resilient equilibrium in a game consider the payoffs of the players not in T, and
I if a - is a k-resilient' equilibrium in the cheap-talk require that they are not worse due to deviation; for
extension of F (where we identify the strategy at in resilience, we consider the payoffs of players in C,
the game F with the strategy in the cheap-talk game and require that they are not better due to deviation.
where player i never sends any messages beyond It is natural to combine both notions. Given a game
those sent according to cr,). F, let I ’.be the game that is identical to r except that
A standard assumption in game theory is that the players in T are fixed to playing strategy x.
utilities are (commonly) known; when we are given
a game we are also given each player’s utility. Definition 4 o ' is a (k, t) -robust equilibrium if is
When players make decisions, they can take other t-immune and, for all T C N such that | T\ < t and
players’ utilities into account. However, in large all joint strategies x , a - _T is a ^-resilient strategy
systems it seems almost invariably the case that of I ’) .
there will be some fraction of users who do not To state the results of Abraham et al. ( )
respond to incentives the way we expect. For and ( ) on implementing mediators, three
example, in a peer-to-peer network games need to be considered: an underlying
Computer Science and Game Theory 2009

game I , an extension T^of T with a mediator, and a • Ifn < 2k + 21, then mediators cannot, in general,
cheap-talk extension Tct of 1 ’. Assume that 1 is a be e-implemented, even with broadcast
normal-form Bayesian game: each player has a type channels. Moreover, even assuming cryptogra-
from some type space with a known distribution phy and polynomially bounded players, the
over types, and the utilities of the agents depend on expected running time of an implementation
the types and actions taken. Roughly speaking, a depends on the utility functions of the players
cheap talk game implements a game with a mediator and e.
if it induces the same distribution over actions in • Ifn S' k ' t, then, assuming cryptography and
the underlying game, for each type vector of the polynomially bounded players, mediators can be
players. With this background, I can summarize the e-implemented using cheap talk, but if n# 2k +
results of Abraham et al. ( ) 2t, then the running time depends on the utilities
and ( ). in the game and s.
• If n <k+3t, then even assuming cryptography,
• If n > 3 k + 31, a (k, t) -robust strategy a=’ with polynomially bounded players, and a (k + f)
a mediator can be implemented using cheap talk -punishment strategy, mediators cannot, in
(that is, there is a (k, /(-robust strategy a=’' in a general, be e -implemented using cheap talk.
cheap talk game such that <r > and a=’' induce • Ifn > k + t, then, assuming cryptography,
the same distribution over actions in the under- polynomially bounded players, and a public- key
lying game). Moreover, the implementation infrastructure (PKI), we can e-implement a
requires no knowledge of other agents’ utilities, mediator.
and the cheap talk protocol has bounded running
time that does not depend on the utilities. The proof of these results makes heavy use of
• If/; < 3/c — 3/, then, in general, mediators techniques from computer science. All the possibil-
cannot be implemented using cheap talk without ity results showing that mediators can be
knowledge of other agents’ utilities. Moreover, implemented use techniques from secure multiparty
even if other agents’ utilities are known, medi- computation. The results showing that if n < 3k +
ators cannot, in general, be implemented with- 3t, then we cannot implement a mediator without
out having a (k +1)-punishment strategy (that is, knowing utilities, and that, even if utilities are
a strategy that, if used by all but at most (k + I ) known, a punishment strategy is required, use file
players, guarantees that every player gets a fact that Byzantine agreement cannot be reached if t
worse outcome than they do with the equilib- < n/3; the impossibility result for n < 2k - 3t also
rium strategy) nor with bounded running time. uses a variant of Byzantine agreement.
• If n > 2k + 3t, then mediators can be A related line of work considers implementing
implemented using cheap talk if there is a mediators assuming stronger primitives (which
punishment strategy (and utilities are known) in cannot be implemented in computer networks); see
finite expected running time that does not Izmalkov et al. ( ) and Lepinski et al.
depend on the utilities. ( ) for details.
• lfn#2k+3t then med i ators cannot, in general, be
implemented, even if there is a punishment
strategy and utilities are known. Other Topics
• If n > 2k + 2t and there are broadcast channels
There are many more areas of interaction between
then, for all e, mediators can be e -implemented
computer science than I have indicated in this brief
(intuitively, there is an implementation where
players get utility within e of what they could survey. I briefly mention a few others here.
get by deviating) using cheap talk, with Interactive Epistemology
bounded expected running time that does not Since the publication of Aumann’s ( ) seminal
depend on the utilities. paper, there has been a great deal of activity in
2010
trying to understand the role of knowledge in
games, and providing epistemic analyses of solution
concepts; see Battigalli and Bonanno (
for a survey. In computer science, there has been a
parallel literature applying epistemic logic to reason
about distributed computation. One focus of this
work has been on characterizing the level of
knowledge needed to solve certain problems. For
example, to achieve Byzantine agreement common
knowledge among the nonfaulty agents of an initial
value is necessary and sufficient. More generally, in
a precise sense, common knowledge is necessary
and sufficient for coordination. Another focus has
been on defining logics that capture the reasoning of
resource-bounded agents. A number of approaches
have been considered. Perhaps the most common
considers logics for reasoning about awareness,
where an agent may not be aware of certain
concepts, and can know something only if he is
aware of it. This topic has been explored in both
computer science and game theory; see Dekel et al.
( ), Fagin and Halpem
( ), Halpem ( ), Halpem and Rego
( ), Heifetz et al. ( ), and Modica and
Rustichini ( , ) for some of the work in
this active area. Another approach, so far considered
only by computer scientists, involves algorithmic
knowledge, which takes seriously the assumption
that agents must explicitly compute what they
know. See Fagin et al. ( ) for an
overview of the work in epistemic logic in computer
science.
Network Growth
If we view networks as being built by selfish
players (who decide whether or not to build links),
what will the resulting network look like? How does
the growth of the network affect its functionality?
For example, how easily will influence spread
through the network? How easy is it to route traffic?
See Fabrikant et al. ( ) and
Kempe et al. ( ) for some recent computer
science work in this burgeoning area.
Efficient Representation of Games
Game theory has typically focused on ‘small’
games, often two- or three-player games, that are
easy to describe, such as Prisoner’s Dilemma, in
Computer Science and Game Theory
order to understand subtleties regarding basic issues
such as rationality. To the extent that game theory is
) used to tackle larger, more practical problems, it
will become important to find efficient techniques
for describing and analysing games. By way of
analogy, 2n — 1 numbers are needed to describe a
probability distribution on a space characterized by
n binary random variables. For n 100 (not an
unreasonable number in practical situations), it is
impossible to write down the probability
distribution in the obvious way, let alone do
computations with it. The same issues will surely
arise in large games. Computer scientists use
graphical approaches, such as Bayesian networks
and Markov networks (Pearl ), for representing and
manipulating probability measures on large spaces.
Similar techniques seem applicable to games; see,
for example, Kearns et al. ( ), Roller and Milch
( ), and La Mura ( ) for specific
approaches, and Kearns ( ) for a recent over
view. Note that representation is also an issue when
we consider the complexity of problems such as
computing Nash or correlated equilibria. The
complexity of a problem is a function of the size of
the input, and the size of the input (which in this
case is a description of the game) depends on how
the input is represented.
Learning in Games
There has been a great deal of work in both com-
puter science and game theory on learning to play
well in different settings (see Fudenberg and Levine
, for an overview of the work in game theory). One
line of research in computer science has involved
learning to play optimally in a rcinforccmcnt-
lcaming setting, where an agent interacts with an
unknown (but fixed) environment. The agent then
faces a fundamental tradeoff between exploration
and exploitation. The question is how long it takes
to learn to play well (to get a reward within some
fixed e of optimal); see Brafman and Tennenholtz (
) and Kearns
and Singh ( ) for the current state of the art.
A related question is efficiently finding a strategy
minimizes regret - that is, finding a strategy that is
guaranteed to do not much worse than the best
strategy would have done in hindsight (that is,
Computer Science and Game Theory
even knowing what the opponent would have done).
See Blum and Mansour ( ) for a recent
overview of work on this problem.
See Also
► Computation oi
► Computational Methods it Econometrics
► Computing in Mechanism E 1 sign
► )ata Mining
► ilectronic Commerce
► ipistemic Game Theory: An Ov TO V
► istemic Game Theory: E ^ ? =
► Mathematics of Networks
► Mechanism Design (New Developments)
► Rationality, Bounded
► /oting P aradoxes
Acknowledgment The work for this article was supported in
part by NSF under grants CTC-0208535 and ITR-0325453,
by ONR under grant N00014-02-1- 0455, by the DoD
Multidisciplinary University Research Initiative (MUR1)
program administered by the ONR under grants N00014-01-
1-0795 andN00014-04-1-0725, and by AFOSR under grant
F49620-02-1-0101. Thanks to Larry Blume, Christos
Papadimitriou, Ilya Segal, Eva Tardos, and Moshe
Tennenholtz for useful comments.
Bibliography
Abraham, I., D. Dolev, R. Gonen, and J. Halpem. 2006.
Distributed computing meets game theory: Robust
mechanisms for rational secret sharing and multiparty
computation. In Proceedings of the 25 th ACM symposium
on principles of distributed computing, 53-62. Abraham, I.,
Dolev, D., and Halpem, J. 2007. Lower bounds on
implementing robust and resilient mediators. Available at
V/:; Ev crg/abs/CTE 35'5. Accessed 19 June 2007.
Adar, E., and B. Huberman. 2000. Free riding on Gnutella.
First Monday 5(10).
Alvisi, L., A.S. Ayer, A. Clement, M. Dahlin, J.P. Martin, and
C. Porth. 2005. BAR fault tolerance for cooperative services.
In Proceedings of the 20th ACM symposium on operating
systems principles (SOSP 2005), 45-58. Archer, A., and E.
Tardos. 2001. Truthful mechanisms for one-parameter
agents. In Proceedings of the 42nd IEEE symposium on
foundations of computer science, 482-491.
Archer, A., and E. Tardos. 2002. Frugal path mechanisms. In
Proceedings of the 13th ACM-SIAM symposium on
discrete algorithms, 991-999.
2011
Aumann, R. 1959. Acceptable points in general cooperative n-
person games. In Contributions to the theory of games,
ed. A. Tucker and R. Luce, vol. TV, 287-324. Princeton:
Princeton University Press.
Aumann, R.J. 1976. Agreeing to disagree. Annals of Sta-
tistics 4: 1236-1239.
Aumann, R.J. 1987. Correlated equilibrium as an expression
of Bayesian rationality. Econometrica 55: 1-18.
Axelrod, R. 1984. The evolution of cooperation. New York:
Basic Books.
Babaoglu, O. 1987. On the reliability of consensus-based fault-
tolerant distributed computing systems. ACM
Translation on Computer Systems 5: 394-416.
Barany, I. 1992. Fair distribution protocols or how the players
replace fortune. Mathematics of Operations Research
17: 327-340.
Battigalli, P., and G. Bonanno. 1999. Recent results on belief,
knowledge and the epistemic foundations of game theory.
Research in Economics 53: 149-225.
Ben-Qr, M. 1983. Another advantage of free choice:
Completely asynchronous agreement protocols. I n Pro-
ceedings of the second ACM symposium on principles
of distributed computing, 27-30.
Ben-Porath, E. 2003. Cheap talk in games with incomplete
information. Journal of Economic Theory 108: 45-71.
Blum, A., and Y. Mansour. 2007. Learning, regret minimi-
zation, and equilibria. In Algorithmic game
theory, ed. N. Nisan, T. Roughgarden, E. Tardos, and V.
Vazirani. Cambridge: Cambridge University Press.
Blum, B., C.R. Shelton, and D. Roller. 2003. A continuation
method for Nash equilibria in structured games. In
Proceedings of the 18th international joint conference
on artificial intelligence (1JCA1 ‘03), 757-764.
Brafman, R.I., and M. Tennenholtz. 2002. R-MAX: A general
polynomial time algorithm for near-optimal
reinforcement learning. Journal of Machine Learning
Research 3: 213-231.
Chen, X., and X. Deng. 2006. Settling the complexity of 2-
player Nash equilibrium. In Proceedings of the 47th
IEEE symposium on foundations of computer science,
261-272.
Chor, B., and C. Dwork. 1989. Randomization in Byzantine
agreement. In Advances in computing research 5:
Randomness and computation, ed. S. Micali, 443-497.
Greenwich: JAI Press.
Chu, F., and J.Y. Halpem. 2001. A decision-theoretic
approach to reliable message delivery. Distributed
Computing 14: 359-389.
Clarke, E.H. 1971. Multipart pricing of public goods. Public
Choice 11: 17-33.
Conitzer, V., and T. Sandhokn. 2003. Complexity results about
Nash equilibria. In Proceedings of the 18 th international
joint conference on artificial intelligence {IJCAI ‘03),
765-771.
Conitzer, V., T. Sandholm, and J. Lang. 2007. When are
elections with few candidates hard to manipulate?
Journal of the ACM 54(3), Article 14.
2012 Computer Science and Game Theory

Cramton, P., Y. Shoham, and R. Steinberg. 2006. Combi- Halpem, J.Y. 2003. A computer scientist looks at game theory.
natorial auctions. Cambridge, MA: MIT Press. Games and Economic Behavior 45: 114-132.
Daskalis, C., P.W. Goldberg, and C.H. Papadimitriou 2006. Halpem, J.Y., and L.C. Rego. 2007. Reasoning about
The complexity of computing a Nash equilibrium. In knowledge of unawareness. Games and Economic
Proceedings of the 38th ACM symposium on theory of Behavior. Also available at
computing, 71-78. . Accessed 24 June 2007.
Dekel, E., B. Lipman, and A. Rustichini. 1998. Standard state- Halpem, J.Y., and V. Teague. 2004. Rational secret sharing
space models preclude unawareness. Econometrica 66: and multiparty computation: Extended abstract. In Pro-
159-173. ceedings of the 36tli ACM symposium on theory of
Eliaz, K. 2002. Fault-tolerant implementation. Review of computing, 623-632.
Economic Studies 69: 589-610. Heifetz, A., M. Meier, and B. Schipper. 2006. Interactive
Fabrikant, A., A. Luthra, E. Maneva, C.H. Papadimitriou, unawareness. Journal of Economic Theory 130: 78-94.
and S. Shenker. 2003. On a network creation game. In Heller, Y. 2005. A minority-proof cheap-talk protocol.
Proceedings of the 22nd ACM symposium on Unpublished manuscript.
principles of distributed computing, 347-351. Hopcrofl, J.E., and J.D. Ullman. 1979. Introduction to
Fagin, R., and J.Y. Halpem. 1988. Belief, awareness, and automata theory, languages and computation. New
limited reasoning. Artificial Intelligence 34: 39-76. York: Addison-Wesley.
Fagin, R., J.Y. Halpem, Y. Moses, and M.Y. Vardi. 1995. Hurwicz, L. 1977. On the dimensional requirements of
Reasoning about knowledge. Cambridge, MA: MIT informationally decentralized Pareto satisfactory pro-
Press A slightly revised paperback version was published cesses. In Studies in Resource Allocation Processes, ed.
in 2003. K.J. Arrow and L. Hurwicz, 413-424. New York:
Feigenbaum, J., C. Papadimitriou, and S. Shenker. 2000. Cambridge University Press.
Sharing the cost of muliticast transmissions (preliminary Izmalkov, S., S. Micali, and M. Lepinski. 2005. Rational
version). In Proceedings of the 32nd ACM symposium secure computation and ideal mechanism design. In
on theory of computing, 218-227. Proceedings of the 46th IEEE symposium foundations
Fischer, M.J. 1983. The consensus problem in unreliable of computer science, 585-595.
distributed systems. In Foundations of computation Kearns, M. 2007. Graphical games. In Algorithmic game
theory, vol. 185, ed. M. Karpinski. Lecture Notes in theory, ed. N. Nisan, T. Roughgarden, E. Tardos, and V.
Computer Science. Berlin/New York: Springer, Vazirani. Cambridge: Cambridge University Press.
pp. 127-140. Kearns, M.J., and M.K. Reiter. 2005. Proceedings of the
Fischer, M.J., N.A. Lynch, and M.S. Paterson. 1985. sixth ACM conference on electronic commerce (EC
Impossibility of distributed consensus with one faulty ‘05). New York: ACM. Table of contents available at
processor. Journal of the ACM 32: 374-382. . Accessed 19 June 2007.
Forges, F. 1990. Universal mechanisms. Econometrica 58: Kearns, M., and S.P. Singh. 1998. Near-optimal reinforcement
1341-1364. learning in polynomial time. In Proceedings of the 15th
Fudenberg, D., and D. Levine. 1998. The theory of learning international conference on machine learning. 260-268.
in games. Cambridge, MA: MIT Press. Kearns, M., M.L. Littman, and S.P. Singh. 2001. Graphical
Gerardi, D. 2004. Unmediated communication in games with models for game theory. In Proceedings of the 17th
complete and incomplete information. Journal of conference on uncertainty in artificial intelligence
Economic Theory 114: 104-131. (UA12001), 253-260.
Gibbard, A. 1973. Manipulation of voting schemes. Kempe, D., J. Kleinberg, andE. Tardos. 2003. Maximizing the
Econometrica 41: 587-602. spread of influence through a social network. In
Gilboa, I., and E. Zemel. 1989. Nash and correlated equi- Proceedings of the ninth ACM S1GKDD international
librium: Some complexity considerations. Games and conference on knowledge discoveiy and data mining, 137-
Economic Behavior 1: 80-93. 146.
Goldreich, O., S. Micali, and A. Wigderson. 1987. How to play Kftr-Dahav, N.E., D. Monderer, and M. Tennenholtz 2000.
any mental game. In Proceedings of the 19th ACM Mechanism design for resource-bounded agents. In Inter-
symposium on theory of computing, 218-229. national conference on multiagent systems, 309-316.
Govindan, S., and R. Wilson. 2003. A global Newton method Roller, D., and B. Milch. 2001. Structured models for
to compute Nash equilibria. Journal of Economic multiagent interactions. In Theoretical aspects of ratio-
Theory 110: 65-86. nality and knowledge: Proceedings of the eighth con-
Groves, T. 1973. Incentives in teams. Econometrica 41: 617- ference (PARK 2001), 233-248.
631. Koutsoupias, E., and C.H. Papadimitriou. 1999. Worst- case
Halldorsson, M.M., J.Y. Halpem, L. Li, and V. Mirrokni. equilibria, hr Proceedings of the 16th conference on
2004. On spectrum sharing games. In Proceedings of the theoretical aspects of computer science, vol. 1563,
23rd ACM symposium on principles of distributed Lecture Notes in Computer Science. Berlin: Springer, pp.
computing, 107-114. 404^413.
Halpem, J.Y. 2001. Alternative semantics for unawareness.
Games and Economic Behavior 37: 321-339.
Computer Science and Game Theory
Kushilevitz, E., andN. Nisan. 1997. C o m m u n i c a t i o n
c o m p l e x i t y . Cambridge: Cambridge University Press.
La Mura, P. 2000. Game networks. In Proceedings of the 16th
conference on uncertainty in artificial intelligence
(UAI2000), 335-342.
Lehmann, D., R. Muller, and T. Sandholm. 2006. The winner
determination problem. In Combinatorial
a u c t i o n s , ed. P. Cramton, Y. Shoham, and R.
Steinberg. Cambridge, MA: MIT Press.
Lemke, C.E., and J.J.T. Howson. 1964. Equilibrium points of
bimatrix games. J o u r n a l o f t h e S o c i e t y f o r
I n d u s t r i a l a n d A p p l i e d M a t h e m a t i c s 12:
413—423.
Lepinski, M., S. Micali, C. Peikert, and A. Shelat. 2004.
Completely fair SFE and coalition-safe cheap talk. In
Proceedings of the 23rd ACM symposium
p r i n c i p l e s o f d i s t r i b u t e d c o m p u t i n g , 1-10.
Linial, N. 1994. Games computers play: Game-theoretic
aspects of computing. In H a n d b o o k o f g a m e
t h e o r y , ed. R.J. Aumann and S. Hart, vol. 2. Amster-
dam: North-Holland.
Modica, S., and A. Rustichini. 1994. Awareness and
partitional information structures. T h e o r y a n d
D e c i s i o n 37: 107-124.
Modica, S., and A. Rustichini. 1999. Unawareness and
partitional information structures. G a m e s a n d E c o -
n o m i c B e h a v i o r 27: 265-298.
Monderer, D., and M. Tennenholtz. 1999a. Distributed games.
G a m e s a n d E c o n o m i c B e h a v i o r 28: 55-72.
Monderer, D., and M. Tennenholtz. 1999b. Distributed games:
From mechanisms to protocols. In P r o c e e d i n g s o f
the 16th national conference on
a r t i f i c i a l i n t e l l i g e n c e (AAAI '99), 32-37.
Moreno, D., and I. Wooders. 1996. Coalition-proof equi-
librium. Games a n d E c o n o m i c B e h a v i o r 17: 80-
112.
Mount, K.., and S. Reiter. 1974. The informational size of
message spaces. Journal of Economic Theory 8: 161-192.
Nash, J. 1950. Equilibrium points in n-person games. Pro-
ceedings of the National Academy of Sciences 36: 48^
Neyman, A. 1985. Bounded complexity justifies cooperation in
finitely repeated Prisoner’s Dilemma. Economic Letters
19: 227-229.
Nisan, N. 2006. Bidding languages for combinatorial auctions.
In Combinatorial auctions. Cambridge, MA: MIT Press.
Nisan, N., and A. Ronen. 2000. Computationally feasible VCG
mechanisms. In Proceedings of the second ACM
conference on electronic commerce (EC '00), 242-252.
Nisan, N., and A. Ronen. 2001. Algorithmic mechanism
design. Games and Economic Behavior 35: 166-196.
Nisan, N., T. Roughgarden, E. Tardos, and V. Vazirani. 2007.
Algorithmic game theory. Cambridge: Cambridge
University Press.
Papadimitriou, C.H. 1994a. Computational complexity.
Reading: Adison-Wesley.
Papadimitriou, C.H. 1994b. On the complexity of the parity
argument and other inefficient proofs of existence. Jour-
nal of Computer and System Sciences 48: 498-532.
2013
Papadimitriou, C.H. 2001. Algorithms, games, and the
internet. In P r o c e e d i n g s o f t h e 3 3 r d A C M
s y m p o s i u m o n t h e o r y o f c o m p u t i n g , 749-
753.
Papadimitriou, C.H. 2005. Computing correlated equilibria in
multiplayer games. In P r o c e e d i n g s o f t h e 3 7 t h
ACM symposium on theory of computing,
49-56.
Papadimitriou, C.H. 2007. The complexity of finding Nash
equilibria. In A l g o r i t h m i c g a m e t h e o r y , ed. N.
Nisan, T. Roughgarden, E. Tardos, and V. Vazirani. Cam-
bridge: Cambridge University Press.
Papadimitriou, C.H., and T. Roughgarden. 2005. Computing
equilibria in multi-player games. In P r o c e e d i n g s o f
the 16th ACM-SIAM symposium on
d i s c r e t e a l g o r i t h m s , 82-91.
Papadimitriou, C.H., and M. Yannakakis. 1994. On com-
plexity as bounded rationality. In P r o c e e d i n g s o f
the 26th ACM symposium on theory of
c o m p u t i n g , 726-733.
Pearl, J. 1988. Probabilistic reasoning in intelligent systems.
San Francisco: Morgan Kaufmann.
Pease, M., R. Shostak, and L. Lamport. 1980. Reaching
agreement in the presence of faults. J o u r n a l o f t h e
A C M 27: 228-234.
Porter, R., E. Nudelman, and Y. Shoham. 2004. Simple search
methods for finding a Nash equilibrium. In P r o -
ceedings of the 21st national conference
o n a r t i f i c i a l i n t e l l i g e n c e ( A A A I ' 0 4 ) , 664-
669.
Rabin, M.O. 1983. Randomized Byzantine generals. In
Proceedings of the 24th IEEE symposium on foundations of
computer science, 403—409.
Roughgarden, T., and E. Tardos. 2002. How bad is selfish
routing? J o u r n a l o f t h e A C M 49: 236-259.
Rubinstein, A. 1986. Finite automata play the repeated
prisoner’s dilemma. J o u r n a l of Economic
T h e o r y 39: 83-96.
Rubinstein, A. 1998. Modeling bounded
r a t i o n a l i t y . Cambridge, MA: MIT Press.
Sandholm, T., and Yokoo, M. 2003. T h e s e c o n d
i n t e r n a tional joint conference on A u t o n o m o u s
A g e n t s a n d Multiagent Systems (A4M4S 2003). Table
of contents available at
. Accessed 25 June 2007.
Satterthwaite, M. 1975. Strategy-proofness and Arrow’s
conditions: Existence and correspondence theorems for
voting procedures and social welfare functions. Journal of
Economic Theory 10: 187-217.
Segal, I. 2006. Communication in economic mechanisms. In
Advances in economics and e c o n o m e t r i c s : T h e o r y
and application, ninth world congress
( E c o n o m e t r i c society monographs), ed. R. Blundell,
W.K. Newey, and T. Persson, 222-268. Cambridge:
Cambridge University Press.
Shamir, A., R.L. Rivest, and L. Adelman. 1981. Mental poker.
In The mathematical gardner, ed. D.A. Klamer, 37-43.
Boston: Prindle, Weber, and Schmidt.
Tennenholtz, M. 2002. Competitive safety analysis: Robust
decision-making in multi-agent systems. J o u r nal o f
2014
Urbano, A., and J.E. Vila. 2002. Computational complexity
and communication: Coordination in two-player games.
Econometrica 70: 1893-1927.
Urbano, A., and J.E. Vila. 2004. Computationally restricted
unmediated talk under incomplete information.
Economic Theory 23: 283-320.
Vetta, A. 2002. Nash equilibria in competitive societies, with
applications to facility location, traffic routing and
auctions. In Proceedings of the 43rd IEEE symposium
on foundations of computer science, 416-425.
Vickrey, W. 1961. Counterspeculation, auctions and
competitive sealed tenders. Journal of Finance 16: 8-37.
Yao, A. 1982. Protocols for secure computation (extended
abstract). In Proceedings of the 23rd IEEE symposium
on foundations of computer science, 160-164.
Computing in Mechanism Design
Tuomas Sandholm
Abstract
Computational issues are important in mecha-
nism design, but have received insufficient
research interest. This article briefly reviews
some of the key ideas. I discuss computing by
the centre, such as an auction server or vote
aggregator, and computing by the agents, be
they human or software. Limited computing
hinders mechanism design in several ways, and
presents deep strategic interactions between
computing and incentives. On the bright side,
novel algorithms and increasing computing
power have enabled better mechanisms. Perhaps
most interestingly, with computationally limited
agents, one can implement mechanisms that
would not be implementable among
computationally unlimited agents.
Keywords
Algorithmic mechanism design; Automated
mechanism design; Borda voting rule; Bounded
rationality; Combinatorial auctions; Complexity
theory; Computing by the centre; Computing in
mechanism design; Deliberation equilibrium;
Elicitor; ex post equilibrium;
Computing in Mechanism Design
Expressive commerce; Fielded combinatorial
auctions; Fielded expressive auctions; Gibbard-
Satterthwaite th; Incentive compatibility;
Maximin voting rule; Non-truth- promoting
mechanism; Mechanism design; Performance
profile tree; Plurality voting rale; Preference
determination; Preference elicitation; Pull-
pushmechanism; Revelation principle; Strategic
computing; Tree search; Vickrey auction;
Vickrey-Clarke-Groves (VCG) mechanism
JEL Classifications
Cl
Introduction
Computational issues in mechanism design are
important, but have received insufficient research
interest until recently. Limited computing hinders
mechanism design in several ways, and presents
deep strategic interactions between computing and
incentives. On the bright side, novel algorithms and
increasing computing power have enabled better
mechanisms. Perhaps most interestingly, limited
computing of the agents can be used as a tool to
implement mechanisms that would not be
implementable among computationally unlimited
agents. This article briefly reviews some of the key
ideas, with the goal of alerting the reader to the
importance of these issues and hopefully spurring
future research.
I will discuss computing by the centre, such as
an auction server or vote aggregator, in Section “
Then, in
Section “ ”, I will
address the agents’ computing, be they human or
software.
Computing by the Centre
Computing by the centre plays significant roles in
mechanism design. In the following three subsec-
tions I will review three prominent directions.
Computing in Mechanism Design
Executing Expressive Mechanisms
As algorithms have advanced drastically and
computing power has increased, it has become
feasible to field mechanisms that were previously
impractical. The most famous example is a com-
binatorial auction (CA). In a CA, there are multiple
distinguishable items for sale, and the bidders can
submit bids on self-selected packages of the items.
(Sometimes each bidder is also allowed to submit 2. Approximate winner determination. Th e
exclusivity constraints of different forms among his
bids.) This increase in the expressiveness of the bids
drastically reduces the strategic complexity that
bidders face. For one, it removes the exposure
problems that bidders face when they have
preferences over packages but in traditional
auctions are allowed to submit bids on individual
items only.
CAs shift the computational burden from the
bidders to the centre. There is an associated gain
because the centre has all the information in hand to
optimize while in traditional auctions the bidders
only have estimated projected (probabilistic)
information about how others will bid. Thus CAs
yield more efficient allocations.
On the downside, the centre’s task of determin-
ing the winners in a CA (deciding which bids to
accept so as to maximize the sum of the accepted
bids’prices subject to not selling any item to more
than one bid) is a complex combinatorial optimi-
zation problem, even without exclusivity constraints
among bids. Three main approaches have been
studied for solving it.
3. Restricting the bidding language s o m u ch
t ha t
1. Optimal winner determination using some form
of tree search. For a review, see Sandholm (
). The advantage is that the
bidding language is not restricted and the opti-
mal solution is found. The downside is that no
optimal winner determination algorithm can run
in polynomial time in the size of the problem
instance in the worst case, because the problem
is A^P-complete (Rothkopf et al.
). (A/’P-complete problems are problems
tor which the fastest known algorithms take
exponential time in the size of the problem
instance in the worst case. V is the class of
2015
easy problems solvable in polynomial time. The
statement of winner determination not being
solvable in polynomial time in the worst case
relies on the usual assumption V / J f P . This is an
open question in complexity theory, but is
widely believed to be true. If false, that would
have sweeping implications throughout
computer science.)
advantage is that many approximation algo-
rithms ran in polynomial time in the size of the
instance even in the worst case. For reviews of
such algorithms, see Sandholm ( ) and
Lehmann et al. ( ). (Other suboptimal
algorithms do not have such time guarantees,
such as local search, stochastic local search,
simulated annealing, genetic algorithms and
tabu search.) The downside is that the solution is
sometimes far from optimal: no such algorithm
can always find a solution that is within a factor
m in | #bidsl~£, items?~£1
(1 )
of optimal (Sandholm ). (This assumes CP'P /
J f V . It is widely believed that these two
complexity classes are indeed unequal.) For
example, with just nine items for sale, no such
algorithm can extract even 33 per cent of the
available revenue from the bids in the worst
case. With 81 items, that drops to 11 per cent.
optimal (within the restricted language) winner
determination can be conducted in worst-case
polynomial time. For a review, see Muller ( ).
For example, if each package bid is
only allowed to include at most two items, then
winners can be determined in worst-case
polynomial time (Rothkopf et al. ). The
downside is that bidders have to shoehorn their
preferences into a restricted bidding language;
this gives rise to similar problems as in non-
combinatorial mechanisms for multi-item
auctions: exposiue problems, need to speculate
how others will bid, inefficient allocation, and
so on.
2016
Truthful bidding can be made a dominant strat-
egy by applying the Vickrey-Clarke-Groves (VCG)
mechanism to a CA. Such incentive com- patibihty
removes strategic complexity of the bidders. The
mechanism works as follows. The optimal
allocation is used, but the bidders do not pay their
winning bids. Instead each bidder pays the amount
of value he takes away from the others by taking
some of the items. This value is measured as the
difference between the others’ winning bids’ prices
and what the others’ winning bids’ prices would
have been had the agent not submitted any bids.
This mechanism can be executed by determining
the winners once overall, and once for each agent
removed in turn. (This may be accomplishable with
less computing. For example, in certain network
auctions it can be done in the same asymptotic
complexity as one winner determination -
Hershberger and Suri •)
Very few canonical CAs have found their way to
practice. However, auctions with richer bid
expressiveness forms (that are more natural in the
given application and more concise) and that
support expressiveness also by the bid taker have
made a major breakthrough into practice (Sandholm
; Bichler et al. ). This is sometimes called
expressive commerce to distinguish it from vanilla
CAs. The widest area of application is currently
industrial soincing. Tens of billions of dollars worth
of materials, transportation and services are being
soinced annually using such mechanisms, yielding
billions of dollars in efficiency improvements. The
bidders’ expressiveness forms include different
forms of flexible package bids, conditional
discounts, discount schedules, side constraints (such
as capacity constraints), and often hundreds of cost
drivers (for example, fixed costs, variable costs,
transshipment costs and costs associated with
changes). The item specifications can also be left
partially open, and the bidders can specify some of
the item attributes (delivery date, insurance terms,
and so on), in alternate ways. The bid taker also
specifies preferences and constraints. Winner
determination then not only decides who wins what,
but also automatically configures the items. In some
of these events it also configures
Computing in Mechanism Design
the supply chain several levels deep as a side effect.
On the high end, such an auction can have tens of
thousands of items (multiple units of each), millions
of bids, and hundreds of thousands of side
constraints. Expressive mechanisms have also been
designed for settings beyond auctions, such as
combinatorial exchanges, charity donations and
settings with externalities.
Basically all of the fielded expressive auctions
use the simple pay-your-winning-bids pricing mle.
There are numerous important reasons why few, if
any, use the VCG mechanism. It can lead to low
revenue. It is vulnerable to collusion. Bidders would
not tell the truth because they do not want to reveal
their cost structures, which the auctioneer could
exploit the next time the auction is conducted, and
so on (Sandholm ; Rothkopf )•
Basically all of the fielded expressive auctions
use tree search for winner determination. In prac-
tice, modem tree search algorithms for the problem
scale to the large and winners can be determined
optimally. If winner determination were not done
optimally in a CA, the VCG mechanism can lose its
truth-dominance property (Sandholm ). In fact, any
truthful sub-
optimal VCG-based mechanism for CAs is unrea-
sonable in the sense that it sometimes does not
allocate an item to a bidder even if he is the only
bidder whose bids assign non-zero value to that
item (Nisan and Ronen ).
Algorithmic Mechanism Design
Motivated by the worry that some instances of NP-
hard problems may not be solvable within
reasonable time, a common research direction in
theory of computing is approximation algorithms.
They trade off solution quality for a guarantee that
even in the worst case, the algorithm runs in
polynomial time in the size of the input.
Analogously, Nisan and Ronen ( ) pro
posed algorithmic mechanism design: designing
approximately optimal mechanisms that take the
centre a polynomial number of computing steps
even in the worst case. However, this is more
difficult than designing approximately optimal
algorithms because the mechanism has to motivate
the agents to tell the truth.
Computing in Mechanism Design 2017

Lehmann et al. ( ) studied this for CAscustom design of mechanisms for every setting. The
with single-minded bidders (each bidder being onlysetting can be described by the agents’ (discretized)
interested in one specific package of items). Theytype spaces, the designer’s prior over types, the
present a fast greedy algorithm that guarantees a desired notion of incentive compatibility (for
example, dominant strategies vs. Bayes-Nash
solution within a factor %/#iterns of optimal. They
implementation), the desired notion of participation
show that the algorithm is not incentive compatible
with VCG pricing, but is with their custom pricingconstraints (for example, ex interim, ex post or
none), whether payments are allowed, and whether
scheme. They also identify sufficient conditions for
any (approximate) mechanism to be incentive the mechanism is allowed to use randomization. )
This can yield better mechanisms for previously
compatible (see also Kfir- Dahav et al. ). There has
been substantial follow-on work on subclasses of studied settings because the mechanism is designed
single- minded CAs. for the specific setting rather than a class of settings.
Lavi and Swamy ( It can also be used for settings not previously
) developed a technique
for a range of packing problems with which any ^- studied in mechanism design.
approximation algorithm (that is, algorithm that For almost all natural (linear) objectives, all
variants of the design problem are J\fV-complete if
guarantees that the solution is within a factor k of
the mechanism is not allowed to use randomization,
optimal) that also bounds the integrality gap of the
linear programming (LP) relaxation of the problem but randomized mechanisms can be constructed for
by k can be used to construct a ^approximation all these settings in polynomial time using linear
mechanism. The LP solution, scaled down by k, can programming. Custom algorithms have been
developed for some problems in each of these two
be represented as a convex combination of integer
categories. (Even the latter category warrants
solutions, and viewing this convex combination as
research. While the linear programme is polynomial
specifying a probability distribution over integer
in the size of the input, the input itself can be
solutions begets a VCG-based randomized
exponential in the number of agents.) Structured
mechanism that is truthful in expectation. For CAs
representations of the problem can also make the
with general valuations, this yields an 0( \/#ilems)
design process drastically faster.
-approximate mechanism.
Beyond the general setting, automated mech-
In a different direction, several mechanisms
anism design has been applied to specific settings,
have been proposed where the agents can help the
such as creating revenue-maximizing CAs (without
centre find better outcomes. This is done either by
the need to discretize types)- Likhodedov and
giving the agents the information to do the centre’s
Sandholm (a recognized problem that eludes
computing (Banks et al. ; Land et al. ; Parkes andanalytical characterization; even the two-item case
Shneidman ), or by allowing the agents to change is open), reputation systems (Jurca and Fairings ),
what they told the mechanism based on the safe exchange mechanisms (Sandholm and
mechanism’s output and potentially also based on Ferrandon ), and supply chain settings
what other agents told the mechanism (Nisan and (Vorobeychik et al. ). Automated mechanism design
Ronen ). In VCG-based mechanisms, an agent software has recently also been adopted by several
benefits from lying only if the lie causes the mechanism design theoreticians to speed up their
mechanism to find an outcome that is better overall.
research.
It turns out that even multistage mechanisms can
Automated Mechanism Design
be designed automatically (Sandholm et al. ).
Conitzer and Sandholm ( ) proposed the idea
Furthermore, automated mechanism design has
of automated mechanism design: having a com-
been applied to the design of online mechanisms
puter, rather than a human, design the mechanism.
(Hajiaghayi et al. ), that is,
Because human effort is eliminated, this enables
2018 Computing in Mechanism Design

mechanisms that execute while the world changes protocols that are easy to manipulate, so that they
-for example, agents enter and exit the system. become hard to manipulate (Conitzer and Sandholm
). Specifically, before the original protocol is
executed, one pairwise elimination round is
Computing by the Agents executed among the candidates, and only the
winning candidates survive to the original protocol.
I will now move to discussing computing by the
This makes the protocols ./UP-hard, # Phard (#P-
agents.
hard problems are at least as hard as counting the
Mechanisms That Are Hard to Manipulate number of solutions to a problem in P), or even
This section demonstrates that one can use the fact PSPACE-hard (PSPACE-hard problems are at least
that agents are computationally limited to achieve as hard as any problem that can be solved using a
things that are not achievable via any mechanism polynomial amount of memory ) to manipulate
among perfectly rational agents. constructively, depending on whether the schedule
A seminal negative result, the GibbardSat- of the pre-round is determined before the votes are
terthwaite th, states that if there are three or more collected, randomly after the votes are collected, or
candidates, then in any non-dictatorial voting the scheduling and the vote collecting are carefully
scheme there are candidate rankings of the other interleaved, respectively.
voters, and preferences of the agent, under which All of the hardness results of the previous
the agent is better off voting manipulatively than paragraph rely on both the number of voters and the
truthfully. One avenue around this impossibility is number of candidates growing. The number of
to construct desirable general non-dictatorial voting candidates can be large in some domains, for
protocols under which finding a beneficial example when voting over task or resource
manipulation is prohibitively hard computationally. allocations.
There are two natural alternative goals of However, in other elections - such as presi-
manipulation. In constructive manipulation, the dential elections - the number of candidates is
manipulator tries to find an order of candidates that small. If the number of candidates is a constant,
he can reveal so that his favourite candidate wins. both constructive and destructive manipulation are
In destructive manipulation, the manipulator tries to easy (in P), regardless of the number of voters
find an order of candidates that he can reveal so that (Conitzer et al. ). This holds even if the voters are
his hated candidate does not win. These are special weighted, or if a coalition of voters tries to
cases of the utility-theoretic notion of improving manipulate. On the bright side, when a coalition of
one’s utility, so the hardness results, discussed weighted voters tries to manipulate, complexity can
below, carry over to the usual utility-theoretic arise even for a constant number of candidates: see
setting. Tables and . Another lesson from that table is that
Unfortunately, finding a constructive manipu- randomizing over instantiations of the mechanism
lation is easy (in V) for the plurality, Borda and (such as schedules of a cup) can be used to make
maximin voting rules (Bartholdi et al. ), which are manipulation hard.
commonly used. On the bright side, constructive As usual in computer science, all the results
manipulation of the single transferable vote (STV) mentioned above are worst-case hardness. Unfor-
protocol is AfP-hard (Bartholdi and Orlin ) (as is tunately, under weak assumptions on the preference
manipulation of the second order Copeland distribution and voting rule, most instances of any
protocol (Bartholdi et al. ), but that hardness is voting rule are easy to manipulate (Conitzer and
driven solely by the tie-breaking rale). Even better, Sandholm ).
there is a systematic methodology for slightly All of the hardness results discussed above hold
tweaking voting even if the manipulators know the non-
manipulators’ votes exactly. Under weak
Number of candidates:
2
Borda V
Computing in Mechanism Design
Veto V
2019
STV V
Computing in
Mechanism
Plurality with runoff Design,
V
Table 1 Complexity of •'V P -complete
Copeland V
Maximin constructive weighted
V Ap-complete Ap-
coalitional manipulation Arp_,complete complete
Randomized cup V A"P-complete
Cup V A"P-complete
Plurality V
■AP-
complete
A'P-
coniplctc
-A^P-

Computing
in Mechanism Design,
Table 2 Complexity of
destructive weighted
coalitional manipulation

assumptions, if weighted coalitional manipulation Theorem 1 (Conitzer and Sandholm )

with complete information about the others’ votes is Suppose the centre is trying to maximize social
hard in some voting protocol, then individual welfare, and neither payments nor randomization is
unweighted manipulation is hard when there is allowed. Then, even with just two agents (one of
uncertainty about the others’ votes (Conitzer et al. ). whom does not even report a type, so dominant
strategy implementation and Bayes-Nash imple-
Non-Truth-Promoting Mechanisms mentation coincide), there exists a family of pref-
A challenging issue is that even if it is prohibitively erence aggregation settings such that:
hard to find a beneficial manipulation, the agents
might not tell the truth. For example, an agent might • the execution of any optimal incentive-
take a chance that he will do better with a lie. The compatible mechanism is AfV -complete for the
following result shows that, nevertheless, center, and.
mechanism design can be improved by making the • there exists a non-incentive-compatible mech-
agents face complexity. (This is one reason why anism which (1) requires the centre to carry out
computational issues can render the revelation only polynomial computation, and (2) makes
principle inapplicable. One of the things the finding any beneficial insincere revelation NP-
principle says is that for any non-truth-promoting complete for the type-reporting agent.
Additionally, if the type-reporting agent man-
mechanism it is possible to construct an incentive-
ages to find a beneficial insincere revelation, or
compatible mechanism that is at least as good. The
no beneficial insincere revelation exists, the
theorem below challenges this.)
social welfare of the outcome is identical to the
social welfare that would be produced by
2020 Computing in Mechanism Design

any optimal incentive-compatible mechanism. If a bidder has the opportunity to approximate

Finally, if the type-reporting agent does not his valuation to different degrees, how much
manage to find a beneficial insincere revelation computing time should the bidder spend on refining
where one exists, the social welfare of the its valuation? If there are multiple items for sale,
outcome is strictly greater than the social how much computing time should the bidder
welfare that would be produced by any allocate on different bundles? A bidder may even
optimal incentive-compatible mechanism. allocate some computing time to evaluate other
bidders’ valuations so as to be able to bid more
An analogous theorem holds if, instead of strategically; this is called strategic computing.
counting computational steps, we count calls to a To answer these qsts, Larson and Sandholm (
commonly accessible oracle which, when supplied ) developed a deliberation control method
with an agent, that agent’s type, and an outcome, called a performance profile tree for projecting how
returns a utility value for that agent. an anytime algorithm (that is, an algorithm that has
an answer available at any time, but where the
Preference (Valuation) Determination Via
quality of the answer improves the more computing
Computing or Information Acquisition
time is allocated to the algorithm) will change the
In many (auction) settings, even determining one’s
valuation if additional computing is allocated
valuation for an item (or a bundle of items) is
toward refining (or improving) it. This deliberation
complex. For example, when bidding for trucking
control method applies to any anytime algorithm.
lanes (tasks), this involves solving two AfV-
Unlike earlier deliberation control methods for
complete local planning problems: the vehicle
anytime algorithms, the performance profile tree is
routing problem with the new lanes of the bundle
a fully normative model of bounded rationality, it
and the problem without them (Sandholm ). The
takes into account all the information that an agent
difference in the costs of those two local plans is the
can use to make its deliberation control decisions.
cost (valuation) of taking on the new lanes.
This is necessary in the game- theoretic context;
In these types of settings, the revelation prin-
otherwise a strategic agent could take into account
ciple applies only in a trivial way: the agents report
some information that the model does not.
their data and optimization models to the centre,
Using this deliberation control method, the
and the centre does the computation for them. It
auction can be modelled as a game where the
stands to reason that in many applications the centre
agents’ strategy spaces include computing actions.
would not want to take on that burden, in which
At every point, each agent can decide on which
case such extreme direct mechanisms are not an
bundle to allocate its next step of computing as a
option. Therefore, I will now focus on mechanisms
function of the agent’s computing results so far
where the agents report valuations to the centre, as
(and in open-cry auction format also the others’
in traditional auctions.
bids observed so far). At every point, the agent can
Bidders usually have limited computing and
also decide to submit bids. One can then solve this
time, so they cannot exactly evaluate all (or even
for equilibrium: each agent’s (deliberation and
any) bundles - at least not without cost. This leads
bidding) strategy is a best- response to the others’
to a host of interesting issues where computing and
strategies. This is called deliberation equilibrium.
incentives are intimately intertwined.
This notion, and the performance profile tree,
For example, in a one-object auction, should a
apply not only to computational actions but also to
bidder evaluate the object if there is a cost to doing
information gathering actions for determining val-
so? It turns out that the Vickrey auction loses its
uations. (In contrast, most of the literature on
dominantstrategy property: whether or not the
information acquisition in auctions does not take
bidder should pay the evaluation cost depends on
the other bidders’ valuations (Sandholm ).

Computing in Mechanism Design
Auctionmechanism Speculation by perfectly rational agents?
Single itemComputingFirst price Yes
in Mechanism Design,
Table 3 Can stra- perfectly rational
computingDutchoccur in deliberationYes
equilibrium? The tions instantly without Yes
results are in bold. As a benchmark
English No from
ing each others’
classicalVickrey
auction theory, the tableNoalso shows whether or not
Multiple First price Yes
items
VCG No
into account that valuations can be determined to
different degrees and that an agent may want to
invest effort to determine others’ valuations as well
- even in privatevalue settings.)
Table shows in which settings strategic
computing can and cannot occur in deliberation
equilibrium. This depends on the auction mech-
anism. Interestingly, it also depends on whether the
agent has limited computing (for example, owning
a desktop computer that the agent can use until the
auction’s deadline) or costly computing (for
example, being able to buy any amount of
supercomputer time where each cycle comes at a
cost).
The notion of deliberation equilibrium can also
be used as the basis for designing new mechanisms,
which hopefully work well among agents whose
computing is costly or limited. Unfortunately, there
is an impossibility (Larson and Sandholm ): there
exists no mechanism that is sensitive (the outcome
is affected by each agent’s strategy), preference
formation independent (does not do the
computations for the agents; the agents report
valuations), non-misleading (no agent acts in a way
that causes others to believe his true type has zero
probability), and deliberation-proof (no strategic
computing occurs in equilibrium, that is, agents
compute only on their own problems). Current
work involves designing mechanisms that take part
in preference formation in limited ways: for
example, agents
Strategic computing
Limited Costly 2021
computing computing
Yes agents, that canYes
determine their valua- tegic
Yes consider- most interesting
cost, would benefit from
N o valuations whenYdeciding
es how to bid
No Yes
Yes Yes
Yes Yes
report their performance profile trees to the centre,
which then coordinates the deliberations incre-
mentally as agents report deliberation results. Cur-
rent research also includes designing mechanisms
where strategic computing occurs but its waste-
fulness is limited.
Preference Elicitation by the Centre To reduce the
agents’ preference determination effort, Conen and
Sandholm ( ) proposed a
framework where the centre (also known as elic-
itor) explicitly elicits preference information from
the agents incrementally on an as-needed basis by
posing queries to the agents. The centre thereby
builds a model of the agents’preferences, and
decides what to ask, and from which agent, based
on this model. Usually the process can be
terminated with the provably correct outcome while
requiring only a small portion of the agents’
preferences to be determined. Multistage mecha-
nisms can yield up to exponential savings in pref-
erence determination and communication effort the
agents need to go through compared to single-stage
mechanisms (Conitzer and Sandholm )•
The explicit preference elicitation framework
was originally proposed for CAs (but the approach
has since been used for other settings as well, such
as voting). For general valuations, an exponential
number of bits in the number of items for sale has
to be communicated in the worst case
2022 Computing in Mechanism Design

Depending on the application, the reasons for this

no matter what queries are used (Nisan and Segal ).
However, experimentally only a small fraction ofpreference may include avoiding a single vulnerable
the preference information needs to be elicited point of failure, distributing the computing effort
before the provably optimal solution is found. (for computational efficiency or because the data is
inherently distributed), and enhancing privacy. The
Furthermore, for valuations that have certain types
preference carries over from traditional computer
of structure, even the worst-case number of queries
science applications to different forms of
needed is small. Research has also been done on the
relative power of different query types. negotiation systems -for example, see Sandholm (
If enough information is elicited to also deter- ) for an early dis
mine the VCG payments, and these are the pay- tributed automated negotiation system for software
ments charged to the bidders, answering the agents.
Feigenbaum et al. (
elicitor’s queries truthfully is an ex post equilibrium ) have studied lowest-
cost inter-domain routing on the Internet, modifying
(a strengthening of Nash equilibrium that does not
a distributed protocol so that the agents (routing
rely on priors). (This assumes there is no explicit
cost or limit to valuation determination; domains) are motivated to report their true costs and
mechanisms have also been designed for settings the solution is found with minimal message passing.
For a review of some other research topics in this
where there is an explicit cost (Larson ).) This holds
space, see Feigenbaum and Shenker ( ).
even if the agents are allowed to answer queries that
One can go further by taking into account the
the elicitor did not ask (for example, queries that
fact that agents might not choose to follow the
are easy for the agent to answer and which the
prescribed protocol. They may cheat not only on
agent thinks will significantly advance the
information-revelation actions, but also on
elicitation process). We thus have a pull-push
message-passing and computational actions.
mechanism where both the centre and the agents
Despite computation actions not being observable
guide the preference revelation (and thus also the
by others, an agent can be motivated to compute as
preference determination/ refinement by the
prescribed by tasking at least one other agent with
agents). For a review, see Sandholm and Boutilier (
the same computation, and comparing the results
). Ascending
(Sandholm et al. ). Careful problem partitioning can
(combinatorial) auctions are an earlier special case,
also be used to achieve the same outcome without
and have limited power compared to the general
redundancy by only requiring agents to perform
framework (Blumrosen and Nisan )• computing and message passing tasks that are in
Preference elicitation can sometimes be com-their own interest (Parkes and Shneidman ).
putationally complex for the centre. It can be Shneidman and Parkes ( )
complex to intelligently decide what to ask next,
propose a general proof technique and instantiate it
and from whom. It can also be complex to deter- to provide a non-manipulable protocol for inter-
mine whether enough information has been elicited
domain routing. Monderer and Tennenholtz ( )
to determine the optimal outcome. Even if the develop protocols for one-item auctions
elicitor knows that enough has been elicited, it can
executed among agents on a communication net-
be complex to determine the outcome - for work. The protocols motivate the agents to correctly
example, allocation of items to bidders in some reveal preferences and communicate. For the setting
CAs. where agents with private utility functions have to
agree on variable assignments subject to side
Distributed (Centre-Free) Mechanisms
constraints (for example, meeting scheduling),
Computer scientists often have a preference for
Petcu et al. ( ) developed a
distributed applications that do not have any
VCG-based distributed optimization protocol that
centralized coordination point (centre).
finds the social welfare maximizing allocation
Computing in Mechanism Design
and each agent is motivated to follow the protocol Acknowledgment This work was funded by the National
in terms of all three types of action. The only Science Foundation under TTR grant HS0427858, and a
Sloan Foundation Fellowship. I thank Felix Brandt, Christina
centralized party needed is a bank that can extract Fong, Joe Halpem, and David Parkes for helpful comments.
payments from the agents.
Cryptography is a powerful tool for achieving
privacy when trying to execute a mechanism in a
distributed way without a centre, using private
communication channels among the agents. Con- Bibliography
sider first the setting with passive adversaries, that Banks, J.S., J. Ledyard, and D. Porter. 1989. Allocating
is, agents that faithfully execute the specified dis-
tributed communication protocol, but who try to
infer (at least something about) some agents’ pri-
Bartholdi, J. Ill, and J. Qrlin. 1991. Single transferable vote
vate information.
• If agents are computationally limited - for
example, they are assumed to be unable to factor
large numbers - then arbitrary functions can be
computed while guaranteeing that each agent
maintains his privacy (except, of course, to the
extent that the answer of the computation says
something about the inputs) (Goldreich et al. ).
Thus the desire for privacy does not constrain
what social choice functions can be
implemented.
• In contrast, only very limited social choice
functions can be computed privately among
computationally unlimited agents. For example,
when there are just two alternatives, every
monotonic, non-dictatorial social choice func-
tion that can be privately computed is constant
(Brandt and Sandholm ). With special structure
in the preferences, this impossibility can
sometimes be avoided. For example, with the
standard model of quasi-linear utility, first- price
auctions can be implemented privately; second-
price (Vickrey) auctions with more than two
bidders cannot (Brandt and Sandholm
)•
A more general model is that of active
adversaries who can execute the distributed
communication protocol unfaithfully in a coor-
dinated way. A more game-theoretic model is that
of rational adversaries that are not passive, but not
malicious either. For a brief overview of such work,
see computer science and game theory.
2023
uncertain and unresponsive resources: An experimental
approach. R A N D J o u r n a l o f E c o n o m i c s 20: 1-
25.
resists strategic voting. S o c i a l C h o i c e a n d
W e l f a r e 8: 341-354.
Bartholdi, J. Ill, C. Tovey, and M. Trick. 1989. The com-
putational difficulty of manipulating an election. S o c i a l
C h o i c e a n d W e l f a r e 6: 227-241.
Bichler, M., A. Davenport, G. Hohner, and J. Kalagnanam.
2006. Industrial procurement auctions. In
C o m b i n a t o r i a l a u c t i o n s , ed. Cramton, Shoham,
and Steinberg. Cambridge, MA: MIT Press.
Blumrosen, L., and N. Nisan. 2005. On the computational
power of iterative auctions. In P r o c e e d i n g s o f t h e
ACM conference on electronic
c o m m e r c e , 29-43. Vancouver: ACM Press.
Brandt F, andSandholm T. 2004. (Im)possibility of uncon-
ditionally privacypreserving auctions. In
Proceedings
of the international conference on autonomous agents and
multi-agent systems. 8 1 0 - 1 7 .
Brandt F., and T. Sandholm. 2005. Unconditional privacy in
social choice. In Proceedings of the conference on
theoretical aspects of rationality and knowledge. 207-18.
Conen W., and T. Sandholm. 2001. Preference elicitation in
combinatorial auctions: Extended abstract. In Pro-
ceedings of the ACM conference on electronic commerce,
256-9. More detailed description of algorithmic aspects in
Proceedings of the IJCAI01 workshop on economic
agents, models, and mechanisms. 71-80.
Conitzer V., and T. Sandholm. 2002. Complexity of mech-
anism design. In Proceedings of the conference on
uncertainty in artificial intelligence. 103-10.
Conitzer V., and T. Sandholm. 2003. Universal voting
protocol tweaks to make manipulation hard. In Pro-
ceedings of the international joint
conference on artific i a l i n t e l l i g e n c e . 781-8.
Conitzer V., and T. Sandholm. 2004. Computational criti-
cisms of the revelation principle. In C o n f e r e n c e o n
logic and the foundations of game and
d e c i s i o n t h e o r y . Earlier versions: AMEC-03, EC-
04.
Conitzer V., and T. Sandholm. 2006. Nonexistence of voting
rules that are usually hard to manipulate. In
Proceedings of the national conference
on artificial intelligence.
2024
Conitzer, V., T. Sandholm, and J. Lang. 2007. When are
elections with few candidates hard to manipulate?
J o u r n a l o f t h e A C M 54(3): 14.
Cramton, P., Y. Shoham, and R. Steinberg, ed. 2006. (\ > m -
b i n a t o r i a l a u c t i o n s . Cambridge, MA: MIT Press.
Feigenbaum, J., C. Papadimitriou, R. Sami, and S. Shenker.
2005. A BGP-based mechanism for lowest cost routing.
D i s t r i b u t e d C o m p u t i n g 18: 61-72.
Feigenbaum J., and S. Shenker. 2002. Distributed algorithmic
mechanism design: Recent results and future directions.
In P r o c e e d i n g s of the international
workshop on discrete algorithms and
methods for mobile computing and
c o m m u n i c a t i o n s . 1 13.
Goldreich O., S. Micali and A. Wigderson. 1987. How to play
any mental game or a completeness theorem for protocols
with honest majority. In P r o c e e d i n g s o f t h e
s y m p o s i u m o n t h e o r y o f c o m p u t i n g . 218-29.
Hajiaghayi, M.T., R. Kleinberg,and T. Sandholm. 2007.
Automated online mechanism design and prophet
inequalities. In P r o c e e d i n g s o f t h e n a t i o n a l
conference on artificial intelligence.
Hershberger J, and S. Suri. 2001. Vickrey prices and shortest
paths: What is an edge worth? In P r o c e e d i n g s o f
the symposium on foundations of
c o m p u t e r S c i e n c e . 252 9.
Jurca R, and B. Fallings. 2006. Minimum payments that
reward honest reputation feedback. In P r o c e e d i n g s
of the At' M conference on electronic
c o m m e r c e . 190 9.
Kfir-Dahav N, D. Monderer, and M. Tennenholtz. 2000.
Mechanism design for resource bounded agents. In
Proceedings of the international
c o n f e r e n c e o n m u l t i - a g e n t s y s t e m s . 309-
315.
Land, A., S. Powell, and R. Steinberg. 2006. PAUSE: A
computationally tractable combinatorial auction. In
C o m b i n a t o r i a l a u c t i o n s , ed. Cramton, Shoham,
and Steinberg. Cambridge, MA: MIT Press.
Larson Is.. 2006. Reducing costly information acquisition in
auctions. In P r o c e e d i n g s o f t h e a u t o n o m o u s
a g e n t s a n d m u l t i - a g e n t s y s t e m s . 1167-74.
Larson Is., and T. Sandholm. 2001. Costly valuation com-
putation in auctions. In P r o c e e d i n g s o f t h e
theoretical aspects of rationality and
k n o w l e d g e . 169-182.
Larson Is, and T. Sandholm. 2005. Mechanism design and
deliberative agents. In Proceedings of the international
conference on autonomous agents and multi-agent systems.
650 656.
Lavi IT, and C. Swamy. 2005. Truthful and near-optimal
mechanism design via linear programming. In P r o -
ceedings of the symposium on
f o u n d a t i o n s o f c o m p u t e r s c i e n c e . 595 604.
Lehmann D., R. Muller, and T. Sandholm. 2006. The winner
determination problem. In P Cramton, Y. Shoham and R.
Steinberg.
Lehmann, D., L.I. O’Callaghan, and Y. Shoham. 2002. Truth
revelation in rapid, approximately efficient combinatorial
Computing in Mechanism Design
Proceedings of the national conference on artificial
intelligence. 267-74.
Monderer D, and M. Tennenholtz. 1999. Distributed games:
From mechanisms to protocols. In P r o c e e d i n g s o f
the national conference on artificial
i n t e l l i g e n c e . 32-7.
Muller R. 2006. Tractable cases of the winner determination
problem. In P Cramton, Y. Shoham and R. Steinberg.
Nisan N, and A. Ronen. 2000. Computationally feasible VCG
mechanisms. In P r o c e e d i n g s o f t h e A C M
c o n f e r e n c e o n e l e c t r o n i c c o m m e r c e . 242-
52.
Nisan. N.. and A. Ronen. 2001. Algorithmic mechanism
design. G a m e s a n d E c o n o m i c B e h a v i o r 35:
166 196.
Nisan. N.. and I. Segal. 2006. The communication require-
ments of efficient allocations and supporting prices.
J o u r n a l o f E c o n o m i c T h e o r y 129: 192-224.
Parkes D.. and J. Shneidman. 2004. Distributed
implementations of generalized Vickrey Clarke Groves
auctions. In P r o c e e d i n g s o f t h e
international conference on autonomous
a g e n t s a n d m u l t i - a g e n t s y s t e m s . 261 -268.
Petcu A., B. Faltings. and D. Parkes. 2006. MDPOP: Faithful
distributed implementation of efficient social choice
problems. In Proceedings of the
international conference on autonomous
a g e n t s a n d m u l t i - a g e n t s y s t e m s . 1397 404.
Rothkopf, M. 2007. Thirteen reasons why the Vickrey Clarke
Groves process is not practical. O p e r a t i o n s
R e s e a r c h 55: 191-197.
Rothkopf, M., A. Pekec, and R. Harstad. 1998. Computa-
tionally manageable combinatorial auctions. M a n a g e -
m e n t S c i e n c e 44: 1131-1147.
Sandholm T. 1993. An implementation of the contract net
protocol based on marginal cost calculations. In P r o -
ceedings of the national conference on
a r t i f i c i a l i n t e l l i g e n c e . 256-62.
Sandholm, T. 2000. Issues in computational Vickrey auctions.
International Journal of Electronic
C o m m e r c e 4 : 107-129 .Early version in ICMAS-96.
Sandholm, T. 2002a. Algorithm for optimal winner deter-
mination in combinatorial auctions. A r t i f i c i a l
I n t e l l i g e n c e 135: 1-54 .Earlier versions: ICE-98
keynote, Washington U. tech report WUCS-99-01 Jan.
1999, IJCAI-99.
Sandholm, T. 2002b. eMediator: A next generation electronic
commerce server. C o m p u t a t i o n a l I n t e l l i g e n c e
18: 656-676 .Earlier versions: Washington U. tech report
WU-CS-99-02 Jan. 1999, AAAI-99 Workshop on AI in
Ecommerce, AGENTS-00.
Sandholm T. 2006. Optimal winner determination algorithms.
In P Cramton, Y. Shoham and R. Steinberg.
Sandholm, T. 2007. Expressive commerce and its application
to sourcing: How we conducted $35 billion of generalized
combinatorial auctions. .1/ M a g a z i n e 28(3): 45-58.
Sandholm T, and Boutilier C. 2006. Preference elicitation in
combinatorial auctions. In P. Cramton, Y. Shoham and
R. Steinberg.
Comte, Isidore Auguste Marie Francois Xavier (1798-1857)
Sandholm T, V. Conitzer, and C. Boutilier. 2007. Automated
design of multistage mechanisms. In P r o c e e d i n g s o f
the international joint con ference on
a r t i f i c i a l i n t e l l i g e n c e . 1500-6.
Sandholm T., and V. Ferrandon. 2000. Safe exchange planner.
In Proceedings of the international conference on multi-
agent systems. 255-62.
Sandholm, T., K. Larson, M. Andersson, O. Shehory, et al.
1999. Coalition structure generation with worst case
guarantees. A r t i f i c i a l I n t e l l i g e n c e 111: 209-238.
Shneidman J, and D.C Parkes . 2004. Specification faithfulness
in networks with rational nodes. In P r o c e e d i n g s o f
the ACM symposium on principles of
d i s t r i b u t e d c o m p u t i n g . 88-97.
Vorobeychik Y, C. Kiekintveld, and M. Wellman. 2006.
Empirical mechanism design: Methods, with
application to a supply chain scenario. In Proceedings of
the ACM conference on electronic commerce. 306-15.
Comte, Isidore Auguste Marie Francois
Xavier (1798-1857)
Robert B. Ekelund Jr.
Auguste Comte, co-founder of sociology and
positivist philosopher, was bom at Montpellier in
1798 and died in Paris in 1857. A student at the
Ecole Polytechnique until he was dismissed for
disobedience and incorrigible behaviour, Comte
became the secretary to Henri de Saint- Simon in
1818, a position he held until 1824. Over this period
Comte developed the kernel of positivist philosophy
and, along with Saint- Simon, modem sociology.
The irascible Comte spent the rest of his life - often
in the face of frequent poverty and desperate
personal circumstances including mental
breakdowns and a failed marriage - establishing,
altering and working out positivism as a
philosophical system of knowledge and as the
foundation for the ‘science of society’. Comte’s
chefs d’oeuvre included the encyclopedic Cours
dephilosophie positive (1830—42) and his System
of Positive Polity (1851-4).
2025
Comte’s social philosophy encompassed all
aspects of life, which he believed to be the har-
monious working together of two inseparable
elements - an organism and its environment. The
‘true’ methods of science were empirical (or
inductive), and Comte, as did his mentor Saint-
Simon, believed that they could be applied, mutatis
mutandis, to all branches of thought including the
social sciences, of which economics was a part. He
succinctly described this basis in the System of
Positive Polity. ‘It [the positivist synthesis of all
knowledge] rests at every point upon the
unchangeable Order of the World. The right
understanding of this order is the principal subject
of our thoughts; its preponderating influence
determines the general course of our feelings; its
gradual improvement is the constant object of our
actions ...’ (vol. 1, p. 21).
‘Gradual improvement’ came about in three
successive stages of understanding the cause of
phenomena. In the ‘Law of Three Stages’, Comte
argued that man’s conception of ‘order’ or the
interrelationships of causes and effects, went first
through a religious or theological stage, the
theological-primitive content of which diminished
progressively from fetishism, polytheism and
monotheism. The metaphysical or a priori stage
followed the theological and represented man’s
attempt to discover ‘order’ by reason. The final
stage in understanding order was the positive stage
wherein science, or the knowledge of relationships
between disparate phenomena, brought man to a
kind of perfection. Within this latter stage, Comte
argued that a second great law obtained - that of
decreasing generality and increasing complexity of
understanding. Here, sciences progressed in a
definite ordering, each dependent on the previous
one, from mathematics to astronomy to physics to
chemistry to biology and, finally, to sociology.
Comte’s system, its broadly empirical method-
ology, and its precept that society was an organism
which evolved under constraints that were
themselves ultimately altered by social activities
and behaviour, had a measurable impact upon
classical economics and upon the form that
2026
economic analysis would take in the neoclassical
period and beyond. J.S. Mill introduced Comte’s
ideas to England and, for a time at least, was deeply
influenced by the French philosopher. Mill’s Logic
(1843) contains copious and favourable references
to Comte’s inverse- deductive (empirical-historical)
method for use in the broader and more important
investigation of integrated social studies. In the end,
however, Mill reaffirmed the a priori ‘Ricardian’
method for the narrower concerns of political
economy. Again, Comte’s influence surfaced in
Mill’s Principles of Political Economy (1848) in the
form of the famous ‘statics and dynamics’
distinction and in Mill’s admissions that the ultimate
aim of social science was a broader conception of
the process than political economy had to offer. In
his Principles, however, Mill reaffirmed the
(admittedly provisional) a priori deductive method
as the essential and proper one for studying political
economy. Ultimately, Mill and a number of Comte’s
followers totally rejected the ‘religion of humanity’
that Comte later made of his positivist principles
(1851-77). Mill was especially aghast at the
infringements on individual liberty that Comte’s
quasi-medieval ‘Catholicism without Christianity’
envisioned. In his Autobiography Mill attacked
Comte’s planned society as ‘the completest system
of spiritual and temporal despotism which ever yet
emanated from a human brain, unless possibly that
of Ignatius Loyola’ ( , p.149).
In the late classical and early neoclassical period
Comte’s ideas received some reinforcement at the
hands of the British historicists, notably John Kells
Ingram. But his impact, especially on methodology
and on his belief that political economy should be
subordinate to sociology, was successfully nullified
by the efforts of John Elliott Caimes (). While
admitting that empiricism
was a necessary adjunct to economic theory, Caimes
defended an essentially abstract and a priori method
in political economy. Caime’s ideas on method were
replicated by the leading methodologist of the
neoclassical and even post- neoclassical periods,
John Neville Keynes. The attempt to infuse Comtian
and other broader methods into political economy,
in other words,
Comte, Isidore Auguste Marie Francois Xavier (1798-1857)
reinforced Ricardo’s method which became the
dominant method of economists in the 20th century.
See Also
Selected Works
1830—42. Cours de philosophic positive, 2 vols,
5th ed. Trans. H. Martineau; 3rd ed. London,
1893.
1851—4. Systeme de politique positive, ou Traite
de sociologie institutant la religion de
l’humanite (Trans: J.H. Bridge et al. as System
of positive polity). London, 1875—7; New
York: Burt Franklin Reprint,
4 vols, 1968.
18 99 . Lettres inedites de John Stuart Mill a
Auguste Comte, publiees avec les reponses de
Comte et un introduction par Lucien Levy-
Bruhl. Pa r i s : F e l i x A l c an .
References
Abrams, P. 1968. T h e origins of British
s o c i o l o g v 1 8 3 4 - 1 9 1 4 . Chicago: University of
Chicago Press.
Adelman, P. 1971. Frederic Harrison and the ‘Positivist’
attack on orthodox political economy. H i s t o r y o f
P o l i t i c a l E c o n o m y 3( 1): 170-189.
Caimes, J.E. 1870. M. Comte and political economy. F o r t -
n i g h t l y R e v i e w 13: 579-602.
Ekelund Jr., R.B., and E. Olsen. 1973. Comte, Mill and
Caimes: The positivist-empiricist interlude in late clas-
sical economics. J o u r n a l o f E c o n o m i c I s s u e s
7(3): 383-416.
Hayek, F.A. 1955. The counter-revolution of science, studies in
the abuse of reason. New York: The Free Press.
Ingram, J.K. 1899. A history of political
e c o n o m y . London: A. and C. Black, 1915.
Mill, J.S. 1865. A u g u s t e C o m t e a n d p o s i t i v i s m .
London: N. Triibner.
Mill, J.S. 1873. A u t o b i o g r a p h y . New York: Columbia
University Press, 1924.
Seligman, B.B. 1969. The impact of positivism on economic
thought. H i s t o r y o f P o l i t i c a l E c o n o m y 1(2):
256-278.
Concentration Measures 2027

important characterization of the interaction of

Concentration Measures
Juan Esteban Carranza
Abstract
Concentration is a characterization of the size
distribution and quantity of competing firms
within a specific market or industry. The most
common concentration measures are the
Herfindahl index and the /j-firm concentration
rate. The Herfindahl index is the sum of the
squared market shares of all the firms in a
market, whereas the /j-firm concentration rate isbiggest firms in the market where size is measured
the sum of the market shares of the n biggest
firms. These measures are a significant reflec- characterize the concentration of any given market,
tion of the underlying degree of competitive-
ness, but are sensitive to the adopted market
definition, and must be interpreted carefiilly
depending on the specifics of the case.
Keywords
Gini coefficient; Herfindahl index; Lemer index; ‘Herfindahl-Hirschman index’; see Hirschman , for
Market definition; Market structure
JEL Classification
D4
firms within a specific market or industry.
The most common concentration measures are
the ‘n-firm concentration rate’ and the ‘Herfindahl
index’. Let S, be the market share of firm i; the ‘n-
firm concentration rate’ is the sum of the market
shares of the n biggest firms within the market:
C(n) = £s,,
;1
As indicated, the summation above is taken over
the set of n biggest firms in the market. So, for
example, the two-firm concentration rate of a given
market is the sum of the market shares of the two
according to observed sales. In order to fully
though, a number of these rates must be used, since
there is no agreed-on value for n. This complicates
its use for comparing concentration over time and
across sectors, and for its use in statistical analysis.
The Herfindahl index, first devised by Albert
Hirschman to measure the concentration of trade
across sectors (so that the index is also known as
its history), is the sum of the squared market shares
of all firms in the market:
N
;1

The term concentration (also firm concentration,
industry concentration or market concentration)
refers to aspects of the distribution of firm size
within a specific market or industry that have
traditionally been used to characterize the degree of
competitiveness in the market. Even though the size
of firms can be measured using many different
variables, such as employment or assets, the sales
level is the most commonly used size measure.
Accordingly, if very few firms serve a very large
portion of the market, it is said that the given
market is highly ‘concentrated’, whereas if no
single firm has a large share of sales it is said that
the market is not ‘concentrated’. Since con-
centration is an important reflection of the under-
lying market structure, its measurement is an
The summation in this case is taken over the set of
all N firms in the market. This index lies between
zero and 1: if there is only one firm in the market,
so that the market has the highest possible con-
centration, the index is 1. If, on the other hand, there
are many equally sized firms in the market, the
index will be close to zero. By squaring the
individual market shares, this index gives relatively
greater weight to the market shares of large firms.
Conversely, the addition of one small firm to the
market dilutes somewhat the market share of larger
firms, and has a marginal negative effect on the
index, which is consistent with any notion of market
concentration. Any value of this index can
correspond to multiple
2028 Concentration Measures

market configurations, being in that sense less with long-distance phone companies and Internet
illustrative of the actual concentration of a companies as providers of communication services.
market than a set of //-firm concentration rates. The Internet, on the other hand, competes in some
On the other hand, this index can be easily instances with cable and satellite companies, radio
correlated with other market characteristics and stations and even newspapers as sources of news
is therefore very useful for statistical analysis. and entertainment. What exactly the relevant market
Other less commonly used concentration faced by mobile phone companies is will depend on
measures include entropy coefficients, the Gini the type of issue being addressed. Accordingly, the
coefficient and measures of the variance of concentration measures will change depending on
market shares across firms within a market. The the adopted market definition.
entropy coefficient is usually computed using the On the other hand, even if the market is well
following formula: defined, computed concentration measures may not
reflect at all the real competitive structure of the
e= (£). market. For example, even in markets as highly
concentrated as the market for computer processors,
the dominant firms have to account for the invisible
This index takes value zero if there is only one
competition of potential entrants. The same happens
firm in the market and grows as market concen-
in regional markets where outside firms are kept at
tration decreases. The interpretation of this coef-
bay by few local films with a combination of low
ficient is complicated, because its formula
prices and high transportation costs. Computed
weights both large and small firms less heavily
concentration measures for specific markets cannot
than mid-size firms and grows rmbormdedly as
account for this unobserved competition and may
market concentration decreases. It is therefore
therefore lead to wrong conclusions regarding the
less commonly used than the Herfindahl index.
underlying behaviour of films. In these instances, a
The Gini coefficient is commonly used to char-
behavioural measure, such as die Lemer index,
acterize the income or wealth inequality within a
which measures the relative size of films' markups,
society. Its drawback as a measure of market con-
may be a better indicator of the competitive
centration is that it is useful only to measure the
structure of the market.
concentration of firms’ sizes within a market, given
There is a body of empirical literature that uses
a number of firms. So according to the Gini
market concentration measures across industries to
coefficient a duopolistic market with two firms of
approximate the underlying differences in indus-
equal size is as concentrated as a market with 100
tries' competitiveness. They were then used to infer
firms with identical size. The same drawback
statistically the relationship between market
applies for the use of measures of the variance of
‘structure’ and market ‘performance’. For example,
firm size - whose definition is simply the sample
correlations of R&D expendirtire and market con-
variance of firm sizes.
centration were computed to investigate whether
A11 the concentration measures mentioned
films in concentrated markets were more or less
above arc vciy sensitive to the acUtal market
likely to innovate than films in more competitive
definition that is used. In markets for differentiated
markets. The value of such correlations is limited
products, for example, products may face a
because the observed concentration may be both a
continuum of similar products, and determining
cause and an effect of individual films' behaviour
which similar products exactly constitute a market
and the relationship is shaped by the specifics of the
is not always easy. Take the specific example of the
industry. In order to avoid the ambiguities of such
market faced by US mobile phone services: with
an inter-industry approach, the more recent empir-
just a handfiil of national providers it is
ical microeconomic literature has generally focused
concentrated given the standard concentration
instead on the understanding of firm behaviour
measures. These national films, nevertheless, are
within specific industries, for which the use of
also competing with local companies in various
concentration measures is less relevant.
segments of the market and even
Concentration Ratios
See Also
► Competition
► run Ratio
► Market Structu
Bibliography
Hirschman, A.O. 1964. The paternity of an index. A m e r i -
c a n E c o n o m i c R e v i e w 54: 761-762.
Ravenscrafl, D. 1983. Structure-profit relationships at the line
of business and industry level. R e v i e w o f E c o n o m -
i c s a n d S t a t i s t i c s 65: 22-31.
Scherer, F.M. 1970. Industrial market structure and economic
performance, 2nd ed. Boston: Houghton Mifflin.
Simon, H., and C.P. Bonini. 1958. The size distribution of
business firms. A m e r i c a n E c o n o m i c R e v i e w ’
48:607-617.
Concentration Ratios
William G. Shepherd
These standard indicators of the degree of oligopoly
in markets are used in studying market conditions.
The ratio is the combined market shares of the
‘oligopoly group’ of firms which, being few, are
closely interdependent. The ratio is usually based on
three, four or five firms.
In theory the ratio indicates the market power
held by the interdependent group. When the ratio is
high, a few firms dominate the market and, with
some degree of collusion, can raise prices and
perform other conventional monopoly actions.
Higher concentration makes effective collusion
more probable. If the oligopolists achieve perfect
joint maximation of profits, then the market power
they exert is as great as if the firms were unified
into one dominant firm. Concentration is therefore
an indicator of diluted monopoly power.
How diluted it is depends on the degree of
firms’ cohesion. The fewer the firms, the more
impact a departure from collusion will have on the
joint outcome. Also, such departures will tend to be
discovered more quickly, and therefore be open to
more effective punishment by the other oligopolists.
Therefore, higher concentration tilts
2029
the oligopolists’ choices away from maverick price-
cutting and toward collusion. Accordingly, high
concentration should avoid the disintegration that
often afflicts efforts to fix prices.
The distinction between tight oligopoly (a four-
firm ratio above 60 per cent) and loose oligopoly (a
ratio below 40 per cent) has come to be regarded as
particularly important. In tight oligopoly, collusion
is likely to crystallize effectively into strong
cooperation, as the oligopolists’ common interests
overwhelm the rewards from cheating. Loose oli-
gopoly, by contrast, is seen as a setting for disinte-
gration, where the many oligopolists with low
market shares are jointly unable to avert the
endemic price cutting. This reasoning, which is
broadly confirmed by the common run of experi-
ence in actual markets, suggests that a threshold
value of concentration in the 50-60 per cent range
should present a clear divide between the effective
competition seen in loose oligopoly and the high
market power that tight oligopoly may create.
Concentration is usually second in importance
to individual market shares, as an element of market
structure. Thus, for example, a General Motors may
have a market share of 50 per cent, while it and the
other three leading firms have a combined
concentration ratio of 90 per cent. GM’s own price-
profit, innovation and other results are likely to be
influenced more closely by its own 50 per cent
share - giving it a dominant-firm position - than by
the fact that it is in a market with high
concentration in four firms.
Concentration’s effects, such as they are, may
be modified by entry conditions. Free and vigorous
entry can limit the tight oligopolists’ ability to exert
market power. This is a controversial area, for there
may be few active entrants actually ready to take
advantage of free-entry conditions. Also, entry may
be slow or marginal, so that it does not strongly
affect the core of the market positions held by the
leading firms.
Starting in the 1930s, the ratios soon gained pre-
eminence in studying the degrees of market power.
One was the then-new focus on oligopoly. The
other was the new availability of the actual ratios
for hundreds of US manufacturing industries for the
year 1935. (US and UK ratios appear about every
four or five years, as part of the
2030
industrial census. Only manufacturing industries are
covered, although ratios on urban US banking
markets are prepared by other sources).
Econometric analysis has made extensive use
of these ratios, with scores of papers reporting
regressions relating concentration to price-cost
patterns, growth rates, efficiency, rates of
innovation, etc. In fact, concentration ratios
reigned as the research focus for analysis of
market power and its effects. Indeed, from 1939
to about 1970 the abundant availability of the
ratios reinforced the tendency to regard (perhaps
over-regard) oligopoly as a central issue. With
the growing focus on individual market shares,
the role of concentration ratios became less
important in the 1970s. Even so, the ratios
continue to be an indispensable descriptive
statistic, used widely in research into the several
elements of structure and their effects.
Before summarizing some of the results, one
needs to note that the ratios arc subject to a
serious technical fault. A correct definition of
the market is important, if true concentration
within the market is to be measured. The US
Bureau of the Census has a detailed system of
standard industrial classification (SIC). All sectors
are grouped by numbers, ranging from SIC 1 to
100. The manufacturing sector covers ‘industry
groups’ 20 through 39, and so forth. Differing
degrees of fineness are given. The five-digit
product level is, on balance, about correct in
fitting the average scope of true markets. Yet most
research has focused on the much broader four-
digit 'industries’, which now number about 450
in the US.
About half of these four-digit census 'indus-
tries’ depart seriously from correct market
boundaries. The use of raw census ratios has
undermined a good deal of the past research on
structure and its effects. Commonly, the ratios
arc too broad, lumping together distinct products
and geographic markets (for example, the
national four-firm US ratio is 14 per cent for
newspapers, but concentration ratios in true local
newspaper markets probably average close to 100
per cent). Some cases go the other way, with the
official ratios too high (for example, imports are not
included in the ratios, and this is important for steel,
television sets, shoes, cameras, automobiles and
many others).
Adjusting the ratios to fit true market conditions
Concentration Ratios
weighted average degree of four-firm concentration
in US manufacturing industries has been actually
about 60 per cent, rather than the 40 per cent
indicated by the raw census ratios.
Properly adjusted to eliminate these biases, a
concentration ratio is an excellent descriptive
statistic. It conveys the degree of oligopoly 'tight-
ness'. It can show changes in structure pretty accu-
rately. Thus the market power indicated by a ratio
of 53 or 63 may be a matter of debate, but a rise in
ratio from 53 to 63 for a given industry strongly
suggests that there has been a rise in market power.
By 1975, the ratios have provided enormously
valuable lessons, in several directions. They were
the workhorse statistic in describing the structure of
hundreds of industries, in case sUidics of industries,
in antitrust investigations and in other straightfor-
ward treatments. They had become a pivotal basis
for antitrust policy choices, such as in merger cases,
where the degree of concentration was one basis for
deciding whether to oppose the merger. During
1960-75 they gave rise to many scores of regression
analyses, which tried to relate concentration to the
possible effects of market power.
The correlation of concentration with price
behaviour drew the largest volume of testing. These
estimations were commonly plagued by the use of
uncorrected, raw census ratios, which introduced
substantial errors. Even so, a broad and significant
correlation did emerge, enough to establish a
presumption that the theoretical effects of oligopoly
market power on pricing activity do occur in
practice. The most successful testing involved
multivariate models of five to eight independent
variables, including filter variables such as capital
intensity and growth, and other structure- related
variables such as market shares, advertising
intensity and capital-requirements barriers to entry.
Another important line of research attempted to
explain concentration as the dependent variable.
Growth rates were one possible causative variable,
while economies of scale were another. Growth
emerged as a very weak influence, and scale econo-
mics Uuncd out to explain only a limited amount of
actual concentration in important national markets.
The weakness of these results may, in part, have
reflected errors in the concentration ratios
themselves.
Finally, concentration's possible effects on
innovation, stability' and wealth distribution
Condillac, Etienne Bonnot de, Abbe de Mureau (1714-1780)
have been explored. These studies often were forced
to adopt quite creative approaches, in light to the
data problems. Again, the general patterns have
confirmed the main predictions of theory, that high
concentration tends to affect results much as high
monopoly power does. High concentration was
associated, on the whole, with slowed innovation,
greater instability of production and the
disequalizing of the distribution of wealth.
Yet testing continues on all of these points, and
the concentration ratios have tended only to suggest
patterns, not resolve them. Studies since 1970 have
focused more on individual market shares, showing
them to have stronger effects than concentration,
just as theory predicts. Though the use of
concentration ratios in regression analysis may have
peaked, the ratios (adjusted as appropriate) will
undoubtedly continue as a main basis for describing
the degree of market power in a wide array of
markets in the US, UK, Canada, Japan and certain
other countries.
See Also
► :: »,,-;:ee o:: l./>z : : j
► ILiz.'r;-. 7 :
► Market Structure
References
Bain, J.S. 1956. B a r r i e r s t o n e w c o m p e t i t i o n .
Cambridge, MA: Harvard University Press.
Bain, J.S. 1968. I n d u s t r i a l o r g a n i z a t i o n , Revisedth
ed. New York: Wiley.
Blair, J.M. 1972. E c o n o m i c c o n c e n t r a t i o n . New
York: Harcourt, Brace, Jovanovich.
Chamberlin, E.H. 1933. The theory of
m o n o p o l i s t i c c o m p e t i t i o n , 8th ed. Cambridge,
MA: Harvard University Press. 1962.
Collins, N., and L.E. Preston. 1968. C o n c e n t r a t i o n a n d
price-cost margins in manufacturing
i n d u s t r i e s . Berkeley: University of California Press.
Comanor, W.S., and R.H. Smiley. 1975. Monopoly and the
distribution of wealth. Q u a r t e r l y J o u r n a l o f
E c o n o m i c s 89: 177-194.
Comanor, W.S., and T. Wilson. 1975. A d v e r t i s i n g a n d
m a r k e t p o w e r . Cambridge, MA: Harvard University
Press.
Goldschmid, H.J., H.M. Mann, and J.F. Weston (eds.). 1974.
Industrial concentration: The new
l e a r n i n g . Boston: Little, Brown.
2031
Mann, H.M. 1966. Seller concentration, barriers to entry, and
rates of return in thirty industries, 1950-1960. R e v i e w
o f E c o n o m i c s a n d S t a t i s t i c s 48: 296-307.
Mansfield, E. 1964. Industrial research and development
expenditures. J o u r n a l o f P o l i t i c a l E c o n o m y
72: 319-340.
Scherer, EM. 1980. I n d u s t r i a l m a r k e t s t r u c t u r e
a n d e c o n o m i c p e r f o r m a n c e , 2nd ed. Boston:
Houghton Mifflin.
Shepherd, W.G. 1969. Market power and racial discrimi-
nation in white-collar employment. A n t i t r u s t
B u l l e t i n 14:141-161.
Shepherd, W.G. 1970. M a r k e t p o w e r a n d e c o n o m i c
w e l f a r e . New York: Random House.
Shepherd, W.G. 1979. T h e e c o n o m i c s o f i n d u s t r i a l
o r g a n i z a t i o n . Englewood Cliffs: Prentice-Hall.
Shepherd, W.G. 1982. Causes of increased competition in the
US economy, 1939-1980. R e v i e w o f E c o n o m i c s
a n d S t a t i s t i c s 64: 613-626.
Stigler, G.J. (ed.). 1955. B u s i n e s s c o n c e n t r a t i o n
a n d p r i c e p o l i c y . Princeton: Princeton University
Press.
Stigler, G.J., and J.K. Kindahl. 1970. T h e b e h a v i o r o f
i n d u s t r i a l p r i c e s . New York: Columbia University
Press for the National Bureau of Economic Research.
US Bureau of the Census. 1980. C o n c e n t r a t i o n r a t i o s
Condillac, Etienne Bonnot de, Abbe de
Mureau (1714-1780)
Peter Groenewegen
Keywords
Baudeau, N.; Condillac, E. B. de; Exchange;
Galiani, F.; Le Trosne, G. F.; Physiocracy;
Price; Turgot, A. R. J.; Value; Verri, P.
JEL Classifications
B31
Philosopher and economist. Bom at Grenoble, the
third son of a well-to-do aristocratic family,
Condillac took his name from an estate purchased
by his father in 1720. As a sickly child with poor
eyesight he had little early education and was
apparently still unable to read by the age of 12.
After his father’s death in 1727 he moved to Lyon
to live with his oldest brother, continuing his
education at its Jesuit college. Through this brother
he may have first met Jean Jacques Rousseau, who
was tutor to
2032 Condillac, Etienne Bonnot de, Abbe de Mureau (1714-1780)
his nephews in 1740 and became a life-long friend.
His second brother, l’Abbe de Mably, took
Condillac to Paris in c. 1733 to study theology at
Saint Sulpice and the Sorbonne. He was ordained in
1740 and for the rest of his life ‘ever faithful to die
Christian church, would always wear his cassock,
always remain l’Abbe’ (Lefevre , p. 11).
For the next 15 years he lived the life of a Paris
intellectual, studying the philosophy of Descartes,
Malebranche, Leibniz and Spinoza, ‘to whose
speculative systems he formed a life-long aversion,
preferring the English philosophers Locke (who
particularly influenced his thinking), Berkeley,
Newton and rather belatedly, Bacon’ (Knight , pp.
8-9). In this period he published the works which
made his philosophical reputation: the Essay on the
Origin of Human Knowledge (1746), the Traite des
Systemes (1749), his most famous philosophical
work Treatise on the Sensations (1754) described as
the ‘most rigorous demonstration of the [18th-
century] sensationalist psychology’ (Knight , p. 12)
and his Traite des Animaux (1755).
Apart from giving him entry to the Paris salons,
where at Mile de Lespinasse’s salon he is reputed to
have first met Turgot, another life-long friend (Le
Roy , p. ix), his intellectual reputation gained him
the position of tutor to Louis XV’s grandson, the
Duke of Parma. From 1758 to 1767 he resided in
Parma. Because of its prime minister’s economic
development policies, inspired by a mixture of
‘mercantilism, physiocracy and the ideas of
Goumay’, Condillac developed an interest in
economic matters, an interest ‘indirectly confirmed
by his known contacts with the Italian political
economists, Beccaria and Gherardo’ (Knight , pp.
231-2). In 1768 he returned to Paris, but by 1773
had retired to his estate of Flux near Beaugency,
where he died in 1780. During the last decade of his
life he published his Cours d’Etudes (1775), his
work on economics (1776), a text on logic (1780)
for use in Polish Palatinate schools, and
commenced the unfinished La Langue des Calcids
(1798). In 1752, he became a member of the Royal
Prussian Academy; in 1768 after his return from
Parma he was elected to the French Academy. His
works have been frequently collected, most recently
by Le Roy ( -51).
The impetus for Condillac’s writing Le Com-
merce et le Gouvernement has been ascribed to a
desire to assist his friend Turgot in the difficulties
he faced in 1775 as finance minister over the grain
riots induced by his restoration of the free trade in
grain (Le Roy , p. xxv; Knight , p. 232). This fits
with the work’s unqualified support for free trade in
general and the grain trade in particular (1776, esp.
pp. 344-5, which seems directly inspired by the
Paris events of 1775). Writing the book may also be
explained as a return favour for Turgot’s assistance
in getting Condillac (1775) published (cf. Knight ,
pp. 13, 232). Despite Condillac’s strong support for
this major part of Physiocratic policy and his close
adherence to other aspects of Physiocracy, his
argument that manufacturing was productive
brought critical replies from Baudeau and Le
Trosne ( ). In
this context it may be noted that his work bears
little direct Physiocratic influence, the major influ-
ence being Cantillon ( ), the only work
directly cited apart from Plumard de Dangeul ( ). It
is, however, possible to detect some
influence from the economics of Turgot, Galiani
and Verri on the theory of value, price and com-
petition (cf. Spengler , p. 212).
As published, the work is divided into two
parts. The first provides the elements of the science.
Its starting point is the foundation of value, which
Condillac finds in the usefulness of an object rela-
tive to subjective needs making relative scarcity the
key variable determining value. Value is distin-
guished from price because price can only originate
in exchange. It is determined by the competition
between buyers and sellers guided by their subjec-
tive estimation of value. Gains from exchange arise
from differences in value; for Condillac, value
cannot exchange for equal value. Although
Condillac did discuss the costs of acquiring com-
modities, his emphasis is on exchange, trade and
price. Exchange presumes surplus production and a
need for consumption. Hence trade inspires and
animates production and is essential to increasing
wealth. Only simple pictures of production are
presented: farm labourers producing prime neces-
sities of food and materials; artisans transforming
raw materials into essentials and luxuries; traders
who circulate these products at home and abroad.
By this circulation trade distributes the annual
Condorcet, Marie Jean Antoine Nicolas Caritat, Marquis de (1743-1794)
product and under competitive conditions settles its
true prices. Condillac is more concerned with
developing the institutions associated with trade:
growth of towns and villages, money, banking,
credit, interest and the foreign exchanges, the
defence of property by government and hence the
need for taxation, and the effects of restraints on
trade, including the grain trade. The second part is
almost completely devoted to examining effects of
specific obstacles to trade ranging from war, tariffs,
taxes, excessive government borrowing to luxury
spending in the capital city and exclusive trading
privileges. Moderate wants combined with complete
freedom constitute his recipe for the best form of
economic development.
Condillac’s economic work received a mixed
reception from later economists. J.B. Say ( , and the French enlightenment. New Haven/London: Yale
p. xxxv) described it as an attempt ‘to found a
system of... a subject which [the author] did not
understand’. Jevons ( , p. xviii) praised
Condillac’s ‘charming philosophic work [because]
in the first few chapters ... we meet perhaps the
earliest distinct statement of the true connections
between value and utility...’. Macleod ( described it
as a ‘remarkable work ... utterly neglected but in
scientific spirit... infinitely superior to Smith’. Since
then, it has remained neglected even though as ‘a
good if somewhat sketchy treatise on economic
theory and policy [it was] much above the common
run of its contemporaries’ (Schumpeter , pp. 175-6).
Selected Works
17 46 . An essay on the origin of human knowledge.
T r an s . Th om a s N u g en t . L on do n: N ou r s e ,
17 56 . 1 74 9. Traite des Sysfemes, oil Von en
demele les inconvenine et les advantages.
P a r i s / A m s t e r d am .
1754. Treatise on the sensations. Trans. B.S. Condorcet, Marie Jean Antoine Nicolas
Geraldine Czar. London: Favill Press, 1930.
1755. Trait e des Animaux, oil apres avoir fain des
observations critique sur le sentiment de Des- H. Moulin and H. Peyton Young
cartes et sur celui de M. Bujfon on entreprend d
’expliquer leurs principales facultes . . ..
Am s t e r d a m.
17 55 . Cours d 'Etudepour l 'instruction du Prince
de Parma, Aujourd’hui Ferdinand, Due de
Parma. P a rm a ( an d P a r i s ) .
2033
17 76 . Le Commerce et le gouvernement
consideres relativement 1’una.l 'autre. In
Oeuvres completes de Condillac. V o l . 4 .
Pa r i s : B r i e r e , 1 82 1.
1780. The logic. Trans. B.S. Joseph Neef.
Philadelphia, 1809.
17 98 . Le langue des Calculs, Ouvrage Posthume
et elementaire. P a r i s .
Bibliography
Cantillon, R. 1755. E s s a y o n t h e n a t u r e o f
c o m m e r c e i n g e n e r a l Trans. B.S.H. Higgs.
London: Macmillan, 1931.
Jevons, W.S. 1871. T h e o r y o f p o l i t i c a l e c o n o m y .
4th ed. London: Macmillan, 1911.
Knight, I.F. 1968. The geometric spirit. The Abbe de Condillac
University Press.
Le Roy, G., ed. 1947-51. O e u v r e s p h i l o s o p h i q u e s d e
C o n d i l l a c . Paris: Press Universitaires de France.
Le Trosne, G.F. 1777. De L 'interet Social, par rapport a la
Valeur, d la circulation, d 1 'Industrie, & au commerce
interieur & exterieur: Ouvrage elementaire dans lequel on
discoute quelques Principes de M. I Abbe de Condillac.
Paris.
Lefevre, R. 1966. C o n d i l l a c o u l a j o i e d e v i v r e .
Paris: Editions Seghers.
Macleod, H.D. 1896. T h e h i s t o r y o f e c o n o m i c s .
London: Bliss, Sands & Co..
Plumard de Danguel, L.J. 1754. Remarques sur les avantages
et les desavantages de la France et de la Gr. Bretagne par
rapport au commerce et aux autres sources de la puissance
des etats. Leyden (and Paris).
Say, J.-B. 1805. A t r e a t i s e o n p o l i t i c a l e c o n o m y
of the production, distribution and
c o n s u m p t i o n o f w e a l t h . Trans. C.R. Prinsep,
New American Edition. Philadelphia, 1880; reissued New
York: Kelley, 1963.
Schumpeter, J.A. 1954. F l i s t o r y o f e c o n o m i c
a n a l y s i s . London: Allen & Unwin, 1959.
Spengler, J.J. 1968. Condillac, Etienne Bonnot de. In E n c y -
Caritat, Marquis de (1743-1794)
Keywords
Black, D.; Borda, J.-C. de; Condorcet, Marquis
de; Condorcet’s paradox; Impossibility
 23 19 16 2
Top A B C C
C C B A
Bottom B A A B

colleague in the Academy of Sciences) raise the

same objection against the plurality mle. Suppose,
says Condorcet (Discours preliminaire, p. lviii) that
60 voters have the opinions shown in Table about
three candidates A, B, C.
In the illustration, candidate A wins by plurality.
Yet if we oppose A against B only, A loses (25 to
35) and in A against C, A loses again (23 to 37).
Thus the plurality mle does not convey accurately
the opinion of the majority. From these identical
premises, Borda proposes his well- known scoring
method (each candidate receives 2 points from a
voter who ranks him first, 1 point from one who
ranks him second, and none from one who ranks
him last; hence C is elected with score 78), whereas
Condorcet opens a quite different route.
Condorcet posits a simple binomial model of
voter error: in every binary comparison, each voter
has a probability 1/2 < P < 1 of ordering the
candidates correctly. All voters are assumed to be
equally able, and there is no correlation between
judgements on different pairs. Thus for Condorcet
the relevant data is contained in the ‘majority
tournament’ that results from taking all pairwise
votes:

B beats A, 35 to 25; C beats A, 37 to 23;

C beats B,41 to 19.

Condorcet proposes that the candidates be

ranked according to ‘the most probable combination
of opinions’ (Essai, p. 125). In modem statistical
terminology this is a maximum likelihood criterion
(see Young ).
In the above example the most probable com-
bination is given by the ranking: CBA since the
three statements C over B, C over A, B over A
agree with the greatest total number of votes.
Condorcet’s ranking criterion implies that an
23 17 2 10 8
A Condorcet,
B MarieBJean Antoine
C Nicolas Caritat,
C Marquis de (1743-1794) 2035
B C A A B
C A C B A
Condorcet, Marie Jean Antoine Nicolas Caritat, Mar- Many more useful insights can be discovered in
quis de (1743-1794), Table 2 the Essai. For instance the issue of strategic
manipulations, which has played a central role in
the theory of elections since the late 1960s, is
suggested in places, although it is never system-
atically analysed. For example, on page clxxix of
the Discours Preliminaire, Condorcet criticizes
Borda’s method as more vulnerable to a ‘cabale His
alternative (such as C) that obtains a majority over argument is supported by the modem game
every other alternative must be ranked first. Such an theoretical approach: whenever the configurations
alternative, if one exists, is known as a ‘Condorcet of individual opinions guarantee existence of a
winner’. Condorcet winner, it defines a strategy-proof voting
As Condorcet points out, some configurations of rule. This is one of the principal arguments in
opinions may not possess such a winner, because favour of Condorcet consistent voting rales, namely,
the majority tournament contains a cycle (a rales electing the Condorcet winner whenever it
situation known as ‘Condorcet’s paradox’). He exists (see, for example, Moulin , eh. 4).
exhibits the example shown in Table .
Here A beats B, 33 to 27; B beats C, 42 to 18; C
beats A, 35 to 25. According to Condorcet’s
maximum likelihood criterion, this cycle should be
broken at its weakest link (A over B), which yields
See Also
the ranking B over C over A. Therefore in this case
B is declared the winner.
► : E : : C : 7 3 3
Somewhat later in the Essai (pp. 125-6), Con-
dorcet suggests that one may compute the maxi-
► kiting Paradoxes
mum likelihood ranking of n candidates by, first,
choosing die n(n l)/2 binary propositions that have
file majority in their favour; then, if there are Selected Works
cycles, successively deleting those with smallest
1785. Essai sur l’application de l’analyse a la
majorities until a complete ordering of the candi-
probability des decisions rendues a la plurality
dates is obtained. Unfortunately, for n > 3 this
voix. Paris.
heuristic algorithm does not necessarily yield the
ranking that accords with the greatest number of
votes. An axiomatic characterization of Condorcet’s Bibliography
rale is given in Young and Levenglick ( ).
Arrow, K. 1951. Social choice and individual values. New York:
Condorcet’s idea of reducing individual opin- Wiley.
ions to all pairwise comparisons between alterna- Black, D. 1958. T h e t h e o r y o f c o m m i t t e e s a n d
tives proved essential to the aggregation of e l e c t i o n s . Cambridge: Cambridge University Press.
May, K. 1952. A set of independent necessary and sufficient
preferences approach initiated by Arrow ( ).
conditions for simple majority decision.
The key axiom independence of irrelevant alter- E c o n o m e t r i c a 20: 680-084.
natives (HA) requires that voting on a pair of Moulin, H. 1983. T h e s t r a t e g y o f s o c i a l c h o i c e .
candidates be enough to determine the collective Amsterdam: North-Holland.
Young, H.P. 1986. Optimal ranking and choice from pairwise
opinion on this pair: this generalizes majority
comparisons. In I n f o r m a t i o n p o o l i n g a n d
tournaments by dropping the symmetry across g r o u p d e c i s i o n m a k i n g , ed. B. Grofman and G.
voters and across candidates. In this sense Arrow’s Owen. Greenwich: JAI Press.
impossibility theorem means that the Condorcet Young, H.P, and A. Levenglick. 1978. A consistent extension
paradox is inevitable in any non- dictatorial voting of Condorcet’s election principle. S I A M J o u r n a l o f
M a t h e m a t i c s 35: 285-300.
method satisfying IIA.
2036
Conflict and Settlement
Jack Hirshleifer
All living beings are competitors for the means of
existence. Competition takes the more intense form
we call conflict when contenders seek to disable or
destroy opponents, or even convert them into a
supply of resources. Conflict need not always be
violent; we speak, for example, of industrial
conflicts (strikes and lockouts) and legal conflicts
(law suits). But physical straggle is a relevant
metaphor for these ordinarily non-violent contests.
The Statics of Conflict
Involved in a rational decision to engage in conflict,
economic reasoning suggests, will be the decision-
maker’s preferences, opportunities and perceptions.
These three elements correspond to traditional
issues debated by historians and political scientists
about the ‘causes of war’: Is war mainly due to
hatred and ingrained pugnacity (hostile
preferences)? Or to the opportunities for material
gain at the expense of weaker victims? Or is war
mainly due to mistaken perceptions, on one or both
sides, of the other’s motives or capacities?
Of course it is quite a leap from the choices of
individuals to the war-making decisions of collec-
tivities like tribes or states. Group choice-making
processes notoriously fail to satisfy the canons of
rationality, most fundamentally owing to disparities
among the interests of the individual members. Thus
the internal decision-making structures of the
interacting groups may also be implicated among
the causes of war.
Setting aside this last complication, Figs, and
are alternative illustrations of how preferences,
opportunities, and perceptions might come together
in a simple dyadic interaction. In each diagram the
curve QQ bounds the ‘settlement opportunity set’ -
what the parties can jointly attain by peaceful
agreement or compromise -
Conflict and Settlement
Conflict and Settlement, Fig. 1 Statics of conflict - large
potential settlement region
drawn on axes representing Blue’s income IB and
Red’s income IR. The points PB and PR, in contrast,
indicate the parties’ separate perceptions of the
income distribution resulting from conflict. The
families of curves labelled UB and UR are the
familiar utility indifference contours of the two
agents.
Figure shows a relatively benign situation:
settlement opportunities are complementary, so
there is a considerable mutual gain from avoiding
conflict; the respective preferences display benev-
olence on each side; and the perceptions of returns
from conflict are conservative and agreed (PB and
PR coincide). The ‘Potential Settlement Region’
PSR (shaded area in the diagram), that is, the set of
income distributions such that both parties regard
themselves as doing better than by fighting, is
therefore large - which plausibly implies a high
probability of coming to an agreement. Figure
shows a less pleasant situation: antithetical oppor-
tunities, mutually malevolent preferences, and
divergently optimistic estimates of the returns from
conflict. The PSR is therefore small, and the
prospects for settlement much poorer.
What might be called the materialistic theory
attributes conflict, ultimately, to competition for
resources. Primitive tribes attack one another for
land, for hoards of consumables, or for slaves.
Similar aims evidently motivated barbarian
Conflict and Settlement
Conflict and Settlement, Fig. 2 Statics of conflict - small
potential settlement region
invasions of civilized cities and empires in ancient
times, and European colonial imperialism in the
modem era. Yet, between contending parties there
will almost always be some element of com-
plementary interests, an opportunity for mutual gain
represented by the potential settlement region PSR.
Orthodox economics has always emphasized the
scope of mutual benefit, even to the point of losing
sight of conflict; certain dissident schools, notably
the Marxists, have committed the opposite error.
While a detailed analysis cannot be provided here,
among the factors underlying the relative material
profitability of fighting versus negotiating are
wealth differentials, Malthusian pressures, military
technology, and the enforceability of agreements.
In contrast with the materialistic approach,
attitudinal theories of conflict direct attention to the
respective preference functions. An issue which has
excited considerable interest concerns the relative
weights assignable to genetic versus cultural
determinants of attitudes. One extreme viewpoint,
for example, regards xenophobic wars of family
against family, of tribe against tribe, or nation
against nation, as biologically ‘normal’ in the
human species. An opposite interpretation pictures
man as an innately compliant being, who has to be
culturally indoctrinated into bellicosity.
2037
Finally, what might be termed informational
theories of conflict emphasize differences of per-
ceptions of beliefs. Neoclassical economics tends to
minimize the importance of such divergences -
partly because they tend to cancel out from a large-
numbers point of view, partly because incorrect
beliefs are adjusted by experience in the process of
establishing an economic equilibrium. But conflict
and war are pre-eminently small- numbers,
disequilibrium problems. Indeed, conflict may be
regarded as in a sense an educational process. The
school of actual straggle teaches the parties to
readjust their perceptions to more realistic levels.
Wars end by mutual consent when the potential
settlement opportunities are seen as more attractive
than continued fighting.
The Dynamics of Conflict
Static and dynamic elements are both importantly
involved in conflict or settlement processes. In
game theory terms, the payoff environment,
represented by the familiar normal-form matrix, is
the static element. The dynamic element may be
called the protocol ofplay, as pictured in the game
tree, the protocol specifies the allowable step-by-
step moves in the light of the players’ information at
each stage.
A few very simple payoff environments are
shown in Matrices 1-4. The numbers in each cell
indicate ordinally ranked payoffs for each player, 1
being the poorest outcome in each case. In Matrix 1,
‘Land or Sea’, the environment is characterized
by completely antithetical
(constantsum) payoffs. The other three matrices -
‘Chicken’, ‘Reciprocity’ and ‘Prisoners’ Dilemma’
- represent several of the many different possible
mixed-motive situations combining an element of
opposition of interests with an opportunity for
mutual gain.
The simplest protocol to analyse is one-round
sequential play, first Row selects one of his options,
then Column makes his move in the light of Row’s
choice, and the game ends. In a sequential-play
protocol it is always possible to find a ‘rational’
solution. If Column can be relied to choose his best
final move then Row, knowing
 Defend by Defend
land by sea Soft Tough
Attack by
land 1,2 2,1 Soft 3,3 2,4
Attack by 2,1 1,2 Tough 4,2 1,1
sea

Matrix 3 Matrix 4 PRISONERS’

RECIPROCITY DILEMMA
Soft Tough Co-operate Defect

Soft 4,4 1,3 Co-operate 3,3 1,4

Tough 3,1 2,2 Defect 4,1 2,2

this, can calculate his best first move accordingly.
(This process results in what is called a ‘perfect
equilibrium’.) In contrast, where the protocol dic-
tates that players in a single-round game choose
simultaneously - or, equivalently, where each
chooses in ignorance of the other’s move - solution
concepts are harder to justify. The most commonly
employed is called the ‘Nash equilibrium’ (or
‘equilibrium point’), a pair of strategies from which
neither player would want to diverge unilaterally.
In the ‘Land or Sea’ payoff environment, under
the one-round sequential-move protocol, it is the
second-mover or defender who has the advantage. If
Row moves first, for example, Column can always
successfully counter; e.g. if Row attacks by land,
Column will defend by land. Hence the (1,2)
payoff-pair is the outcome regardless of Row’s
initial move. In military terms the defence has an
intrinsic advantage whenever the attacker must
visibly commit his forces to one or another line of
attack. And, of course, where the defence has such
an advantage neither party is motivated to initiate
warfare through aggression. But if ‘Land or Sea’ is
played under the simultaneous- move protocol, both
parties are groping in the dark and little can be said
with confidence. (Here the Nash equilibrium would
have each side choosing its move at random, in
effect tossing a coin.)
In the payoff environment of ‘Chicken’ (Matrix
2), while the opportunities remain highly
antithetical there is now a mutual interest in
avoiding the disastrous (1,1) outcome that comes
about when both play Tough. In contrast with ‘Land
or Sea’, in the ‘Chicken’ payoff environment the
advantage lies with tht first-mover. Specifically,
Row should rationally play Tough, knowing that
Column then has to respond with Soft. For, Column
must accept the bad (payoff of 2) to avoid the worst
(payoff of 1). If the protocol dictates simultaneous
moves, however, once again the players are groping
in the dark. Under the Nash equilibrium concept
they choose probabilistically, which implies that the
disastrous (1,1) outcome will indeed occur a
percentage of the time. There is a suggestive
application of this model to industrial conflict. If
union (or management) becomes committed to play
Soft, it will be at a disadvantage in negotiations -
the other side will then surely play Tough. But if
both play Tough, there is no hope for peaceful
settlement. Hence each side should rationally adopt
a ‘mixed’ strategy, with the consequence that strikes
and lockouts will occur in a certain fraction of the
dealings.
The ‘Reciprocity’ payoff environment (Matrix
3) is more rewarding to cooperative behaviour. The
idea is that each player would
 Fold Retaliat
e
Refrai Conflict and Settlement
n 2,3 2,3
Attack 3,1 1,2
Conflict and Settlement, Matrix 5
Fold
Fig. 4 Retaliat DETERRENCE
e
Refrai WITHOUT
n 2,3 2,3 COMMITMENT
Attack 3,2 1,1
answer Soft with Soft - leading to the mutually
preferred (4,4) payoffs - but failing this, would
respond to Tough with Tough. If the sequential-
move protocol applies, the first-mover would then
always rationally choose Soft, and so the ideal (4,4)
payoff-pair should be achieved. But under the
simultaneous-move protocol, with each party in the
dark about the other’s move, again the outcome is
quite unclear. In fact there are three Nash equilibria:
pure-strategy solutions at (4,4) and (2,2), and a
mixed-strategy solution as well.
Finally, in the famous ‘Prisoners’ Dilemma’
payoff environment (Matrix 4) the parties are likely
to find themselves in the Defect-Defect ‘trap’ with
(2,2) payoffs, even though (3,3) could be achieved
were each to play Cooperate. Here the ‘trap’ takes
hold under both sequential- move and
simultaneous-move protocols.
The preceding discussion could only be sug-
gestive, limited as it was to 2-player single-round
games, within that category to only a few two-
strategy symmetrical payoff environments, and
finally to the very simplest protocols - excluding,
for example, all negotiations and communications
between the parties. Space limitations permit
comment upon only a few additional points:
Perceptions
Standard game models assume that players know
not only their own payoffs but also their oppo-
nents’. Unintentional error on this score, or else
deliberate deception, may play a crucial role. Sup-
pose two parties in the ‘Reciprocity’ payoff envi-
ronment of Matrix 3 find themselves initially
playing Tough-Tough with outcome (2,2).
2039
Matrix 6
DETERRENCE
REQUIRING
COMMITMENT
Imagine now they are given a chance to shift
strategies under a sequential-move protocol. As
first-mover, Row would be happy to change from
Tough to Soft if only he could rely upon Column to
respond in kind. But Row may, mistakenly, believe
that Column’s payoffs are as in ‘Chicken’, from
which he infers that Column would stand pat with
Tough. Row would therefore not shift from Tough,
hence Column in his turn would not change either.
(Some authors have gone so far as to attribute all or
almost all of human conflict to such mistaken ‘self-
fulfilling beliefs’ about the hostility of opponents,
but of course this pattern is only one of many
possibilities.)
Commitment and Deterrence
In some circumstances the second-mover in point of
time (Column) may be able to commit himself to a
given response strategy before Row makes his first
move. While Column thereby surrenders freedom of
choice, doing so may be advantageous. Consider
threats and promises. A threat is a commitment to
undertake a second-move punishment strategy even
where execution thereof is costly. A promise
similarly involves commitment to a costly reward
strategy. Matrices 5 and 6 illustrate how a threat
works. Row’s choices are Attack or Refrain, while
Column’s only options are to Retaliate or Fold if
Row attacks. Column’s problem, of coiuse, is to
deter Row’s attack. In Matrix 5 Column prefers to
Retaliate if attacked, a fact that - given Row’s
preference - suffices for deterrence. Commitment is
not required. (Since Column prefers to Retaliate,
there is no need to commit himself to do so.) In
Matrix 6 the Column player prefers to turn the other
cheek; if attacked,
2040
he would rather Fold than Retaliate. Unfortunately,
this guarantees he will be attacked! (Note that here
it is not excessive hostility, but the reverse, that
brings on conflict.) But if Column could commit
himself to Retaliate, for example by computerizing
the associated machinery beyond the possibility' of
his later reneging, then deterrence succeeds. In
short, if a pacific player can reliably threaten to do
what he does not really want to do, he won’t have to
do it! (Needless to say, so dangerous an
arrangement is not to be casually recommended.)
The Technology of Struggle
Conflict is a kind of ‘industry’ in which different
'films' compete by attempting to disable opponents.
Just as the economist without being a manager or
engineer can apply certain broad principles to the
processes of industrial production, so, without
claiming to replace the military commander, he can
say something about the principles governing how
desired results are ‘produced ’ through violence.
Buttles typically proceed to a definitive outcome
- victory or defeat. Wars on the whole tend to be
less conclusive, often ending in a compromise
settlement. These historical generalizations reflect
the working of increasing versus decreasing rettims
applied to the production of violence:
1. Within a sufficiently small geographical region
such as a battlefield, there is a critical range of
increasing returns to military strength - a small
increment of force can make the difference
between victory and defeat.
2. But there are decreasing returns in projecting
military power away from one's base area, so
that it is difficult to achieve superiority over an
enemy's entire national territory. The increasing-
returns aspect explains why there is a ‘natural
monopoly' of military force within the nation-
state. The diminishing-returns aspect explains
why a multiplicity of nation-states have
remained militarily viable to this date.
Conflict and Settlement
(However, there is some reason to believe, the
technology of attack through long-range
weapons has now so come to prevail over the
defence that a single world-state is indeed
impending.)
Going into the basis for increasing returns, at
any moment the stronger in battle can inflict a
more-than-proportionatc loss upon his opponent,
thus becoming progressively stronger still. Impor-
tant special cases of this process arc modelled via
Lanchester’s equations. In combat, in the ideal case
where all the military units distribute their fire
equally over the enemy’s line, the process equations
arc:
AB/At = - ARR
AR/At =k B
B
Here B and R arc the given force sizes for Blue
and Red, and the per-unit military efficiencies arc
given by the kB and kR coefficients. It follows that
military strengths arc equal when:
kB B : = -kRR:
But even where military strength varies sensitively
than as the square of force size, it remains quite
generally the case that in the combat process the
strong become stronger and the weak weaker,
leading to ultimate annihilation unless flight or
surrender intcivcnc. (Of course, a skilful com-
mander finding himself with an adverse force
balance will attempt to change the tactical situation
- by timely withdrawal, deception, or other
manoeuvre.)
One implication of increasing returns may be
called the ‘last-push principle’. In the course of a
conflict each side will typically not be fully aware
of the force size and strength that the opponent is
ultimately able and willing to put in the field.
Hence the incentive to stand fast, even at high cost,
lest a potentially won battle be lost. (Foch: ‘A battle
won is a battle in which one will not confess
oneself beaten’.) This valid point unfortunately
tends to lead to battlefield carnage
Conflict and Settlement
beyond all reasonable prior calculations, as expe-
rienced for example at Verdun.
On the other hand, an effective substitute for
force size is superior organization. An integrated
military unit is far more powerful than an equally
numerous conglomeration of individual fighters,
however brave. Organizational superiority, far more
than superiority in weapons, explains why small
I-’ll top can expeditionary contingents in early
modem times were able to defeat even vast
indigenous forces in America, Africa and Asia.
Battles are thus often a contest of organizational
forms; the army whose command structure first
cracks under pressure is the loser.
As for diminishing returns, in the simplest case
an equilibrium is achieved at a geographical
boundary such that:
~ SB*B = — SRXR
Here MB and MR are military strengths at the
respective home bases, Sfi and .SR are decay gra-
dients, and xB and xR are the respective distances
from base. The condition of equality determines the
allocation of territory.
The ‘social physics’ of straggle is of course far
more complex than these simplistic initial models
suggest. There are more or less distinct offence and
defence technologies, first-strike capability is not
the same as retaliatory strength, countering
insurgency is a different problem from central land
battle, etc.
Conflict, Society and Economy
Conflict theory can help explain not only the size
and shape of nations, but the outcomes of com-
petition in all aspects of life: contests among social
classes, among political factions and ideologies,
between management and labour, among
contenders for licences and privileges (‘rent-
seeking’), between plaintiffs and defendants in law
suits, among members of cartels like OPEC,
between husband and wife and sibling and sibling
within the family, and so on. Whenever resources
can be seized by
2041
aggression, invasion attempts can be expected to
occur. Invasive and counter-invasive effort absorb a
very substantial fraction of society’s resources in
every possible social structure, whether egalitarian
or hierarchical, liberal or totalitarian, centralized or
decentralized. Furthermore, every form of human
social organization, whatever else can be said for or
against it, must ultimately meet the survival test of
internal and external conflict.
Notes on the Literature of Conflict (of
Special Relevance for Economists)
Classical military thought from Machiavelli to
Clausewitz to Liddell Hart, though rarely analytical
in the economist’s sense, remains well worth study.
An excellent survey is Edward Mead Earle ( ).
Modem work in this classical genre
understandably concentrates upon the over-
whelming fact of nuclear weaponry and the problem
of deterrence; the contributions of Herman Kahn (
, ) are notable. There is of course
a huge historical literature on conflict and war. An
interesting economics-oriented interpretive history
of modem warfare is Geoffrey Blainey ( ).
William H. McNeill ( ) examines the
course of military organization and technology from
antiquity to the present, emphasizing the social and
economic context. On a smaller scale John Keegan (
) provides a valuable picture of
how men, weapons, and tactics compete with and
complement one another on the battlefield. There is
also a substantial body of statistical work
attempting in a variety of ways to summarize and
classify the sources and outcomes of wars; the best
known is Lewis F. Richardson ( ).
Mathematical analysis of military activity, that is,
quantifiable modelling of the clash of contending
forces, is surprisingly sparse. The classic work is
Frederick William Lanchester ( [1956]).
The modem analysis of conflict, typically com-
bining the theory of games with the rational-
decision economics of choice, is represented by
three important books by economists: Thomas C.
Schelling ( ), Kenneth E. Boulding
2042 Congestion

( ), and Gordon Tullock ( ). Works by

non-economists that are similar in spirit include Congestion
Glenn H. Snyder and Paul Diesing ( ) and
Bruce Bueno de Mesquita ( ). A tangentially Richard Amott and Marvin Kraus
related literature, making use of the rather
mechanical psychologistic approach of Richardson (
), includes a very readable book by
Anatol Rapoport ( ). Abstract
‘Congestion’ is die phenomenon whereby the
quality of service provided by a congestible
facility degrades as its aggregate usage
See Also increases, when its capacity is held fixed. Here,
the economic theory of congestion is developed
► in the context of road traffic. The primary
questions of interest are how the capacity of a
► f;:z.e ' ice congestible facility and its usage fee should be
chosen. This leads naturally to the question of
whether the usage fees collected will be
References sufficient to cover capacity costs at the
Blainey, G. 1973. The causes of war. New York: The Free optimum.
Press.
Boulding, K.E. 1962. Conflict and defense: A general theory.
Keywords
New York: Harper & Brothers.
Bueno de Mesquita, B. 1981. The war trap. New Haven: Yale Clubs; Congestion; Externality cost; First-best
University Press. pricing; Local public goods; Ramsey pricing;
E a r l e , E . M . ( e d . ) . 1 9 4 1 . Makers of modern strategy: Mili- Second-best theory
tary thought from Machiavelli to Hitler. P r i n c e t o n :
Princeton University Press.
Kahn, H. 1960. On thermonuclear war. Princeton: Princeton
University Press.
Kahn, H. 1962. Thinking about the unthinkable. New York: JEL Classifications
Avon Books. R41
Keegan, J. 1976. The face of battle. New York: Viking.
Lanchester, F.W. 1916. Aircraft in warfare: The dawn of the ‘Congestion’ is the phenomenon whereby the
fourth arm. London: Constable. Extract reprinted in the quality of service provided by a congestible facility
world of mathematics, ed. James R. Newman, vol. 4. New
degrades as its aggregate usage increases, when its
York: Simon & Schuster, 1956, 2138-2157.
M c N e i l l , W . H . 1 9 8 2 . The pursuit of power: Technology, capacity is held fixed. We shall develop the
armed force, and society since AD 1000. C h i c a g o : economic theory of congestion in the context of
University of Chicago Press. road traffic, but congestion is pervasive: more
Rapoport, A. 1960. Fights, games, and debates. Ann Arbor:
telephone usage increases the probability of
University of Michigan Press.
R i c h a r d s o n , L . F . 1 9 6 0 a . Arms and insecurity: A encountering a busy line; higher electricity demand
mathematical study of the causes and origins of war. may lead to voltage fluctuations, brownouts and
Pittsburgh: Quadrangle. eventually blackouts; more swimmers in a pool
Richardson, L.F. 1960b. Statistics of deadly quarrels. Pitts-
make comfortable swimming more difficult; more
burgh: Quadrangle.
Schelling, T.C. 1960. The strategy of conflict. Cambridge, patients visiting a medical clinic results in longer
MA: Harvard University Press. waits and lower-quality care; in a more crowded
S n y d e r , G . H . , a n d P . D i e s i n g . 1 9 7 7 . Conflict among classroom, students receive less individual
nations: Bargaining, decision making, and system structure attention, and more time is wasted on administration
in international crises. P r i n c e t o n : Princeton
University Press. and discipline; and so on. The economic theory of
T u l l o c k , G . 1 9 7 4 . The social dilemma: The economics of congestion identifies how the capacity of a
war and revolution. B l a c k s b u r g : University congestible facility and its usage fee should be
Publications.
Congestion
chosen. Some degree of congestion is typically
socially optimal.
The economic theory of congestion has much in
common with the theory of clubs and local public
goods (Scotchmer ). The two literatures examine
similar issues, but the economic theory of
congestion has a policy perspective, while the
theory of clubs and local public goods focuses on
decentralized provision.
Formally, we may define congestion as follows.
Consider a congestible facility in a steady state, that
comprises I congestible elements. (Congestible
elements for a sports stadium, for example, include
nearby roads, parking facilities, the ticket office,
washrooms, concessions, and seating.) Element i is
characterized by a flow capacity, kh and a stock
capacity, Ki: the flow capacity is the maximum
throughput per unit time, the stock capacity the
maximum number of users at a point in time.
Similarly, the level of usage is described in terms of
the throughput of congestible element /, nh and the
number of users at a point in time, Nt. The
congestible facility provides J dimensions of quality
of service, with the level of dimension j indicated
by sj.
Letting k, K,n, N and s denote the corresponding
vectors,
s = S(k,K,n,N). ( 1)
Congestion occurs when there is at least one com-
bination of j and i for which sj is monotone
decreasing in n[ (flow congestion) or N] (stock
congestion), that is, when some dimension of
quality of service falls as the throughput or stock of
users of some congestible element of capacity
increases. This is the static or steady-state definition
of congestion. The dynamic definition of congestion
adds time subscripts to s, k, K, n and N, and appends
equations of motion relating stocks and flows for
the various elements of capacity.
For some congestible elements, such as a turn-
stile, the bottleneck in the Vickrey ( ) bottle
neck model of traffic congestion, or a switching
circuit, the flow capacity constraint is the more
important; for others, such as a telephone line, an
elevator, a swimming pool, or seating at a football
2043
stadium, the stock capacity constraint is the more
important. It should also be noted that a congestible
facility can take the form of a network of
congestible elements of capacity; a natural dis-
tinction is then between link congestion (for
example, highway links) and nodal congestion (for
example, traffic intersections).
To develop the theory, we consider a particular
congestible facility having a single element of
capacity and identical users, that is in a steady state:
a road of uniform width connecting a single entry
point A and a single exit point B, for which an
increase in traffic flow increases travel time and an
increase in road width reduces it. In this context, the
deterioration of quality of service with an increase in
usage is the increase in travel time from an increase
in traffic flow.
We start with the short-run problem of deter-
mining optimal flow and its decentralized attain-
ment, holding road width fixed. Let/denote flow, w
road width, t = t( f w) the travel time function with
(functional subscripts denote partial derivatives) tf >
0 and < 0 and p the value of time. Then the cost to
an individual driver of travelling from A to B, the
user cost, is p t( f w). Total user costs per unit time
equal flow times user cost: pft{ f w). The social cost
per unit time from increasing flow by one unit, with
capacity held fixed, the short-run marginal social
cost, is pt( f, w) + fifty (f, w). The first term is the
user cost of the extra driver; the second, the
congestion externality cost. A driver imposes a
congestion externality by slowing other drivers
down; increasing steady- state flow by one car
increases each car’s travel time by tf(f,w) and social
cost by pfiy (f w).
Figure displays short-run equilibrium. p denotes
trip price, D(p) the aggregate trip demand function,
and uc(f) and snnsc(f ) the user cost and short-run
marginal social cost as a function of / holding w
fixed. With no toll, a user’s trip price equals his user
cost, and equilibrium occurs where the demand and
user cost functions intersect, with flow fe. Assuming
that the marginal social benefit from a trip equals the
corresponding marginal willingness to pay, the
optimum occurs where the demand and short-run
marginal social cost curves intersect, with flow/*.
Thus, with no toll, equilibrium flow is excessive.
2044
Congestion,
Fig. 1 Demand and supply
on a congested highway
Efficiency obtains when economic agents face the
social costs of their decisions and derive the social
benefits from them. In the notoll case, the price of a
trip falls short of its marginal social cost since a
driver does not pay for slowing down other drivers.
Following Pigou ( ), the standard
remedy for internalizing the congestion externality
is to impose a toll equal to the congestion
externality cost, evaluated at the social optimum: t*
in Fig. . This causes the trip price function to shift
up from uc (/) to iic( f ) + t* and equilibrium flow to
fall to the optimal level.
The above argument illustrates the general
principle that efficient utilization of a congestible
facility requires that the price equal short-run
marginal social cost and the toll the congestion
externality cost. Different user types - for example,
cars and trucks - may impose different congestion
externality costs. Efficiency then requires that the
toll be differentiated according to user type.
We now turn to the long-run planning problem
in which both road width and flow are choice
variables. We then consider decentralization of the
optimum. Let B (/) denote the social benefit
Congestion
per unit time from flow/ and C(w) the amortized
capital cost of road width w. (We ignore the com-
plications that arise when the congestible facility is
sufficiently large that its construction alters factor
prices.) The social surplus generated by the road
(per unit time) equals social benefit minus social
cost, and social cost equals total user cost plus
amortized capital cost:
SS(f, w) = B(f) — pftif, w ) - C{w) .
(2 )
It is easily seen from ( ) that the road width that
maximizes social surplus is that which minimizes
social cost. This means that, when the long-run
planning problem is solved, production is carried
out according to the long-run cost structure, and the
short-run marginal cost pricing (which is again)
required for optimal flow is equivalent to long-run
marginal cost pricing:
p = LRMC.
(3 )
Now, recall the basic result of production theory
Congestion
whether LRAC is constant, decreasing or increasing.
Combining this with ( ), we have the result that,
when LRAC is constant,/? = LRA C holds at a long-
run optimum. This is equivalent to equality between
the total value of output and the total cost of output.
Since total user cost is a component of both, this
equality implies equality between toll receipts and
amortized capital cost. Thus, in the case of constant
long-run average cost, the revenue raised from the
optimal toll exactly covers the capital cost
ofproviding a road of optimal width. This is known
as the ‘self-financing’ result. It was first derived by
Mohring and Harwitz ( ) and
subsequently generalized by Strotz ( ) (For a
geometric derivation, see Amott and Kraus ).
The self-financing result extends to congestible
facilities with multiple elements of capacity, mul-
tiple dimensions of quality of service, and multiple
user groups. If a congestible facility exhibits
constant long-run average costs, provision of the
facility can be decentralized via competing ‘clubs’;
competition will result in each club charging each
user a fee for use of its congestible facility equal to
the congestion externality cost he imposes, and
choosing optimal capacity.
The above theory was developed on the
assumption of a steady state. In the extension to
treat nonstationary dynamics, which is conceptually
straightforward, the distinction between flow
externalities and stock externalities becomes
sharper.
The theory relates to first-best pricing and
capacity choice when congestion is the only exter-
nality. When usage entails other externalities, such
as pollution, firstbest pricing should take these into
account. In any policy context, additional practical
constraints that mle out attainment of the full first-
best allocation need to be considered. These are
treated by applying second-best theory (Diamond
and Mirrlees ). Consider, for example, the pricing
problem facing a public transit authority. The
underpricing of urban auto travel may call for the
underpricing of mass transit (Levy-Lambert ;
Marchand ); since optimal lump-sum redistribution
is infeasible for informational reasons, the authority
may choose to sacrifice some efficiency to improve
equity by charging lower fares to needy groups
(Atkinson
2045
and Stiglitz ), rationing, or nonlinear pricing
(Wilson ); administrative costs may preclude fine-
tuning the fare according to distance travelled or
time of day, leading to variants of Ramsey pricing
(Mohring ); the authority may face a deficit
constraint, requiring it to price above marginal
social cost (Boiteux ); with
distortionary taxation, the social cost of financing
an extra dollar of transit authority deficit may
significantly exceed one dollar (Vickrey ); and the
government may choose to deviate from marginal
social cost pricing to provide the public transit
authority with higher-powered incentives (Laffont
and Tirole ) or to achieve political objectives. These
considerations will also cause second-best capacity
to deviate from first-best capacity.
See Also
► iViwi
Bibliography
Amott, R., and M. Kraus. 2003. Principles of transport
economics. In Handbook of transportation science, ed.
R. Hall, 2nd ed. Boston: Kluwer.
Atkinson, A., and J. Stiglitz. 1980. Lectures on public
economics. New York: McGraw-Hill.
Boiteux, M. 1956. Sur la gestion des monopoles publics
astreints a l’equilibre budgetaire. Econometrica 24: 22-
40.
Diamond, P., and J. Mirrlees. 1971. Optimal taxation and
public production I: Production efficiency and II: Tax
rules. American Economic Review 61 (8-27): 261-278.
Laffont, J.-J., and J. Tirole. 1993. A theory of incentives in
procurement and regulation. Cambridge, MA: MIT
Press.
Levy-Lambert, H. 1968. Tarification des services a qualite
variable: Application aux peages de circulation.
Econometrica 36: 564-574.
Marchand, M. 1968. A note on optimal tolls in an imperfect
environment. Econometrica 36: 575-581.
2046
Mohring, H. 1970. The peak-load problem with increasing
returns and pricing constraints. American Economic
Review 60: 693-705.
Mohring, H., andM. Harwitz. 1962. Highway benefits: An
analyticalframework. Evanston: Northwestern Univer-
sity Press.
Pigou, A. 1947. A study in public finance. 3rd ed. London:
Macmillan.
Scotchmer, S. 2002. Local public goods and clubs. In
Handbook of public economics, ed. A. Auerbach and M.
Feldstein, vol. 4. Amsterdam: North-Holland.
Strotz, R. 1965. Urban transportation parables. In The public
economy of urban communities, ed. J. Matgolis.
Washington, DC: Resources for the Future.
Vickrey, W. 1959. Statement on the pricing of urban street
use. Hearings: US Congress, Joint Committee on
Metropolitan Washington Problems, 11 Nov, 454—477.
Vickrey, W. 1969. Congestion theory and transport invest-
ment American Economic Review Proceedings 59: 251-
260.
Wilson, R. 1993. Nonlinear pricing. Oxford: Oxford Uni-
versity Press.
Conglomerates
Alan Hughes
The overall output of a firm may be composed of
activity in more than one product market. The
growth of individual firms will be composed of
changes in the scale of their activities in each of the
markets in which they operate and in the numbers
of those markets. In any period these changes will
consist of horizontal expansion in the market(s) in
which they operated at the beginning of the period
and entry into new markets; where there is a
supplier or buyer relationship with the original
market then this expansion will be vertical
integration. Expansion which fits neither of these
categories is termed diversifying or conglomerate
expansion. Growth in any of these directions may
be by the purchase of new assets (internal growth),
or by the purchase of existing assets through
takeover or merger (external growth). Although it is
common to refer to non-horizontal and non-vertical
expansion as diversified, or conglomerate, the latter
term also has a more specific connotation
emphasizing
Conglomerates
particularly diverse external expansion. It has in
particular been used to mean a company which has
by a deliberate strategy of external growth, often
away from declining sectors, developed a highly
diversified product range which cannot easily be
characterized in terms of a single, or well defined,
group of production technologies, a single set of
major competitors, or a stable place in a well
defined industry group (Steiner 1975; Weston ).
(And in the US context to have financed that
expansion with issues of paper rather than cash,
accompanied by accounting techniques for
consolidating acquired companies designed to boost
earnings per share and make future paper issues
even more profitable (Blair ; Steiner 1975).)
Although diversified businesses predate World
War II (often in the form of financial holding
companies in Europe and Japan) conglomerate
companies in the above sense are essentially a
postwar phenomenon, and have been associated
with the widespread adoption of decentralized
divisionalized management structures. In particular,
in the United States the growth of merger activity in
the 1960s was dominated by diversified
acquisitions, the most spectacular of which were
associated with the emergence of a group of par-
ticularly aggressive conglomerates. For instance,
between 1961 and 1969 ITT, already a very large
multinational telecommunications company,
acquired amongst other concerns, the largest US
bakery, the largest US hotel chain, and the largest
US house builder, the second largest US car rental
service and a number of large insurance and finance
companies. Gulf and Western over a similar period
acquired companies in sugar, tobacco, steel, paper,
banking, insurance and motion pictures (Blair ).
These are extreme examples of a general longer
run tendency for diversification to increase in the
post-war period in all the major industrial
economies (Berry ; Jacquemin and De Jong ;
Utton ). Case studies of corporate growth strategies,
and estimates of levels of, and changes in,
diversification in the 1960s and 1970s reveal that
rapid unrelated product expansion is outweighed by
expansion based on a related product, or ‘narrow
spectrum’ diversification strategy
Conglomerates
(e.g. outside a fairly finely defined (say 3 or 4 digit)
primary product group but within a broader (say 2
digit) industry, of which the primary group is a part
(Wood ; Channon ; Rumelt ; Berry ; Biggadike ;
Utton ; Caves et al. ; Spruill ). Nevertheless, these
studies also suggest an increase in the importance of
unrelated product expansion and a situation has now
been reached where the largest companies in the
major industrial countries have, by the long-term
pursuit of such strategies, come to occupy leading
positions in many different industries (Shepherd ;
Blair ; Utton ) and where the market in corporate
takeovers, and divisionalized management
structures, permit the easy pursuit of further
conglomerate growth by external means (Mueller ).
This has inevitably raised questions about the
relationship between market competition and
conglomerate growth, and about the effects of
conglomerate merger upon corporate performance.
Estimating the impact of conglomerate activity
in an industry upon levels of, and trends in, its
concentration are surrounded by empirical prob-
lems. These arise from lack of precise data on the
market shares of individual firms in different
industries, and on the evolution over time of those
shares. Similar problems limit attempts to measure
the impact of conglomerate entry, either by merger,
or new investment. The evidence suggests that
conglomerate mergers have had little impact on
levels of, or trends in market concentration in the
US (Goldberg , ) whilst in
the UK and the US neither the presence of diver-
sified firms, nor new entry by diversification, seems
to lead to increased levels of concentration. If
anything the reverse seems to be the case (Berry ;
Utton ). Effects on competitive behaviour rather
than market structure are a little better documented.
Here the argument is that operating over many
markets enhances power in each of them
individually (Edwards ). Thus it is argued that
conglomerate firms may impose reciprocal buying
pressures upon suppliers to encourage them to use,
as inputs, the products of other divisions of the
parent conglomerate; may employ predatory pricing
in newly entered markets, cross- subsidized by
activities elsewhere; and practice
2047
mutual forebearance with accepted spheres of
influence agreed with other conglomerates. More
difficult to detect may be other effects claimed to
arise from reductions in potential competition,
where it is argued, for instance, that entry by a large
conglomerate may deter other likely entrants, or
lead to subsequent anti-competitive behaviour,
which could not, or otherwise would not, occur.
Where entry is by merger it may also be argued that
this is at the expense of new investment in the
market, either by incumbents, or the new entrant
itself (Markham ; Steiner ; Scherer ). Examples can
be found in the US, and in the UK and Europe, of
most these practices. There is little to suggest
however that these are persistent, typical, or
pervasive features of conglomerate behaviour.
Where they have been most prominent they appear
to have been due at least as much to individual
market power as to overall conglomerate strength
(Markham ; Scherer ; Utton ).
Corresponding to claims of the anticompetitive
losses which may follow from the spread of
conglomeration are claims of likely benefits. Here it
is claimed that such firms may allocate more
resources per unit of sales to research and
development, since the chances of utilizing spinoffs
and unexpected findings within the organization are
higher (Nelson ); may experience economies of
scope (Panzar and Willig ) and may enjoy lower
costs of raising capital on the stock market in
response to more stable earnings streams, and the
reduced risk of bankruptcy that conglomerate
spread may bring. There is no consistent evidence
to suggest that these effects lead to any superiority
in profit performance, either for the individual firm,
or for industries in which conglomerate firms play
an important role, some studies finding positive,
and others negative, effects on profit levels or
stability (Rhoades , ; Utton ;
Caves et al. ; Kelly ). There is, however, some
evidence to support the view that diversified firms
have higher R&D inputs and patent outputs than
specialized firms, although there is an obvious
problem of causation involved (Wood ; Kennedy
and Thirlwall ; Scherer ). The effects of
conglomerate merger have been much
2048
more extensively investigated than the impact of
conglomerate firms as such. Those studies for the
US which examine periods beyond the stock market
conglomerate boom years of the late 1960s show
such merging conglomerate companies either doing
less well, or at best as well, as other companies (or
portfolios of shares) in terms of profitability, profit
stability and shareholder returns. This is especially
so when profitability measures are used which
allow for the loan financing techniques used to
build up the most spectacular conglomerate empires
in the 1960s merger boom (Mueller ). Evidence for
other countries is more fragmentary but that for the
UK, for instance, suggests that conglomerate
mergers perform relatively well in terms of
profitability compared to mergers in general, though
neither outperform companies relying on internal
growth (Meeks ; Cosh et al. ).
Most of this work may be set in the context of a
static neoclassical perspective to trade off monopoly
welfare losses against efficiency gains. On this basis
it would be hard to mount a significant case for or
against conglomerates. To look at the problem in
this way is however, to distract attention from what
may be regarded as more fundamental questions
about the working of the economic system as a
whole. There are three interrelated issues here.
What are the comparative advantages of organizing
economic activity on the basis of interfirm market
processes as opposed to intrafirm administrative and
organizational processes, what is the impact of the
spread of conglomeracy upon the flow of economic
resources between alternative prospective uses, and
how does diversification and the spread of
conglomeracy affect the flow of information upon
which a market economy is based.
On one view of the world the growth of
divisionalized conglomerate companies may be
regarded, in many instances, as a superior resource
allocation mechanism to interfirm market
transactions. Internal administrative allocative
decisions to move resources between the company’s
individual markets and divisions, it is argued, are
based upon more and better information than that
available to outsiders in the stock market
(Williamson , ; Chandler ,
Conglomerates
). From this point of view appropriately orga-
nized conglomerates are efficient mini-capital
markets, and represent that potent source of new
entry by mobile capital, and adjustment away from
declining sectors which lies at the heart of the
competitive adjustment process (Clifton , and, from
a different perspective, Weston ). Moreover, the
wider the spread of industries covered by any firm,
and the wider the threat of takeover of the
inefficient, or sleepy, or of entry by new
investment, the more forceful this argument
becomes. On the other hand conglomerate firms are
not all embracing in their industrial coverage, their
acquisitions may not be especially driven by
industrial logic or production efficiency, and it may
be argued that in addition they reduce the efficiency
of operation of the capital market itself, given that
financial reporting by large firms is notoriously
aggregative.
In this sense the defining characteristics of
conglomerates are to be found in their internal
financial and administrative arrangements, which
have in principle freed them from the particular
constraints of individual product markets and pro-
duction technologies. They represent the latest
boundary between organization and the market
(Coase ) and as such fit uneasily into any generally
accepted model of the workings of the macro or
micro economy as a whole.
See Also
Bibliography
Berry, C.H. 1975. Corporate growth and diversification.
Princeton: Princeton University Press.
Biggadike, E.R. 1979. Corporate diversification entry
strategy and performance. Boston: Harvard University,
Graduate School of Business.
Blair, J.M. 1972. Economic concentration. New York:
Harcourt Brace Jovanovich.
Caves, R.E. 1980. Industrial organisation, strategy and
structure. Journal of Economic Literature 18(1): 64-92.
Caves, R.E., et al. 1980. Competition in the open economy.
Cambridge, MA: Harvard University Press.
Conjectural Equilibria
Chandler, A. 1962. Strategy and structure: Chapters in the
history of the industrial enterprise. New York: Anchor
Books.
Chandler, A. 1977. The visible hand: The managerial revolution
in American business. Cambridge, MA: Harvard University
Press.
Channon, D.F. 1973. The strategy and structure of British
enterprise. London: Macmillan.
Clifton, J.A. 1977. Competition and the evolution of the
capitalist mode of production. C a m b r i d g e J o u r n a l
o f E c o n o m i c s 2: 137-151.
Coase, R.H. 1937. The nature of the firm. E c o n o m i c a 4 :
386-405.
Cosh, A., A. Hughes, and A. Singh. 1980. The causes and
effects of takeovers in the United Kingdom. In T h e
d e t e r m i n a n t s a n d e f f e c t s o f m e r g e r s , ed.
D.C. Mueller. Cambridge, MA: O.G.H. Publishers.
Edwards, C.D. 1955. Conglomerate bigness as a source of
power. In B u s i n e s s c o n c e n t r a t i o n a n d p r i c e
p o l i c y , ed. G. Stigler. New York: National Bureau for
Economic Research.
Goldberg, L.G. 1973. The effect of conglomerate mergers on
competition. J o u r n a l o f L a w a n d E c o n o m i c s
16(1): 137-158.
Goldberg, L.G. 1974. Conglomerate mergers and concen-
tration ratios. R e v i e w o f E c o n o m i c s a n d
S t a t i s t i c s 56(2): 303-309.
Hughes, A., and A. Singh. 1980. Mergers concentration and
competition in advanced capitalist economies: An
international perspective. In D e t e r m i n a n t s a n d
e f f e c t s o f m e r g e r s , ed. D.C. Mueller. Cambridge,
MA: O.G.H. Publishers.
Jacquemin, A.P., and H.W. De Jong. 1977. E u r o p e a n
i n d u s t r i a l o r g a n i s a t i o n . London: Macmillan.
Kelly, M. 1980. The effects of diversification on market
structure and monopoly power. In M e r g e r s a n d
e c o n o m i c p e r f o r m a n c e , ed. K. Cowling et al.
Cambridge: Cambridge University Press.
Kennedy, C., and A.P. Thirlwall. 1972. Technical progress: A
survey. E c o n o m i c J o u r n a l 82: 11-72.
Markham, J.W. 1973. C o n g l o m e r a t e e n t e r p r i s e
a n d p u b l i c p o l i c y . Boston: Harvard Graduate
School of Business.
Meeks, G. 1977. Disappointing marriage.
Cambridge: Cambridge University Press.
Mueller, D.C. 1969. A theory of conglomerate mergers.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 83(4): 643-
659.
Mueller, D.C. 1977. The effects of conglomerate mergers. A
survey of the empirical evidence. J o u r n a l o f
B a n k i n g a n d F i n a n c e 1(4): 315-347.
Mueller, D.C. (ed.). 1980. The determinants and effects of
mergers: An international comparison. New York: O.G.H.
Publishers.
Nelson, R. 1959. The simple economics of basic scientific
research. J o u r n a l o f P o l i t i c a l E c o n o m y 67:
297-306.
Panzar, J.C., and R.D. Willig. 1981. Economies of scope.
American Economic Review: Papers and
P r o c e e d i n g s 71(2): 268-272.
2049
in 1963. Review of Economics and Statistics 55(2): 146-155.
Rhoades, S.A. 1974. A further evaluation of the effect of
diversification on industry profit performance. R e v i e w
o f E c o n o m i c s a n d S t a t i s t i c s 56(4): 557-559.
Rumelt, R.P. 1974. S t r a t e g y structure and
economic performance. Boston: Harvard
Graduate School of Business.
Scherer, F.M. 1979. I n d u s t r i a l m a r k e t s t r u c t u r e
a n d e c o n o m i c p e r f o r m a n c e , 2nd ed. Chicago:
Rand McNally, 1980.
Shepherd, W.G. 1970. M a r k e t p o w e r a n d e c o n o m i c
w e l f a r e . New York: Random House.
Spruill, C.R. 1981. C o n g l o m e r a t e s and the
e v o l u t i o n o f c a p i t a l i s m . Carbondale: Southern
Illinois University Press.
Steiner, P.O. 1976. M e r g e r s : M o t i v e s , e f f e c t s ,
p o l i c i e s . Ann Arbor: University of Michigan Press.
Utton, M.A. 1979. Diversification and
c o m p e t i t i o n . Cambridge: Cambridge University
Press.
Weston, J.F. 1980. Industrial concentration, mergers and
growth. In Mergers and economic
e f f i c i e n c y , vol. I. Washington, DC: U.S. Government
Printing Office.
Williamson, O.E. 1975. Markets and hierarchies: Analysis and
anti-trust implications. New York: Free Press.
Williamson, O.E. 1985. T h e e c o n o m i c i n s t i t u t i o n s
o f c a p i t a l i s m . New York: Free Press.
Wood, A.J.B. 1971. Diversification, merger and research
expenditure: A review of empirical studies. In T h e
c o r p o r a t e e c o n o m y , ed. R. Harris and A.J.B.
Wood. London: Macmillan.
Conjectural Equilibria
F. H. Hahn
Abstract
In imperfectly competitive economies, agents
must take note of the effects of their decisions
on the market environment. Such effects, being
uncertain, are the subject of conjecture. Even if
conjectures are not derivable from some first
principles of rationality, conjectural theories are
of interest because they attempt a general
equilibrium analysis of non-perfect competition.
The conjectural approach takes proper and
explicit note of the perceptions by individuals of
their market environment; it is possible that
what is the case may depend on what agents
believe to be the case.
2050 Conjectural Equilibria

Keywords for some agent(s) to change prices. This causes

Bootstrap equilibria; Conjectural equilibria; difficulties when the economy is perfectly com-
Duopoly; Extensive form games; Fixprice petitive (Arrow ) since it implies that the agent can
equilibria; Fixprice models; Game theory; influence his market environment by his own
General equilibrium; Imperfect competition; No actions. That is one reason why a fictitious
surplus condition; Perfectly competitive auctioneer has been introduced to account for price
equilibrium; Rational conjectural quilibrium; changes.
Reasonable conjectures; Sequential equilibrium When the economy is not perfectly competitive
these difficulties are avoided. A price will be
changed if some agent conjectures that such a
change would be to his advantage. As a corollary
JEL Classifications then a conjectural equilibrium must be a state from
D5 which it is conjectured by each agent that it would
be disadvantageous to depart by actions which are
In an economy with very many agents the market under the individual agent’s control. (For a formal
environment of any one of these is independent of definition see below.)
the market actions he decides upon. More generally But there are other difficulties. In particular,
one can characterize an economy as perfectly there is the question of the source of conjectures. If
competitive if the removal of any one agent from the these are taken as given exogenously then there are
economy would leave the remaining agents just as many states which could be conjectural equilibria
well off as they were before his removal. (The for some conjectures. It should be noted that a
economy is said to satisfy a ‘no surplus’ condition; similar objection can be raised in conventional
see Makowski ; and Ostroy . ) When an economy is equilibrium analysis. There it is the preferences of
not perfectly competitive, an agent in making a agents which are taken as exogenous and there too
decision must take note of its effect on his market there are many equilibria which are compatible with
environment, for example, the price at which he can some (admissible) preferences. However, while
sell. This effect may not be known (or known with conjectures may turn out to be false and this may
certainty) and will therefore be the subject of occasion a change in conjectures, it is less easy to
conjecture. A conjecture differs from expectations point to equally simple and convincing endogenous
concerning future market environments which may, mechanisms of preference change. For that reason
say, be generated by some stochastic process. It is one may feel that conjectural equilibrium requires
concerned with responses to the actions of the agent. that conjectures are in some sense correct
In the first instance then the topic of conjectural
(‘rational’). For if they are not they will change in
equilibria is that of an economy which is not the light of experience. This argument is considered
perfectly competitive by virtue of satisfying a no below.
surplus condition. But, as we shall see, an economy The reason why the idea of conjectural equi-
could fail to satisfy this condition and yet have a libria is of interest is that economies which are not
perfectly competitive equilibrium. intrinsically perfectly competitive (for example,
By an equilibrium in economics we usually because of the large number of agents) are of
mean an economic state which is a rest (critical) interest and because it allows one to study price
point of an (implicit) dynamic system. For instance, formation without an auctioneer.
it is postulated in the textbooks that, when at going
prices the amount agents wish to buy does not equal
the amount they wish to sell, prices will change. An Illustration
Strictly this should mean that there would, in such a
situation, be an incentive Consider two agents each of whom can chose an
action at from a set of action A,. Let A = Ax / A2
Conjectural Equilibria 2051

with elements a («,, a,). Then a conjecture Cj is a conjectures are less arbitrarily imposed on the theory
map from A x A, to AT written as not much has been gained - almost any pair of
actions could be a conjectural equilibrium. A first
Ci = 6i(a, a')- attempt to remedy this is to ask that conjectures be
correct (rational). If that is to succeed in any simple
Its interpretation is this: given the actions ofthe fashion it will be necessary to suppose that each
two agents (a), C, is the action of j conjectured by i agent has a unique best action under this conjecture.
to be result from his choice of <1 ,. (In a more This is very limiting and it means that some of the
general formulation the conjecture can be a prob- classical duopoly problems cannot be resolved in
ability distribution but that is not considered here.) this way.
We require conjectures to be consistent'. Let the status quo again be a*. Then if 0j and 02
are correct conjectures it must be that

6i(a, at) = aj ( 1)
V 2{e* 1 (a*,a 2 ),e* 1 [(a u a* 1 ) ; e* 2 (a*,a 2 )\}
This says that if agent i continues in his action at x > vijflj, 6'1 [(aj,a 2 ), d\\ } all flj = A2.
then he conjectures that j will do likewise. (This use (3)
of the word ‘consistent’ is not that of Bresnahan ,
and others who use it to mean ‘correct’.) vi { 6 * 2 ( a * , a 2 ) , [(aj, a 2 ) , 0*2{a*, a2)] }
Suppose now that there is a function v from A
to R2, written as v(a) = \\’\{a), v2(a)], which gives > v i | a s , 0 [ ( a [ , a 2 ), a j | a l l f l j = A i. ( 4 )
the payoffs to the agents as a function of their joint
action a. Consider a* to be one such joint action.
A rational conjectural equilibrium is then a con-
One says that a* is a conjectural equilibrium for the
jectural equilibrium a* (with conjectures 9\(j, 0*2 (•)
two agents if
which satisfy ( ) and ( )). It must be re-emphasized
that such an equilibrium may not exist for some A
v1 [«/, 6j(a*,aj)\ < Vi[a*,6i(a*)] all a; eA,-,/ =
1,2 and v (see Gale ; Hahn )•
However, the idea is simple and, where appli-
(2)
cable, coherent. It has however been criticized (in a
That is, the joint action a* is a conjectural equi- somewhat intemperate and muddled paper) by
librium if no agent, given his conjecture, believes Makowski ).This criticism appears to have had some
that he can improve his position by deviating to a appeal to some game theorists who like to think of
different action. games in extensive form (which they sometimes like
It is not the case that conjectural equilibrium, as to call dynamic). The criticism is this: when agent 1
defined, always exists. For instance in the case of a deviates from a* he is interested in the payoffs
duopoly in a homogeneous product where the action which he will get given this deviation and agent 2’s
is ‘setting the price’, v may not be concave and a response. This payoff Makowski thinks of as
sensible conjecture may have discontinuities. One accruing in the ‘period’ after agent l’s deviation. But
thus needs special assumptions to ensure existence when agent 2 responds in that period he is interested
or one must face the possibility that agents do not in this payoff in the period following this response.
chose actions but probability distributions over So the agents expect ‘the game to end’ in different
actions (mixed strategies); for example, Kreps and periods (Makowski , p. 8). Moreover, after agent 2
Wilson ( ) in has responded, agent 1, in his turn, will again want
their work on sequential equilibrium employ con- to respond, that is, deviate from the deviation he
jectures which are probability distributions. started with. This criticism is then illustrated with an
Supposing that a conjectural equilibrium exists, example in which one agent
one may reasonably argue that until
2052 Conjectural Equilibria

expects the other to return to the status quo after he > u u(0/)
*($)'*{* yj
has deviated from it. (6)
All of this, however, is wrong. Firstly, if one i,j=l,2,6le6.
wants to give a time interpretation to conjectures
and so forth, then actions must be thought of as This is just a Nash equilibrium where conjectures
strategies. That is, the deviating agent deviates in are interpreted as strategies (Hart ).
one or more elements of his plan over the whole While this is still quite demanding, it is signif-
length of the game (perhaps infinite). Under correct icantly weaker than ( ). If equilibria exist they may
conjectures responses and counter-responses are be ‘bootstrap equilibria’, that is, they will depend on
taken into account in evaluating the benefits of beliefs about the actions of others, which beliefs
deviation. Hence, and secondly, a deviating agent is may be incorrect. There is certainly no ground for
in this situation never surprised by the response of believing that they will be efficient.
the other, which therefore does not lead him to One can go one step further in the direction of
further revise his deviation. On the definition, agent plausibility by requiring that conjectures be rea-
1 expects the response to his deviation to be Oja*, sonable only for small, or infinitesimal, deviations
ax). Suppose this gives a2 which is correct. Then that from the status quo. After all, large experiments are
agent knows that the new status quo will be (ai_ a f) likely to be costlier than small ones. This will allow
= a and if he has calculated benefits correctly he a larger class of reasonable conjectures and
will not wish to deviate again. equilibria.
However, there is the following to be said in
favour of Makowski’s criticism. Deviations in General Conjectural Equilibrium
strategies may not be observable by the other agent.
It is fair to say that at present general equilibrium
Therefore in traditional duopoly models with a
theory is in some way complete only for a perfectly
sequential structure the re-interpretation of actions
competitive economy, that is, one where the returns
as strategies may be inappropriate. There is some
to an individual agent are just equal to the contri-
evidence that in the duopoly literature with
bution which he makes (Makowski ; Ostroy ). In
conjectures the consequent difficulties have not
general (although there are exceptions) such an
always been appreciated. It is also the case that too
economy exists when it is large (for example, it
little attention has been paid to the assumption of a
consists of a non-atomic continuum of agents). But
unique best response on which the above
there is now another possibility: an economy can be
formulation depends.
perfectly competitive if agents conjecture that their
An alternative to rational conjectures are rea-
market actions will have no effect on the prices at
sonable conjectures (Hahn ). A conjecture is
which they can trade.
reasonable if acting on any other conjecture would
The following assertion will be clear from what
lower profits given the conjectures of other firms.
has already been discussed. Let us say that an
Suppose that 6 is the set of all possible consistent
economy is intrinsically perfectly competitive if it
conjectures. For any 9, c d; assume that there is a
satisfies the no-surplus condition. Then perfectly
unique optimizing choice of output by firm i of
competitive conjectures are rational if an economy
yjOj. Then z’s conjecture (fj c 9 is reasonable if
is intrinsically perfectly competitive. But perfectly
given/th conjecture Of.
competitive conjectures can be reasonable even
when the economy is not intrinsically perfectly
competitive. That is, conjectures may be such that,
if an agent acts on any conjecture other than the
v, a i i ei e e.
perfectly competitive one, his profits will be lower.
(5) For instance, this may even be the case for two
duopolists with constant marginal costs whose
But then a reasonable conjectural equilibrium is a conjectures refer to the price
pair (dp d!}) each in 0 such that
Conjectural Equilibria
charged by the rival firm. It will also be clear that if
we do not require conjectures to be either rea-
sonable or rational then, in general, conjectures can
be found to support a competitive equilibrium in an
economy which is not intrinsically perfectly
competitive.
In a general equilibrium context it is not clear
what it is that firms are supposed to conjecture. In
some sense the conjecture must refer to the reaction
of the whole economy to the action of the
conjecturing agent. In other words, it is not obvious
how to define a game which adequately represents
the economy. But in what sense?
Consider an economy with n produced goods
and m non-produced goods. For simplicity suppose
that all firms are single-product firms and that all
firms producing the same good are alike, including
their conjectures. There are very many households
whose reasonable conjectures are always the
competitive one. Households receive the profits of
firms. Since the action of any one firm can affect the
prices at which households can trade it is not at all
clear what it is in the households’ interest that the
firms should maximize (Gabszewicz and Vial ). If
all households are alike it could be their common
utility function, but that seems far removed from the
world. I shall arbitrarily assume that firms maximize
then- profits in terms of one of the non-produced
goods, say the first. This is arbitrary but it seems to
me equally dubious to suppose that firms always
choose in the ‘best interests of shareholders’,
especially when that interest is often difficult and
sometimes impossible to define.
Let p e R", w e R™ 1 be the price vectors in
terms of good m of produced and non- produced
goods respectively (so wm= 1). Let yk G Yj c R"+m be
the production of firm j where y,; > 0 is its output of there is scant justification (Hahn ).
good j, ytl < 0 is an input of good i, produced or non-
produced. Let y Ly;-, where yy C_ all j. Let z c
R™ be the endowment
of non-produced goods and
F = {y| y > (0, -z)}
so that F is the set of feasible net production vectors ( ) is not imposed for firm k. One then requires that
Y. Let 0h/ be the share of household h in firm/.
2053
Given any y £ F we think of each household as
endowed with a certain strictly positive stock of
non-produced goods and 0;,;T; of the production of
firm j. To avoid unnecessary complications assume
0;,; (j = 1, ... , n). to be such that if zh is the stock of
non-produced goods owned by household h:
For ally eF : zh + e'vyj ^ 0 aH h- (7)
j
Households consume both types of goods. Hence
for any y e F there is now an associated pure
exchange economy where each household’s
endowment is given by ( ). Making the usual
assumptions there will exist at least one equilibrium
\p(y), w(y)]. Suppose for the moment that there is
only one for each y e F.
Now firm j in this equilibrium observes \p(y),
w(y)] and will deviate from y} (if it deviates at all) if
it can thereby increase its conjectured profits. Let
p{y)My)Jj
be the conjectural profit function of firm j. Then y°,
p(y°),w(y°) is a conjectural equilibrium if for ally =
1,. •n:
p(y°),w(y°),y0
(8)
x > 7ij ' p ( y ° ) My ° ) , y j all y,eF.
Such a conjectural equilibrium will exist if all
TIJ(-) are quasi-concave, an assumption for which
If we demand that conjectures be rational then
conjectured and actual profit must coincide for all
yt(thc two coincide for yk = yk by the requirement
that conjectures be consistent). One proceeds as
follows. Let yk y{)k . Given the conjectures of the
remaining firms find the conjectural equilibrium of
the economy p{y*(k), w\y*(k), y*(kj\), where y(k) is
the vector y with yk in the Ath place and condition
for allyk> 0
2054 Conjectural Equilibria

P(y°),w(y°),yk X k { p \ y * { k ) \ , w \ y * ( k ) \ , y k } (J = 1, ... , /?). The islands are small replicas of

71 k
(8a) the whole economy. But households have shares in
firms on all islands so that if there are enough
islands their share in any firm on their own island is
where nk{-) is actual profit. For rational conjectures very small. That means that any firm can disregard
this should be true for all k. the effect of a change in its own profits on the
It will be seen that rational conjectural equilib- demand for the good it produces. To make this work
rium is very demanding. For a certain class of one supposes that produced goods are totally
conjectures it will not even exist (Gale ; Hahn ). immobile between islands while non- produced
More importantly, the whole procedure breaks down goods are totally mobile. By an appropriate
if given a deviation by k, the conjectural assumption on consumers on each island one
equilibrium, is not unique. Lastly, even if by ensures that they all have the same demand. Lastly,
sufficient assumptions one overcomes these since shares in a firm are held on many different
difficulties, it is not agreeable to common sense to islands the firm, in acting in the shareholder’s
suppose that firms can correctly calculate general interest is justified in neglecting the effect of its
equilibrium responses to their actions, nor is it actions on relative prices on its own island and so is
obvious that they should always be concerned only justified in maximizing profits.
with equilibrium states. From the point of view of conjectural equilib-
Reasonable conjectures do not fare much better, rium the island assumption allows firms (both
although a notable contribution to their study has reasonably and rationally) to ignore effects of their
recently been made by Hart ( ). Hart own actions on w - the price vector of non-
notices that conjectures of firms induce a supply produced goods. These will be determined by
correspondence (not generally convex) on their part. demand and supply over all islands and in this
Here let us suppose that we can in fact speak of determination any one firm can be regarded as
supply functions. These can be thought of as playing a negligible role. This is some gain in
strategies in a manner already discussed. A realism. But after all allowances have been made it
reasonable conjectural equilibrium then satisfies the is still hue that (a ) the assumptions required for the
condition that, given the supply functions of other existence of reasonable conjectural equilibrium are
firms, no deviation by firm k to another supply uncomfortably strong and (b) even when that is
response can increase its profits. In ( ) one then neglected such an equilibrium seems to have small
substitutes on the right-hand side fory(, T}'j\p(y°), descriptive power.
w(y°)],an admissible supply function (see Hart ) of j
and requires the inequal
ity to hold for all such functions. Of course, one has Simpler Approaches

Negishi ( ) made the first, justifiably cele

y° = vMy°)My0)]
for a reasonable conjectural equilibrium.
To show existence of such an equilibrium will
require strong assumptions. The technicalities will
be found in Hart ( ). However, one of
the assumptions which he makes is not only tech-
nically useful but economically sensible since it
leads firms to face a simpler task in forming con-
jectures. Hart supposes the economy to consist of a
number of islands each of which has many
consumers and one firm of each type
brated, attempt to incorporate imperfect competition
in general equilibrium analysis. He did this by
letting single product firms have consistent inverse
demand conjectures (the case he studies most
thoroughly makes these linear). Consistency is all
he asked for of conjectures but he also needed the
uncomfortable postulate that the resulting
conjectural profit functions be quasiconcave. Later
Hahn ( ), Silvestre ( )
and others added the requirement that, besides being
consistent, the conjectured demand functions have,
if differentiable, the correct slope at
Conjectural Equilibria
equilibrium (that is, that the conjecture be infini-
tesimally or ‘first order’ rational). It turns out that
this extra requirement does not much restrict con-
jectures, nor thus the set of equilibria which can be
generated by some conjectures. The reason roughly
is this: in conjectural equilibrium, when conjectured
profit functions are twice differentiable, the partial
derivative of the conjectured profit function of firm
j with respect to its own output much vanish.
Suppose the economy to be in such an equilibrium
and consider an infinitesimal output deviation by
firm k. To find the equilibrium which ensues,
differentiate all equilibrium relations, other than that
for firm k, with respect to the output of firm k.
Amongst these will be the condition that the
marginal profit conjectured of every firm (other than
k), be zero. Hence differentiation of that condition
will yield second-order terms. But we can choose
these arbitrarily since we are requiring only first-
order rationality. One can show in fact that these
second-order terms can be chosen so as to make the
first-order conjectured change in profit of any firm k
correspond to the actual change. (Details in Hahn .)
Hence first-order rationality imposes few
restrictions.
Both Hahn ( ) and Negishi ( ) have
also considered kinked conjectures. The idea is this.
If an agent can transact at the going price as much
as he desires his conjectures are competitive. If he is
quantity constrained (for example, if a firm cannot
sell an amount determined by equality between
marginal cost and price) his conjectures are non-
competitive. That is, he considers that a price
change is required to relax the quantity constraint.
The fixprice methods of Dreze ( ) and
others can be interpreted as an extreme form of such
conjectures - for instance to relax a constraint on
sales, price, it is conjectured, must be reduced to
zero.
To such conjectures there have been two objec-
tions. Firstly, they assume that an agent’s conjec-
tures are not influenced by constraints on others. For
instance, a firm which can hire as much labour as it
wants at the going wage while workers cannot sell
as much as they like does not conjecture that it
could have the same amount of labour at a lower
wage. To this one can answer that it is not easy for
an agent to observe the quantity
2055
constraints on others. For instance, unemployment
statistics do not tell us whether workers have chosen
not to work or whether they are constrained in their
sale of labour. None the less, this objection has
some force and needs further study with proper
attention to the information of agents.
The other objection is that these kinked con-
jectures are not explained. That is tme if explanation
turns on what an agent knows or can learn. None the
less, the hypothesis seems to be to have
psychological verisimilitude. If I can always sell my
labour at the going wage there is little occasion for
the difficult conjecturing of what would happen if I
raised my wage. This is not so if I find that I cannot
find employment at the going wage.
In any event these simpler approaches allow one
to incorporate traditional monopolistic competition
in a general equilibrium framework. Of course,
some of the assumptions such as concave
conjectured profit functions are strong. On the other
hand, one can now allow for a certain amount of
increasing returns (Silvestre ).
Some Conclusions
The conjectural approach has this merit: it takes
proper and explicit note of the perceptions by
individuals of their market environment. Economic
theory perhaps too often neglects the possibility that
what is the case may depend on what agents believe
to be the case. Historians and others have long since
studied the intimate mutual connection between
beliefs and events but economists have not made
much headway here. The conjectural approach is
perhaps a small beginning. For it deals with the
theories agents hold and this must plainly enter into
our theory of agents.
In particular one should not pay too much
attention to the objection that conjectures may not
be derivable from some first principles of
rationality. It seems to me quite proper to find their
description in history. Nor, as has been argued, will
an appeal to learning render conjectures in some
sense objectively justifiable. This is clear from the
discussion of reasonable
2056
conjectures and from the costs of experimentation.
For hundreds of years witches were burned in the
light of a reasonable theory which few would now
regard as having proper objective correlatives.
There is no reason to suppose that it is possible for
businesses or governments now to do better than
some of the best minds of the past.
From a more immediately relevant standpoint,
conjectural theories are of interest because they
attempt a general equilibrium analysis of nonperfect
competition. It is good to know that in a proper
sense perfectly competitive economies can be
viewed as limiting Cournot conjectural equilibrium
economies (Novshek and Sonnenschein ). But this
knowledge does not contribute to the study of
properly imperfectly competitive economies. Again
the study of fixprice equilibria has borne some
fruits, but not those which were first sought by
Trillin ( ) when he proposed a
framework for general equilibrium with monopo-
listic competition. If it is the case that actual
economies are not perfectly competitive nor that
they behave ‘as if’ they were, then the task set by
Triffin requires serious attention, and it is likely that
con jectural theories will have a role to play.
Recent developments in game theory (for
example, Kreps and Wilson ) suggest that these two
conjectures will have to play a part. Indeed, quite
generally in that theory players conjecture that their
opponent is ‘rational’ in an appropriate sense. It is
not the case that the conjectural equilibrium
approach is an alternative to the game theoretic one.
See Also
Bibliography
Arrow, K.J. 1959. Toward a theory of price adjustment.
In M. Abramovitz et al., The allocation of economic
resources. Stanford: Stanford University Press.
Bresnahan, T.F. 1981. Duopoly models with consistent
conjectures. American Economic Review 71: 934—945.
Dreze, J. 1975. Existence of equilibrium under price rigid-
ity and quantity rationing. International Economical
Review 16: 301-320.
Conspicuous Consumption
Gabszewicz, J.J., and J.D. Vial. 1972. Oligopoly ‘a la
Cournot’ in general equilibrium analysis. Journal of
Economy Theory 4: 381-400.
Gale, D. 1978. A note on conjectural equilibria. Review of
Economic Studies 45 (1): 33-38.
Hahn, F.H. 1977. Exercise in conjectural equilibria.
Scandinavian Journal of Economics 79: 210-226.
Hahn, F.H. 1978. On non-Walrasian equilibria. Review of
Economic Studies 45: 1-17.
Hart, O. 1982. Reasonable conjectures, Theoretical eco-
nomics paper no. 61. STICERD, London School of
Economics.
Kreps, D.M., and R.B. Wilson. 1982. Sequential equilibria.
Econometricas 50: 863-894.
Makowski, L. 1980. A characterisation of perfectly com-
petitive economies with production. Journal of Eco-
nomic Theory 22: 208-221.
Makowski, L. 1983. 'Rational conjectures' aren't rational
and 'reasonable conjectures ' aren't reasonable , Dis-
cussion paper no. 60. SSRC Project on isk,
Information and Quantity Signals, Cambridge
University.
Negishi, T. 1960. Monopolistic competition and general
equilibrium. Review of Economic Studies 28: 196-202.
Negishi, T. 1979. Micro-economic foundations of Keynesian
macro-economics. Amsterdam: North- Holland.
Novshek, W., and H. Sonnenschein. 1978. Cournot and
Walras equilibrium. Journal of Economic Theory 19:
223-266.
Ostroy, J. 1980. The no-surplus condition as a characteri-
sation of perfectly competitive equilibrium. Journal of
Economic Theory 22: 183-207.
Silvestre, J. 1977. A model of general equilibrium with
monopolistic behavior. Journal of Economic Theory 16:
425-442.
Triffin, R. 1940. Monopolistic competition and general
equilibrium theory. Cambridge, MA: Harvard Univer-
sity Press.
Ulph, D. 1983. Rational conjectures in the theory of oli-
gopoly. International Journal of Industrial Organization
1 (2): 131-154.
Conspicuous Consumption
F. Stankovic
Keywords
Conspicuous consumption; McCulloch, J. R.;
Rae, J.; Say, J.-B.; Scarcity; Smith, A.; Vanity;
Veblen effect; Veblen, T
Constant and Variable Capital 2057

JEL Classifications life is the passion for effective accumulation:

D1 ‘Articles of which consumption is conspicuous, are
incapable of gratifying this passion’ (Rae, )•
Conspicuous consumption means the use of con-
However, it was Veblen who introduced the
sumer goods in such a way as to create a display for
concept of conspicuous consumption as a
the purpose of impressing others rather than for the
phenomenon important for the understanding of
satisfaction of normal consumer demand. It is
consumption as a whole. He gave Rae no reference
consumption intended chiefly as an ostentatious
at all.
display of wealth. The concept of conspicuous
Veblen’s historical and socio-economic explana-
consumption was introduced into economic theory
tion of this institution gave as a result the so-called
by Thorstein Veblen ( ) in the context of
‘Veblen effect’. This is the phenomenon whereby as
his analysis of the latent functions of ‘conspicuous
the price of an article falls some consumers constme
consumption’ and ‘conspicuous waste’ as symbols
this as a reduction in the quality of the good or loss
of upper-class status and as competitive methods of
of its ‘exclusiveness’ and cease to buy it.
enhancing individual prestige.
Veblen argued that the leisure class is chiefly
interested in this type of consumption, but that, to a See Also
certain degree, it exists in all classes. The leisure
class undoubtedly has much more opportunity for ► : . Job 845 ; >
this kind of consumption. The criterion as to ► .den 7: 7 unde (1857-18 5)
whether a particular outlay fell under the heading of
conspicuous consumption was whether, aside from Bibliography
acquired tastes and from the canons of usage and
conventional decency, its result was a net gain in Mason, R.S. 1981. Conspicuous consumption: A study of
exceptional consumption behaviour. New York: St.
comfort or in fullness of life.
Martin’s Press.
It is widely though that Veblen introduced the Rae, J. 1834. The sociological theory of capital, ed. C.
concept of conspicuous consumption into economic Mixter. New York: Macmillan.
literature, but it was known much earlier. Adam Smith, A. 1776. An inquiry into the nature and causes of the
wealth of nations, ed. E. Cannan. London: Methuen,
Smith ( , Book I, Ch. 11) wrote about
1981.
people who like to possess those distinguishing Sweezy, P. 1952. Veblen and Marx. In Socialism and
marks of opulence that nobody but themselves can American life, ed. D.D. Egbert and S. Persons, 2 vols.
possess. In the eyes of such people the merit of an Princeton: Princeton University Press.
object that is in any degree either useful or beautiful Veblen, T. 1899. The theory of the leisure class. London:
George Allen & Unwin.
is greatly enhanced by its scarcity, or by the great
amount of labour required to accumulate any
considerable quantity of it. This is the labour for
which nobody but themselves can afford to pay.
Smith concluded that this domain was ruled by
fashion. J.-B. Say and McCulloch wrote about this
Constant and Variable Capital
issue in a similar way. But the author who first used
the term ‘conspicuous consumption’ was the N. Okishio
Canadian economist John Rae (1796-1872). His
explanation of the nature and effects of luxury was
based on the meaning of vanity in human life. He
understood vanity to be the mere desire for supe- Definition
riority over others without any reference to merit.
The aim is to have what others cannot have, In Das Kapital Marx defined Constant Capital as
whereas the stimulus to productivity in economic that part of capital advanced in the means of
production; he defined Variable Capital as the
2058 Constant and Variable Capital

part of capital advanced in wages (Marx Vol. I, ch.
6). These definitions come from his concept of
Value: he defined the value of commodities as the
amount of labour directly and indirectly necessary
to produce commodities (Vol. I, ch. 1). In other
words, the value of commodities is the sum of C
and N, where C is the value of the means of
production necessary to produce them and N is the
amount of labour used that is directly necessary to
produce them. The value of the capital advanced in
the means of production is equal to C.
However, the value of the capital advanced in
wages is obviously not equal to N, because it is the
value of the commodities which labourers can buy
with their wages, and has no direct relationship with
the amount of labour which they actually expend.
Therefore, while the value of the part of capital that
is advanced in the means of production is
transferred to the value of the products without
quantitative change, the value of the capital
advanced in wages undergoes quantitative change in
the process of transfer to the value of the products.
This is the reason why Marx proposed the
definitions of constant capital C and variable capital
V.
The definition of constant capital and variable
capital must not be confused with the definition of
fixed capital and liquid capital. Fixed capital is a
part of constant capital which is totally used in
production process but transfers its value to prod-
ucts only partially. Liquid capital is a part of
constant capital which is totally used up and trans-
fers its whole value within one production process.
So constant capital is composed of both fixed
capital and liquid capital, and on the other hand
liquid capital belongs partly to constant capital and
partly to variable capital.
Marx introduced the concept ‘value-
composition of capital’, /t, which is defined as the
ratio of constant capital C to variable capital V:
relationship between capitalists and labourers. In
fact definition ( ) can be rewritten as
C/N reflects the character of the process of pro-
duction and N/V reflects the class relationship
between capitalists and labourers. C/N is the ratio of
the amount of labour necessary to produce the
means of production to the amount of labour
directly bestowed, which is completely determined
by the material condition in the process of
production, while N/V is the ratio of the amount of
labour which labourers actually expend to the
amount of labour that is necessary in order to
produce commodities which labourers can purchase
with their wages. If labourers are forced to work
longer with less wages, this ratio must rise.
Marx proposed to call the value-composition of
capital, insofar as it is determined by the material
condition of the process of production, ‘the organic
composition of capital’. More explicitly, ‘The
value-composition of capital, inasmuch as it is
determined by, and reflects, its technical com-
position, is called the organic composition of
capital’ (Capital, Vol. Ill, ch. 8). However, as
shown above, the value composition of capital is not
determined by the material condition of the process
of production alone. So it is better to introduce the
ratio C/N in the place of the organic composition of
capital, which is determined only by the material
condition in the process of production. In order to
avoid confusion, I call this ratio the ‘organic
composition of production’. This is the ratio of dead
labour to living labour, which Marx himself
frequently used in Das Kapital.
Variable Capital and Source of Profit

In contrast to Smith, Ricardo and others, Marx

"4 - (i) attached great importance to analysis to find the
source of profit. He found that source in surplus
Marx knew well that the value composition of labour, which is the excess of labour expended by
capital reflects not only material characteristics of labourers over the value of commodities which
the process of production but also the social labourers can obtain with their wages (Capital,
 B
t\ > a\t\ + Tj —12 (9)

B (10)
h > aiti + x2 — t2

By equations ( ) and ( ) and the above inequalities,

we have

u>o (11)

hfl aj > 0 (12)

Consequently we arrive at the conclusion

T > Bt2.

(1 3)

This inequality implies the existence of surplus

value, because surplus value is the excess of
working hours T over the amount of labour nec-
essary to produce commodities which labourers can
receive with wages B. If the number of workers
employed is n, then total expended labour is nT and
variable capital measured in terms of value is Bt2n.
So the inequality ( ) can be rewritten as

N>V (14)

This is the reason Marx called capital advanced in

wages variable capital.
As shown above, Marx proved the theorem of
the source of profit under the assumption of equiv-
alent exchange. Though this is a clear-cut way to
show the results, it has induced various critiques.
Many critics have said that Marx’s theorem would
be right if all exchanges were equivalent
2060 Constant and Variable Capital

exchange, but that in reality exchanges are seldom

equivalent so his theorem cannot be valid. In order
to refute such a criticism we must prove the theo-
rem without the assumption of equivalent exchange By comparing ( ) and ( ), and ( ) and ( ),
(see Okishio ,
; ; Morishima ). Mathematically,
our task is to find necessary and sufficient condi-
tions for inequalities ( ), ( ) and ( ) to have non-
negative solutions for pi, p2. From ( ) we know
easily that the condition
l-«i>0 (15) value (necessary labour). In the famous controversy
is necessary for pj to be positive. This condition
ensures that the society will obtain net output. Next,
substituting ( ) into ( ), and from ( ) we
have
P\ Tl# p2 T{\-a{)'
(16)
Pi Pi . (22)
— a2-----1-12
ww
we get
->ti, O'=1,2) (23)
w
Equation ( ) implies that if positive profit exists,
then the price-wage ratio (the amount of
commanded labour) is greater than the amount of
with Ricardo, Malthus pointed out this difference
between labour commanded and labour embodied.
Though he wrongly thought that this difference
injured the validity of the labour theory of value, he
had come near to the Marxian theory of the source
of profit (see Malthus , pp. 61-3, 120).
Condition ( ) is rewritten as

On the other hand, from ( ) and ( ) we get

1/ti > w/Pi
P\ ; T ~ xiB
(17)
PI Tai This condition shows that if positive profit exists,
We can easily get from ( ) and ( ) then the productivity of labour (1 It,) must be
greater than the rate of real wages (w/p,).

a2T\B
x2B.
(1 - ) Organic Composition and Production
Price
Inequality ( ) is rewritten as
The concept of organic composition of capital (19 )
plays an important role in Marx’s analysis of prices.
T>B The price of production (Ricardo’s ‘natural
price’) that gives every industry the equal rate of
By ( ), ( ) and ( ) the above becomes profit is determined by the following equations:
( 20 )
' ' Pl = (1 + r)(alPl + n w) (24)
T > Bt2.

Thus we can arrive at Marx’s result. Pi = { 1 + 7~) (p p\ + Xiw) (25) w = Rp2

2

For later convenience we show another expres-

sion for the existence of surplus value. Dividing ( ) (26)
and ( ) by w, we get
where r is the general (equal) rate of profit.
Pl Pl, The first problem is to examine the relationship
(21)
—
w w -----F between
 [a\h , , a
\h , 1
------h 1 ------h VT
h ti Tl and Variable Capital
Pi Constant Tl 2061
a
h Pi t2 a
2h ^ lPl .
-------h VI ’
L r2 T 2 Jfl~£i production is api/T, and the organic composition of
h Pi' capital is u,t2 h, Rt2.
If they are equal then we have equivalent exchange, The second problem is to examine the influence
if not we have non-equivalent exchange from the of the change in real wage rate on the relative prices
point of view of the labour theory of value. The determined by ( ), ( ) and ( ):
values of the commodities are determined by ( ) and
( ). The ratio of the value of production-goods to
consumption-goods t\lt2 is given as

Denoting the relative price of production-goods to

consumption-goods as p, from ( ), ( ) and ( ) we
h obtain
h a
(27)
ih
*2 +1 /(p)=u2 P2 + (T2R -ai)p- TIR = 0. (32)
T2

The relative price of production-goods to Differentiating ( ) with respect to R, we have

consumption-goods determined by ( ) and ( ) is
given as dp _ Tl ~ ?2p
(33)
d R 2a2p + toR ~
a\
h The denominator above is positive, because from
_ (28)
“2 ()
h *2 P\ + VT
T2
denominator x p = a2 p2 + i\R > 0.
Comparing ( ) with ( ), we obtain

We shall show that the sign of the numerator

depends on the comparison between the organic
(29) composition of capital in both sectors.
The function ftp) in ( ) is drawn in Fig. .The
meaningful solution of the equation ( ) is given
The expression in brackets on the RHS of ( ) is at p*. Substituting t,/t2 into ftp), we get
given by

[] = ( h w ~ P i ) ( — A>0. (30)
\T1 T2 /

If profit is positive, from ( ) /, vv is negati ve.

So we can conclude

fi| P-AppaAf-l.

(3i)
t2<P2 TI>T2

The RHS of the above means the comparison of the

organic composition of production and also the
organic composition of capital, because as shown
above the organic composition of
2062 Constant and Variable Capital

Since the mass of the employed living labour is

T\(d2T\ — a2I2). continually on the decline as compared to the mass of
materialized labour set in motion by it, i.e., to the
productively consumed means ofproduc- tion, it
Therefore if a 2 i i > 0 then/(i1/T2) > 0, so follows that the portion of living labour, unpaid and
congealed in surplus-value, must also be continually
considering the graph of f(p) we know that ij/ x2 > p
on the decrease compared to the amount of value
. In the same way we can conclude that ifa 2X| represented by the invested total capital. Since the
«it2=0, then t\/T2=/>. Consequently, from ( ) we ratio of the mass of surplus-value to the value of the
invested total capital forms the rate of profit, this
can conclude
rate must constantly fall (C a p i t a l , Vol. HI, Chap.
13, p. 213).
d(P-±\/&R% 0^ ! ^ .
XPiJ t2 Therefore Marx’s true intention is to insist that
if the organic composition of production v = C/N
This proposition is first established in Ricardo’s (the ratio of the mass of materialized labour to the
Principles ( , p. 43). mass of living labour) increases sufficiently, the
rate of profit must fall.
This can be proved as follows (Okishio ). From
Organic Composition and the ( ) and ( ), and
Rate of Profit
v = C/N (36)
The concept of organic composition of capital plays
an important role in Marx’s analysis of the
we have
movement of the rate of profit (Fig. ).
Marx defined the rate of profit as
St+ I
rt+1 - r, rt
C,+1 + Vt+i
£t+1
- r,
'7+1 (1 + £/+t) + 1
1
- r, (37)
v,+i(l/e,+i + 1) + l/^+i
By ( ), equation ( ) is rewritten as

r
where suffixes t, t + 1 denote periods.
= JTv e = s/v (35
> The RHS of ( ) is an increasing function of e. If
we take the limiting value as e tends to infinity, we
where e is the rate of exploitation.
have
He asserted that if the organic composition of
capital n increases sufficiently then the rate of profit
r must inevitably decrease. This is the famous Taw 1
rt+1 - t , <----------r,.
of the tendency for the rate of profit to fall’ '7+1
(Capital, Vol. Ill, Chap. 13).
Many people have criticized this theorem. They Therefore we conclude, if vt+\ > 1 !ru then
have said that if the rate of exploitation e increases rt+\ - r t < 0 .
sufficiently, r may increase in spite of the increase The above reasoning can be restated. The
of p So r does not necessarily decrease, even if ji reciprocal of the organic composition of production
increases sufficiently (Robinson ; Sweezy ). Such a sets an upper limit to the rate of profit, because
critique overlooks the logic of Marx’s argument.
Marx stated:
5 ^S + V N
(38)
c+vK c ~c
Constant and Variable Capital 2063

(1 + r)fr = (ai + Z 2 R ) { 1 + r ) - 2(41)

From ( ), ( ), ( ), we know

1 - (1 + r)ax >0 1 - (1 + r)x2R > 0 (42)

From ( )fr < 0.fR is rewritten as

/« = (!+ r){[(! + r)a\ - 1]T2 - (1 + i^aiZ!}

So by ( ) f/t < 0, from which dr/AR < 0. As R goes

to zero r tends to its upper limit, which is obtained
from ( )

rmax = -—— ■ (43)

a\

Since the value of the means of production is

Constant and Variable Capital, Fig. 2
determined by ( ), we have
If this upper limit decreases sufficiently, the rate of 1
- a \ = (1 -a i ) h = J i _ = N± (44) d\ CL\
profit must eventually decrease, as shown in Fig. .
t\ Cl\t\ C1
In response to criticisms of this view we must
say that as far as we accept Marx’s assumption that
Thus the upper limit of the general rate of profit is
the inverse of the organic composition (N/C) tends
given by the reciprocal of the organic composition
toward zero, Marx’s conclusion inevitably follows. of production in the means of production sector.
So far we have defined the rate of profit as ( ) Therefore if the organic composition in that sector
and C, V, S are all measured in terms of labour rises sufficiently, the general rate of profit must fall.
value. However, the general rate of profit r must be
determined by ( ), ( ) and ( ). Can we derive
the same conclusions for such a redefined r?
Eliminatingp\,p2, «' from ( ), ( ) and ( )
we have
Organic
f(r,R)=( 1 + r)2/e(<3!T2 - a2z) Composition and
— (1 + r)(a i + z2R) + 1=0
(39) Unemployment

Differentiating/^, R) we have The concept of organic composition of capital plays

an important role in Marx’s analysis of the
/,dr +/sdfl = 0 (40) movement of employment (Capital, vol. I, ch. 23).
Marx assumed a rise in labour productivity to
where accompany the rise in the organic composition of
production C/N. If C/N rises then from the defini-
/, = 2(1 + r ) R ( a \ x 2 - a 2 z 1) - (a \ + x 2 R )
tion of organic composition the amount of
f R = (1 + c)2(aiT2 - a 2 z ) - (1 + r ) z 2
employment must decrease relative to constant
capital.
Considering ( )
However, how does the increase in the organic
composition influence the absolute level of
employment?
2064 Constant and Variable Capital

Many people thought that even if C/N rises
sufficiently, still if constant capital C also increases
then the absolute level of employment can also
increase, though less than proportionately to
constant capital (Oppenheimer ). But by reasoning
similar to that used for ‘the tendency of the rate of
profit to fall’, we can prove that if organic
composition rises sufficiently, then the absolute
level of employment must actually decrease.
amount of employed labourer, Nt + i must inevi-
tably become less than Nt, however high the rate of
accumulation of capital may be (Okishio ). The rate
of accumulation of capital AC/C itself is bounded
by the reciprocal of the organic composition. From (
) and ( )
AC N _ 1
~C< V

The organic composition of production in the fth so that, because it is reasonable to assume that the
period vt is defined as growth rate of labour supply is non-negative, we
can say that if the organic composition rises
sufficiently the rate of unemployment inevitably
rises. Though Marx did not state this explicitly, we
think that this is what he wanted to say.
In analysing Marx’s theorem on the movement
The accumulation of constant capital AC = Ct+1 - C, of the rate of profit and employment, we have
is financed from surplus value S. accepted his central assumption that the organic
composition of production rises sufficiently over
C,+1 — C t < S t . (46) time. However, there arises the problem: under
what conditions do capitalists choose techniques
The surplus value A is a part of the amount of living
that have sufficiently high organic compositions of
labour which labourers expend
production?
Marx seemed to think that the rise in labour
S, < Nt. (47)
productivity and the rise in the organic composition
are two aspects of the same thing. But these two do
By ( ), we obtain, not always go together. Marx himself knew that if
labour productivity in the means of production
sector rises very high then even if technical
Af/+i — N t —- - -C,+i------C t
V/+1 v, composition rises, still the value composition may
-----(C/+i — C , ) + C , (------- remain constant or decrease.
V/+1 \V, | | v , As to the capitalists’ introduction of new tech-
niques we have the following propositions:

From ( ) and ( ) we get (1) If the real wage rate remains constant and
capitalists introduce new techniques which
N , + i - N , < — S t + C A — --------- raise the rate remains of profit (calculated at the
V,+ l \V/+1 V , current prevailing prices and wage) then the
N, ^ ( 1 1
<--------Y C , I----------- new general rate of profit does not decrease,
V/+1 \v,+i v, whatever the organic composition may be.
= —!—(1 + v,-v(+1)0 (2) If the real wage rate rises and capitalists adapt
V/+1V, to this situation with the introduction of new
techniques, then the new general rate of profit
we can say, if (1 + v, - v, +1) < 0 then Nt+l - Nt < 0.
Therefore, if the organic composition of production
in the t + 1th period, vt+i, increases sufficiently so as
to exceed 1 + vt, then the
Constitutional Economics
does is higher than the one which would be
expected if such a new technique were not
introduced.
For the proofs of these propositions, see
See Also
► Marxian Value Analysis
► Irganic Composition of Capital
► ui piu! Value
Bibliography
Dmitriev, V.K. 1898. The theory of value of David Ricardo. In
V.K. Dmitriev, economic essays on value,
competition a n d u t i l i t y , ed. D.M. Nuti.
Cambridge: Cambridge University Press, 1974.
Malthus, R. 1820. Principles of political economy considered
with a view to their practical application, 1st ed. London:
John Murray.
Marx, K. 1867-94. C a p i t a l . Translated from the third
German edition by Samuel Moore and Edward Aveling,
ed. Frederick Engels. New York: International Publishers,
1967.
May, K. 1949. The structure of classical theories. R e v i e w
o f E c o n o m i c S t u d i e s 17(1): 60-69.
Morishima, M. 1973. M a r x ’ s e c o n o m i c s : A d u a l
t h e o i y o f v a l u e a n d g r o w t h . Cambridge:
Cambridge University Press.
Okishio, N. 1955a. Kachi to Kakaku (Value and production
price). In K e i z a i g a k u K e n k y u N e m p o (The
Annals of Economic Studies). Kobe University. No. 19.
Okishio, N. 1955a. Monopoly and the rates of profit. K o b e
U n i v e r s i t y E c o n o m i c R e v i e w 1: 71-88.
Okishio, N. 1963. A mathematical note on Marxian theorems.
W e l t w i r t s c h a f t l i c h e s A r c h i v 91 (pt. 2): 287-
298.
Okishio, N. 1972. A formal proof of Marx’s two theorems.
Kobe University Economic Review 18: 1-6.
Okishio, N., et al. 1978. Three topics on Marxian fundamental
theorems. K o b e U n i v e r s i t y E c o n o m i c R e v i e w
24: 1-18.
Oppenheimer, T. 1903. D a s Grundgesetz der
M a r x s c h e n G e s e l l s c h a f t s l e h r e . Book II, ch.
25. Berlin: Refiner.
Ricardo, D. 1821. On the principles of political economy and
taxation. In W o r k s a n d c o r r e s p o n d e n c e o f
D a v i d R i c a r d o , vol. 1, ed. P. Sraffa. Cambridge:
Cambridge University Press, 1951-73.
Robinson, J. 1942. A n essay on Marxian
e c o n o m i c s . London: Macmillan.
2065
Sweezy, P.M. 1942. The theory of capitalist development:
Principles of Marxian political economy. New York:
Oxford University Press.
Constitutional Economics
James M. Buchanan
The term Constitutional Economics (Constitutional
Political Economy) was introduced to define and to
classify a distinct strand of research inquiry and
related policy discourse in the 1970s and beyond.
The subject matter is not new or novel, and it may
be argued that ‘constitutional economics’ is more
closely related to the work of Adam Smith and the
classical economists than its modem ‘non-
constitutional’ counterpart. Both areas of inquiry
involve positive analysis that is ultimately aimed at
contributing to the discussion of policy questions.
The difference lies in the level of or setting for
analysis which, in turn, implies communication with
different audiences.
Orthodox economic analysis, whether this be
interpreted in Marshallian or Walrasian terms,
attempts to explain the choices of economic agents,
their interactions one with another, and the results
of these interactions, within the existing legal-
institutional-constitutional structure of the polity.
Normative considerations enter through the
efficiency criteria of theoretical welfare economics,
and policy options are evaluated in terms of these
criteria. The policy analyst, building on the analysis,
presents his results, whether explicitly or implicitly,
to the political decision-makers, who then make
some ultimate determination from among the
available set. In this role the policy analyst directly,
and the theorist indirectly, are necessarily advising
governmental decision-makers, whoever these may
be.
By both contrast and comparison, constitutional
economic analysis attempts to explain the working
properties of alternative sets of legal- institutional-
constitutional rales that constrain
2066 Constitutional Economics

the choices and activities of economic and political agents is incorporated in the analysis of alternative
agents, the rules that define the framework within sets of constraints.
which the ordinary choices of economic and
political agents are made. In this sense, con-
stitutional economics involves a ‘higher’ level of Constitutional Economics and
inquiry than orthodox economics; it must incor- Classical Political Economy
porate the results of the latter along with many less
As suggested, Constitutional Economics is related
sophisticated subdisciplines. Normative consider-
to classical political economy and it may be con-
ations enter the analysis in a much more complex
sidered to be an important component of a more
manner than through the artificially straightforward
general revival of the classical emphasis, and par-
efficiency criteria. Alternative sets of rules must be
ticularly as represented in the works of Adam
evaluated in some sense analogously to ranking of
Smith. (The closely related complementary com-
policy options within a specified institutional
ponents are discussed briefly in section “
structure, but the epistemological content of the
”.) One obvious aim of the
‘efficiency’ criteria becomes more exposed.
classical political economists was to offer an
The constitutional economist, precisely because
explanation and an understanding of how markets
the subject matter is the analysis of alternate sets of
operate without detailed political direction. In this
rales, has nothing to offer by way of policy advice
respect, orthodox neoclassical economics follows
to political agents who act within defined rales. In
directly in the classical tradition. But the basic
this sense, constitutional economics is not
classical analysis of the working of markets was
appropriately included within ‘policy science’ at all.
only a necessary step toward the more compre-
At another level, however, the whole exercise is
hensive purpose of the whole exercise, which was
aimed at offering guidance to those who participate
that of demonstrating that, precisely because mar-
in the discussion of constitutional change. In other
kets function with tolerable efficiency indepen-
words, constitutional economics offers a potential
dently of political direction, a powerful normative
for normative advice to the member of the
argument for constitutional structure exists. That is
continuing constitutional convention, whereas
to say, Adam Smith was engaged directly in
orthodox economics offers a potential for advice to
comparing alternative institutional structures,
the practising politician. In a real sense,
alternative sets of constraints within which
constitutional economics examines the choice of
economic agents make choices. In this comparative
constraints as opposed to the choice within
analysis, he found it essential to model the working
constraints, and as this terminology suggests, the
properties of a non-politicized economy, which did
disciplinary attention of economists has almost
not exist in reality, as well as the working properties
exclusively been placed on the second of these two
of a highly politicized mercantilist economy, which
problems.
could be directly observed.
A preliminary illustration of the distinction may
There is no need here to enter the lists on either
be drawn from the economics of monetary policy.
side of the ‘ideas have consequences’ debate. We
The constitutional economist is not directly
know that the economy of Great Britain was
concerned with determining whether monetary ease
effectively de-politicized in the late 18th and early
or monetary restrictiveness is required for furthering
19th centuries, and from the analysis of Smith and
stabilization objectives in a particular setting. On
his classical fellow travellers there emerged both
the other hand, he is directly concerned with
positive understanding of economic process and
evaluating the properties of alternative monetary
philosophical argument for a particular regime. The
regime (e.g. rale-directed versus discretionary, fiat
normative argument for laissez faire was, perhaps
versus commodity standards). The ultimate
inevitably, intermingled with the positive analysis
objective of analysis is the choice among the
of interaction within a
institutions within which political agents act. The
predicted behaviour of these
Constitutional Economics
particular structure of constraints, essentially those
that describe the minimal, protective, or night-
watchman state. Economics, as a social science,
emerged, but in the process attention was diverted
from the institutional structure. Even the predicted
normative reaction against the overly zealous
extension of the laissez faire economics argument
was couched in ‘market failure’ terms, rather than
in the Smithian context of institutional comparison.
The early socialist critique of market order, both in
its Marxist and non-Marxist variants, was almost
exclusively negative in that it elaborated putative
failures of markets within an unexamined set of
legal-political rules while it neglected analysis of
the alternative rules that any correction of the
alleged failures might require. Only with the
debates on socialist calculation in the decades prior
to World War II did the issues of comparative
structure come to be examined.
It was only in the half-century after these
debates that political economy, inclusively defined,
returned, in fits and starts, to its classical tradition.
Given the legal order of the protective state (the
protection of property and the enforcement of
contracts ), we now know that under some
conditions ‘markets fail’ when evaluated against
idealized criteria, whether these be ‘efficiency’,
‘justice’, or other abstract norms. We also know
that ‘politics fails’ when evaluated by the same
criteria. Any positive analysis that purports to be of
use in an ultimate normative judgment must reflect
an informed comparison of the working properties
of alternative sets of rules or constraints. This
analysis is the domain of Constitutional Economics.
Constitutional Economics and Social
Philosophy
Classical political economy emerged from moral
philosophy, and its propounders considered their
efforts to fall naturally within the limits of philo-
sophical discourse. As a modem embodiment,
Constitutional Economics is similarly located,
regardless of disciplinary fragmentation. How can
persons live together in liberty, peace and
2067
prosperity? This central question of social philos-
ophy requires continuing contributions from many
specialists in inquiry, surely including those of the
constitutional economists. By their focus directly on
the ultimate selection of a set of constraining rules
within which ordinary social interaction takes place,
constitutional economists remove themselves at
least one stage further from the false position of
‘social engineer’ than their counterparts in orthodox
economics. Precisely because there is no apparently
simple evaluative criterion analogous to ‘allocative
efficiency’ at hand, the constitutional economist is
less tempted to array alternatives as if an
unexamined criterion commands universal assent.
The artificial abstraction of ‘ social utility’ is likely
to be less appealing to those who concentrate on
choices among constraints than to those who
examine choices within constraints.
If, however, there is no maximand, how can
ultimate normative consequence emerge? In this
respect, one contribution lies at the level of positive
analysis rather than in a too-hasty leap into
normative evaluation. Classical political economy
contains the important principle of spontaneous
coordination, the great discovery of the 18th cen-
tury. This principle states that, within the legal
umbrella of the minimal state and given certain
conditions, the market ‘works’. Even if in the
principle’s modern embellishment we must add
‘warts and all’, we still have come a long way
toward a more comprehensive understanding of the
alternatives for social order. To economics the
extent that his efforts expand the public under-
standing of this principle, in application to all
institutional settings, the constitutional economist
remains under less apparent compulsion to advance
his own privately preferred ‘solutions’ to the
ultimate choice among regimes.
The New Political Economy
Care should be taken not to claim too much for
Constitutional Economics, especially if a narrow
definition is used. As noted earlier, this research
programme, by designation, emerged in the 1970s
to describe efforts at analysing the effects of
2068 Constitutional Economics

alternative sets of rales, as opposed to analyses of closely parallel Constitutional Economics, with, of
choices made within existing and unexamined course, an historical rather than a comparative
structures. In a more comprehensive overview of emphasis (North and Thomas ).
developments after World War II, Constitutional
Economics takes its place among an intersecting set
of several research programmes, all of which have Presuppositions
roots in classical political economy. Critical
Constitutional Economics, along with the related
emphases differ as among the separate programmes,
research programmes mentioned above, shares a
but each reflects efforts to move beyond the
central methodological presupposition with both its
relatively narrow confines of orthodox neoclassical
precursor, classical political economy, and its
economics.
counterpart in modem neoclassical microeconom-
In continental Europe, the whole set of sub-
ics. Only individuals choose and act. Collectivities,
disciplines is included under the rubric ‘The New
as such, neither choose nor act and analysis that
Political Economy’. Within this set we can place (1)
proceeds as if they do is not within the accepted
Public Choice, from which Constitutional
scientific canon. Social aggregates are considered
Economics emerged; (2) Economics of Property
only as the results of choices made and actions
Rights; (3) Law and Economics or Economic
taken by individuals. The emphasis on explaining
Analysis of Law; (4 ) Political Economy of Regu-
non-intended aggregative results of interaction has
lation; (5) the New Institutional Economics, and (6)
carried through since the early insights of the
the new Economic History. Defined imperial-
Scottish moral philosophers. An aggregative result
istically, Constitutional Economics would parallel
that is observed but which cannot, somehow, be
the inclusive term and embrace all of these pro-
factored down and explained by the choices of
grammes, since some attention is drawn in each
individuals stands as a challenge to the scholar
case to the legal-political constraints within which
rather than as some demonstration of non-
economic and political agents choose. Differences
individualistic organic unity.
can be identified, however, and it may be useful to
Methodological individualism, as summarized
summarize some of these here, even if detailed
above, is almost universally accepted by economists
discussion of the other research programmes cannot
who work within mainstream, or non-Marxian,
be attempted.
traditions. A philosophical complement of this
Public Choice, in its non-constitutional aspects
position that assumes a central role in Constitutional
of inquiry, concentrates attention on analyses of
Economics is much less widely accepted and is
alternative political choice structures and on
often explicitly rejected. A distinction must be
behaviour within those structures. Its focus is on
drawn between the methodological individualism
predictive models of political interactions, and is a
that builds on individual choice as the basic unit of
preliminary but necessary stage in the more general
analysis and a second presupposition that locates
constitutional inquiry. The economics of property
the ultimate sources of value exclusively in
rights, law and economics, and the political
individuals.
economy of regulation remain somewhat closer to
The first of these presuppositions without the
orthodox economic theory than Constitutional
second leaves relatively little scope for the
Economics or Public Choice. The standard
derivation of constitutional structures from indi-
efficiency norm remains central to these
vidual preferences. There is no conceptual nor-
subdisciplines, both as an explanatory benchmark
mative bridge between those interests and values
and as normative ideal. The new institutional eco-
that individuals might want to promote and those
nomics is directed more toward the interactions
non-individualistic values that are presumed to
within particular institutional forms rather than
serve as ultimate normative criteria. The whole
toward the comprehensive structure of political
constitutional exercise loses most if not all of its
rales (Furabotn and Richter ; Frey ). Some elements
raison d'etre in such a setting. If the ultimate
of the new economic history
Constitutional Economics
values which are to be called upon to inform the
choices among institutions are non- individualistic,
then there is, at best, only an instrumental argument
for using individually expressed preferences in the
process of discovering those values.
On the other hand, if the second presupposition
concerning the location of the ultimate sources of
value is accepted, there is no other means of
deriving a ‘logic of rules’ than that of utilizing
individually expressed interests. At base, the second
presupposition implies democracy in governance,
along with the accompanying precept that this
structure of decision-making only takes on
normative legitimacy with the prefix ‘constitu-
tional’ appended to it.
Wicksell as Precursor
The single most important precursor to Constitu-
tional Economics in its modem variant is Knut
Wicksell, who was individualist in both of the
senses discussed above. In his basic work on fiscal
theory (Finanztheoretische Untersuchungen,
), Wicksell called attention to the significance
of the rales within which choices are made by
political agents, and he recognized that efforts at
reform must be directed toward changes in the rules
for making decisions rather than toward modifying
expected results through influence on the behaviour
of the actors.
In order to take these steps, Wicksell needed
some criterion by which the possible efficacy of a
proposed change in rules could be judged. He
introduced the now-familiar unanimity or consensus
test, which is carried over into Constitutional
Economics and also allows the whole research
programme to be related closely to the contractarian
tradition in political philosophy. The relationship
between the Wicksellian and the Paretian criteria is
also worthy of note. If only individual evaluations
are to count, and if the only source of information
about such evaluations is the revealed choice
behaviour of individuals themselves, then no
change could be assessed to be ‘efficient’ until and
unless some means could be worked out so as to
bring all persons (and
2069
groups) into agreement. If no such scheme can be
arranged, the observing political economist remains
silent. The Wicksellian contribution allowed the
modem economist to bring the comparative analysis
of rules or institutions within a methodological
framework that utilizes and builds on the efficiency
criterion, which, when interpreted as indicated, does
not require departure from either of the
individualistic presuppositions previously
discussed.
Homo Economics in Constitutional
Choice
Constitutional Economics, as distinct from the
complementary research programme on political
constitutions that are within the boundaries of law,
political science, sociology and other disciplines,
goes beyond the logical presuppositions of indi-
vidualism to incorporate nontautological models of
individual utility maximization. Homo economicus
takes a central role in comparative institutional
inquiry. Individuals are assumed to seek their own
interests, which are defined so as to retain
operational content.
Two quite different arguments can be made in
support of this postulate in Constitutional Eco-
nomics. The first is based simply on methodological
consistency. To the extent that individuals are
modelled as utility maximizers as they participate in
market relationships, there would seem to be no
basis for postulating a shift in motivation as they
behave within non-market constraints. There is at
least a strong presumption that individuals do not
undergo character transformation when they shift
from roles as buyers or sellers in the market-place
to roles as voters, taxpayers, beneficiaries, politi-
cians, or bureaucrats in the political process. A
more sophisticated reason for postulating con-
sistency in behaviour lies in the usefulness of the
model for the whole exercise of institutional com-
parison. If the purpose is to compare the effects of
alternative sets of constraints, some presumption of
behavioural consistency over the alternatives is
necessary in order to identify those differences in
results that are attributable to the differences in
constraints.
2070
A second argument for introducing homo
economicus in Constitutional Economics is both
more complex and more important. It is also the
source of confusion because it is necessary to
distinguish carefully between the use of homo
economicus in predictive social science, specifically
in positive Public Choice and in neoclassical
economics, and in Constitutional Economics. There
is an argument for using the construction in the
latter, even if there are demonstrated empirical
limits on the explanatory power of the model in the
former.
The argument is implicit in the work of the
classical economists. It was stated as a methodo-
logical principle by both David Hume and J.S. Mill:
In constraining any system of government, and fixing
the several checks and controls of the constitution,
each man ought to be supposed a knave, and to have
no other end, in all his actions, than private interest.
(Hume [ ], 1963, pp. 117—18)
The very principle of constitutional government
requires it to be assumed that political power will be
abused to promote the particular purposes of the
holder; not because it is always so, but because such
is the natural tendency of things, to guard against
which is the special use of free institutions. (Mill [ ],
1977, p. 505)
The ultimate purpose of analysing alternative sets
of rules is to inform the choice among these sets.
The predicted operating properties of each alter-
native must be examined, and these properties will
reflect the embodied models of individual behav-
iour within the defined constraints. Behavioural
departures from the presumptive models used in
deriving the operating properties will, of course, be
expected. But the costs of errors may not be
symmetrically distributed around the single best
predictive model. The predicted differential loss
from behavioural departures from a model that
involves ‘optimistic’ motivational assumptions may
be much larger than the predicted differential gain
if the model is shown to be an accurate predictor.
Hence, comparative evaluation of an institution
based on an altruistic model of behaviour should
take into account the possible non-linearity in the
loss function that describes departures from the best
estimates. (In legal practice, formal contracts
include protections against
Constitutional Economics
worst-case behaviour patterns.) In constitutional
choice, therefore, there is an argument for incor-
porating models of individual behaviour that pre-
sume more narrowly defined self-interest than any
empirical record may warrant (Brennan and
Buchanan ).
Applications
Applications of Constitutional Economics, as a
research programme, have emerged in several
settings. First, consider taxation. Post-Marshallian
economic theory, either in its partial or general
equilibrium model, was often applied to tax inci-
dence. Analysis was directed toward predicting the
effects of an exogenously imposed tax on the
private economizing behaviour of persons in their
varying capacities as demanders and suppliers of
goods and services in the market-place. Building on
this base of positive analysis, normative welfare
economics allows a ranking among alternative equi-
revenue tax instruments in terms of the Paretian
standard. In both the positive and normative
aspects, neoclassical tax theory embodies the
presumption that taxes, as such, are exogenous to
the choice process.
The major contribution of modem Public
Choice, as a subdiscipline in its own right, has been
that of endogenizing political decisionmaking. In its
direct emphasis, public choice theory examines the
political decision mles that exist with a view toward
making some predictions about just what sort of tax
institutions or tax instruments will emerge.
Constitutional Economics, as an extended research
programme that emerges from Public Choice, goes
a step further and uses the inputs from both
neoclassical economics and public choice theory to
analyse how alternative political rales might
generate differing tax rules.
The relevant constitutional choice may be that
of granting government authority to levy taxes on
Tax Base A or Tax Base B. Suppose that under the
neoclassical equi-revenue assumption, analysis
demonstrates that the taxing of A generates a lower
excess burden than the taxing of
B. Analysis of the political choice process may
Constitutional Economics
demonstrate, however, that government, if given the
authority to tax A, will tend to levy a tax that will
generate more revenue than would be forthcoming
under an authority to tax B. The equi- revenue
alternatives may not be effective political
alternatives under any plausibly acceptable
modelling of the behaviour of political agents. Once
this simple point is recognized, the normative
significance of the neoclassical ranking of tax
instruments is reduced. Discussion shifts necessarily
to the level of interaction between political decision
structures and fiscal institutions.
A second application of Constitutional Eco-
nomics is found in the post-Keynesian discussion of
budgetary policy. The Keynesian advocacy of the
use of governmental budgets to accomplish
macroeconomic objectives was based on a neglect
of the political decision structure. The proclivity of
democratic governments to prefer spending over
taxing, and hence to bias budgets toward deficit, is
readily explained in elementary public choice
theory (Buchanan and Wagner ). This essential step
in public choice reasoning leads naturally to inquiry
into the relationships between the constraints that
may be placed on political choice and predicted
patterns of budgetary outcomes. Out of this
intensely practical, and important, application of
Constitutional Economics emerged the intellectual
bases for the normative argument that, in the post-
Keynesian era when moral constraints on political
agents have lost much of their previous
effectiveness, formal rules limiting deficit financing
may be required to insure responsible fiscal
decisions. In the modem setting, such rales would
limit spending rates. But it is perhaps worth noting
that, in the political environment of Sweden in the
1890s, Wicksell advanced analytically similar
proposals for reform in the expectation that, if the
suggested reforms should be implemented, public
sector outlay would increase.
The analysis of alternative rules for ‘the transfer
constitution’ represents a third application of
constitutional economics. With the 1971 publication
of John Rawls’s A Theory of Justice, renewed
attention came to be placed on principles of
distributive justice. Although explicitly pre-
constitutional, Rawls’s work has a close
2071
relationship with the efforts to derive criteria for
political and economic rules of social interaction.
Economists, as well as other social scientists and
social philosophers, have come increasingly to
recognize that the untrammelled interplay of
interest-group politics is unlikely to further objec-
tives for distributive justice. Analysis of how this
politics operates in the making of fiscal transfers
suggests that principled adjustments in the posttax,
post-transfer distribution of values is only likely to
be achieved if the institutional rules severely restrict
the profitability of investment in attempts to subvert
the transfer process.
Further applications include the regulatory
constitutions, along with the organization of public
enterprises. In its inclusive definition, Constitutional
Economics becomes the analytical route through
which institutional relevance is reintroduced into a
sometimes sterile social science. In its less inclusive
definition, Constitutional Economics, along with its
related and complementary research programmes,
restores ‘political’ to ‘economy’, thereby bringing a
coherence that was absent during the long hiatus
during which ‘economics’ made putative claims to
independent status.
See Also
►
References
Brennan, G., and J.M. Buchanan. 1980. T h e p o w e r t o
tax: Analytical foundations of the fiscal
c o n s t i t u t i o n . Cambridge: Cambridge University
Press.
Brennan, G., and J.M. Buchanan. 1985. T h e r e a s o n o f
rules: Constitutional political economy.
Cambridge: Cambridge University Press.
Buchanan, J.M. 1974. T h e l i m i t s o f l i b e r t y :
B e t w e e n a n a r c h y a n d l e v i a t h a n . Chicago:
University of Chicago Press.
Buchanan, J.M., and G. Tullock. 1962. T h e c a l c u l u s o f
consent: Logical foundations of
constitutional democracy. Ann Arbor:
2072 Constitutions, Economic Approach to

Buchanan, J.M., and R.E. Wagner. 1977. D e m o c r a c y i n normative vs positive;; Corruption; First
deficit: The political legacy of Lord
amendment; Harsanyi, J; Landes, W; Negative
K e y n e s . New York: Academic.
Frey, B. 1984. A new view of economics: Comparative externalities; Party systems; Philadelphia con-
analysis of institutions. S c e l t e P u b b l i c h e 1: 17-28. vention; Posner, R; Rawls, J; Rent seeking;
Furubotn, E.G, Richter, R. (eds). 1980. T h e n e w i n s t i t u - Social contract theory; Social welfare function;
t i o n a l e c o n o m i c s - A s y m p o s i u m . Zeitschrift
Tullock, G
fur die gesamte Staatswissenschaft, 140.
Flayek, F.A. 1973-1979. L a w , l e g i s l a t i o n , a n d
l i b e r t y , 3 vols. Chicago: University of Chicago Press.
Flume, David. 1741. On the interdependency of parliament. In JEL Classification
Essays, moral, political and literary. K1
London: Oxford University Press, 1963.
McKenzie, R. 1982. B o u n d t o b e f i ' e e . Palo Alto:
Hoover Press.
The economic approach to constitutions applies the
McKenzie, R. (ed.). 1984. C o n s t i t u t i o n a l e c o n o m i c s . methodology of economics to the study of
Lexington: Lexington Books. constitutions, just as public choice applies this
Mill, J.S. 1861. Considerations on representative government. methodology to the full range of topics of political
In E s s a y s o n p o l i t i c s a n d s o c i e t y , vol. XIX of
Collected works of J.S. Mill. Toronto: University of
science.
Toronto Press, 1977. The economic approach to constitutions began
North, D.C., and R.P. Thomas. 1973. T h e r i s e o f t h e with The Calculus of Consent by James Buchanan
western world: A new economic history. and Gordon Tullock ( , hereafter B&T).
Cambridge: Cambridge University Press.
Theirs was largely a normative analysis of what
Rawls, J. 1971. A t h e o r y o f j u s t i c e . Cambridge, MA:
Harvard University Press. ought to go into a constitution. Their main findings
Wicksell, K. 1896. Finanztheoretische and the literature that grew out of their work are
U n t e r s u c h u n g e n . Jena: Gustav Fischer. Central reviewed first, after which the positive stream of the
portions of this work were published in English
literature is discussed.
translation as ‘A new principle of just taxation’. In
Classics in the theory ofpublic finance,
ed. R.A. Musgrave and A.T. Peacock. London:
Macmillan, 1959.
Normative Research on Constitutions

Arguably the most important contribution of The

Constitutions, Economic Approach to
Dennis C. Mueller
Abstract
The economic approach to constitutions applies
the methodology of economics to the study of
constitutions. This entry reviews the normative
literature on constitutions, which assumes a two-
stage collective decision process, and the
positive literature that examines the decisions
made by constitutional conventions and their
economic consequences.
Keywords
Beard, C; Bentham, J; Buchanan, J; Budget
deficits; Collective choice; Constitutionalism
Calculus was to view democracy as a two-stage
process. In stage one, institutions to make future
collective decisions are placed into the constitution.
In stage two, collective decisions are made using
these rules. The long-run nature of the choices at the
first stage creates considerable uncertainty about the
consequences of different voting rules. This
uncertainty makes unanimous agreement on the
rules of the political game likely, even though
individuals would disagree in stage two about the
outcomes of the game. This unanimity at the
constitutional stage provides the normative
underpinning for the constitution (B&T, p. 7).
Harsanyi ( ) also used uncer
tainty over future positions to produce unanimity
and to provide a normative argument for a Ben-
thamite social welfare function (SWF), as did Rawls
( ) in his ethical theory of a social
contract. Mueller ( ) discussed conditions
Constitutions, Economic Approach to
under which a B&T constitution maximizes a
Harsanyian SWF.
Another innovation in The Calculus was to
introduce the external costs of collective decisions
(B&T, pp. 63-8). When a collective choice is made
without the consent of all members of the
community, the decision can make some members
worse-off. The votes of those favouring the decision
thus impose a negative externality on those
opposing it. The smaller the majority required to
pass an issue, the more likely it is that an individual
is on the losing side. However, the amount of time
required to make a collective decision is also likely
to increase with the required majority. The optimal
majority minimizes the sum of collective decisions’
external and decision-making costs.
There is nothing in B&T’s costs-minimization-
approach that implies that the optimal majority is
likely to be a simple majority, and thus their
approach does not account for this rule’s ubiquitous
use. The approach does imply the widespread use of
the simple majority rule, if one of the two cost
curves - most plausibly decision-making costs - has
a sharp discontinuity at 50% (Mueller , pp. 76-8).
Rae ( ) used the two-stage approach to
provide a completely different normative justifi-
cation for the simple majority rale. At the consti-
tutional stage, each individual is uncertain of
whether he will favour x or - x in future votes on
these binary issues. The expected gain if an indi-
vidual favours x and x wins equals the expected loss
if x wins and the individual favours ~ x. Rae further
assumed that the probability of favouring x equals
the probability of favouring ~ x. An egoist chooses
the voting rule that minimizes the probability that
she favours x in the future and ~ x is imposed, or
that she favours ~ x and x is imposed. The simple
majority is the only rule satisfying this condition.
(For additional discussion and references see, Rae
and Schickler .)
Mueller ( ) generalized the two-stage
approach to show that the optimal majority for
binary choices depends on the relative payoffs from
the two issues. (Riley , presents a game theoretic
analysis of a two-stage constitutional process.) As
the loss to those favouring x rises relative to the
gain to those favouring ~ x,
2073
higher required majorities become optimal to
implement ~ x, with unanimity being optimal when
the asymmetry in payoffs is very large. Mueller (
, , ch. 14) employed this anal
ysis to explain why placing rights to act into a
constitution would maximize the expected utilities
of those writing it.
Positive Research on Constitutions
The positive literature of constitutions falls into two
categories: studies of constitutional conventions
and of the consequences of constitutions. The
second category is obviously very large, and so I
provide only the flavour of this type of work.
Charles Beard’s work ( ) might well be
regarded as the first economic analysis of the Phil-
adelphia Convention. Beard stressed the self-
interest of the participants, and claimed that the
final product reflected the interests of the landown-
ing aristocracy. In an equally cynical analysis,
Landes and Posner ( , p. 893) claimed that the
First Amendment was a result of pressure from
‘publishers, journalists, pamphleteers, and others
who derive pecuniary and nonpecuniary income
from publication and advocacy of various sorts’.
Case studies of constitutional conventions confirm
the importance of the self-interest of the
participants in determining the constitution’s
content. For example, representatives from small
parties favour mles that produce proportional
representation and low percentage thresholds for
taking seats in the parliament. Representatives from
large parties favour the reverse. If delegates are
selected geographically, the constitution protects
geographic interests. (For further discussion and
references to the literature, see Elster , and
Mueller , ch. 21). Econometric analyses confirm
these findings. McGuire and Ohlsfeldt ( )
and McGuire ( ) concluded
that the votes of delegates to the Philadelphia con-
vention reflected both their personal interests and
those of their constituencies. Eavey and Miller (
) reached the same conclusion from the vot
ing patterns of those who ratified the Pennsylvania
and Maryland constitutions.
A key decision facing any constitutional convention
is whether to design institutions that will
2074
produce a two-party system or a multiparty system.
In practice, this choice appears to rest upon the
number of representatives elected from each
electoral district (Taagepera and Shugart ; Lijphart ;
Mueller , chs. 8-10). Recent theoretical and
empirical work by Persson and Tabellini ( , ,
, ) and
Persson et al. ( ) demonstrates the economic
importance of this choice. They find more rent
seeking, more cormption, more redistribution and
larger deficits in multiparty systems. Presidential
systems lead to smaller governmental sectors
because they generally contain stronger checks and
balances than parliamentary systems. (For a review
and references to other contributions, see Persson
and Tabellini .)
Conclusions
There are two kinds of people in the world: those
who believe that constitutions matter and those who
do not. The contributors to the literature reviewed
here fall into the former category. Their work helps
illustrate why and in what way constitutions matter,
and further illustrates the fruitfulness of
undertaking an economic approach to the study of
constitutions.
See Also
► Indiana . Jame M. (Born 1919)
► V. -: - Ratioi
Bibliography
Beard, C. 1913. An economic interpretation of the constitution
of the United States. New York: Macmillan, 1941.
Buchanan, J., and G. Tullock. 1962. The calculus of consent.
Ann Arbor: University of Michigan Press.
Eavey, C., and G. Miller. 1989. Constitutional conflict in state
and nation. In The federalist papers and the new
institutionalism, ed. B. Grofman and D. Wittman. New
York: Agathon Press.
Elster, J. 1991. Arguing and bargaining in two constituent
assemblies. Mimeo. Storrs Lectures, Yale Law School.
Constitutions, Economic Approach to
Harsanyi, J. 1955. Cardinal welfare, individualistic ethics, and
interpersonal comparisons of utility. Journal of Political
Economics 63: 309-321.
Landes, W., and R. Posner. 1975. The independent judiciary
in an interest-group perspective. Journal of Law and
Economics 18: 875-901.
Lijphart, A. 1990. The political consequences of electoral laws,
1945-85. American Political Science Review 84: 481-
496.
McGuire, R. 1988. Constitution making: A rational choice
model of the Federal Convention of 1787. American
Journal of Political Science 32: 483-522.
McGuire, R., and R. Ohlsfeldt. 1986. An economic model of
voting behavior over specific issues at the constitutional
convention of 1787. Journal of Economic History 46: 79-
111.
Mueller, D. 1973. Constitutional democracy and social
welfare. Quarterly Journal of Economics 87: 60-80.
Mueller, D. 1991. Constitutional rights. Journal of Law,
Economics, and Organization 7: 313-333.
Mueller, D. 1996. Constitutional democracy. Oxford:
Oxford University Press.
Mueller, D. 2001. The importance of uncertainty in a two-
stage theory of constitutions. Public Choice 108: 223-258.
Mueller, D. 2003. Public choice III. Cambridge: Cambridge
University Press.
Persson, T., and G. Tabellini. 1999. The size and scope of
government: Comparative politics with rational politi-
cians. European Economic Review 43: 699-735.
Persson, T., and G. Tabellini. 2000. Political economics -
Explaining economic policy. Cambridge, MA: MIT
Press.
Persson, T., and G. Tabellini. 2003. Economic effects of
constitutions. Cambridge, MA: MIT Press.
Persson, T., and G. Tabellini. 2004a. Constitutions and
economic policy. Journal of Economic Perspectives 18:
75-98.
Persson, T., and G. Tabellini. 2004b. Constitutional rules and
fiscal policy outcomes. American Economic Review 94:
25-45.
Persson, T., G. Roland, and G. Tabellini. 2000. Comparative
politics and public finance. Journal of Political
Economy 108: 1121-1161.
Rae, D. 1969. Decision-rules and individual values in
constitutional choice. American Political Science
Review 63: 40-56.
Rae, D., and E. Schickler. 1997. Majority rule. In Perspec-
tives on public choice, ed. D.C. Mueller. Cambridge,
MA: Cambridge University Press.
Rawls, J. 1971. A theory of justice. Cambridge, MA:
Belknap.
Riley, J. 2001. Constitutional democracy as a two-stage game.
In Constitutional culture and democratic rule, ed. J.
Ferejohn, J. Rakove, and J. Riley. Cambridge:
Cambridge University Press.
Taagepera, R., and M. Shugart 1989. Seats and votes. New
Haven: Yale University Press.
Consumer Durables 2075

divisible and no discreteness issues arise: thus they

Consumer Durables need not be owned in integer amounts and the
question of scrapping never arises though durables
John Muellbauer can be traded on the second-hand market. Stating
the assumptions this baldly when it is plain how
counter-factual many of them are serves to
anticipate some of the discussion below.
Applied work on the demand for durable goods has On these ‘neoclassical’ assumptions, the con-
usually analysed two kinds of data. The first is sumer maximizes
time-series data on purchases, aggregated over
consumers and typically with different kinds of u , = V ( q t , S „ q t + 1 , S t + u - - ; q T , S T , A T ) (1)
durables aggregated into one or two groups. The
second is cross-section data on the ownership of subject to a sequence of period to period budget
different kinds of durables. There has been a constraints of the form, for example, at t,
corresponding specialization in economic theory
with that appropriate to the first type of data At = At+1 ( 1 + rt) + y, — q,
neglecting the issues of discreteness of ownership -(P?/Pt)(St-( 1 - f>)5f_,) (2)
emphasized in the second and instead focusing on
the dynamics of investment, expectations and where q is the flow of non-durable consumption, S
adjustment costs, these being neglected in the is the stock of durables and A is the stock of real
theory of discrete choice at the level of individual financial assets defined as the nominal stock
households. The discussion below is in this tradi- deflated by the price index for non-durable goods.
tion. In the first part, the focus is on the dynamics of Its presence in ( ) reflects the bequest motive. The
piuchases and in the second on the microeconomics budget constraint ( ) which looks slightly formidable
of discrete choice. just expresses the fact that the change in financial
A good which is durable yields a flow of ser- assets equals financial saving, i.e. income minus
vices into the future. Whether other issues arise, the expenditure. The change in real financial assets is A,
analysis of the demand for durables must take into — A,_\ and income is rtAt_1 + yt, where r is the real
account the distinction between stocks of goods and rate of return andy is real non-property income.
flows of services and the intertemporal character of Note that the real rate of return includes any capital
the decision to purchase or own a durable good. The gains or losses so that property income r,A, _ \ is an
simplest coherent model which captures these two economist’s rather than a national income
essential features was first exposited by Cramer ( accountant’s measure. Expenditure in real terms, i.e.
) though somewhat analogous ana in terms of the flow of non-durable consumption,
lyses of the demand for investment goods had been consists of q, and money expenditure on durables
in the literature for some time (Fisher ). deflated by the price index for non-durables pt.
The assumptions made in this model are the Money expenditure on durables is p?[st — {1 —
following. Consumers maximize utility through 5)s,_i], where pD is the price index for durables and
time. They can lend and borrow at the same interest where the stock of durables S, = qf + (1 — S)s,-\
rate. They can instantly buy or sell a durable good where cff is the flow of purchases and <5 is the rate
at the same price. There are no psychic adjustment of deterioration of durables so that (1 — d)S,_j is
costs or habits associated with purchasing or the amount of durables owned at t - 1 that survives
owning a durable good. The service flow from into period t. The generalization of ( ) and ( ) when
owning a durable is proportional to the stock. q and S are vectors is easy.
Deterioration of the service flow through time is It is conventional in intertemporal consumer
geometric. New vintages of durables are exactly the theory under point expectations to convert the
same as the old, when converted into efficiency
units. Durable goods are perfectly
2076 Consumer Durables

period to period budget constraints into the present q, = g{p, pn,w,)

S, = f{p,pn,W,). (8)
value form by eliminating financial assets from the
sequence of constraints ( ) for t, t+1, t + 2, . . . .
Life-cycle wealth is defined as the value of initial where p , are the vectors (pj) , (pj\\j) which appear
durables plus financial assets plus human wealth in ( ). In the above, p,, }’j for j > t are,
(the present value of current and expected non- of course, forecasts made at / of the respective real
property income): interest rates, relative user costs of durables and real
non-property incomes prevailing in the future. In
Wt=p°/p,{l - S)S,- 1 + A t _ ,( l + r,) empirical work, as well as choosing tractable
restrictions on preferences and so for ( ), some
(3) assumptions need to be made about how consumers
j=‘ make these forecasts.
Considering the forecasts necessary to construct
where the real discount factor pj is defined by
the relative user cost of durables ["[> immediately
raises a potential problem. In recent experience, the
1
P j = n ^ + n T , t + i £ j Z T (4 ) variability in ex post ["[> has been tremendous. In
i=t+l the 1970s inmost Western economies, the ex post
Pj = l.
real rate of interest r,+ 1 was frequently negative and,
The budget constraint for the decision variables particularly when accompanied by relative price
qt, St, qt + i , St + i etc. correspondingly is: declines in durable goods, it is quite likely that ex
post the user cost ["[> was sometimes negative. It is
true that there are expectations mechanisms such as
w = Y i i + Of /p,)S
p q
t t
the adaptive one which, combined with a low
M
r (5) adjustment parameter, might have made the e.r ante
+ Y PM/ Pi) isi - ( 1 - S)SJ-1 ] ■ user cost less variable than the ex post and perhaps
j=t +1
prevented it from becoming negative. However, in
The effective relative price associated with St practice, even fairly crude extrapolations of past
can then be written in the ‘user cost’ form experience would have led to a highly variable
series for the relative user cost j\t, much more so
than is plausibly consistent with the relative smooth
behaviour of purchases of durables if ( ) were the
n
t=(p?/p<) true model. In Muellbauer ( ), I tested a model
A
(p?+i/Pt+i)
based on ( ) for British quarterly data. The evidence
X 1 — (1 — (5) 1+ / I + rt+i strongly rejects such a model both because of the
P?+i/Pt+i failure of crossequation restrictions and because of
X the obvious statistical mis- specification of the
'■pf/Pt^-^i^+i/Pt+i)
+L+1 durables equation.
(6) It seems that one needs a theory which gives
rise to much more sluggishness or ‘persistence’ in
which is the approximation familiar from purchases of durables. The standard way of building
Jorgenson’s ( ) neoclassical theory of invest in such persistence is to posit adjustment costs for
ment. Thus ( ) becomes durables. There has been much work over the years
(see Stone and Rowe ; Chow ; and Nerlove ), on the
^ E/H • i> - stock adjustment model which is an ad hoc way of
//s

building in sluggish adjustment. Weissenberger (

(7) ) has estimated a
j=i j=i model for quarterly British data on non-durable

Maximizing ( ) subject to ( ) gives demand

functions
Consumer Durables 2077

and durable demands based on a quadratic utility of the type defined in ( ) above. As long as these are
function with quadratic adjustment costs and linear functions of (q,p qy _ b qy _ 2, ...) a user cost
rational expectations. Apart from the coefficients on price py can be defined. Then, since the
financial assets he finds a reasonable degree of intertemporally separable form of the utility func-
coherence between the two equations and the tion permits two-stage budgeting (see Gorman, on
durables equation fits well and its residuals are well SEPARABILITY below), there exist demand
behaved. Muellbauer and Pashardes ( , functions
revised 1987) build persistence effects into
zpreferences in a rather more general way though, Z
U = SiiPnA) (11)
in the context of a quadratic utility function, the
effect is similar to Weissen-berger’s. Since their Where
empirical results on annual British data for a system
of nine demand functions, one of which is for
A=
durables, are rather satisfactory, it is worth exam- 2=1
ining the approach.
From ( ), purchases of good i
They assume that the utility function is
intertemporally separable in transformed quantities
Zjf. Qu = \ai ~ (1 - Si)]Sit-i + 1 , a‘ tit- (12)
Oi

U, = V[vt(zlt,...,znt),...,vT(z1T,...,zl,T),AT]. ( 9) wh e r e This can also be written in the form

=
%ij ($ij {Sij 1 tf/)
= [<hj + (1 - ^ (<5,-/l - at).
(10) Mit = ~~§~Azi'

and where die stock 5^ = qy + (1 — Siiqy _ i + (1 + (1 - Oi)[za-i - qlt i). (13)

— St)2qy _ 2 + • • • . In a steady state S; = q/dj and This elegantly expresses an extended kind of
so z, qt In the case of a non-durablc good (for which partial adjustment model termed an ‘error correction
<5,- 1), model’ by Hendry. Changes in q respond to changes
in z but with a stabilizing feedback to last period’s
_ Vij - Wij-\ deviation between q and z. The lower is durability,
i.e. the higher is St and the higher is at, slower is the
speed of adjustment. This makes it clear that in the
absence of persistence effects, one should expect
greater volatility of purchases for durables than for
non-durables. In the empirical results reported for
annual British data by Muellbauer and Pashardes,
there are strong persistence effects both for durables
The parameter a, can be thought of as a ‘habit’ and for non-durables though generally a little larger
or ‘persistence’ parameter when 0 < a, < 1. With for durables. Thus the volatility of purchases for
diminishing marginal utility, the more the consumer durables is higher than for non-durables but much
consumed of good i last period the greater is his or lower than it would be for a zero persistence
her marginality utility and so ‘need’ for the good parameter a,.
this period. When a, < 0, it can be interpreted as a Equation ( ) can be estimated in two ways.
shortlived durability parameter since, with d, 1, a The one adopted by Muellbauer and Pashardes uses
purchase last period but not earlier gives utility this the identity
period.
It was Spinnewyn ( , ) who first
extended the user cost concept of price of a durable
good to more general transformed quantities
2078 Consumer Durables

n in a used one, the differential of new and used car

X
°t - YjPitZil-
i= 1 prices could be added to ( ) as an extra, some
what ad hoc explanatory variable.
There are alternative ways of modelling the
Potentially it suffers from the possible correla-
effects of prices of complementary goods such as
tion between a disturbance term added to ( ) and
the gasoline needed to ran a car. One way is just to
the zu s embodied in the budget af The alternative is
include this price in the vector of prices p°t as long
to solve the intertemporal optimization problem to
as the form of preferences embodied ing,() is quite
give a solution for xlt as a function of life-cycle
general, e.g. a flexible functional form (Diewert ).
wealth and price and rate of return expectations.
Under this alternative some assumptions, as in the But imposing rather more structure often yields
study of the life-cycle consumption function, need rewards. If it is car services that appear in the utility
to be made to model expectations empirically and function and the cost of these consists both of a user
the results are likely to be sensitive to which cost or rental equivalent and a running cost that
assumptions are made. depends on the fuel efficiency of the car, price
Models of this type give good results for aggre- effects are likely to be modelled more accurately.
gate time series data, at least relative to formula- This kind of approach also suggests vintage effects
tions that ignore persistence or durability. But one that could be observable even on aggregate data. In
may well wonder about the source of persistence to response to the fuel price increases in the 1970s
which the above analysis gives simple expression. manufacturers eventually brought models to the
In many ways the most plausible explanation is the market which were much more fuel efficient than
gap between buying and selling prices of durables previous vintages. The effective price differential
due to installation costs, transactions costs or to between new models and the existing stock is likely
information asymmetries. According to Akerlof’s ( to be an important element in the replacement
) ‘lemon effect’ the potential decision. As the fuel efficiency gap between the
buyer of a used durable fears that the reason the existing stock and new models narrows, ceteris
owner wishes to sell is that the durable is a lemon. paribus, one would expect replacement demand to
Since there may be no way that the seller can fall off. Similar vintage effects can arise through
other quality improvements in new durables. The
convince the buyer that it is not, a car that is only a
few weeks old and has suffered no physical only way quality improvements could show up in
deterioration is likely to be saleable only at a pricethe simple model leading to ( ) is through falls in the
substantially below the showroom price. Thus even quality adjusted price pD.
if a consumer expects substantial capital gains on a Among the advantages of the assumption that
durable, there is little incentive to buy in order to new and used durables, when converted into effi-
sell at a profit. The most that can then be expected ciency units, are perfect substitutes is that it sim-
is some advancement of purchases that are likely to plifies the analysis of the interaction of supply and
have been made soon anyway. demand. Total market supply of stock is the sum of
At the individual level, such a gap between the surviving stock (1 — bi)Sit-\, which is given,
buying and selling prices makes comer solutions plus new supply, which if firms are competitive is a
likely (see Deaton and Muellbauer ( ),
function of current and expected values of pf and
pp. 360-64 for a discussion). It is then very difficult
costs of production. Aggregate demand for stock
to derive tractable econometric models that reflect given ( ) can be written in the form
the theory at all precisely. One might take
something like ( ) as a starting point and try to
build in additional elements such as the asymmetric Su = aiSit-1 + 1 d a‘ 8i(Pnx")-
response that larger restrictions on selling than on
buying suggest. For specific durables such as If prices clear the market, equating supply and
automobi les where most buyers of new cars trade demand determines pf, which is thus endogenous
Consumer Durables
and ought to be treated as such in econometric
work. In practice, however, new and used durables
may not be perfect substitutes, new prices may be
set by oligopolistic producers and the second-hand
market, despite its imperfections, may be more
demand responsive in its prices. Nevertheless, the
above model is a bench-mark that raises issues
which applied economists working on the demand
for durables need to face.
The other source of data on the demand for the
durables is household surveys. In such cross-
sections, the discreteness of ownership must be
explicitly recognized. The classic paper by Farrell (
) was one of the first to analyse the threshold
effects which govern ownership. The modem eco-
nomic treatment in its simplest form can be
explained as follows. Let P° be the rental price of a
durable good and the budget constraint be
pq + p DS = x
(1 4)
Suppose S = 1 if the durable is owned and S 0 if
not. Let the single period utility function be
u = v(q,S,b, e )
where b is a vector of observable household char-
acteristics and s summarizes unobservable ones in a
scalar. If S = 0, we can solve for q = x/p from the
budget constraint and if .S' I , q (x - l>n)/P. Thus the
durable is owned if the utility from owning v[(x
pD)/p, 1, b, e ] exceeds that from not owning v[xip,
0, b, C ]. These solved out utility functions are
termed ‘indirect utility functions’. This ownership
criterion is still valid if, in fact, durables of this type
vary in size, performance, luxuriousness or other
characteristics that can be summarized in a quality
index. 5=1 then refers to the minimum quality
available while .S' > 1 for higher qualities.
Maximizing (15) subject to ( ) then gives
conventional demand functions,
given S ^ l , q = g{x,p,pP; b, e ), S = g°(x, p, //’; b,
C ) and the durable is owned if
gD(x,p,pD-,b,(E:) >\.
(1 6)
Here pD is a quality corrected price index, which
2079
assumed to be the same for all consumers. Equation
( ) suggests the use of Probit or Logit
analysis to examine ownership variations on micro-
data (see McFadden 1973, McFadden
). Given information on rental expenditure
defined as pDS = pDgD(x, p, pD\ b, e) if gD( ) > 1,
Tobit analysis is the appropriate technique (see
Tobin ).
Equation ( ) also has implications for
‘quasi’-Engel curves which link ownership in an
income bracket to the income level. Suppose that (
) holds if € < 6(x, p, pD, b). Then the pro
portion of households with budget .r and charac-
teristics b owning the durable is
'Q /»oo
f(x,b,e) de/ f ( x , b , e ) d e (17)
—oo —oo
where /() is the joint probability density of x, b and
e. Provided the budget .r can be plausibly linked to
observable income, this provides a justification for
the ‘quasi’-Engel curves estimated, for example, by
Aitchison and Brown ( ),
Cramer ( ), Pyatt ( ), and Bonus ( ).
These studies have not always, however, given as
much attention to the household characteristics b as
they might have done. If b and e were inde-
pendently distributed of x and 0 monotonic in x,
then there must be a sigmoid relationship between
the level of x and the proportion owning the durable
given the sigmoid shape of the cumulative
distribution function for x. In practice, though there
is quite a high correlation between such household
characteristics as size and income, empirical
‘quasi’-Engel curves are usually sigmoid in shape.
In aggregate, as average income rises over time
there is also a sigmoid relationship between the
average income level and the aggregate proportion
owning a durable. As Deaton and Muellbauer ( , pp.
370-1) note, this sigmoid
shape could be partly due to the sigmoid shape of
the cumulative distribution function of income and
partly due to other causes of diffusion such as
epidemic models of the spread of a disease which
may be appropriate when new goods such as
television are introduced.
For simplicity the discussion above has taken the
case of a single type of durable. But as
2080
McFadden ( ) and Dubin and McFadden
( ) demonstrate, the generalization to a port
folio of different kinds of durables still results in
tractable models which can be estimated by max-
imum likelihood techniques. All these models of
ownership, however, need to be given a long-run
interpretation which abstracts from transactions
costs and imperfections in second-hand markets.
The latter may cause specific households to have
ownership patterns that differ from those they
would have in a steady state. In cross-sections,
information on past decisions and past income is
usually missing, so that the kind of dynamic ele-
ments discussed in the earlier part of this entry
cannot be analysed empirically.
See Also
► : :vvA- lg
► Acvctng iLwfsU'z
► i'v 7
Bibliography
Aitchison, J., and J.A.C. Brown. 1957. T h e l o g n o r m a l
d i s t r i b u t i o n . Cambridge: Cambridge University
Press.
Akerlof, G. 1970. The market for ‘lemons’. Q u a r t e r l y
J o u r n a l o f E c o n o m i c s 84(3): 488-500.
Bonus, H. 1973. Quasi-Engel curves, diffusion, and the
ownership of major consumer durables. J o u r n a l o f
P o l i t i c a l E c o n o m y 81(3): 655-677.
Chow, G. 1957. Demand for automobiles in the US: a study in
consumer durables. Amsterdam: North-Holland.
Cramer, J.S. 1957. A dynamic approach to the theory of
consumer demand. R e v i e w of Economic
S t u d i e s 24: 73-86.
Cramer, J.S. 1962. A statistical model of the ownership of major Consumer Expenditure
consumer durables. Cambridge: Cambridge University
Press.
Deaton, A., and J. Muellbauer. 1980. E c o n o m i c s a n d
c o n s u m e r b e h a v i o u r . Cambridge: Cambridge
University Press.
Diewert, W.E. 1971. An application of the Shephard duality
theorem: A generalized Leontief production function.
J o u r n a l o f P o l i t i c a l E c o n o m y 79(3): 481-507.
Dubin, J.A., and D. McFadden. 1984. An econometric analysis
of residential electric appliance holdings and
consumption. E c o n o m e t r i c a 52(2): 345-362.
Farrell, M.J. 1954. The demand for motor cars in the United
States. J o u r n a l o f t h e R o y a l S t a t i s t i c a l
S o c i e t y , S e r i e s A 117(2): 171-201.
Consumer Expenditure
Fisher, I. 1930. T h e t h e o r y o f I n t e r e s t . New York:
The Macmillan Company.
Jorgenson, D.W. 1963. Capital theory and investment
behaviour. A m e r i c a n Economic Review,
P a p e r s a n d P r o c e e d i n g s 53: 247-259.
McFadden, D. 1974. Conditional logit analysis of qualitative
choice behavior. In F r o n t i e r s i n E c o n o m e t r i c s ,
ed. P. Zarembka. New York: Academic Press.
McFadden, D. 1981. Econometric models of probabilistic
choice. In S t r u c t u r a l a n a l y s i s o f d i s c r e t e
d a t a , ed. C. Manski, and D. McFadden. Cambridge,
MA: MIT Press.
Muellbauer, J. 1981. Testing neoclassical models of the
demand for durables. In E s s a y s i n t h e t h e o r y
and measurement of consumer
b e h a v i o u r , ed. A.S. Deaton. Cambridge: Cambridge
University Press.
Muellbauer, J. and P. Pashardes 1982, T e s t s o f d y n a m i c
specification and homogeneity in a
d e m a n d s y s t e m . (Discussion paper 125, Birkbeck
College, 1982; revised as Institute of Fiscal Studies,
discussion paper, 1987).
Nerlove, M. 1957. A note on long-run automobile demand.
Journal of Marketing 21(July): 57-64.
Pyatt, G. 1964. Priority patterns and the demand for household
durable goods. Cambridge: Cambridge University Press.
Spinnewyn, F. 1979. The cost of consumption and wealth in
models with habit formation. E c o n o m i c s L e t t e r s
2(2): 145-148.
Spinnewyn, F. 1981. Rational habit formation. E u r o p e a n
E c o n o m i c R e v i e w 15(1): 91-109.
Stone, R., and D.A. Rowe. 1957. The market demand for
durable goods. E c o n o m e t r i c a 25(July): 423-443.
Tobin, J. 1958. Estimation of relationships for limited
dependent variables. E c o n o m e t r i c a 26(January): 24-
36.
Weissenberger, E. 1984. A n i n t e r t e m p o r a l s y s t e m
of dynamic consumer demand functions.
Centre for Labour Economics, London School of
Economics, Discussion paper No. 186.
Angus Deaton
Abstract
Consumers’ expenditure is a central concern of
economics, both in microeconomic terms (the
relationship between prices, expenditure and
welfare) and in macroeconomic terms (the rela-
tionship between expenditure and income).
Consumer Expenditure
This article examines the interplay between
theory and evidence in the study of consumers’
expenditure and its composition. Although
models have been developed from the theory of
consumption that illuminate much of the
available data, many standard presumptions of
economics lack substantial bodies of evidence
such as central theories in the natural sciences
enjoy.
Keywords
Almost ideal demand system (AIDS); ARIMA
process; Bernoulli utility functions; Compen-
sated demand; Consumer expenditure; Con-
sumers’ expenditure; Consumption function;
Convexity; Demand curve; Demand functions;
Denver Income Maintenance Experiment;
Econometrics; Elasticity; Engel curves; Engel’s
law; Euler equation; Fisher, I.; Fixed needs
model; Flexible functional forms; Friedman, M.;
Generalized axiom of revealed preference
(GARP); Generalized Leontief system; Hicks,
J.; Household budgets; Inflation; Intertemporal
utility functions; Keynes, J.; King, G.; Kuznets,
S.; Law of demand; Life cycle hypothesis; Life-
cycle income model; Linear expenditure system;
Lucas, R.; Marginal rate of substitution;
Michigan Panel Study of Income Dynamics
(PSID); Modigliani, F.; Non-linear optimization;
Optimal taxation; Permanent income model;
Prices; Ramsey, F.; Random walk model;
Revealed preference theory; Risk aversion;
Roy’s identity; Samuelson, P.; Saving ratio;
Shephard’s lemma; Slutsky matrix; Stigler, G.;
Stone, J.; Stone-Geary utility function;
Substitution effect; Tax reform analysis; Unit
root model; Utility maximization
JEL Classifications
E2
The study of consumers’ expenditure, both in total
and in composition, has always been of major
concern to economists. Neoclassical economics sees
the delivery of individual consumption as the main
object of the economic system, so that
2081
the efficiency with which the economy achieves this
goal is the criterion by which alternative systems,
institutions and policies are to be judged. Within a
capitalist economy, such considerations lead to an
examination of the relationship between prices and
consumption behaviour, and theoretical
development and empirical analysis have been a
major continuous activity since the middle of the
last century. Even older is the tradition of using
individual household budgets to dramatize poverty,
and the relationship between household incomes and
household expenditure patterns has occupied social
reformers, statisticians and econometricians since at
least the 18th century. In more modem times, it has
been recognized that the study of public finance and
of taxation depends on a knowledge of how price
changes affect the welfare and behaviour of
individuals, and the recent development of optimal
tax theory and of tax reform analysis has placed
additional demands on our understanding of the
links between prices, expenditures and welfare.
In the last fifty years, aggregate consumption
has become as much of an object of attention as has
its composition, and in spite of a common
theoretical structure, there has been a considerable
division of labour between macro economists,
interested in aggregate consumption and saving, and
micro economists whose main concern has been
with composition, and with the study of the effects
of relative prices on demand. The interest of
macroeconomics reflects both long-term and short-
term interests. What is not consumed is saved,
saving is thrift and the basis for capital formation, so
that the determinants of saving are the determinants
of future growth and prosperity. More immediately,
aggregate consumption accounts for a large share of
national income, typically more than three-quarters,
so that fluctuations in behaviour or ‘consumption
shocks’ have important consequences for output,
employment, and the business cycle. Since Keynes’s
General Tlieoiy, the consumption function, the
relationship between consumption and income, has
played a central role in the study of the macro-
economy. Since the 1930s, there has been a con-
tinuous flow of theoretical and empirical
developments in consumption function research,
2082
and some of the outstanding scientific achievements
in economics have been in this field.
In this essay, the major themes will be the inter-
play between theory and evidence in the study of
consumers’ expenditure and its composition. If
economists have any serious claim to being scien-
tists, it should be clearly visible here. The best
minds in the profession have worked on the theory
of consumption and on its empirical implementa-
tion, and there have always been more data avail-
able than could possibly be examined. I hope to
show that there have been some stunning successes,
where elegant models have yielded far from obvious
predictions that have been well vindicated by the
evidence. But there is much that remains to be done,
and much that needs to be put right. Many of file
standard presumptions of economics remain just
that, assumptions unsupported by evidence, and
while modem price theory is logically consistent
and theoretically well developed, it is far from
having that solid body of empirical support and
proven usefulness that characterizes similar central
theories in the natural sciences.
A Simple Theoretical Framework
Almost all discussions of consumer behaviour begin
with a theory of individual behaviour. I follow
neoclassical tradition by supposing that such
behaviour can be described by the maximization of
a utility function subject to suitable constraints. The
axioms that justify utility maximization are mild,
see any microeconomic text such as Varian (
/1984) or Deaton and
Muellbauer ( ), so that utility maximization
should be seen as no more than a convenient
framework that rules out the grossest kind of
behavioural inconsistencies. The assumptions that
have real force are those that detail the constraints
facing individuals or else put specific structure on
utility functions. Perhaps the most general
specification of preferences that could be considered
is one that is written
Ui=El{f{ql,q2,...,qt,...,qT)} ( 1)
Note that by writing utility in the form ( ), since the
where u, is utility at time t, Et is the expectation
expectation operator is additive over states of the
operator for expectations formed at time /, q! to qT
world preferences are in effect assumed to be
Consumer Expenditure
are vectors of consumption in periods 1 to T, and /
(•) is a quasi-concave function that is nondecreasing
in each of its arguments. Several things about this
formulation are worth brief discussion. The function
/(•) yields the utility that would be obtained from the
consumption vector under certainty, and it
represents the utility from a life-time of
consumption; the indices 1 to T therefore represent
age with 1 the date of birth and T that of death. The
expectation operator is required because choice is
made subject to uncertainty, not about the choices
themselves, which are under the consumer’s control,
but about the consequences of current choices for
future opportunities. It is not possible to travel
backward through time, so that choices once made
cannot be undone, and yet the cost of current
consumption in terms of future consumption
foregone is uncertain, as is the amount of resources
that may become available at future dates. The
consumer must therefore travel through life, filling
in the slots in ( ) from left to right as best as he or
she can, and at time (or age) t, everything to the left
will be fixed and unchangeable, whether now seen
to be optimal or not, while everything ahead of t is
subject to the random buffeting of unexpected
changes in interest rates, prices, and incomes. The
solution to this sort of maximization problem has
been elegantly characterized by Epstein ( );
here I shall work
with something that is more restrictive but more
useful and note in section “
” below some phenomena that are better
handled by the more general model.
Intertemporal utility functions are frequently
assumed to be intertemporally additive, so that the
preference rankings between consumption bundles
in any two periods or ages are taken to be
independent of consumption levels in any third
period. If so, the utility function ( ) takes the more
mathematically convenient form
T
* = (2 )
r= 1
Consumer Expenditure
simultaneously additive over both states and
periods, an assumption that can be formally
defended, see Gorman ( ) and Browning
et al. ( ). It has the consequence that risk aver
sion and intertemporal substitutability become two
aspects of the same phenomenon. Individuals that
dislike risk, and will pay to avoid it, will also
attempt to smooth their consumption over time and
will require large incentives to alter their preferred
consumption and saving profiles. Note also that the
additive structure of ( ) means that, unlike the case
of ( ), previous decisions are irrelevant for current
ones. For decision-making at time t, bygones are
bygones, and conditional on asset and income
positions, future choices are unaffected by what has
happened in the past. There can therefore be no
attempt to make up for lost opportunities, nor can
such phenomena as habit formation be easily
modelled.
Because utility in ( ) is intertemporally sepa-
rable, maximization of life-time utility implies that,
within each period, the period subutility function v,
( •) must be maximized subject to whatever total it
is optimal to spend in that period. The period by
period allocation of consumption expenditure to
individual commodities need not, therefore, be
planned in advance, but can be left to be determined
when that period or age is reached, and period t
allocation will follow according to the rule
maximize v, (qt) subject to pt ■ qt= xh (3)
where pt is the price vector corresponding to q, and
xt is the total amount to be spent in t. Problem ( ) is + 1), nor that in (t + 1) without thinking about (t +
one of standard (static) utility maximization, though
note that xt is not given to the consumer, but is
determined by the wider intertemporal choice
problem. Nevertheless, not the least advantage of
the intertemporally additive formulation is its
implication that the composition of expenditure
follows the standard utility maximization rale. It
allows separate attention to be given to demand
analysis on the one hand, i.e. to the problem ( ), and
to the consumption function on the other hand, this Given this, the consumer can look ahead from
being understood to be the intertemporal allocation period t to period (T — 1) and foresee that the
of resources, i.e. the determination of xt.
2083
Write the maximized value of utility from the
period t problem as ij/t(xt, pt), where i]/(■) is a
standard indirect utility function. The original
intertemporal utility function then takes the form
T-t
n, =E,^21J/r (xl+r,pl+r). ( 4)
i=0
The constraints under which this function is max-
imized are most conveniently analysed through the
conditions governing the evolution of wealth from
period to period. If A, is the (ex-dividend) value of
assets at the start of period t, N,t is the nominal
holdings of asset i with price Pit, d„ is the dividend
on i paid immediately before the beginning of t, and
v, is income in period t, then
A,1 i = 'y ' Nit(Pi, 11 I dlt 11) (5)
i
Y] NitPit = A,+ yt-xt. ( 6)
i
Conditions ( ) and ( ) determine how wealth evolves
from period to period, and the picture is completed
by requiring that the consumer’s terminal assets be
positive, i.e.
dr+i > 0 (7)
To solve this problem, the technique of backward
recursion is used. This rests on the observation that
it is impossible to know what to do in period t
without taking into account the problem in period (t
2), and so on. However, in period T there is no
future, so that looking ahead from date t, we can
write subutility in period T in terms of that period’s
price and inherited assets, and we write this as v- /;
i.e.
vj = vj'(Aj) = i / ' T{AT +yT>Pt)- ( 8)
problem then will be to choose the composition of
assets N so as to maximize vr_i, where
2084 Consumer Expenditure

= max [^r_j(Ar_! +yT_l-N- PT-1 , pT-i) +Ej-I {i;r[N many of the implications of the theory of con-
sumption. Note first that it is little more than the
■ (Pj + dj)}} 1.
standard result that the marginal rate of substitution
(9)
between today’s and tomorrow’s consumption
At the next stage, assets in (T — 2) will be allocated should be equal to the relative price. However, the
so as to trade off the benefits of consumption in (T equation is set in a multiperiod framework, not a
— 2) versus the benefits of AT_X in vT_\ in ( ) above two-period one, and it explicitly recognizes the
and again yielding a maximized value vT_2. As we uncertainty in both asset returns and in the value of
follow this back through time, the consumer finally money in subsequent periods. The equation also
reaches the current period t, where he or she faces holds for all i, i.e. for all assets, so that the result
an only slightly complicated version of the usual also has implications for asset pricing as well as for
‘today tomorrow’ trade-off; the asset vector N must consumption and saving, and for this reason the
be chosen to solve the problem, model is often referred to as the consumption-asset
pricing model. I shall return to these implications
u, = ma x i (i,{A, + y, - N ■ Pt,pt) below.
N The theory as presented above is the modem
+ E,{vl+l[N •( / > , + , +dt+1)}}. equivalent of the life-cycle theory of consumption
that dates back to Irving Fisher ( ) and Frank
(1 0) Ramsey ( ), and that had its modem genesis in
the papers by Modigliani and Brumberg ( )
From this sequence of problems, several and (1954, published ). Modigliani and Brumberg’s
important results readily follow. First, consider the treatment differs from the above only in not
derivatives of each of the functions vr(Ar) which explicitly modelling uncertainty, and by including
represent the marginal value of an extra unit of only a single asset. The modem version appears
currency for the remaining segment of life time first in Breeden ( ) and in
utility from r through to T. By the envelope theorem Hall ( ), see also Grossman and Shiller ( ).
(see for example Dixit ( ) for a
good exposition), it is legitimate to differentiate
through the maximization problem, from which Predictions and Evidence

v'(A, ) = d\j/r/dxr = Xr, say, (11)
so that Xr is the marginal utility of money in period
r. Secondly, the maximization of ( ) with respect to
portfolio choice gives the relationship, for each
asseti,
Pud^Jdx, =E,{(Pit+1 + dil+i)d\jjl+Jdxt+\}
(12)
which, defining the asset return Rh , as (Pit+1 +
du+\)/Pit< and using ( ) can be rewritten in the
simple form
X, = E,(X,+iRj,+1).
(1 3)
This equation, in current parlance often referred to
as the ‘Euler equation’, can be used to derive
One of the most important implications of the
theory above, and of Eq. ( ) in particular, is that the
evolution of consumption over the fife- cycle is
independent of the pattern of income over the life-
cycle. The asset evolution Eqs. ( ) and ( ) allow
consumers to borrow and lend at will, so that the
only ultimate constraint on their consumption is one
of life-time solvency. In consequence, consumption
patterns are free to follow tastes, the evolution of
family structure, or the different needs that come
with ageing, provided that in the end total life-time
expenditure lies within (total) life-time resources,
whether from inherited wealth or from labour
income. It is often assumed that tastes are such that
consumers prefer to have a relatively smooth
consumption stream, and this can be illustrated
from a special case of Eq. ( ).
Assume that the within-period utility function is
Consumer Expenditure
homothetic so that if/(x, p) is <p(x/a(p)) for some
linearly homogeneous function a(-), and that </?(•)
has the isoelastic from with elasticity (1 — <r).
Life-time utility takes the form
T-t
M, = ^ (i + S)~r [.xt+r/a(pt+r)]1_<r (14)
r=0
where <5 is the rate of pure time preference, and a
> 0 is the coefficient of relative risk aversion and
the reciprocal of the intertemporal elasticity of
substitution. Equation ( ) can be used to evaluate ( ),
and gives immediately
£[{(1 + r*+i)/(l + 8)}{ct/ct+iy] = 1 (15)
where rt+ \ is the real after tax rate of interest from /
to / + 1 on any asset, and ct is real consumption, x/a
(/;,). Equation ( ) shows that, if expectations are
fiilfilled, consumption will grow over the life- cycle
if the real rate of interest is greater than the rate of
pure time preference, and vice versa, while with i~f
= 8, consumption is constant with age. These results
are of course an artefact of the specific assumptions
about utility, and for any real household
consumption can be expected to vary predictably
with age according to patterns of family formation,
growth, and ageing; Modigliani and Ando ( ) recent econometric evidence.
have suggested that consumption
per ‘equivalent adult’ might be constant over the
life-cycle. But whatever the shape of preferences,
there need be no relationship between the profiles of
consumption and of income; income can be saved
until it is needed, or borrowed against if it is not yet
available.
Independent of the life-time pattern of con-
sumption is its level, which under the life-cycle
model is determined by the level of total life-time
resources, so that individuals with the same tastes
but with higher incomes or higher inherited assets
will have higher levels of consumption throughout
their lives. If the future were entirely predictable,
the consumption plan at any point in time could be
decided with reference to the level of total wealth,
this being the value of financial assets and the
discounted present value of current and future
incomes. In this sense, the life-cycle model
2085
is a permanent income theory of consumption,
where permanent income is the annuity value of
lifetime wealth, though the lifetime interpretation is
only one of the many that are offered in Friedman’s
( ) original statement. Whether
life-cycle or not, linking consumption to future
incomes has important consequences. First, con-
sumption will respond only to ‘surprises’ or
‘shocks’ in income; changes in income that have
been foreseen are already discounted in previous
behaviour and should not induce any changes in
plans. Of course, this does not mean that con-
sumption will not change along with changes in
income; a change may have been planned in any
case, and some proportion of any actual change may
well have been unforeseen. However, if a
substantial fraction of the regular changes in income
over the business cycle are foreseen by consumers,
or if unanticipated fluctuations in income are
regarded as only temporary with limited
consequences for total life time resources, then
consumption will not respond very much to cyclical
fluctuations in income. Aggregate consumption is
indeed much smoother than is aggregate income,
and this has been traditionally accepted as an
important piece of confirmatory evidence. I shall
take up the matter again below when I deal with the
The distinction between measured income and
permanent income is also important for the inter-
pretation of cross-sectional evidence. Since mea-
sured income can be regarded as an error-ridden
proxy for permanent income, the regression of
consumption on measured income will be biased
downward (rotated clockwise) compared with the
true regression of consumption on permanent
income. Crosssectional regressions, or time-series
regressions of simple Keynesian consumption
functions will therefore tend to understate the long-
run marginal propensity to consume. Well before
the work on life-cycle models, Kuznets ( )
showed that the long-run saving ratio in
the United States had been roughly constant in spite
of repeated cross-sectional analyses showing that
the saving ratio rose with income, and the life- cycle
theory could also readily account for these findings.
It is interesting to note that the constancy of the
saving ratio is far from being well
2086
established as an empirical fact; the evidence for
other countries with long-run data is very mixed,
and even the United States saving ratio is clearly
influenced in the long-run by technical change,
migration patterns, and demographic shifts, see
Kuznets ( ) and Deaton ( ). Life-cycle
and permanent income theories also predict that
households with atypically high income will tend to
save a great deal of it, a prediction which explained
the apparently anomalous finding that black
households tend to save more than white
households at the same level of measured income;
since blacks typically have lower household income
than whites, those with the same measured income
can be expected to have a higher transitory
component.
The Modigliani and Brumberg life-cycle story
was also important because it offered a story of
capital accumulation in society as a whole that
relied on the way in which people made preparation
for their own futures, particularly for their future
retirement. In a stationary life-cycle economy, in
which there is neither economic nor population
growth, aggregate saving is zero, and the old, as
they dissave, pass on the ownership of the capital
stock to the next generation who are, in turn, saving
for their own retirement. With either population or
income growth, the aggregate scale of saving by the
young would be greater than that of dissaving by
the old, so that, to a first approximation, the
aggregate saving ratio, while in the long ran
independent of the level of national income, would
depend on the sum of its population and per capita
real income growth rates. Modigliani ( ), in
his Nobel address, has
given an account of how very simple stylized
models of saving and refinement yield quite accu-
rate predictions of the saving ratio and of the ratio
of wealth to national income, and the predictions
about the growth effects have been repeatedly borne
out in international comparisons of saving rates, see
Modigliani ( ), Houthakker ( , the return to wealth was included with labour
), Leff ( ) and Surrey ( ). Perhaps the
only problem with these interpretations is that there
is little evidence that the old actually dissave,
except by running down state social security or
pension schemes; see for example Mirer ( ).
Partly, this may be a rational response to
Consumer Expenditure
uncertainty about the date of death and about
possible medical expenses near the end of life
(Davies ), partly there may be statistical problems
of measurement (Shorrocks ), and partly consumers
may wish to leave bequests. However, most
countries’ tax systems penalize donors who do not
pass on assets prior to death, so the reason for the
size of actual bequests remains something of a
mystery. Bemheim et al. ( ) have gone so far as to
suggest that parents
retain their wealth until death in order to control
their heirs and to solicit attention from them. They
claim empirical support for a positive relationship
between visits by children to their parents and
parents’ bequeathable assets; visits are apparently
especially frequent to rich sick parents, but not at all
frequent to poor sick parents. Related to the dispute
about the reason for bequests is a parallel dispute on
their importance in the transmission of the capital
stock, see the original contribution by Kotlikoff and
Summers ( ) and Modigliani's
reply, summarized in his ( ) Nobel lecture.
The life-cycle and permanent income models
also provided the econometric specifications for a
generation of macro econometric models. Ando and
Modigliani ( ) suggested a simple form
for the aggregate consumption function in which
real aggregate consumption was a linear function of
expected real labour income, YL, and of the real
value of financial assets, i.e.
c, = acE,(YL) + 8W,.
( 16 )
In practical econometric work, the expectation was
typically replaced by a linear function of current
and past values of labour income, a procedure that
can be formally justified by modelling labour
income as a linear ARIMA process, a topic to which
I shall return below. Wealth or a subset of wealth
was included as data allowed, although sometimes
income which could then be replaced by total
income, so that, with smoothing, ( )
becomes a permanent (total) income model of
consumption. A favoiuite variant, suggested in
Friedman ( ), was to model permanent
income as an infinite moving average of current
income with geometrically declining weights,
Consumer Expenditure
>f = (l-A)^AVr. (17)
r=0
so that if current consumption is proportional to
permanent income, substitution yields
c, = kct-1 + k(\ - X)y„
( 18 )
a formulation that is also easy to defend if con-
sumers ‘partially adjust’ to changes in current
income. Models like ( ), possibly with additional
lags, and with the occasional appearance of more or
less ‘exotic’ regressors, such as wealth, interest
rates, inflation rates, money supply, as well as
various dummy variables for ‘problem’
observations, were the standard fare of macro-
econometric models in their heyday, from the early
sixties for about a decade and a half. They fit the
data well, they accounted for the smoothness of
consumption relative to income, and they accorded
at least roughly with the general features of the life-
cycle and permanent income formulations which
provided them with pedigree and general theoretical
legitimacy. Dozens of papers could be cited within
this tradition; those by Stone ( , ),
Evans ( ), and Davidson et al.
( ) will perhaps stand as good examples.
Recent Econometric Experience
In the mid-1970s, the general state of complacency
of macroeconomic modelling was rapidly eroded,
largely by the apparent inability of the standard
models to explain, let alone to predict, the
coexistence of unemployment and inflation. The
relationship between consumption and income did
not escape some of the blame, although the main
focus of attack was elsewhere. Standard
consumption functions, which had worked well into
the early seventies, seriously under-predicted
aggregate saving during the period of (at least
relatively) rapid inflation that characterized most
Western economies in the middle of the decade.
The implementation of the theory of the
consumption function was also singled out for
discussion in Lucas’s famous ( )
2087
essay that became known as the Lucas ‘critique’. As
Lucas forcefully argued, if consumption is
determined by the discounted present value of
expected future incomes, the response of con-
sumption to a change in income is not well- defined
until we know how expectations of income are
formed. Each observed realization will cause a re-
evaluation of future prospects in accordance with
formulae that depend on the nature of the stochastic
process governing income. If the nature of the
stochastic process is changed, for example by a
fundamental change in the tax code, then the way in
which information is processed will change, and
new information about incomes will have different
implications for future expectations and for future
consumption. This insight is of great importance,
although its implications for econometric modelling
were initially taken much too negatively; if the rules
keep changing, econometric models will be
inherently unstable (as evidenced by their
performance in the mid-seventies) and we should
give up trying to find stable relationships. Instead,
as events have shown, the introduction of rational
expectations has given a whole new lease of life to
the study of consumption, with developments as
positive as anything that has happened since the
life-cycle and permanent income models were the
‘new’ theories in the mid-fifties. Lucas’s critique
suggested at least two lines for research. First, could
the failure of consumption functions, or indeed of
macroeconometric models in general, really be
traced to a change in the way expectations were
formed? If so, it ought to be possible to detect
changes in the stochastic process generating real
income. Second, and more generally, if expectations
are important, there ought to be high returns to the
simultaneous modelling of consumption and
income, so that knowledge of the structure of the
latter can be used either to estimate the consumption
function or to test for the validity of the
expectations mechanism. My own reading of the
evidence is that the Lucas critique is not capable of
explaining the failure of the empirical consumption
function, but that the underprediction of saving
resulted from ignorance of the fact that saving
appears to respond positively to inflation, or at least
to unanticipated inflation.
2088
There is overwhelming evidence from a large
number of countries, see in particular Koskela and
Viren ( , ), that saving increased with
inflation in the 1970s, even when we allow for real
income and its various lags. Such a finding is also
consistent with the life-cycle theory since unan-
ticipated inflation imparts a negative shock to real
assets, so that risk-averse, low inter-temporal elas-
ticity consumers will save to replace the lost assets
so as to avoid the chance of low consumption later.
It is also possible to explain the relationship
through the confusion between relative and absolute
price changes that is engendered by unanticipated
inflation in an environment in which goods are
bought sequentially, see Deaton ( ), but it
would be hard to devise a test that would separate
this from the life-cycle explanation. But if inflation
was indeed the cause of the failure of the empirical
consumption functions, then it is a standard enough
story. An important variable was omitted from the
analysis, it had not been very variable in the past so
that its omission was hard to detect, and economists
had not been imaginative enough to perceive its
importance in advance. The Lucas critique is only
one of the many problems that can beset an
econometric equation, and it does not seem to have
been the fatal one in this case.
The second research direction, the joint exam-
ination of income and consumption, has proved
more productive. The first important step was taken
by Hall ( ), who pointed out that
Eq. ( ) implies that, as an approximation con
sumption should follow a random walk with drift.
To see why, assume that the real interest rate r is
constant and known, and write ( ) in the form
c
t+ I = {(1 + + r )} c t a + e<+1 (19)
where the expectation at t of £t+\ is zero. Equation (
) is exact, but a convenient expres
sion can be reached by factoring ct out of the right
hand side, taking logarithms, and approximating.
This gives
lnc,+i = Inc, + g + v,+i (20)
where g is positive or negative as r is greater than
or less than S, and the ‘innovation’ v,_ b like e,+,
Consumer Expenditure
has expectation zero at time t. Equation ( )
shows that, in the absence of ‘news’, consumption
will grow or decline at a steady rate g, so that
nothing that is known by the consumer at time / or
earlier should have any value for predicting the
deviation of the rate of change of consumption from
its constant mean. The result is often referred to as
the ‘random walk’ property of consumption, though
the theory does not predict that vt+\ has constant
variance, so that, strictly speaking, the stochastic
process is not a random walk.
For someone used to thinking about the con-
sumption function as the relationship between
consumption and income, Eq. ( ) is notable for the
apparent absence of any reference to income. But of
course income can appear through the stochastic
term vt +1 if current income contains new
information about its own value or about future
values of income, and this will generally be the
case. The random walk model does not predict that
consumption should not respond to current income.
It does however predict that, conditional on lagged
consumption, past income or changes in income
should not be correlated with the current change in
consumption, and a considerable amount of effort
has recently gone into testing this proposition. In
Hall’s ( ) original
paper, to the surprise of the author and of much of
the profession, the model worked well for an
aggregate of United States consumption of non-
durables and services. The level of consumption
certainly depends on its own lagged value, but the
addition of one or more lagged values of income or
of further lagged values of consumption did not
significantly add to the explanatory power of the
model. Hall examined the role of the number of
other lagged variables and discovered that lagged
stockmarket prices had predictive power for the
change in consumption, so that he concluded by
formally rejecting the model. However, the over-
whelming impression was favourable, at least rel-
ative to expectations.
Hall’s test procedures are attractive because
they do not depend on the properties of the income
process, and focus only on consumption and its
lags. But robustness comes at the price of power,
and later work has devoted considerable attention to
the joint properties of consumption and real
Consumer Expenditure
income. Perhaps the natural route to modelling is to
find a representation of real income as a stochastic
process, typically as some sort of ARIMA. Once
this is known, changes in income can be
decomposed into anticipated and unanticipated
components using the standard forecasting formulae
from statistical time series analysis, so that it
becomes possible to test whether consumption
responds to one but not to the other. The random
walk model seemed not to survive these tests so
well. Papers by Flavin ( ) and by Hayashi
( ) showed that, for United States data, con
sumption is sensitive to anticipated changes in
income, something that should not be the case in a
thoroughgoing life-cycle model in which consumers
are efficiently looking into the future. The
phenomenon became known as the ‘excess sensi-
tivity’ result, and was typically ascribed to the
existence of a substantial number of consumers who
wish to borrow against future income but are unable
to do so. Such liquidity constrained consumers can
be expected to consume all their available income,
so that their consumption will increase one for one
with all income changes, whether anticipated or not.
However, it is not clear that the excess sensi-
tivity finding is itself robust. First, it is becoming
increasingly recognized that the problems of
econometric testing in the time-series models are
more severe than had been generally supposed. The
time series of both consumption and income are
non-stationary, and it sometimes seems as if
hypothesis testing in models involving non-
stationary variables is like building on shifting
sands; see Mankiw and Shapiro ( , )
and Durlauf and Phillips ( ) for some of the
problems. Second, there are a large number of
variables other than income which can affect con-
sumption, so that, according to ( ), surprises in
wealth and in inflation should affect consumption,
as should the level of real interest rates. Adding
even a few of these variables reduces degrees of
freedom and diminishes the probability of being
able to reject the basic model. Both Bean ( ) be consistent with a model in whi ch about one fifth
and Blinder and Deaton ( ) find that time-
series models of consumption with several variables
are more easily reconciled with the theory than are
the simple two variable models. Not all of
2089
this should be ascribed to lack of degrees of free-
dom; for example Blinder and Deaton consistently
find that unanticipated changes in wealth affect
consumption and that anticipated changes do not.
Third, even in a bivariate income- consumption
model, Campbell ( ) has found
that the model is largely consistent with the time-
series evidence. Campbell recognizes the possibility
of time-series feedback from lagged consumption to
income, and models saving and the change in
income as a bivariate vector- autoregressive system
in which each series is regressed on lagged values
of both. The structure of this representation then
turns out to be very close to what it would have to
be if the life-cycle rational expectations model were
correct. The conflict between Campbell’s results
and the excess sensitivity findings are presumably
accounted for by the feedback from saving to
changes in labour income, since his model is
otherwise compatible with the earlier ones.
Similarly mixed findings are also being uncov-
ered from longitudinal panels that follow individual
households over time. In contrast to the situation
with labour supply, there are few panel data in the
United States that cover household consumption,
and most work has used the data on expenditure on
food that is contained in the Michigan Panel Study
of Income Dynamics (PSID). In an elegant paper,
Hall and Mishkin ( ) found results that
were in accord with the
excess sensitivity results; there is a strong negative
correlation in their data between changes in
consumption and changes in lagged income that is
inconsistent with the view that only surprises in
income should matter. However, since in their data
changes in income are negatively correlated over
time, a negative correlation between the lagged
income change and the change in consumption can
be interpreted as a positive correlation between
consumption changes and changes in actual income,
as predicted by the model of liquidity constraints.
Hall and Mishkin conclude that these results would
of consumers were unable to borrow as much as
they wished. Once again, these results were
supported by other similar evidence, see in
particular Zeldes ( ) and
2090
Bemanke ( ), also using the PSID, Runkle
( ), using data from the Denver Income Main-
tenance Experiment, and Hayashi ( ) using
panel data from Japan. However, one potential
problem with the use of panels is the importance of
errors of measurement in such data. There is a
considerable body of evidence that PSID income
changes are subject to very substantial reporting
errors, see in particular Altonji ( ), Duncan
and Hill ( ), and Abowd and Card ( ).
Altonji and Siow ( ) have recently estimated
a model similar to Hall and Mishkin’s using the
PSID but with allowance for measurement error,
and they find little conflict with the view that
consumption responds only to news. However, it is
unclear, at least to this reader, whether the
acceptance of the model represents low power once
errors of measurement are allowed for, or whether
such errors really offer a plausible explanation for
Hall and Mishkin’s findings.
A more formal line of research has attempted to
estimate the Euler condition ( ) directly, thus
avoiding the approximations made by Hall and by
others. Rewrite (15) once more, this time as
(1 + 7"f+l)(c*+l) a ~ (l+^)(c;) a = £;+l (21)
where, as before £,+! is orthogonal to any variable
known in period t or earlier. Hansen and Singleton (
) proposed that the parameters in ( ) be
estimated by a generalized methods of moments
scheme. Suppose that we have two variables or
instruments zu and z2t, each known at time t, so that
we have Et(zit £, + j) 0 for i 1,2. We can
then estimate the two unknown parameters, a and
A, by equating sample and theoretical moments,
and solving the two equations, i = 1, 2
T '^[z„{(l +/V+iKo+i) a — (1 + £>XC<) °1] = 0-
1=0
(22)
If, as is typically the case, we have more than two
z-variables, then it will not generally be possible to
choose the two parameters so that ( ) is exactly
zero. Instead, the vector can be made as small as
possible, or more specifically, the
Consumer Expenditure
parameters can be estimated by minimizing a qua-
dratic form that can be thought of as a weighted
sum of squares of the left-hand side of ( ); see
Hansen and Singleton for details. If the model were
true, this minimized value ought to be small, so that
with more instruments than parameters, the
generalized method of moments procedure yields a
test-statistic that is diagnostic for model adequacy.
Test procedures based directly on the Euler
conditions have several notable advantages. As was
the case for Hall’s procedures, few assumptions
have to be made about the structure of the income
process, and the model satisfies the best
professional standards of seeking a direct confron-
tation between theory and data with as few
approximations and supplementary assumptions as
possible. The model can also be readily extended to
test the implications of the consumption asset
pricing model by repeating the tests using the
returns on a range of alternative assets, see ( )
above. Hansen and Singleton’s study, as well as
several others, find that the test statistics are much
too large to be consistent with the theory and so
reject the intertemporal model implied by the Euler
conditions. Given the apparent superiority of the
tests, these results have been accorded a great deal
of weight in the literature. However, while I believe
that Hansen and Singleton’s work represents a very
important methodological advance, I think that
there are good reasons for not treating their results
as a definitive rejection of life-cycle theory. The
high level of technique that is embodied in deriving
the Euler equation, not to mention the complexity of
generalized methods of moments estimation, should
not blind us to the very simple, even simple-
minded, economic story that underlies these models.
Fundamentally, the Euler equation says that the
marginal rate of substitution between today’s and
tomorrow’s consumption should be equal to the rate
of return on assets between today and tomorrow, so
that estimation of the Euler equation, unlike the
Hall or excess-sensitivity tests, focuses very directly
on the relationship between real interest rates and
changes in real consumption, and the model will not
fit the data if there is no close association between
the two. And it only takes a very cursory
Consumer Expenditure
inspection of United States time-series data to see
that there is no such association. Real consumption
grew in all but one year between 1954 and 1984,
while real after-tax interest rates were as often
negative as positive, so that consistency with the
theory would require that the pure rate of time
preference be negative. Nor is there any association
between the rate of growth of consumption and the
level of real after-tax interest rates, see Deaton (
) for some data. But this
in no way reflects badly on the life-cycle theory. As
was made perfectly clear in the original Modigliani
and Brumberg papers, and it is the essence of the
life-cycle model, aggregate consumption cannot be
expected to behave like individual consumption.
Imagine a stationary economy with neither
population nor real income growth, in which there
is an excess of real interest rates over the rate of
pure time preference, and in which all consumers
have identical additive lifetime preferences with
isoelastic subutility functions. In such an economy,
each individual has a consumption path that is
growing over time, but aggregate consumption is
constant, a result that is achieved by old people
dying and being replaced by young people who
have much lower consumption levels relative to
their incomes. Unless we believe that there is some
automatic and immediate relationship between real
interest rates, time preference and growth, as would
obtain for example along a ‘golden age’ growth
path, or unless we believe that consumers have
infinite lives, then there is no reason at all to
suppose that aggregate consumption should look at
all like the life-cycle path of a representative
consumer. Representative agent models are
frequently useful, and it is not very constructive to
dismiss macroeconomics because it requires
implausible aggregation assumptions. However, the
life-cycle model provides a well-worked-out
account of individual and aggregate saving, an
account that is consistent with a good deal of other
evidence and theory, and it does not predict that
aggregate consumption should be consistent with
the intertemporal optimization conditions for a
single individual. The general question of the
effects of interest rates on consumption is
something that has remained in dispute for a long
time, and in spite of repeated
2091
attempts to isolate the effect, careful studies have
tended to be unable to do so, or at least to find
effects that are at all robust, or that can be replicated
on even slightly different data sets or data periods.
Economic theories or policy prescriptions that rely
on intertemporal substitution of consumption in
response to changes in real interest rates are not
well-buttressed by any solid body of empirical
evidence.
Another useful approach to testing the life- cycle
model is to consider the stylized facts of the income
and consumption processes, and to see whether
consumption behaves in the way that is to be
expected given the stochastic process of income.
Most people who have studied the time series for
quarterly real disposable income in the United
States agree that, like GDP, the series can be
parsimoniously described by a model that is linear
in its first two lags, i.e. an autoregression of the
form
y, = «i + + oc3y;_2 + ut (23)
where u, is the income innovation, that part of
current income that cannot be anticipated from
previous observation of the series. Of course, real
income is not a stationary series, but has a strong
upward trend, and there is considerable
disagreement about the nature of this trend, what is
the economic story behind it, and how it should be
modelled. One possibility is that real income
contains a deterministic time trend, so that there is
some sort of equilibrium growth path that cannot be
altered by shocks to the economy. Shocks certainly
exist, but they cause only short term temporary
deviations from the path and have little or no long-
term temporary deviations from the path and have
little or no long-term significance. In this view, Eq.
( ) applies to the deviations of income from trend,
not to income itself; equivalently, ( ) can be
modified by including a linear or quadratic time
trend. The alternative view is that there is no
deterministic trend, but that the rate of change of
income is a stationary stochastic series with
constant mean. In practice, this can look very like
the previous model, but there is the vital conceptual
difference that in the second, non-deterministic
model, there is nothing that will
2092
ever bring income back to any deterministic path. In
consequence, shocks to current income have
permanent and long-lasting effects. The version of (
) that corresponds to this view can be written.
(Or - y - 1) -1’} = p { ( y t-1 - ^-2) - y }
t + u,
(24)
which can readily be seen to be a special case of (
), though note that it is the case where the time
series possesses a unit root, or is stationary in first
differences. For ( ) to be a valid specialization of
( ), the quadratic equation with the a’s of ( ) as
coefficients must have a unit root, hence the term.
Equation ( ) appears to fit the data well and the
parameter p turns out to be around 0.4, so that ( )
says that if the increase in real income in one
quarter is greater than its long term mean, then the
next quarter’s increase is also likely to be above the
mean, though by less. While the longterm mean of
the rate of change of income is constant and equal
to y, good fortune (positive w’s) and bad fortune
(negative m’s) never have to be paid for (or made
up), since shocks are immediately consolidated into
the income level, and growth goes on in the same
way as before, but from the new base. As Campbell
and Mankiw ( ) have emphasized, the unit root
model
exhibits shock persistence, while the deterministic
trend model does not; they suggest that shock
persistence is what we should expect if supply
shocks predominate over demand shocks, with the
reverse in standard Keynesian models where shocks
are typically attributed to fluctuations in aggregate
demand.
It turns out that it is almost impossible to tell
these two processes apart on United States time-
series data. Processes with unit roots are inherently
difficult to tell apart from processes that are
stationary around deterministic trends, and the tests
that are available, Dickey and Fuller (), Phillips and
Perron ( ), certainly can-
not reject the hypothesis that ( ) is a valid spe-
cialization of ( ). Nor would the
tests convince a
believer in the deterministic model that income does
not have a deterministic trend, even though it wi 11
readily be recognized that the deviations from
Consumer Expenditure
trend are themselves close to non-stationarity. Since
both process are special cases of ( ) with
the inclusion of a trend, and since each assumes
parameter values that are very close to one another,
one might think (and hope) that the two models
would have very similar implications. But it is easy
to see this is not true. If permanent income is taken
as the annuity value of discounted future incomes,
then ( ) implies that any innovation u, to current
income, because it will persist forever, and because
it can be expected to be followed by another
infinitely persistent innovation of the same sign,
will change permanent income by more than the
amount of the innovation. Equation ( ) below gives
the formula for the change in permanent income, if
the real interest rate is r, and if real income follows
( ), see Flavin ( ) or Deaton ( ),
+
Ay? (25)
>■)
+1 -
so that the change in permanent income is between
one and a half and twice as large as the innovation
in current income. By contrast, fitting the
deterministic model yields a much smaller effect,
with the change in permanent income about one
fifth of the shock in measured income. Since
consumption should change by about the same
amount as does permanent income, the fife- cycle
model, together with the unit root formulation,
yields the uncomfortable prediction that con-
sumption should be more variable than income over
the business-cycle, not less. If the unit root model is
correct, then the life-cycle and permanent income
models can be rejected because they predict what
they were designed to predict, that consumption is
smooth relative to real income! The deterministic
model gives no such problems, but as yet we have
no way of being sure that it is correct, unless, of
coinse we assume from the start that the life-cycle
story is true.
There is insufficient space in this essay to fol-
low these issues further, or to discuss in detail the
evidence for and against the two formulations of the
stochastic process governing real income; the
interested reader can refer to Deaton ( ) and
Consumer Expenditure
to the evidence on persistence in GDP presented by
Campbell and Mankiw ( ) and by
Cochrane ( ). There are a number of possible
solutions to these puzzles, and a great deal of
empirical work remains to be done, though I suspect
that the time-series data on income are
insufficiently long to allow the isolation of the very
long-run properties on which the permanent income
theory rests, see in particular the interesting paper
by Watson ( ).
Variations on the Basic Theme
There exist many interesting developments of the
basic life-cycle model, and I have space to discuss
only a few. I have already mentioned the role of
liquidity constraints, and many people would take it
as transparent that many consumers do not have
access to unlimited credit, or else face borrowing
rates that are higher than the rates at which they can
lend. Of course, many consumers may be able to
smooth their consumption without recourse to
borrowing, and the borrowing needs of many others
may be met by the typically rather good markets in
home mortgages. For consumers who nevertheless
wish to borrow but cannot, then- spending will be
closely tied to their actual income. For some of the
theoretical and empirical literature on this point see
Flemming ( ),
Dolde and Tobin ( ), and Hayashi ( ).
The theoretical consequences of uncertainty about
the date of death have been worked out by Yaari (
), and as argued above, play a possibly
important part in the explanation of the saving
behaviour of the elderly.
Another line of research is the possible relaxa-
tion of the assumption that preferences are
intertemporally additive. Allowing all periods (or
ages) to interact with all other periods in an
unrestricted way, as in Eq. ( ), would be much too
general to be useful, and the search has been for
simple models that break the restriction in a natural
and straightforward way. One useful analogy is
with the theory of durable good purchases, where
utility depends on the stock of assets possessed, the
stock in turn being the integral of past purchases
less depreciation. Purchases in one period
2093
therefore have consequences for utility in subse-
quent periods, something that will be taken into
account by a forward looking consumer. In the case
of durable goods, the assumption of perfect capital
markets effectively converts durable into non-
durable goods, with the price of a unit of stock for
one period being the implicit rental or user cost, the
latter being defined as the sum of interest cost,
depreciation, and expected capital loss, see for
example Diewert ( ) or Deaton
and Muellbauer ( , ch. 13).
However, various authors, Houthakker and
Taylor ( ) perhaps being the first, have
extended the durable model to encompass ‘psychic’
stocks which, like physical stocks, are augmented
by purchases and diminished by depreciation, but
unlike physical stocks, can either increase or
decrease utility. The latter case covers habit
formation; consumption of an addictive good
generates pleasure now, but engenders a hungry
habit that is pleasureless but costly in the future.
The model has been given an elegant formulation in
two papers by Spinnewyn ( , ).
As an example, see also Muellbauer ( ), take
the utility function
T-t
ut = ^2 + d)~kv(ct+k - <xct+k-1)
( 26 )
k=0
where a is a measure of habit formation. Spinnewyn
maximizes this function with respect not to c,, but
with respect to the ‘net’ quantities z, = c, — y.c, _ i,
and shows how to rewrite the budget constraint so
as to define corresponding prices of the z s that
reflect not only market prices of the goods, but also
the costs of consumption now in terms of pleasure
foregone later. Under certainty, and looking ahead
from time t, the full shadow price of an additional
unit of consumption now is
T-t
Pz = Yl [“/(1 + r)fPt+k ( 27
)
k=0
because the habits that are built up now have to be
paid for later. Note that this sort of formulation also
predicts that it is ct — y.c, , not c, that is
2094
proportional to permanent income, so that con-
sumption itself will adjust only sluggishly to
changes in permanent income with habits causing a
drag. Other formulations of non-separable pref-
erences can be found in the papers by Kydland and
Prescott ( ), and by Eichenbaum et al.
( ), both of which are concerned to reconcile
fluctuations in the aggregate economy with the
behaviour of a single representative agent.
Many of the models discussed so far assume that
the consumption function actually exists, hence
taking for granted the essentially keynesian
assumption that income is given to the consumer,
and is not chosen together with consumption. A
considerable body of work has grown up in the last
ten years that is concerned with the simultaneous
choice of labour supply and consumption in a life-
cycle setting. Heckman ( ) and Ghez
and Becker ( ) are among the pioneers of this
approach. Unlike the price of goods, the price of
leisure tends to show a systematic pattern over the
life-cycle, so that, if consumers are free to choose
their hours, and if they can freely borrow and lend
so as to transfer resources between periods, it will
pay them to work hardest during those periods in
their life-cycles when the rewards for doing so are
highest, and to take their life-time leisure when
wage rates are low and leisure is cheap. There is
superficial evidence in favour of this story, and
Ghez and Becker, followed by Smith ( ) and
Browning et al. ( ), all find that workers tend
to work longest hours in middle age when wage
rates are high and the lowest number of hours at the
beginning and end of the economically active life,
when wage rates are relatively low. Consumption
also tends to peak in the middle age, and this can be
brought into the story by assuming that
consumption and leisure are complements, so that
the lack of leisure in middle age is partially
compensated by high levels of expenditure. This
elegant fable has also been made much of in
equilibrium theories of the business cycle, which
accounts ‘unemployment’ as a voluntary vacation
taken when the real wage is low and leisure is on
sale, see in particular Lucas and Rapping (
and Lucas ( ).
There now exists a growing volume of literature
that shows just how much violence to the
Consumer Expenditure
facts is done by this story. All the evidence quoted
above looks across different individuals at different
points in their life-cycles, while the theory says that
the same individual will change his or her hours of
work along with changes in the real wage over the
life-cycle. Time-series and panel data from the
United States and time-series of cross-sections from
the United Kingdom suggest that this is simply not
the case, see for example Mankiw et al. ( ),
Ashenfelter and Ham
( ), Ashenfelter ( ), and Browning et al.
( ). Even MaCurdy’s ( ) more postive
study provides only very weak evidence, see in
particular Altonji ( ). The joint consumption
and labour supply story fares even less well than the
labour supply model alone, and there is clear
evidence that the way in which consumption and
hours fluctuate over the cycle (sometimes together
and sometimes in opposite directions) is not con-
sistent with the way in which they move together
over the life-cycle. The attempt to provide a unified
theory of business and life-cycles has been an
interesting and important one, but it cannot be said
to have been successful.
I have been somewhat cavalier in my treatment
of aggregation issues, choosing to emphasize them
when I believe them to be important, for example in
the fitting of Euler conditions, and ignoring them
when it has been convenient to do so. Attempts to
do better than this have not been notably successful.
Formal conditions that allow aggregation in
consumption function models are typically too
restrictive to be useful, so that, in theory, changes in
the distribution of income should have detectable
effects on aggregate consumption. However,
attempts such as that by Blinder ( ) to link the
distribution of income
to consumption have not been notably successful,
perhaps because the income distribution is not
variable, or because it changes smoothly enough
over time to preserve a stable relationship between
average income and average consumption. There is
also an issue of aggregation over goods in order to
define real consumption at all, even at the level of
) the individual agent. In the derivation in section “
” above,
I made the convenient assumption that within-
period preferences were homothetic, so that an
Consumer Expenditure
index number of real consumption could be formed.
But homotheticity, although very convenient for
studying the consumption function, is very
inconvenient for studying the allocation of
expenditure among goods since it implies that the
within-period total expenditure elasticities of each
good are all equal to unity. Fortunately, there are
aggregation results of Gorman’s ( ), see also
Deaton and Muellbauer ( , eh. 5) for an
exposition that allows us to have the best of both
worlds, at least if we remain with intertemporally
additive preferences. If the single-period indirect
utility function iJ/(x, p) takes the form known as the
‘generalized Gorman polar form’
' K X , P) = F[x/a(p)\ + b(p) ( 28 )
where a(p) and b(p) are linearly homogeneous
functions of prices and F(-) is monotone increasing,
then the real expenditure index x/a(p) can serve as
an indicator of real consumption just as in the
homothetic case. This happens because when the
consumer chooses the allocation of life-time
expenditure over periods so as to maximize the
intertemporal sum of terms like ( ), the
b(p) terms are irrelevant. However, the intraperiod
demand functions that correspond to ( ) traditionally it is treated as if it were a datum by the
do not display unitary elasticities unless the b(p) is
identically equal to zero, and quite general func-
tional forms are permitted. There is therefore no
real conflict between the analysis of the consump-
tion function on the one hand, and the analysis of
demand on the other. It is to the latter that I now
turn.
Theoretical and Empirical Demand
Functions
Demand functions are the relationships between the
purchase of individual goods, income or total
expenditure, prices, and a variety of other factors
depending on the context. Economists have
attempted to make empirical links between demand
and price since Gregory King’s famous demand
curve for wheat, see Davenant ( ), As any intermediate text shows, see for example
and since the middle of the 19th century, there has Deaton and Muellbauer (
been a great development in the theory of Slutsky matrix must be symmetric and negative
2095
consumer behaviour. Much practical work con-
tinues in the tradition of King, paying little attention
to formal theory, concerning itself instead with
finding empirical regularities. For a firm studying
the demand for its product, or for anyone interested
in establishing a single price elasticity, this
probably remains the best approach; the major
developments in econometric technique and
empirical formulation have not been much
concerned with, or relevant to, these very practical
questions. The pragmatic approach (the term comes
from Goldberger’s famous but unpublished ( )
study ), probably reached its peak with the
publication of Richard Stone’s great monograph,
(Stone ), and much is still to be learned by a careful
study of Stone’s procedures for measuring income
and price elasticities. However, in this essay, I shall
follow the literature, and follow its more
methodological approach.
The theory outlined in section “
” above suggests that the
demand functions of an individual consumer can be
derived by maximizing a utility function v(q)
subject to a budget constraint p ■ q = x, where x is
total expenditure. In the analysis here, x is chosen at
some previous level of decision making, but
consumer, the utility maximization yields a vector q
that is some function g(x, p), say, of total
expenditure and prices. These demand functions
cannot simply be any functions, but must have
certain properties as a result of their origins in
utility maximization. Obviously, the total value of
the demands should be equal to total outlay x, the
‘adding-up’ property, and it must be true that
proportional changes in x and in p do not have any
effect on quantities demanded, the ‘homogeneity’
or ‘absence of money illusion’ property. Somewhat
less obvious are the famous symmetry and
negativity properties. These apply to the Slutsky (
) matrix, S, the typical element of which is
defined as
Sij = dqi/dpj + qpjqj/dx. (2 9)
, ch. 2), the
2096
semi-definite. The symmetry property is not readily
turned into simple intuition; negativity implies that
the diagonal elements of the matrix are nonpositive,
a proposition often referred to as ‘the law of
demand’. The four properties, adding-up,
homogeneity, symmetry and negativity, essentially
exhaust the implications of utility maximization, so
that any empirical demand functions that satisfy
them can be regarded as having been generated by
utility maximization, or by rational choice, with
‘rational’ defined, following Gorman ( ), as
‘having smooth strictly quasi-concave
preferences, and being greedy’.
Stone ( ) was the first to attempt to use
this theory directly to confront the data. He started
from a (general) linear expenditure system of the
form
Pfli = X! avPj + b‘x
j total outlay increases), nor can it allow goods to be
where a,, and bt are unknown parameters. Stone
showed that, in general, the system ( ) does not
satisfy the four requirements, but will do so if, and
only if, the parameters are restricted so that the
model can be written in the form
Pfli = Pill + lh( x
P ■ y) (31)
with the /(-parameters summing to unity. In this
form the model is known as the linear expenditure
system. As Samuelson ( -8) and Geary
( -50) had earlier shown, the utility function
corresponding to ( ) has the form
u
= X) X) AM?/ “ ?/) >
(32)
i generalizing Stone’s procedure of writing down
sometimes refened to (somewhat inappropriately)
as the Stone-Geary utility function. It can be
thought of as a sum of Bernoulli utility functions of
the quantity of each good above the minimal y’s.
Stone’s achievement lay not in deriving the
demand functions, but in thinking to estimate them.
The demand functions ( ), even if fitted to the data
by least-squares, require non-linear optimization,
and Stone invented a simple and not very efficient
scheme, but one that allowed
Consumer Expenditure
him to obtain parameter estimates and a good fit to
interwar British data for a six commodity disag-
gregation of expenditures. This was a major
breakthrough, not only in demand analysis, but also
in applied econometrics in general. Indeed, much of
demand analysis for a decade or so after Stone’s
paper consisted of applying better algorithms and
faster computers to the fitting of Stone’s model to
different data sets.
The linear expenditure system offers a demand
model for a system of, say n goods, and requires
only 2« — 1 parameters, a degree of parsimony that
was very important in allowing the model to be
estimated on very short time-series data. However,
such economy brings its own price, and the linear
expenditure system is very restrictive in the sort of
behaviour that it can allow. In particular, and
pathological cases apart, the model cannot allow
(3
°) inferior goods (goods the demand for which falls as
complements rather than substitutes. (As defined by
Hicks ( ) goods i and
j are complements if the (/', /jth term in the Slutsky
matrix is negative, so that the utility compensated
cross-price response of i to an increase in the price
of j is positive.) Normal (non-inferior) goods that
are substitutes for one another may be the most
important case, but they do not encompass every-
thing that we might want to study. The linear
expenditure system also implies that the marginal
propensity to consume each good is the same no
matter what is the total to be spent, and many cross-
section studies of household budgets have suggested
that this is not in fact the case.
Unfortunately, it is quite difficult to write down
utility functions that will lead to more general
demand functions than those of the linear expen-
diture system, nor is there any obvious way of
functions and making them consistent with the
theory. Progress was only really made once applied
demand analysis started using ‘dual’ formulations of
preferences to specify demands. In the demand
context, duality refers to a switch of variables, from
quantities to prices, so that utility becomes a
function, not directly of quantities consumed, but
indirectly of prices and total expenditure. This
indirect utility formulation is given by
Consumer Expenditure
the function ij/(x, p), already used above, and this is
simply the maximum attainable utility from total
outlay x at prices p. Since ij/(x, p) = u, and the
function is monotone increasing in x, it can be
inverted to give x c(u, p), known as the ‘cost
function’, since it gives the minimum necessary
cost that is required to reach the utility level u. By a
theorem usually attributed to Shephard ( ) (
and to Uzawa ( ), these two functions contain
a complete representation of preferences; provided
preferences are convex, and provided the functions
satisfy homogeneity and convexity (or concavity)
conditions, preferences can be reconstructed from
knowledge of either of the two functions. It is also
very easy to move from either cost or indirect
utility functions to the demand functions. For the
indirect utility function, we have Roy’s identity
(Roy ).
q = - V p i j/{x,p)/\i/x{x,p) = g ( x, p ) (3 3) ‘almost ideal demand system’ (AIDS) has been
which immediately yields demand functions from
preferences in a form that are suitable for estima-
tion, while for the cost function, we have
Shepard’s Lemma ( ),
q = Vpc(u,p) = Vpc[il/(x,p),p] = g ( x, p ) (3 4)
where, as in ( ), the operator V denotes a vector of
partial derivatives.
Demand analysis now had a high road to spec-
ification. Think of some quasi-convex decreasing
function of the ratios of price to total outlay and call
it an indirect utility function, or think of some
function of utility and prices that is increasing in its
arguments and linearly homogeneous and concave
in prices and call it a cost function. Either way, and
with only simple differentiation, new (and
sometimes) interesting demand functions will be
generated. Alternatively, and even more
importantly, it is possible to use theory to aid and
check out empirical knowledge. If it is known that
the marginal propensity to spend on food is a
declining function of total expenditure, or if it is
thought likely that some goods do not depend very
directly on the prices of other goods, it is relatively
straightforward to find out what preferences (if any)
will yield the result. It becomes possible,
2097
not just to generate demand functions serendipi-
tously, but to generate good and useful ones
deliberately.
There are many examples that could be cited
from the literature. One of the most widely used in
the translog model which was first proposed in
1970 by Jorgenson and Lau, see Christensen et al.
) for a convenient reference. To derive the
translog, write the indirect utility function in terms
of the ratios of prices to outlay, r = p/x, and
approximate the indirect utility function as a second
order polynomial in the logarithms of r. Application
of Roy’s identity yields demand functions in which
the budget share of each good is the ratio of two
functions, each of which is linear in the logarithms
of the price to outlay ratios. Estimation of these
rational functions, like estimation of the linear
expenditure system, requires the use of non-linear
maximization techniques. A related model, the
proposed by Deaton and Muellbauer ( ),
and I use this to illustrate some of the
issues that arise with the current generation of
demand models. The AIDS is specified by the
logarithm of its cost function which takes the form
In c{u,p) = ao + o^ln pk + 0.5
k k
X ^Jkm^ /frln Pm
m
+ u exp P* j> (35)
so that, applying Shephard’s lemma and
rearranging, we have demand functions
Ptq jx EE Wi
= ai + PM X /P) + 7'7ln Pj (36)
j
where P is a linearly homogeneous price index, the
form of which can readily be inferred from ( ). The
parameters of the model must satisfy certain
restrictions if ( ) is to be a proper (log) cost
function, and ( ) a proper system of demand
functions. The matrix of y-parameters can be
2098
taken to be symmetric in ( ), but must be so in
( ), and its rows and columns must add to zero
for the homogeneity and adding-up properties to be
satisfied. The /(-parameters can be positive or
negative, with positive values indicating luxury
goods, and negative values necessities. The main
advantage ofthe AIDS model in time-series appli-
cations is that the price index P can typically be
approximated by some known price index selected
before estimation, so that the demand system is
linear in its parameters. In consequence, it can be
estimated by ordinary least squares on an equation
by equation basis, at least if the symmetry of the y-
mattix is ignored. The homogeneity restrictions can
be tested equation by equation using a t- or F-test,
and while imposing or testing symmetry requires an
iterative procedure, estimation can be done by
straightforward iterated restricted generalized least-
squares, see Barten (1979) or Deaton ( ) for
further discussion.
The results of estimating the AIDS model are
sufficiently similar to those from other models and
other studies, see e.g. Barten ( ), Deaton
( ), Christensen et al. ( ), and many
others, that perhaps they can be taken as represen-
tative. What typically seems to happen is that the
homogeneity restrictions appear not to be satisfied,
so that in the application of AIDS to British data,
Deaton and Muellbauer found, for example, that the
F-test for transport had a value of 172 compared
with the 5 per cent critical value of 4.8. Results on
symmetry from AIDS and other systems are more
mixed, and it now seems clear that testing
symmetry is not usually possible given the amount
of data typically available in time series, or put
more positively, that there is no convincing
evidence against symmetry. The difficulty is that
symmetry involves a set of restrictions across
different equations, so that unlike homogeneity,
which involves tests within each equation, exact,
small sample tests are not available. Researchers
have therefore fallen back on asymptotically valid
tests, and it turns out that these work very badly for
the usual sort of samples, especially when there are
more than a very small number of goods in the
demand system. The papers by Laitinen (
and Meisner ( ) further details. The AIDS, by contrast, has exactly
first established the problem, see also Evans and
Consumer Expenditure
Savin ( ) and Bera et al. ( ) for further
evidence.
The AIDS model, like the translog and several
others, e.g. Diewert’s ( ) ‘generalized
Leontief’ system, fall into the class of ‘flexible
functional forms’. This criterion of flexibility, first
proposed by Diewert ( ), is an important guar
antee that the model is sufficiently richly parame-
trized so as to allow estimation of what are thought
to be the main parameters of interest, typically the
total expenditure elasticities, and the matrix of own
and cross-price elasticities. A ‘second order’
flexible functional form is one that has sufficient
parameters, so configured, that it is possible to set
the value of the function, and of its first and second
partial derivatives to any arbitrary set of
(theoretically permissible) values. By applying
Roy’s identity or Shephard’s lemma, it is clear that
a cost or indirect utility function that is a second
order flexible functional form will yield demand
functions that are first-order flexible, so that it is
possible for estimation to yield any set of price and
expenditure elasticities that are consistent with
utility theory. For empirical work, such a guarantee
is important, because it ensures that the elasticities
are being measured, not assumed. Contrast, for
example, the linear expenditure system ( ) with the
AIDS model ( ). Both could be fitted to the same set
of data, and the parameter estimates of each could
be used to generate a complete set of expenditure
and price elasticities. But the linear expenditure
system is not a flexible functional form, and so its
estimated elasticities are not independent of one
another, as is apparent from the fact that there are
2n — 1 parameters compared with the total number
of potentially independent elasticities, which is (n
— 1)(1 + n/l). (There are n 1 independent demand
equations, each of which has an expenditure
elasticity, and n price elasticities; however, one
price elasticity per equation is lost to homogeneity,
and symmetry imposes a further (n l)(n 2)/2
constraints.) The linear expenditure system does not
therefore measure all the price and income
elasticities, but determines them by a mixture of
measurement and assumption, the main assumption
) being that of additive preferences, see Deaton () for
Consumer Expenditure
the right number of parameters to allow for inter-
cepts and a full set of elasticities, so that when it (or
the translog, or the generalized Leontief) is
estimated, so is the full set of elasticities.
Being able to do this is a great step forward in
methodology, but just as the linear expenditure
system probably asks too little of modem data,
(although not of the data available to Stone and the
early pioneers of the systems approach), the second-
order flexible functional forms probably ask too
much, or equivalently, put too little structure on the
problem. The consequences show up in large
standard errors, a high frequency of apparently
chance correlations, and a lack of robustness to
functional form changes within the class of flexible
functional forms, in other words, in all the standard
symptoms of over-parametrization. These problems
are particularly acute for the measurement of price
elasticities, because in most time-series data,
commodity prices tend to move together with
relatively little variation in relative prices. And
although the focus of most research on demand
analysis over the last thirty years has been on the
estimation and testing of price responses, there is
certainly no consensus on what numbers, if any, are
correct. Estimates obtained from the linear
expenditure system are not credible because they
are forced to satisfy an implausibly restrictive
structure, while those from flexible functional forms
are not credible because the data are not informative
enough to supplement the lack of prior structure.
Some intermediate forms are clearly required.
One of the attractions of flexible functional
forms is their ability to approximate quite general
forms for preferences. However, the models so far
considered offer only approximations, and there is
no guarantee that they have satisfactory global
properties. Partly this is the standard problem that a
fitted model will be forced to give a reasonable
account of the data over the sample used for
estimation, but may predict very badly elsewhere.
But there are other deeper issues. Taking the AIDS
as an example, estimation of ( ) subject to sym
metry and homogeneity will produce a system of
estimated demand functions that will satisfy adding-
up, homogeneity and symmetry for all values of x
and p. However, there are two other
2099
important properties that are not assured. First, there
is no guarantee that the predicted budget shares will
necessarily lie between zero and one, so that there
may be regions of price space in which the
estimated model yields nonsensical predictions.
Second, there is no way that the AIDS can be
guaranteed to have a negative semi-definite Slutsky
matrix for all prices, at least not without restricting
parameters to the point where the model ceases to
be a flexible functional form. The parameters could
be chosen so as to satisfy negativity for some
particular combination of prices and outlay, but
there will be no guarantee that the law of demand
will be satisfied elsewhere. In the translog model, it
is possible to impose a restriction that guarantees
negativity everywhere, but the model with the
restriction has the property that all estimated own
price elasticities must be less than minus one,
independently of whether this is in fact true, and it
almost certainly is not, see Diewert and Wales (
). A demand system
is described as ‘regular’ if it has anegative definite
Slutsky matrix and predicts positive demands, and
several empirical studies, see e.g. Wales ( )
for one of the first, found that estimated flexible
functional forms were not regular over disturbingly
large regions of even the parameter space used to
estimate them. Caves and Christensen ( ), and
later Barnett and Lee ( ) and
Barnett et al. ( ), investigated the same prob
lem theoretically by taking a known utility function,
choosing the parameters of flexible functional forms
to match its level and derivatives at a point, and
then mapping out the regions of price space in
which the systems remained regular. The results at
least for the translog and the generalized Leontief
model, were not good.
These regularity issues may seem of limited
importance in practice, but this is far from being the
case. One of the major reasons for being interested
in complete empirical demand systems is to be able
to examine the consequences of price changes,
particularly of price changes that follow changes in
government policy. The United States relies
relatively little on indirect taxation as a source of
public finance, but such is not the case in most of
Europe, and the vast majority of developing
countries maintain complex systems of
2100
price wedges, particularly for foods and for agri-
cultural production. The effects of such systems
cannot be predicted without good information on
how demands respond to price changes, nor can
reforms be intelligently discussed. However, esti-
mated demand systems that are not regular are not a
great deal of help. All of the theory of welfare
economics, of consumer surplus, of optimal taxation
and of tax reform, assumes that demand behaviour
is generated by utility maximization at the
individual level, and implementation without
regularity risks internal contradiction. For example,
if compensated demand functions slope upwards,
the government can generate a deadweight gain by
imposing a distortionary tax. Of course, it may not
be the empirical work that is wrong, but the theory
that we used to try to model behaviour. If so, the
estimated demand functions are still not useful,
since we now have no idea what to do with them.
But I doubt that evidence goes so far; it is not that
behaviour itself is irregular, but that we have not yet
found a good modelling strategy that contains a
reasonable amount of prior information to
supplement the paucity of data, and at the same time
can deliver global regularity if it is warranted by the
evidence.
A number of interesting experiments are cur-
rently under way that involve new modelling
techniques. One possibility is that the Taylor series
expansions that motivate most flexible functional
forms are themselves inadequate to the task. In
particular, Taylor approximations lose their ability
to approximate if they are also asked to possess
other properties of the functions that they are
approximating. For example, we might want to test
whether or not preferences are additively separable,
as in the linear expenditure system. One strategy
would be to write down some second-order
approximation to preferences, estimate the resulting
demand model, and then test whether or not the
conditions imposed on the demands by additivity
are satisfied. But this will not work in general,
because there may be no additive system of demand
equations that has the precise functional form
demanded by the approximation. The same
phenomenon is well illustrated by Stone’s
derivation of the linear expenditure system itself.
The original general linear
Consumer Expenditure
expenditure Eq. ( ) can clearly be justified as a
Taylor approximation to any set of homogeneous
demand functions, and yet the imposition of only
symmetry generates the demand system ( ) which
comes from the additive utility function ( ).
Additivity is not imposed, but linear expen
diture systems are only symmetric if they are
additive. Similarly many flexible function forms are
only globally regular if they are homothetic, see for
example, Blackorby et al. ( ). Several
recent studies have proposed alternative ways of
making functional approximations. Gallant ( ) has
proposed using Fourier series approximations while
Barnett ( ) has suggested that
Laurent series can be used to generate demand
models with good properties. Gallant’s models are
even more heavily parametrized than standard
flexible functional forms, and there must be some
question as to the suitability of trigonometrical
functions for demand functions. Barnett’s ‘miniflex
Laurent’ model does not use the full flexibility of
the Laurent series, but appears to have quite good
approximation and regularity properties in practice,
see Barnett and Lee ( ) and Barnett
et al. ( ); even so, its estimation is complex,
and many of the parameters have to be estimated
subject to inequality constraints.
A second line of current research has abandoned
the standard approach of econometric analysis,
taking instead a completely non-parametric
approach. Since many of the difficulties discussed
above arise from choice of functional form, it is
useful to ask how far it is possible to go without
assuming any functional form at all. We know from
standard revealed preference theory that two
observed vectors of prices and quantities can be
inconsistent with utility maximization; if bundle one
is chosen when bundle two is available, so that
bundle one is revealed preferred to bundle two, then
no subsequent choice should reveal bundle two to
be preferred to bundle one. Before embarking on
the exercise of fitting some specific utility function
to any finite collection of price and quantity pairs,
one might then ask whether the collection is
conceivably consistent with any set of preferences.
If it is, then contradictions between an estimated
system and the theory must be a matter of
inappropriate functional form. The
Consumer Expenditure
conditions for utility consistency of a finite set of
data were originally derived by Afriat ( ), who
proposed a condition called cyclical consistency.
Much later Varian ( ) not only provided an
accessible and clear account of Afhat’s results, but
also recast the cyclical consistency condition into a
‘generalized axiom of revealed preference (GARP)’
that runs as follows. A bundle q is strictly directly
revealed preferred to a bundle q if p'q' > p'q, while
q‘ is revealed preferred to q, if there exists a
sequence, j, k,..., m such that p'q1 > p‘q>, p’q* >
p>qk,..pmqm > pmq, so that c{ is directly or indirectly
(weakly) revealed preferred to q. GARP is satisfied
if for all q revealed preferred to cj, it is not true that
cj is strictly directly revealed preferred to q\ and
given GARP the data can be rationalized by a
continuous, strictly concave, and non-satiated utility
function. Differentiability can also be ensured by a
sight strengthening of GARP, see Chiappori and
Rochet ( ). GARP is readily tested for any
given set of data by checking the pairwise inequal-
ities and using a simple algorithm provided by
Varian to map out the patterns of indirect revealed
preference. Repeated applications of the method to
time-series data have nearly always confirmed the
consistency of the data with the theory. In
retrospect, it is clear that violations of GARP cannot
occur unless some budget lines intersect, so that if,
over time, economic growth has resulted in the
aggregate budget line moving steadily outward with
little change in slopes, GARP is bound to be
satisfied. (However, post-war United States data
budget planes do occasionally intersect, and Bronars
( ) has recently shown that hypothet
ical demands generated by selecting random points
on the actual budget lines would more often than not
fail GARP. )
The contradictions between the parametric and
non-parametric approaches can perhaps be resolved
by thinking of the latter as a modelling technique
that uses a very large number of parameters, so that
the failure of the parametric models to fit theory to
data can be thought of as failure to parametrize the
models sufficiently richly. But I have already argued
that these models already have too many
parameters, and adding more would only exacerbate
the already serious
2101
problems of measurement. For many purposes, the
theory is only useful if it is capable of delivering a
description of the data that is reasonably
parsimonious. There is also something rather simple
minded about non-parametric techniques that tends
to be disguised by the sophisticated and elegant
expositions that have been given them by Varian
and others. Consider a very simple theory that says
variable x should move directly with variable y as,
for example, in the Euler Eq. ( ) above which says
that, under certainty consumption should grow from
period t to t + 1 if and only if the real interest rate
from t to t + 1 is greater than some fixed constant.
A non- parametric test on a finite set of data would
accept the theory if, in fact, x, and y always did
move together, and reject it if x and y ever moved in
opposite directions. That such testing procedures
are widely employed in the press and by the
uninformed public is no reason for treating them
seriously in economics.
I have so far discussed the formulation and
estimation of demand functions, meaning the rela-
tionships between quantities, outlay, and prices, and
this has been the topic of most applied demand
analysis over the last thirty years. However, there is
an older tradition of demand analysis, in which the
object of attention is household budget data, and
this literature has recently been enjoying something
of a revival. Since household budget data typically
come from a cross-section of households over a
short period of time, usually within a single year,
prices are treated as common to all sample points,
so that the focus of attention becomes the
relationship between demand and outlay and the
influence of household composition on the pattern
of household expenditures. The oldest, and perhaps
only law of economics, Engel’s Law that the share
of food in the budget declines as total outlay
increases, comes from Engel’s (1857, published )
study of Belgian working-class families, and early
empirical studies of demand were almost inevitably
based on household surveys (see Stigler ( ) for
a mas
terly review). The modem study of Engel curves,
the relationships between expenditure and total
outlay, begins (and almost ended) with Prais and
Houthakker ( ). Prais and Houthakker studied
2102
the shapes of Engel curves, the relationship between
demand and households, particularly in relation to
the choice of quality, a topic that has subsequently
been unjustly neglected. The functional forms for
Engel curves that Prais and Houthakker examined
became the staple menu for most subsequent
studies, even though only one of their forms, the
linear Engel curve, is capable of satisfying adding-
up, and the linear form typically performs very
badly on the data. Since 1955 a number of other
Engel curves have been proposed, notably the
lognormal Engel curve of Aitchison and Brown (
), and Leser’s
( ) revival of the form suggested much earl i er
by Holbrook Working ( ). Working’s form,
which apparently escaped the attention of Prais and
Houthakker, makes the budget share of each
commodity a linear function of the logarithm of
total outlay. The formulation is particularly useful,
for not only is it capable of accounting for most of
the curvature that is discovered in empirical Engel
curves, but it is also consistent with utility theory,
and corresponds to the case where the welfare
elasticity of the cost of living is independent of
income. Gorman ( ) has provided a general
characterization theorem for Engel curves of the
form
PMi = X!atk (?) ^ (*) (37)
k equivalent adults for food, and six equivalent adults
and has shown that the £*(•) functions can be
powers of .r (polynomial Engel curves), or .r
multiplied by powers of log x (Engel curves relating
budget shares to powers of the logarithm of outlay ),
or have trigonometric forms. This last form includes
Fourier representations of Engel curves, while the
first two allow Taylor or Laurent expansions for the
expenditure/outlay and for the share/log-outlay
forms. The Working-Engel curve is the first
member of Gorman’s ‘share to log’ class, and the
theorem tells us that we may add quadratic or
higher order terms to improve the fit. However,
Gorman’s paper contains a remarkable result; the
matrix of the a-coefficients in ( ) prices of goods, so that demographic changes will
has rank at most equal to three. In consequence, the
share to log and log-squared Engel curves are as
general as any, as are the Engel curves of the
Consumer Expenditure
quadratic expenditure system, see Howe et al. ().
Given Gorman’s results, and the empirical
success of the Working form, it and its quadratic
generalization deserve wide use in the analysis of
budget studies. There is also accumulating evidence
that such forms are indeed necessary. Thomas (
), in a wide-ranging examination
of household survey data from developing coun-
tries, has shown that Engel’s Law itself does not
appear to hold among the very poor, so that, in
many cases, the share of the budget devoted to food
at first rises with total outlay before falling in
conformity with the Law.
Prais and Houthakker also proposed a much-
used formulation for the effects of household
composition on behavioiu. It can be written
AwMO) = f i { x / m 0 {a ) }
(3 8)
where a is a vector of household demographic
characteristics (perhaps a list of numbers of people
in each age and sex category) and /M, and m0 are
scalar valued functions known as the ‘specific’ and
‘general scales’ respectively. In this literature,
scales are devices that convert family structure into
numbers of equivalent adults, so that a family of
two adults and two children might be two
equivalent adults for theatre entertainment, three
for milk. The general scale is supposed to reflect the
overall number of equivalent adults, so that the
Prais and Houthakker model is a simple
generalization of the idea that per capita demand
should be a function of per capita outlay. Barten (
), in a very important paper, took up the
Prais-Houthakker idea of specific scales, but
assumed that the arguments of the household utility
function were the household consumption levels
each deflated by the corresponding specific scale.
The consequences of Barten’s formulation are
similar to those of Prais and Houthakker, but
embody the additional insight that changes in
family composition affect the effective shadow
exercise, not only income, but also substitution
effects on the pattern of demand. The story is often
summarized by the phrase, ‘if you have a wife and
child, a penny bun costs three-pence’,
Consumer Expenditure
quoted in Gorman ( ), but the really far-
reaching substitution effects of children are prob-
ably on time use and labour supply, particularly of
women.
Since household surveys typically contain large
samples of households, there is less need for theory
to save degrees of freedom, and it is possible to
estimate quite general functional forms that link
expenditures to household composition patterns and
then to interpret the results in terms of the various
models. In addition, neither the Prais-Houthakker
nor the Batten model seem to yield easily
implemented functional forms, e.g. linear ones, nor
is it clear that either model is even identified on a
single cross-sectional household survey in which all
prices are constant, see for example Muellbauer (
) and Deaton
( ). However, some empirical results for the
two models can be found in Muellbauer ( , Cvcle
) and in Poliak and Wales ( , ) who
also examine Gorman’s ( ) extension of
Barten’s model in which additional people are
supposed to bring with them fixed needs for par-
ticular commodities. The fixed needs model is close
to the formulation proposed by Rothbarth ( ) for
measuring the costs of children.
Rothbarth pointed out that there are certain com-
modities, adult goods, that are not consumed by
children, so that when children are added to a
household, the only effects on the household’s
consumption of adult goods will be the income
effects that reflect the fact that, with unchanged
total resources, the household is now poorer.
Deaton et al. ( ) have recently attempted to
test Rothbarth’s contention, and in their Spanish
data it seems possible to identify a sensible group of
adult goods, the expenditure on each of which
changes with additional children in the same way as
they change in response to changes in outlay.
Studies of the effects of family composition on
household expenditure patterns have frequently
been concerned, not only with estimating demands,
but also with attempts to measure the ‘cost’ of
children. It would take me too far afield to do
justice to this topic here. Readers interested in this
controversial area should perhaps start with
Rothbarth ( ), who in a few pages makes a
very simple and quite convincing case, and look
2103
also at Nicholson ( ). Poliak and Wales ( )
weigh in on the opposite side, and claim that it is
impossible to measure child costs from expendi-
ture data. My own position is argued in Deaton
and Muellbauer ( ); there are certainly grave
problems to be overcome in moving from the
analysis of household survey data to the measure-
ment of the costs of children, and it is clear that
identifying assumptions must be made that are
more severe and more controversial than those
required, for example, to go from demand func-
tions to consumer surplus. But that does not mean
that it is not possible for such assumptions to be
proposed and to be sensibly discussed.
See Also
^ Beauests 3-HQ. the Life
► Consumer Expenditure lTe
and. the State of Research)
^ ) ernann Tiieorv
► Eder Equations
► Eiedman, Milton •• 2 200
► Rational Expectations
Bibliography
Abowd, J.M., and D. Card. 1985. The covariance structure
ofearnings and hours changes in three panel data sets.
Princeton University, mimeo.
Afriat, S.N. 1967. The construction of a utility function
from expenditure data. International Economic Review
8: 67-77.
Aitehison, J., and J.A.C. Brown. 1957. The lognormal
distribution. Cambridge: Cambridge University Press.
Altonji, J. 1986. Intertemporal substitution in labor
supply: Evidence from micro data. Journal of Political
Economy 94: S176-S215.
Altonji, J., and A. Siow. 1985. Testing the response of
consumption to income changes with (noisy) panel
data. Industrial Relations Section Working Paper No.
186. Princeton University, mimeo.
Ando, A., and F. Modigliani. 1963. The life-cycle hypoth-
esis of saving: Aggregate implications and tests. Amer-
ican Economic Review 53: 55-84.
Ashenfelter, 0.1984. Macroeconomic analyses and micro-
economic analyses of labor supply. Presented to
Camegie-Rochester Conference, Bal Harbor, Florida,
November 1983.
Ashenfelter, O., and J. Ham. 1979. Education, unemploy-
ment, and earnings. Journal of Political Economy 87:
S99-116.
2104
Barnett. W.A. 1983. New indices of money supply and the
flexible Laurent demand system. J o u r n a l of
B u s i n e s s a n d E c o n o m i c S t a t i s t i c s 1:7 23.
Barnett. W.A.. and Y.W. Lee. 1985. The global properties of
the miniflex Laurent, generalized Leontief, and translog
flexible functional forms. E c o n o m e t r i c a 53: 1421-
1437.
Barnett. W.A.. Y.W. Lee, and .VI. Wolfe. 1985. The three
dimensional global properties of the miniflex Laurent,
generalized Leontief, and translog flexible functional
forms. J o u r n a l o f E c o n o m e t r i c s - . 3-31.
Barten. A.P. 1964. Family composition, prices, and expen-
diture patterns. In E c o n o m e t r i c a n a l y s i s f o r
n a t i o n a l e c o n o m i c p l a n n i n g , ed. P.E. Hart, G.
Mills, and J.K. Whitaker. London: Butterworth.
Barten. A.P. 1969. Maximum likelihood estimation of a
complete system of demand equations. E u r o p e a n
E c o n o m i c R e v i e w 1: 7-23.
Bean. C.R. 1985. The estimation of surprise models and the
surprise consumption function. Centre for Eco- nomic
Policy Research (London). Discussion Paper
No.54, mimeo.
Bora, A.K., R. Byron, and C.M. Jarque. 1981. Further
evidence on asymptotic tests for homogeneity in large
demand systems. E c o n o m i c s L e t t e r s 8:101 105.
Bemanke, B.S. 1984. Permanent income, liquidity, and
expenditure on automobiles: Evidence from panel data.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 99: 587-
614.
Bcrnhcim, B.D., A. Schleiffer, and L.H. Summers. 1985.
Bequests as a means of payment. J o u r n a l o f
P o l i t i c a l E c o n o m y : 1045 1076.
Blackorby, C., D. Primont, and R.R. Russell. ll>77. On testing
separability restrictions with flexible functional forms.
J o u r n a l o f E c o n o m e t r i c s 5: 195 209.
Blinder, A.S. 1975. Distribution effects and the aggregate
consumption function. J o u r n a l o f P o l i t i c a l
E c o n o m y 83: 447 475.
Blinder, A.S., and A.S. Deaton. 1985. The time series
consumption function revisited. B r o o k i n g s P a p e r s
o n E c o n o m i c . i c t i v i t y 2: 465 5 1 1 .
Breeden, D. ll>7‘>. An intertemporal asset pricing model with
stochastic consumption and investment opportunities.
J o u r n a l o f F i n a n c i a l E c o n o m i c s 7: 265-296.
Bronars, S.G. 1987. The power of non-parametric tests of
preference maximization. E c o n o m e t r i c a 55: 693 698.
Browning, M.J., A.S. Deaton, and M.J. Irish. 1985. A
profitable approach to labor supply and commodity
demands over the life cycle. E c o n o m e t r i c a 53: 503-
543.
Campbell, J.Y. 1687. Does saving anticipate declining labor
income? An alternative test of the permanent income
hypothesis. E c o n o m e t r i c a 55: 1249-1273.
Campbell, J.Y., and N.G. Mankiw. 1986. Are output fluc-
tuations transitory? National Bureau of Economic
Research Working Paper 1916, processed.
Consumer Expenditure
Caves, D.W., and L.R. Christensen. 1980. Global properties of
flexible functional forms. A m e r i c a n E c o n o m i c
R e v i e w 70: 422-432.
Chiappori. P.-A.. and.I.-C. Rochet. 1987. Revealed prefer-
ences and differentiable demand. E c o n o m e t r i c a 55:
687-691.
Christensen, L.R.. D.W. Jorgenson, and L.J. Lau. 1973.
Transcendental logarithmic production frontiers.
R e v i e w o f E c o n o m i c s a n d S t a t i s t i c s 55: 28-
45.
Cochrane, J.H. 1986. How big is the random walk in GNP?
Department of Economics. University of Chicago,
processed.
Davenant, C. 1699. Essay upon the probable methods of making
a people gainers in the balance of trade. London.
Davidson. J.E.H.. et al. 1978. Econometric modelling of the
aggregate time-series relationship between consumers
expenditure and income in the United Kingdom.
E c o n o m i c J o u r n a l 88: 661 692.
Davies, J.B. 1980. Uncertain lifetime, consumption and
dissaving in retirement. J o u r n a l o f P o l i t i c a l
E c o n o m y 89:561-577.
Deaton, A.S. 1974a. The analysis of consumer demand in the
United Kingdom. 1900 1970. E c o n o m e t r i c a 42: 341
367.
Deaton, A.S. 1974b. A reconsideration of the empirical
implications of additive preferences. E c o n o m i c
J o u r n a l 84: 338 348.
Deaton, A.S. 1975. The structure of demand in Europe 1920
1970. In T h e F o n t a n a e c o n o m i c h i s t o r y o f
E u r o p e , ed. C.M. Cippola, vol. 5. London: Collins-
Fontana.
Deaton, A.S. 1977. Involuntary saving through unanticipated
inflation. A m e r i c a n E c o n o m i c R e v i e w 6 1 :
81)1) 910.
Deaton, A.S. 1986a. Demand analysis. In H a n d b o o k o f
e c o n o m e t r i c s , ed. Z. Griliches and M. Intriligator,
vol. 3. Amsterdam: North-Holland.
Deaton, A.S. 1986b. Life-cycle models of consumption: Is the
evidence consistent with the theory? N B I I R Working
Paper No. 1910, processed.
Deaton, A.S., and J. Muellbauer. 1980a. An almost ideal
demand system. A m e r i c a n E c o n o m i c R e v i e w
1 0 : 312-326.
Deaton, A.S., and J. Muellbauer. 1980b. E c o n o m i c s a n d
c o n s u m e r b e h a v i o u r . New York: Cambridge
University Press.
Deaton, A.S., and J. Muellbauer. 1986. On measuring child
costs, with applications to poor countries. J o u r n a l o f
P o l i t i c a l E c o n o m y 94: 720-744.
Deaton, A.S., J. Ruiz-Castillo, and D. Thomas. 1985. The
influence of household composition on household
expenditure patterns: Theory and Spanish evidence.
Woodrow Wilson School, Princeton University,
processed.
Dickey, D.A., and W.A. Fuller. 1981. Likelihood ratio statistics
for autoregressive time series with a unit root.
E c o n o m e t r i c a 49: 1057-1072.
Consumer Expenditure
Diewert, W.E. 1971. An application of the Shephard duality
theorem: A generalized Leontief production function.
Journal of Political Economy 79: 481-507.
Diewert, W.E. 1973. Functional forms for profit and trans-
formation functions. Journal of Economic Theory 6: 284-316.
Diewert, W.E. 1974. Intertemporal consumer theory and the
demand for durables. Econometrica 42: 497 516.
Diewert, W.E., and T.J. Wales. 1987. Flexible functional
forms and global curvature conditions. Econometrica 55: 43-
68.
Dixit, A.K. 1976. Optimization in economic theory. Oxford: Oxford
University Press.
Dolde, W., and J. Tobin. 1971. Monetary and fiscal effects on
consumption in consumer spending and monetary policy:
The linkages. Boston: Federal Reserve Bank of Boston,
Conference Series no. 5.
Duncan, G.J., and D.H. Hill. 1985. An investigation of the
extent and consequences of measurement error in labor
economic survey data. Journal of Labor Economics: 508-532.
Durlauf, S.N., and P.C.B. Phillips. 1986. Trends versus
random walks in time-series analysis. Cowles Foundation
Discussion Paper No.788. Yale University, New Haven,
processed.
Eichenbaum, M.S., L.P Hansen, and K. Singleton 1984. A
time series analysis of representative agent models of
consumption and leisure choice under uncertainty.
Graduate School of Industrial Administration, Camegie-
Mellon University, Pittsburgh, mimeo.
Engel, E. 1895. Die lebenkosten Belgischer Arbeiter- Familien
frixher und jetzt International Statistical Institute Bulletin 9: 1-74.
Epstein, L. 1975. A disaggregate analysis of consumer choice
under uncertainty. Econometrica 43: 877-892.
Evans, M.K. 1967. The importance of wealth in the con-
sumption function. Journal of Political Economy 75: 335-351.
Evans, G.B.A., and N.E. Savin. 1982. Conflict among the
criteria revisited; the W, LR, and LM tests. Econometrica 50:
737-748.
Fisher, I. 1930. The theory of interest. New York: The Macmillan
Company.
Flavin, M. 1981. The adjustment of consumption to changing
expectations about future income. Journal of Political Economy
89: 974—1009.
Flemming, J.S. 1973. The consumption function when capital
markets are imperfect: The permanent income hypothesis
reconsidered. Oxford Economic Papers 25: 160-172.
Friedman, M. 1957. A theory of the consumption function. Princeton:
Princeton University Press.
Gallant, A.R. 1982. Unbiased determination of production
technologies. Journal of Econometrics 20: 285-323.
Geary, R.C. 1949-50. A note on ‘a constant utility index of the
cost of living’. Review of Economic Studies 18,65-66.
2105
Ghez, G., and G.S. Becker. 1975. The allocation of time and goods over
the life-cycle. New York: Columbia University Press.
Goldberger, A.S. 1967. Functional form and utility: A review
of consumer demand theory. Social Systems Research
Institute, University of Wisconsin, processed.
Gorman, W.M. 1959. Separable utility and aggregation.
Econometrica 27: 469-481.
Gorman, W.M. 1976. Tricks with utility functions. In Essays in
economic analysis, ed. M. Artis and A.- R. Nobay. Cambridge:
Cambridge University Press.
Gorman, W.M. 1981. Some Engel curves. In Essays in the theory and
measurement of consumer behaviour in honour of Sir Richard Stone, ed. A.S.
Deaton. Cambridge: Cambridge University Press.
Gorman, W.M. 1982. Facing an uncertain future. IMSS
Technical Report No.359, Stanford University, processed.
Grossman, S.J., and R.J. Shiller. 1981. The determinants of
the variability of stock market prices. American Economic
Review, Papers and Proceedings 71: 222-227.
Hall, R.E. 1978. Stochastic implications of the life cycle-
permanent income hypothesis: Theory and evidence.
Journal of Political Economy 86: 971-987.
Hall, R.E., and F.S. Mishkin. 1982. The sensitivity of
consumption to transitory income: Estimates from panel
data on households. Econometrica 50: 461-481.
Hansen, L.P, andKJ. Singleton. 1982. Generalized instru-
mental variables estimation of non-linear rational
expectations models. Econometrica 50: 1269-1286.
Hayashi, F. 1982. The permanent income hypothesis: Esti-
mation and testing by instrumental variables. Journal of
Political Economy 90: 895-916.
Hayashi, F. 1985a. Permanent income hypothesis and
consumption durability: Analysis based on Japanese
panel data. Quarterly Journal of Economics: 183-206c.
Hayashi, F. 1985b. Tests for liquidity constraints: A critical
survey. Osaka University and NBER, processed. Pre-
sented at the Fifth World Congress of the Econometric
Society, Cambridge, MA, August 1985.
Heckman, J.J. 1971. Three essays on the supply of labor and
the demand for goods. Unpublished PhD thesis, Princeton
University.
Hicks, J.R. 1939. Value and capital. Oxford: Oxford University
Press.
Houthakker, H.S. 1961. An international comparison of
personal saving. Bulletin of the International Statistical Institute 38: 55-
70.
Houthakker, H.S. 1965. On some determinants of saving in
developed and underdeveloped countries. In Problems in
Economic Development, ed. A.G. Robinson. London: Macmillan.
Houthakker, H.S., and L.D. Taylor. 1970. Consumer demand in the
United States: Analysis and projections. 2nd ed. Cambridge, MA:
Harvard University Press.
2106
Howe, H., R.A. Poliak, andT.J. Wales. 1979. Theory and time
series estimation of the quadratic expenditure system.
Econometrica 47: 1231-1247.
Koskela, E., and M. Viren. 1982a. Saving and inflation: Some
international evidence. Economics Letters 9: 337 344.
Koskela, E., and M. Viren. 1982b. Inflation and savings:
Testing Deaton’s hypothesis. Applied Economics 14: 579 590.
Kotlikoff, L.J., and L.H. Summers. 1981. The role of
intergenerational transfers in aggregate capital accumu-
lation. Journal of Political Economy 89: 706-732.
Kuznets, S. 1946. National income: A summary of findings. National
Bureau of Economic Research. New York: Arno Press.
Kuznets, S. 1962. Quantitative aspects of the economic growth
of nations: VU: The share and structure of consumption.
Economic Development and Cultural Change 10: 1 92.
Kydland, F.E., and E.C. Prescott. 1982. Time to build and
aggregate fluctuations. Econometrica 50: 1345 1370.
Laitinen, Is.. 1978. Why is demand homogeneity so often
rejected? Economics Letters 1: 187-191.
Leff, N. 1969. Dependency rates and saving rates. Economic
Journal 59: 886-896.
Leser, C.E.V. 1963. Forms of Engel functions. Econometrica 31:
694-703.
Lucas, R.E. 1976. Econometric policy evaluation: A critique.
In The Phillips curve and labor markets, Camegie-Rochester
Conference Series on Public Policy 1, ed. K. Brunner and
A. Meltzer. Amsterdam: North-1 lolland.
Lucas, R.E. 1981. Introduction. In Studies in business cycle theory, ed.
R.E. Lucas. Cambridge, MA: MIT Press.
Lucas, R.E., and L. Rapping. 1969. Real wages, employment,
and inflation. Journal of Political Economy 77: 721 754.
MaCurdy, T.E. 1981. An empirical model oflabor supply in
a life-cycle setting. Journal of Political Economy 89: 1059 1085.
Mankiw, N.G., and M. Shapiro. 1985. Trends, random walks,
and tests of the permanent income hypothesis. Journal of Prais, S.J., and H.S. Houthakker. 1955. The analysis of family budgets.
Monetary Economics 16: 165-174.
Mankiw, N.G., and M. Shapiro. 1986. Do we reject too Ramsey, F.P. 1928. A mathematical theory of saving. Eco-
often? Small sample properties of tests of rational
expectations models. Economics Letters 20: 139 145.
Mankiw, N.G., J.J. Rotenrberg, and L.H. Summers. 1985.
Intertemporal substitution in macroeconomics. Quarterly
Journal of Economics 100: 225-251.
Meisner, J.F. 1979. The sad fate of the asymptotic Slutsky
symmetry test. Economics Letters 2: 231-233.
Mirer, T.W. 1979. The wealth-age relationship among the
aged. American Economic Review 69: 435-443.
Modigliani, F. 1970. The life-cycle hypothesis of saving and
inter-country differences in the saving ratio. In Induction, Samuelson, PA. 1947 8. Some implications of linearity.
growth and trade: Essays in honour of Sir Roy Harrod, ed. W.A. Eltis et
al. Oxford: Clarendon Press.
Consumer Expenditure
Modigliani, F. 1986. Life cycle, individual thrift, and the
wealth of nations. American Economic Review 76: 297 313.
Modigliani, F., and A. Ando. 1957. Tests of the life-cycle
hypothesis of savings. Bulletin of the Oxford Institute of Economics and
Statistics 19: 99-124.
Modigliani. F.. and R. Brumberg. 1954. Utility analysis and
the consumption function: An interpretation of cross-
section data. In Post-Keynesian economics. ed. K.K. Kurihara.
New Brunswick: Rutgers University Press.
Modigliani. F.. and Brumberg. 1979. Utility analysis and
aggregate consumption functions: An attempt at inte-
gration. In The collected papers of Franco Modigliani, ed. A. Abel.
vol. 2. Cambridge. MA: MIT Press.
Muellbauer. J. 1977. Testing the Barten model of household
composition effects and the cost of children. Economic Journal
87: 460M87.
Muellbauer. J. 1980. The estimation of the Prais-Houthakker
model of equivalence scales. Econometrica 48: 153-176.
Muellbauer. J. 1985. Habits, rationality and the life-cycle consumption function.
Nuffield College, Oxford, mimeo.
Nicholson, J.F. 1976. Appraisal of different methods of
estimating equivalence scales and their results. Review of
Income and Wealth 22: 1 11.
Phillips, P.C.B., and P. Perron. 1986. Testing for a unit root in
a time series regression. Cowles Foundation Discussion
Paper No.795. Yale University, New Haven, processed.
Poliak. R.A.. and T.J. Wales. 1979. Welfare comparisons and
equivalent scales. American Economic Review 69: 216-221.
Poliak. R.A.. and T.J. Wales. 1980. Comparisons of the
quadratic expenditure system and translog demand sys-
tem with alternative specifications of demographic effects
.Econometrica 48: 595-612.
Poliak, R.A., andT.J. Wales. 1981. Demographic variables in
demand analysis. Econometrica 49: 1533-1551.
Cambridge: Cambridge University Press.
nomicJournal 38: 543 559.
Rothbarth, E. 1943. Note on a method of determining
equivalent income for families of different composition.
Appendix 4 in War-time patterns of saving and spending, ed. C.
Madge, Occasional Paper 4, National Institute of
Economic and Social Research, London.
Roy, R. 1943. De I’utilitee: contribution a la theorie des choix. Paris:
Herman.
Runkle, D.E. 1983. Liquidity constraints and the permanent income hypothesis:
Evidence from paneI data. Mi l", processed.
Review of Economic Studies 15: 88 90.
Consumer Expenditure (New Developments and the State of Research) 2107

Shephard, R. 1953. Cost and production functions. Princeton:

Princeton University Press.
Shorrocks, A.F. 1975. The age-wealth relationship: A cross-
section and cohort analysis. Review of Economics and
Statistics 57: 155-163.
Slutsky, E. 1915. Sulla teoria del bilancio del consumatore.
Giornale degli Economisti 15: 1-26. English translation in
Readings in price theory, ed. G.J. Stigler and K.
Boulding. Chicago: Chicago University Press, 1952.
Smith, J.R 1977. Family labor supply over the life cycle.
Explorations in Economic Research 4: 205-276.
Spinnewyn, F. 1979a. Rational habit formation. European
Economic-Review 15: 91-109.
Spinnewyn, F. 1979b. The cost of consumption and wealth in
a model with habit formation. Economics Letters 2: 145-
148.
Stigler, G.J. 1954. The early history of empirical studies of
consumer behavior. Journal of Political Economy 62: 95-
113.
S t o n e , J . R . N . 1 9 5 4 a . The measurement of consumers’
expenditure and behaviour in the United Kingdom, 1920-
1938, v o l . 1 . C a m b r i d g e : C a m b r i d g e U n i v e r s i t y
Press.
Stone, J.R.N. 1954b. Linear expenditure systems and demand
analysis: An application to the pattern of British demand.
Economic Journal 64: 511-527.
Stone, J.R.N. 1964. Private saving in Britain, past, present,
and future. The Manchester School 32: 79-112.
Stone, J.R.N. 1966. Spending and saving in relation to income
and wealth. L T n d u s t r i a : 471-499.
Surrey, M.J.C. 1974. Saving, growth, and the consumption
function. B u l l e t i n o f t h e O x f o r d I n s t i t u t e
o f S t a t i s t i c s 36: 125-142.
Thomas, D. 1986. Essays on the analysis of Engel curves in
developing countries. PhD thesis, Princeton University.
Uzawa, H. 1964. Duality principles in the theory of cost and
production. I n t e r n a t i o n a l E c o n o m i c R e v i e w
5: 216-220.
Varian, H.R. 1978. M i c r o e c o n o m i c a n a l y s i s . 2nd ed,
1984. New York: Norton.
Varian, H.R. 1982. The non-parametric approach to demand
analysis. E c o n o m e t r i c a 50: 945-973.
Wales, T.J. 1977. On the flexibility of flexible functional
forms: An empirical approach. J o u r n a l
E c o n o m e t r i c s 5: 183-193.
Watson, M.W. 1986. Univariate detrending method with
stochastic trends. Journal of Monetary
E c o n o m i c s 18: 49-75.
Working, H. 1943. Statistical laws of family expenditure.
Journal of the American Statistical
A s s o c i a t i o n 38: 43-56.
Yaari, M.E. 1965. Uncertain lifetime, life insurance, and the
theory of the consumer. R e v i e w o f E c o n o m i c
S t u d i e s 32: 137-150.
Zeldes, S. 1985. Consumption and liquidity constraints: An
empirical investigation. The Wharton School, University
of Pennsylvania, processed.
Consumer Expenditure (New
Developments and the State of
Research)
Orazio P. Attanasio and Guglielmo Weber
Abstract
We provide an overview of recent developments
in the life-cycle permanent income model under
uncertainty, starting from the certainty
equivalence case, and considering precautionary
saving, the workings of insurance and credit
markets, and non-standard preference structures.
Keywords
Aggregate consumption; Aggregate saving;
Asset pricing; Asymmetric information; Buffer
stocks; Certainty equivalence; Commitment;
Consumer expenditure; Consumer expenditure
(New developments and the state of research);
Consumption function; Durable goods; Elas-
ticity of intertemporal substitution; Equity pre-
mium puzzle; Euler equation; Excess
sensitivity; Excess smoothness of consumption;
Expected utility; Financial asset accumulation;
Financial market imperfections; Habits;
Impatience; Interest rate; Intertemporal optimi-
zation; Intertemporal prices; Labour supply;
Lagrange multipliers; Life cycle hypothesis;
Liquidity constraints; Marginal utility of con-
of
sumption; Marginal utility of wealth; Myopia;
Non-convexity; Perfect insurance; Permanent
income life-cycle model; Power utility; Pre-
cautionary savings; Preferences; Private infor-
mation; Prudence; Public finance; Quadratic
utility; Quasi-hyperbolic discounting; Real
business cycles; Reduced form equations;
Retirement; Risk; Risk aversion; Saving; Social
insurance; Social Security in the United States;
Stone-Geary utility; Subjective discount rate;
Time preference; Transaction costs; Uncertainty
2108 Consumer Expenditure (New Developments and the State of Research)
JEL Classifications
E22
The state of research on consumer expenditure up
to the mid-1980s is described in consumer expen-
diture. Here, we provide an overview of recent
developments on the intertemporal model of con-
sumer behaviour under uncertainty. We organize
our discussion around what has been the workhorse
model for the analysis of dynamic consumption
behaviour - the life-cycle permanent income model.
Although our discussion of the intertemporal model
is self-contained, it is not meant to be an exhaustive
survey of this large literature. We do not cover
demand analysis, despite the many exciting
developments that have occurred in recent years.
The permanent income life-cycle (PILC) model,
introduced during the 1950s by Modigliani and
Brumberg ( ) and Friedman ( ), still plays
an important role in the consumption literature. The
PILC model can be loosely defined as a framework
where individuals maximize utility over time given
a set of intertemporal trading opportunities. Con-
sumption at different points in time is treated as
different commodities, so that, given intertemporal
trading opportunities, consumption in a given
period depends on total (life-cycle) resources and
(intertemporal) prices. Optimal consumption
choices are such that the ratio of (expected) mar-
ginal utilities of consumption at different times
equals the ratio of intertemporal prices. Therefore,
the relationship between consumption and total
resources is likely to depend on preferences (and in
particular on the elasticity of intertemporal sub-
stitution and the rate at which the future is
discounted) and on interest rates (as they represent
intertemporal prices). If we allow for uncertainty, as
we discuss below, risk will also enter as a poten-
tially important determinant of consumption.
This model can generate implications and
insights for many important questions not only in
macroeconomics but also in public finance, and has
therefore attracted much attention, both
theoretically and empirically. Recent research has
stressed the need to look at preferences on the one
hand and markets on the other, as the policy
implications are the result of both.
The Permanent Income Life-Cycle
Model
In its simplest incarnation the PILC model considers
a finite horizon, no uncertainty and very simple
preferences. In such a situation, it is simple to
translate the basic intuition of the model, to which
we referred above, into a closed form solution for
consumption that depends not just on current
income but on the total amount of resources
available to an individual and intertemporal prices.
The problem of this specification, of course, is its
lack of realism. Not only do consumers in reality
face much more complicated intertemporal envi-
ronments, but it is likely that these complications
have a first-order effect on consumption choices.
Therefore, the simplest version of the model is a
useful way to convey the main ideas behind PILC,
but it needs to be complicated considerably to be of
use for policy analysis.
The introduction of uncertainty in the model,
which makes it much more realistic, complicates the
problem enormously. The first formalizations of the
life-cycle model under uncertainty date back to the
1970s (Bewley ). Typically, one assumes that
consumers maximize expected lifecycle utility
choosing consumption and, in more general
settings, leisure and financial asset holdings.
Consumers are assumed to know the stochastic
nature of their environment. Even with many
simplifications on the nature of preferences, the
model does not yield closed form solutions for
consumption, except in the most special cases.
MaCurdy ( , see also ) uses dynamic
optimization techniques to derive necessary con-
ditions for the optimal solution of the intertemporal
optimization problem faced by consumers. The
attractiveness of this approach lies in the fact that it
cuts through the necessity of solving the model
completely, which is a very hard task indeed, to
focus on some useful implications of the model. In
particular, these contributions focus on the basic
first order condition, the so-called Euler equation,
that equates the ratio of marginal utilities to
intertemporal prices.
The first macro paper to take this approach is
Hall ( ): under strong assumptions on prefer
ences and returns, (non-durable) consumption is a
random walk, that is:
Consumer Expenditure (New Developments and the State of Research) 2109

E{Ct+\\It) = C, completely the stochastic environment in which the

consumer lives and use particularly simple
(1 ) preferences. In particular, Flavin ( ) assumes
where It denotes information available at time /. that the only stochastic variable is labour income,
This remarkable proposition requires that utility be that preferences are quadratic and that the consumer
quadratic in consumption (and additively separable can save or borrow in a single asset with a fixed
over time, states of nature and in its other rate of interest. Under these conditions, Flavin
arguments, notably male and female leisure and shows that consumption is set equal to permanent
durable goods). It also requires that there is at least income, and this is in turn defined as the present
one financial asset with fixed real return, and that value of current and expected future incomes:
this equals the time-preference parameter. If con-
sumers have rational expectations, then:
Ct+1 — Ci + fij+i-E (fi/+i | Y oo

Z,) — 0 c* = T—^+T— ££M7') (3)

r r
” ” i=0
(2)
where A denotes financial wealth and y is labour
for all variables Z known at time t. A notable income. In this model, the first difference in con-
feature of Hall’s model is that the Euler equation sumption equals the present value of income revi-
for consumption aggregates perfectly, because it sions, due to the accrual of new information
involves linear transformations of the data. Hall between periods t and (/ + 1):
used the Euler equation to test for the prediction
implied by ( ): no variable known to the consumer
at time t should help predict the change in 7S
l+r
k=o (1 +r)
consumption between t and (t + 1). AC,
k=0
Hall’s paper was the first of many contributions (4)
(yy+k\it-i)
that exploited the Euler equation and the fact that
such an approach does not require the complete
specification of the environment in which the
Equation ( ) makes clear the main implications
consumer lives, or even the complete budget con-
of the model: consumption depends on the present
straint. Moreover, the approach is robust to the
discounted value of future expected income. The
presence of various imperfections in some
interest rate plays the important role of converting
intertemporal markets. And while the specification
future resources to present ones and therefore
with quadratic utility yields a linear equation for
constitutes an important determinant of consump-
consumption, alternative specifications, with more
tion. Flavin ( ) noticed that Eq. ( ) imposes
plausible preferences, are easily introduced. For
cross-equation restrictions on the joint time series
instance, in the case of power utility, an expression
process for income and consumption. A similar
similar to ( ) can be obtained for the log of
approach had been followed by Sargent ( )
consumption.
and, subsequently, by Campbell ( ) who
The price that one pays in using the Euler
noticed that an implication of ( ) is that saving
equations approach, which we discuss below, is that
predicts future changes in income, the so-called
one does not obtain a closed form solution for
‘saving for a rainy day’ motive.
consumption.
One of the main implications of the PILC
An approach that goes beyond the consideration
model, particularly evident in Eq. ( ), is that, in
of the Euler equations is taken up in an important
appraising the effects of a given policy, for instance
paper by Flavin ( ).
a tax reform that affects disposable income, a
Flavin ( ) adopts the same theoretical
distinction must be drawn between permanent and
framework as Hall ( ), and assumes that no
temporary changes (Blinder and Deaton ; Poterba ).
other asset is available to the consumer (as in
Bewley ). However, Flavin develops a solution for
consumption. To do so, she has to specify
= It, (5)

1+8' (6) Equation ( ) says that, at each point in time, the

marginal utility of consumption equals the Lagrange
multiplier associated with the budget constraint
relevant for that period, which is sometimes referred
to as the marginal utility of wealth. The second
condition, Eq. ( ), that is derived from intertemporal
optimality, dictates the evolution of the marginal
utility of wealth (<5 is a subjective discount rate).
An equation of this type has to hold for each asset k
for which the consumer is not at a comer. This is
because the consumer is exploiting that particular
intertemporal margin.
The attractiveness of Euler equations is that one
can be completely agnostic about the stochastic
environment faced by the consumer, about the time
horizon, about the presence of imperfections in
financial markets (as long as there is at least one
asset that the consumer can freely trade), about the
presence of transaction costs in some component of
consumption or labour supply. All relevant
information is summarized in the level of the
marginal utility of wealth. The approach is con-
ceptually similar to the use of an (unobservable)
fixed effect in econometrics. By taking first dif-
ferences, one eliminates the unobservable marginal
utility of wealth and is left only with the innovations
to Eq. ( ).
Early papers along these lines were Hansen and
Singleton ( , ), who used power util
ity (also known as isoelastic, isocurvature or CR
RA) as it has more appealing theoretical properties
(relative risk aversion is constant in wealth or
consumption, the elasticity of intertemporal
substitution is also a constant). If we substitute Eq. (
) into ( ) and using the properties of the CR RA
utility function, the Euler equations for consumption
conesponding to each asset (Ic) will be:

where y is a curvature parameter (equal to the

relative risk aversion parameter and to the recip-
rocal of the elasticity of intertemporal substitution)
and <5, the subjective discount rate, measures
impatience.
Consumer Expenditure (New Developments and the State of Research) 2111

An equation such as ( ) can be log-linearized to The great merit of the model with prudence is
obtain (see Hansen and Singleton ): that it highlights the need to save for rainy days
even if sunny days are equally likely. An increased
AlogC,+i = k + ^log(l + rf,,) + e(+1. (8) variance in the shocks to income reduces
consumption even if expected income does not
change. In the case of discrete variables, such as
Although consumption appears on the left- hand
unemployment or illness, changes in first and
side of Eq. ( ), this equation is not a consumption
second moments occur simultaneously, but this is
function, but an equilibrium condition. It cannot
not the case for continuous variables. The ability to
explain or predict consumption levels: consumption
distinguish between first and second moment
is crucially determined by the residual term £,+1 and
effects is of crucial importance in the analysis of
there is nothing that tells us what determines such a
public policy, because of its social insurance
term or how this term changes with news about
characteristics.
income, interest rates or any other relevant variable.
The solution of the Bewley model with more
The Euler equation for a single asset can iden-
general utility functions has to be computed
tify the elasticity of intertemporal substitution, a key
numerically or rely on approximations. Several
parameter for the evaluation of the welfare costs of
studies in the early 1990s took up the challenge of
interest taxation (Boskin ; Summers ) and for the
characterizing such solutions. Deaton ( )
analysis of real business cycles (King and Plosser ).
studied a model with power utility and infinite life.
The joint estimation of several Euler equations can
Deaton considered the existence of liquidity or no-
help identify the pure discount rate parameter
borrowing constraints, and showed that impatient
(governing patience), but also shed light on risk
consumers would hold limited assets to insure
aversion, given that different assets typically have
against low income draws. Carroll ( ) instead
different risk characteristics.
covered the case of finite lives,
The derivation of a closed form solution for
and showed that, if consumers are sufficiently
consumption when certainty equivalence does not
impatient and their labour income is subject to both
hold was first successfully tackled by Caballero (
permanent and temporary shocks, they set
, ). Caballero ( ) took the Flavin
consumption close to income. The model with
model with known finite life, and constant absolute
impatient consumers under labour income uncer-
risk aversion (CARA) preferences, and showed that,
tainty has been labelled ‘the buffer stock model’,
when the optimal consumption age profile is flat
because saving is kept to the lowest level compat-
with no uncertainty, it is increasing with income
ible with the need to buffer negative income shocks.
uncertainty. This change in the slope of the
Later work by Attanasio et al. ( )
consumption profile was described as precautionary
and Gourinchas and Parker ( ) clarifies the
saving, because early in life consumers save more if
role played by age-related changes in demographics
labour income is more uncertain. Later work by
and the hump-shape age profile of labour income in
Gollier ( ) and Carroll and
generating income tracking for relatively young
Kimball ( ) established that a similar result
consumers (micro data show that financial asset
holds whenever the third derivative of the utility
accumulation starts around age 40). Hubbard et al. (
function is positive, a feature of preferences labelled
, ) show instead
prudence. Both CARA and power utility exhibit
how precautionary motives interact with the insur-
prudence. The presence and size of precautionary
ance properties of Social Security in the United
savings is a matter of great relevance for public
States.
policy, in so far as public insurance schemes
Many of the papers cited in the preceding par-
covering such risks as unemployment, health and
agraph consider relatively simple versions of the
longevity should reduce the need for consumers to
life-cycle model. In particular, a single non-durable
accumulate assets.
commodify is assumed and preferences are
assumed to be additively separable
2112 Consumer Expenditure (New Developments and the State of Research)
with leisure and over time. While this greatly
simplifies the solution, the construction of a more
realistic and complex model has become an
important area of research. This development
follows from the recognition that, for many pur-
poses, and in particular for policy analysis, a model
that delivers consumption as a function of
exogenous variable is a very useful tool indeed.
This area of research has to deal with two
important issues. First of all, the model can become
very quickly, from a numerical point of view, very
difficult to solve. The large number of state
variables that characterize the solution of
reasonably realistic models and the consideration of
discrete choices and non-convexities linked to
transaction costs can push the numerical capabilities
of even very powerful computers. Second and even
more importantly, if one wants to obtain solutions
for consumption in a dynamic context, one has to
characterize completely the stochastic environment
in which the consumer lives. This contrast sharply
with the Euler equation approach that allowed the
researcher to be agnostic about most aspects of the
environment and, under certain conditions, avoid
solving difficult problems, such as labour supply,
housing and other durable choices and so on. The
Euler equation would hold regardless of the
presence of non-convexities and other type of
difficulties connected with these choices. These,
instead have to be frilly specified if one wants to
work with a model that delivers a solution for
consumption. These two difficulties constitute
limits for the research in this area that, in all
likelihood will not be overcome in the near future.
The Empirical Evidence on the PILC
Model
Since its introduction in the 1950s, there is no
consensus about the empirical relevance of the
PILC model. While the model it is one of the main
tools in modern macroeconomics and public
finance, its empirical performance is mixed. In this
section, we discuss two branches of the literature.
The life-cycle model with various sources of
uncertainty and generic preferences generates
decision rules and behaviour of great complexity.
Consumption and saving choices depend in an
unknown fashion on eveiy single aspect of the
stochastic environment faced by the consumer, for
instance on the entire distribution of fitUtre wages
and earnings opportunities, on pension
arrangements, on the asset markets the consumer
can access, on mortality risks and so on and so
forth. The Euler equation approach allows
researchers to deal in a rigorous fashion with
extremely rich models and yet derive relatively
simple implications to test some aspects of the
model and, with the help of additional assumptions,
to identify some of the structural parameters that
inform individual behaviour. We now understand
that Euler equations can be used to determine what
type of preferences fits the available data and can
therefore provide one of the building blocks
(preferences) in the study of the questions above.
We also know that the presence of liquidity
constraints does not necessarily produce violations
of Euler equations because, even when liquidity'
constraints are present, they might be rarely
binding.
The Euler equation is robust to a number of
market imperfections, but is silent about how con-
sumption or its growth reacts to specific news about
shocks, changes in interest rates, taxation and so on.
It is therefore useless for specific policy analysis. In
other words, while the parameters of an Euler
equation can be estimated in a wide set of
circumstances, and one can use the equation to test
the specification of the model, none of these results
will provide an answer to questions like what is the
effect of a change in taxation or interest rates on the
level of consumption and saving?
This important shortcoming of the Euler equa-
tion approach explains why such an approach,
which has informed and dominated the large
empirical literature on the validity of the life- cycle
permanent income model is virtually absent in the
public economics literature on, say, the effect of
pension reforms on saving or on the effect of
changes in the taxation of interest on saving. And
yet the conceptual framework that is
Consumer Expenditure (New Developments and the State of Research) 2113

behind the study of these issues is the same as that • What is the magnitude of the relevant preference
used to study consumption behaviour. parameters?
Policy analysis requires instead the availability
of a consumption function, that is, a relation that Tests of the Model
explains consumption as a function of those As mentioned above, a prediction of the model is
variables that the consumer can take as exogenous that changes in consumption cannot be predicted by
at any given moment. Only in the simplest versions expected changes in income or any other variable
of the life-cycle model is it possible to derive an known to the consumer at time t — 1. This is the
analytical expression for the consumption function. essence of the Hall ( ) test and of many
In general, given a set of assumptions on preferenceothers. Evidence that consumption can be predicted
parameters and market and non-market by lagged variables has been interpreted as
indicative of liquidity constraints, myopic
opportunities, one has to rely on numerical solutions
and/or approximations. behaviour, misspecification of preferences and so
A less ambitious but potentially profitable on. The relationship between consumption and
approach that does not require numerical methods income has received considerable attention. The
or incredibly rich data-sets is the estimation of first to observe that the life-cycle model predicts no
reduced form equations, whose specification is relation between the life-cycle profile of income
informed by the life-cycle model. These are par- and consumption was Thurow ( ).
ticularly useful in situations in which one analysesThurow argued that the fact that consumption
large (and possibly exogenous) changes to some of tracked income over the life-cycle was a rejection of
the likely determinants of consumption or saving. the main implications of the PILC model. To this
Such studies can address substantive issues and argument, essentially identical to many others
even test some aspects of the life-cycle model. proposed subsequently, Heckman ( ) replied
that non-separability between consumption and
Examples of studies of this kind include the reaction
of consumption (and saving) to changes in pension leisure could explain such a relationship.
entitlements (Attanasio and Brugiavini ; Attanasio Despite this early exchange, after Hall ( )
and Rohwedder ; Miniaci and Weber ), to swings in a
the value of important wealth components (such as large fraction of the literature based on consumption
housing, Attanasio and Weber ) and to changes in Euler equations focused on the relationship between
specific taxes (Parker ; Souleles ; Shapiro and predictable changes in income and expected
Slemrod ). consumption growth. Hall and Mishkin ( ),
Below we review the empirical evidence on the as well as Campbell and Mankiw ( ,
PILC model, organizing it in two subsections. First ) all report violations of this prediction, and
we start with the empirical evidence derived from label this finding ‘excess sensitivity’. Excess sen-
Euler equations. We then move on to evidence that sitivity can be explained by the presence of liquidity
constrained consumers, or of rule-of-thumb
considers the levels of consumptions, rather than its
changes. consumers, that is, consumers who let their expen-
diture track their income as a way to avoiding the
complexities of choosing the optimal consumption
Evidence from Euler Equations path. However, consistently with Heckman’s ( )
Two important empirical issues can be addressed argument, excess sensitivity can be recon
with the study of Euler equations: • ciled with the intertemporal optimization model if
more general, and sensible, utility functions are
used. In particular, if one assumes that leisiue
affects utility in a non-additive way, consumption
• What is the empirical relevance of the model? changes respond to predictable labour income
Is there a sensible specification of preferences changes, whether or not leisiue is a freely chosen
that fits the observed data? variable. Finally, and importantly, the aggregation
2114 Consumer Expenditure (New Developments and the State of Research)
issue proves to be important. Attanasio and Weber (
) show that results obtained with improperly
aggregated micro data are consistent with results
obtained with aggregate data and indicate rejections
of the model that instead disappear with properly
aggregated data and rich enough preference
structures.
To summarize the discussion so far, it seems
that while simple tests of the life-cycle model seem
to reject the implications from the model and in
particular those derived from Euler equations, it is
possible to find specification of preferences that do
a good job at fitting the available data, especially for
households that are headed by prime-aged
individuals. Aspects that are crucial for fitting the
data are the use of household level data, allowing
for changes in consumption needs induced by
changes in family composition and the use of
preferences specifications that allow for the
marginal utility of consumption to depend on labour
supply.
Estimation of Preference Parameters
Recent research on consumption and saving has
singled out three preference parameters for atten-
tion: the elasticity of intertemporal substitution, the
relative risk aversion parameter and the subjective
discount rate. The size of these parameters has
important implications in many applications of the
model, ranging from macroeconomics to public
finance to financial economics.
Perhaps surprisingly, not much evidence has
been accumulated on the discount factor from the
estimation of Euler equations. This can be explained
by the fact that in log-linearized versions of the
Euler equation, the parameter is not identified, while
non-linear versions of the model are ridden by a
number of econometric problems, particularly in
relatively small samples of the type used in Euler
equation estimation (see Attanasio and Low ).
As for the distinction between the elasticity of
intertemporal substitution (EIS) and the coefficient
of risk aversion, it is absent in the most popular
specifications used in the literature: a model where
consumers maximize expected utility and Evidence from the Levels of Consumption
preferences are iso-elastic and additively separable As stressed above, the Euler equation imposes some
over time. In such a situation, the EIS is
the reciprocal of the coefficient of relative risk
aversion. Not many empirical papers have worked
with preferences that allow for these two parameters
to be disjoint.
An influential paper by Hall ( ) claimed
that this parameter is close to zero. This finding has
been challenged on various grounds. Attanasio and
Weber ( , ) point out that
aggregation bias could be responsible for such a low
estimate: they estimated a much higher elasticity
(around 0.8) using UK and US cohort data (that is,
data from repeated cross-sections, consistently
aggregated over individuals bom in the same years).
In the macro literature little attention has been
paid to the possibility that the EIS may differ across
consumers, particularly as a function of their
consumption. A simple way to capture the notion
that poor consumers may be less able to smooth
consumption across periods and states of nature is
to assume that the utility function does not depend
on total (non-durable) consumption, but rather on
the difference between consumption and needs.
Thus we could retain the analytical attraction of
power utility, but have (C - C*) as its argument,
where C* is an absolute minimum that the
consumer must reach in each and every period. This
functional form is known as Stone-Geary utility in
demand analysis, and is the simplest way to
introduce non-homotheticity in a demand system.
One could interpret ‘external habits’ (Abel ;
Campbell and Cochrane ) as a special way to
parameterize C* (by making it a fraction of past
consumption of other consumers). Attanasio and
Browning ( ), Blundell et al. () and
Atkenson and
Ogaki ( ) are among the few examples of
papers that explicitly allow for wealth- dependent
EIS.
Demographics might also affect preferences,
and might explain consumption changes and the
shape of the consumption age profile, as argued by
Attanasio et al. ( ) as well as Browning and
Ejmaes ( ).
restrictions on the dynamics of consumption
Consumer Expenditure (New Developments and the State of Research)
but, on its own, does not determine the level of
consumption. If one neglects numerical compli-
cations, a solution for consumption can be obtained
by considering jointly the Euler equation and the
sequence of budget constraints faced by the
consumer as well as his or her initial wealth and a
terminal condition. As noted by Sargent ( ),
Flavin ( ) and later by Campbell
( ), the Euler equation and the intertemporal
budget constraint imply a number of crossequation
restrictions for the joint time series processes of
consumption and income. When one is able to
obtain a closed form solution for consumption, as is
the case with quadratic utility, these restrictions can
be easily expressed in terms of a linear time series
model, and tested.
Some of these restrictions are also implied by
the Euler equation, while others are not. In partic-
ular, the restrictions on the contemporaneous cor-
relation between income and consumption are not
implied: as we stressed above, the Euler equation is
silent about how news about income is translated
into news about consumption.
Campbell and Deaton ( ) and West (
proposed a test that links the innovation to perma-
nent income to consumption and presented evidence
that aggregate consumption seems to be
‘excessively smooth’ in that it does not react
enough to news about income. Campbell and
Deaton make a connection between excess sensi-
tivity and excess smoothness. Within the certainty
equivalent model, they jointly model the con-
sumption and income processes as a vector auto-
regression, assuming that income has a unit root
plus some persistence. In this context, consumption
changes reflect the permanent income innovation
more than one-to-one: not only is the income shock
permanent, but it also predicts future, smaller
shocks of the same sign. This implies that over the
business cycle consumption should be more volatile
than income. But in actual aggregate data
consumption is smoother than income: this is
labelled ‘excess smoothness’, and is shown to be
exactly equivalent to excess sensitivity.
Clearly the implications of a given set of
intertemporal preferences for policy relevant
questions depend crucially on the markets
2115
individuals have access to, on their imperfections
and on the nature of the equilibrium they give rise
to. The implications of complete markets would be
very different from those one would derive if
liquidity constraints or other markets imperfections
were prevalent.
Insurance and Credit Markets
So far we have taken the assets the consumer can
use to move resources over time as given and, in the
simplest versions of the model, we have made very
strong assumptions on this crucial aspect. For
instance, we have assumed that consumers can
borrow and lend at a fixed interest rate. The reality
is, obviously, much more complex and, from a
theoretical point of view, very many different
environments have been studied. In particular, the
possibilities open to a consumer depend on the
market arrangements available. Below we discuss
several of these market arrangements and briefly
mention their implications for the determination of
) consumption.
Perfect Insurance Markets
If markets are complete and consumers can trade a
full set of contingent claims without cost, individual
risk will be completely diversified. In such a
situation, a number of results deliver very useful
predictions. In particular, it can be shown that a
competitive equilibrium is symmetric and it is
therefore possible to characterize the properties of
competitive equilibria by considering the problem
of a fictitious social planner, which, given a set of
Pareto weights, maximizes social welfare. A strong
implication of perfect markets is that the mar ginal
utility of different consumers will move
proportionally over time. The implication is very
intuitive: the social planner faces a unique resource
constraint, and marginal utility of all individuals,
multiplied by the appropriate (and ar bitrary) Pareto
weight, will be equal to the multiplier associated to
this unique constraint. As a consequence, marginal
utility will move proportionally. If utility is
isoelastic, consumption moves proportionally.
These implications, stressed by Townsend ( ),
have been tested in several
2116 Consumer Expenditure (New Developments and the State of Research)

papers (Cochrane ; Attanasio and Davis ; Hayashi mentioned that, in a model with finite lives and a
et al. ). non-zero probability that income would be zero in
each time period, standard regularity conditions on
Many Assets the utility function imply that a consumer will never
When there are many assets, one can derive an want to borrow. If income is bounded away from
Euler equation such as ( ) for each of the assets for zero, then the maximum the consumer will want to
which the consumer is not at a comer. The Euler borrow is the present discounted value of the
equations for consumption with different assets minimum value of income repeated in the future.
naturally ties up with asset pricing equations. This This type of constraint has been sometimes referred
approach to asset pricing was developed by to as a ‘natural’ liquidity constraint. Notice that
Breeden ( ) and Lucas ( ), and extended such a constraint does not imply a violation of the
to the case of non-additive separability of con- Euler equation. If the restriction to borrowing is
sumption and leisure in an incomplete markets tighter, the Euler equation will instead be
setting by Bodie et al. ( ). The model we occasionally violated. And, even in periods in which
sketched above is quite restrictive: the relative riskit is not violated, the level of consumption will be
aversion parameter is inversely related to the affected by the possibility that the constraint will be
elasticity of intertemporal substitution: Epstein and binding in the future. As Hayashi ( ) explains, the
Zin ( ) show how this restriction can be presence of an
relaxed in a more general model with power utility operative, albeit not binding, liquidity constraint is
where the timing of uncertainty resolution matters equivalent to a shortening of the planning horizon
(see also Epstein and Zin ; Attanasio and Weber ). or an increase in the discount rate. Evidence can be
Interestingly, an Euler equation for an asset obtained by noting that consumers who are liquidity
holds even if there are important imperfections in constrained will not be sensitive to changes in the
some other assets. As long as the consumer is level of the interest rate. As they will be at a kink of
exploiting a given margin to move resources over an intertemporal budget constraint, the demand for
time, an equation such as 7 will apply. If the loans will be inelastic to changes in the slope of
interest rate for a given asset changes with the levelsuch an intertemporal budget constraint: the interest
of the asset, then the Euler Eq. ( ) will have to be rate.
augmented with a term reflecting this effect
(Pissarides ). Endogenous Liquidity Constraint
In recent years, several studies have tried to model
Liquidity Constraints the shortcomings of credit and insurance markets by
The Euler equation will be violated when the allowing for specific imperfections and frictions
consumer is able, for some reason, to borrow explicitly. The two main causes of imperfections
against future income. In such a situation, Eq. ( ) that have been considered are: (a) private and
will hold as an inequality and the marginal utility of asymmetric information and (b) the inability to
current consumption will be higher than the present perfectly enforce contracts. Models of this type can
discounted value of future consumption. Consumers be seen as ways to endogenize specific market
who are liquidity constrained will be very sensitive structures (such as one where consumers have
to changes in current income. This case has access to a single asset in which they cannot
received a considerable amount of attention in the borrow). In an influential paper, for instance, Cole
literature. Many of the tests of violation of the Euler and Kocherlakota ( ) show that an econ
equation, such as Zeldes ( ), have focused on the omy where individuals have a single bond in which
so-called they can borrow can be derived as a constrained
‘excess sensitivity’ of consumption changes to equilibrium outcome where individuals have private
predictable changes in income. It should be information both on their income and on their
savings.
Consumer Expenditure (New Developments and the State of Research) 2117

Further Extensions and Alternative goods). For this reason micro data seem preferable.
Models The simplest way to introduce habits (or
durability) of consumption is to write the utility
While the evidence on the relevance of the life- function as follows:
cycle model is still inconclusive, a number of
empirical puzzles have directed attention to more - y'x,-i\z,)
complex preference structures. In particular, the
equity premium puzzle and the evolution of (9)
aggregate saving rates in high-growth economies where x is a vector of goods or services and z is any
(South East Asia) has led macroeconomists to other variable that affects marginal utility
incorporate habits into the model. However, there is (demographics, leisure, other goods that are not
still little formal evidence on the empirical explicitly modelled). The y parameters are positive
relevance of habits in micro data. The widely for goods that provide services across periods
documented retirement consumption puzzle (that is, (durability), negative for goods that are addictive
a sudden drop of consumption at retirement ) as (habit formation) or zero for goods that are fully
well as a number of more or less anecdotal pieces of non-durable, non-habit forming (Hayashi ).
evidence on the inadequacy of saving for retirement The Euler equations corresponding to ( )
and other forms of ‘irrational’ behaviour, have been involves x at four different periods of time, and
interpreted as potentially supportive of time- their estimation typically requires panel data. High-
inconsistent preferences. The most elegant way to quality consumption panel data are rare, and this has
introduce time-inconsistent preferences is provided limited the scope for empirical analysis. Meghir and
by the hyperbolic discounting assumption Weber ( ) have used Consumer
(Laibson ). Expenditure Survey (CEX ) quarterly data on food,
transport and services (and a more flexible
Habits
specification of intertemporal non-separabilities
Habits cause consumers to adjust slowly to shocks
than is implied by Eq. ), and found no evidence of
to permanent income, thus potentially explaining
either durability or habits once leisure, stock of
the excess smoothness of aggregate consumption,
durables and cars as well as other conditioning
but also increase the utility loss associated with
variables are taken into consideration.
consumption drops, and may therefore help explain
Similarly negative evidence on habits has been
the equity premium puzzle.
reported by Dynan ( ), using Panel Study of
Habits can take various forms: today’s marginal
Income Dynamics (PSID) annual food at home data.
utility may depend on the consumer’s own past
Carrasco et al. ( ) use Spanish panel data
consumption level (internal habits ) or the past
and find some evidence for habits.
consumption level of other consumers (external
The few studies that have used micro data on
habits). This latter model seems to work better on
non-durable consumption items to investigate the
aggregate data (Campbell and Cochrane ), even
issue find little or no evidence of habits, at least
though a recent survey by Chen and Ludvigson (
once preferences capture the presence of non-
) challenges this conclusion.
separabilities between goods and leisure.
Empirical macro-evidence on the presence of
habits is mixed, and this may be due to the very
Durable Goods
nature of aggregate consumption data, as stressed in
The presence of durable goods has received less
Dynan ( ). The serial correlation of aggre
attention in the micro-based literature than in the
gate consumption growth is affected by time
macro-literature, which has stressed the importance
aggregation (Heaton ), by aggregation over
consumers, and by data construction methods of their high volatility to explain business cycle
(particularly for the services from durable fluctuations (Mankiw ; Chah
et al. ).
2118 Consumer Expenditure (New Developments and the State of Research)

The simplest way to introduce durable goods role, if they can be used as collateral to obtain a
loan that pays for current consumption (Alessie et
into the analysis is to let the stock of durables affect
utility (on the assumption that services are al. ). A typical example could be the ability to
proportional to the stock), and to posit a relation remortgage a house, or to borrow 100 per cent of
between current stock, St, previous stock, St i, and the value of a newly purchased car.
current purchases q, (or maintenance and repairs) in Even if one is not interested in modelling dura-
physical terms like: ble goods, the existence of a stock of durables
should not be neglected when estimating preference
S; = (l-p)S,_,+<7; (10) parameters if utility is not additive in non-durable
goods and durable services. Significant effects of
where p is a constant depreciation rate. This leads durable goods (cars) on the Euler equation for non-
to the standard first-order condition for the durable durables have been found in UK data (Alessie et
good, according to which the relevant price is the al. ), and US data (Padula ).
user cost.
Typically, durable goods are costly to adjust, Quasi-Hyperbolic Discounting
because of transaction costs (resale markets are The widely documented consumption puzzle (that is
dominated by information problems, known as the the sudden drop of consumption at retirement, see
‘lemon’ problem, and search costs are non- Hamermesh ; Banks et al. ; and Bemheim et al. ), as
negligible). Sometimes these costs are modelled as well as a number of more or less anecdotal pieces of
a convex, differentiable function (Bemanke ), but evidence on the inadequacy of saving for retirement
the recent literature has stressed the need to take and other forms of ‘irrational’ behaviour, have been
into account their non-differentiable nature interpreted as potentially supportive of time-
(Grossman and Laroque ; Eberly ; Attanasio ; inconsistent preferences. The most elegant way to
Bertola et al. ). This generates infrequent introduce time-inconsistent preferences is provided
adjustment: consumers do not adjust continuously by the quasi-hyperbolic discounting assumption
in response to depreciation, or income and price (Laibson ). Consumers maximize the expected
shocks, but wait until the actual stock hits value of the following life-time utility index:
either a lower limit, s, or an upper limit, S,
and then adjust it to a target level. An T-t
interesting feature of this literature is that u(ct)+^Mcl+d (11)
aggregate behaviour reflects changes in both T= 1

the number of consumers that adjust and in

the target level. This implies that a different, lower discount
Durable goods might also play an insurance factor is used to choose between this period and the
role, because they can be used to sustain con- next (the product of [i and 8) and between any two
sumption when times are bad. Postponing the other periods (5). This generates time- inconsistent
purchase of food, or clothing, is certainly harder plans, with too little saving for retirement. For this
than failing to replace an old refrigerator or car, reason, consumers may choose to enter long-term
and housing maintenance can be put off for very commitment plans, such as 401(k) s in the United
long periods before structural damage occurs States.
(Browning and Crossley ). Durable goods also The quasi-hyperbolic discounting model lends
play a more specific insurance role, against itself to estimation and testing, but requires solving
changes in the price of the corresponding for the consumption function numerically. Even
services. This is particularly relevant in the case though an Euler equation for this model has been
of housing, where owning your home may be the derived, its empirical use is limited, because it
best way to hedge the risk of future increases in involves the marginal propensity to consume out of
the market price of housing services (Sinai and wealth (Harris and Laibson
Souleles ). Durable goods can also play a
liquidity
Consumer Expenditure (New Developments and the State of Research)
). It also suffers from some potential difficul-
ties related to the definition of the time period,
which crucially affects the properties of the solu-
tion, the length of which is arbitrarily set by the
researcher.
A more tractable specification of preferences
that may be used to model quasi-rational impatience
has been put forward by Gul and Pesendorfer ( ,
), who stress the impor
tance of self-control problems leading to the post-
ponement of saving.
Where Do we Stand?
Since the 1970s we have learned much about the
empirical implications of the life-cycle model and
about the details of the model that need to be
modified to fit the available evidence. Much work,
however, remains to be done. In particular, there is
scope to develop more complex numerical models
that incorporate several realistic features. The areas
of labour supply and housing are, in our opinion,
particularly important. We also need to develop our
understanding of the empirical implications of
alternative models, such as hyperbolic discounting
and check the extent to which they are empirically
distinguishable from more standard models with
complex preferences. Finally, it is important to
stress the need for more and better data. One of the
lessons learned from the development of new
surveys that have been used to measure household
wealth is that with enough ingenuity and creativity
one can measure several of the variables that are
relevant for our understanding of consumption and
saving behaviour.
Our analysis of consumption and saving requires
that more comprehensive measures of consumption
are included in existing surveys, and that we learn
to make systematic use of records on expectations,
perceived uncertainty and so on.
See Also
► Consumer Sxpend.mre
► Consumption Cased Asset Pricing Models
2119
►
Bibliography
Abel, A.B.. 1990. Asset prices under habit formation and
catching up with the Joneses. American Economic
Review 80: 38-42.
Alessie, R., M.P. Devereux, and G. Weber. 1997.
Intertemporal consumption, durables and liquidity
constraints: A cohort analysis. European
Economic Review 41:37-59.
Atkenson, A., and M. Ogaki. 1996. Wealth varying
intertemporal elasticities of substitution: Evidence
from panel and aggregate data. Journal of
Monetary Economics 38: 507-534.
Attanasio, O.P. 2000. Consumer durables and inertial
behaviour: Estimation and aggregation of (S, s) rules
for automobile purchases. Review of Economic
Studies 67: 667-696.
Attanasio, O.P., and M. Browning. 1995. Consumption
over the life cycle and over the business cycle. Ameri-
can Economic Review 85: 1118-1137.
Attanasio, O.P., and A. Brugiavini. 2003. Social security
and households’ saving. Quarterly Journal of
Economics 118: 1075-1119.
Attanasio, O.P., and S.J. Davis. 1996. Relative wage move-
ments and the distribution of consumption. Journal
of Political Economy 104: 1227-1262.
Attanasio, O.P., and H. Low. 2004. Estimating Euler equa-
tions. Review of Economic Dynamics 7: 405-435.
Attanasio, O.P., and S. Rohwedder. 2003. Pension wealth
and household saving: Evidence from pension reforms
in the United Kingdom. American Economic
Review 93: 1499-1521.
Attanasio, O.P., and G. Weber. 1989. Intertemporal
substitution, risk aversion and the Euler equation for
consumption. Economic Journal 99: 59-73.
Attanasio, O.P., and G. Weber. 1993. Consumption
growth, the interest rate and aggregation. Review of
Economic Studies 60: 631 -649.
Attanasio, O.P., and G. Weber. 1994. The UK
consumption boom of the late 1980s: Aggregate
implications of microeconomic evidence. Economic
Journal 104: 1269-1302.
Attanasio, O.P., and G. Weber. 1995. Is consumption
growth consistent with intertemporal optimization?
2120 Consumer Expenditure (New Developments and the State of Research)
Evidence from the consumer expenditure survey.
Journal of Political Economy 103: 1121 1157.
Attanasio. O.P.. J. Banks. C. Meghir, and G. Weber. 1999.
Humps and bumps in life-time consumption. Journal
of Business and Economic Statistics 17: 22 35.
Banks. J., R. Blundell, and S. Tanner. 1998. Is there a
retirement-savings puzzle? American Economic
Review 88: 769-788.
Bemanke. B. 1985. Adjustment costs, durables and aggre-
gate consumption. Journal of Monetary Economics 15:
41-68.
Bemheim, B.D.. J. Skinner, and S. Weinberg. 2001. What
accounts for the s anation in retirement wealth among
U.S. households1? American Economic Review 91: 832-
857.
Bertola. G.. L. Guiso. and L. Pistaferri. 2005. Uncertainty
and consumer durables adjustment. Review of Eco-
nomic Studies 72: 973-1007.
Bewley, T.F. 1977. The permanent income hypothesis: A
theoretical formulation. Journal of Economic Theory
16: 252 259.
Blinder. A., and A. Deaton. 1985. The time series con-
sumption function revisited. Brookings Papers on Eco-
nomic Activity 1985(2): 465-521.
Blundell. R., .VI. Browning, and G. Meghir. 1994. Con-
sumer demand and the life-cycle allocation of house-
hold expenditures. Review of Economic Studies 61: 57-
80.
Bodie, Z., R.C. Merton, and W.F. Samuelson. 1992. Labor
supply flexibility and portfolio choice in a life-cycle
model. Journal of Economic Dynamics and Control 16:
427 449.
Boskin, M.J. 1978. Taxation, saving and the rate of interest
Journal of Political Economy. S3 S28.
Breeden, D.T. 1979. An intertemporal asset pricing model with
stochastic consumption and investment opportunities. Journal
of Financial Economics 7: 265 296.
Browning, M., and T.F. Crossley. 2000. Luxuries are easier to
postpone: A proof. Journal of Political Economy 108: 1022-
1026.
Browning, M., and M. Ejmaes. 2002. Consumption and children.
Working Paper No. 2002 6, Centre for Applied
Microeconometrics, University of Copenhagen.
Caballero, R.J. 1990. Consumption puzzles and precautionary
savings. Journal of Monetary Economics 25: 113-136.
Caballero, R.J. 1991. Earnings uncertainty and aggregate wealth
accumulation. American Economic Review 81: 859-871.
Campbell, J.Y. 1987. Does saving anticipate declining labor
income? An alternative test of the permanent income
hypothesis. Econometrica 55: 1249 1273.
Campbell, J.Y., and J. Cochrane. 1999. Force of habit: A
consumption-based explanation of aggregate stock market
behavior. Journal of Political Economy 107: 205-251.
Campbell, J.Y., and A. Deaton. 1989. Why is consumption so
smooth? Review of Economic Studies 56: 357 373.
Campbell, J.Y., and N.G. Mankiw. 1990. Permanent
income, current income, and consumption. Journal of
Business and Economic Statistics 8: 265 279.
Campbell, J.Y., and N.G. Mankiw. 1991. The response of
consumption to income: A cross-country investigation.
European Economic Review 35: 723 756.
Carrasco, R., J.M. Labeaga. and J.D. Lopez-Salido. 2005.
Consumption and habits: Evidence from panel data.
Economic Journal 115: 144-165.
Carroll. C.D. 1992. The buffer-stock theory of saving:
Some macroeconomic evidence. Brookings Papers on
Economic Activity 1992(2): 61-156.
Carroll. C.D. 1997. Buffer-stock saving and the life cycle/
permanent income hypothesis. Quarterly Journal of
Economics 112: 1-55.
Carroll, C.D.. and VI.S. Kimball. 1996. (On the concavity
of the consumption function. Econometrica 64: 981-
992.
Chah. E.Y.. V.A. Ramey, and R.M. Starr. 1995. Liquidity
constraints and intertemporal consumer optimization:
Theory and evidence from durable goods. Journal of
Money, Credit and Banking 27: 272 287.
Chen, X.. and S.C. Ludvigson. 2004. A land of addicts? An
empirical investigation of habits-based asset pricing
models. New York: New York University Press.
Choi, J.J., D. Laibson, B.C. Madrian, and A. Metrick.
2006. Saving for retirement on the path of least
resistence. In Behavioral public finance: Toward a New
Agenda, ed. Is. McCaffrey and J. Slemrod. New York:
Russell Sage Foundation.
Cochrane, .1.11. 1991. A simple test of consumption insur-
ance. Journal of Political Economy 99: 957 976.
Cole. H.L.. andN.R. Kocherlakota. 2001. Efficient alloca-
tions with hidden income and hidden storage. Review
of Economic Studies 68: 523 542.
Deaton. A. 1991. Saving and liquidity constraints.
Econometrica 59: 1221 1248.
Dynan. K. 2000. I labit formation in consumer
preferences: Evidence from panel data. American
Economic Review 90:391 406.
Eberly, J.C. 1994. Adjustment of consumers' durables
stocks: Evidence from automobile purchases. Journal
of Political Economy 102: 403-436.
Epstein, L.G., and S.E. Zin. 1989. Substitution, risk aver-
sion, and the temporal behavior of consumption and
asset returns: A theoretical framework. Econometrica
57: 937 969.
Epstein, L.G., and S.E. Zin. 1991. Substitution, risk aver-
sion, and the temporal behavior of consumption and
asset returns: An empirical analysis. Journal of Politi-
cal Economy 99: 263-286.
Flavin, M.A. 1981. The adjustment of consumption to
changing expectations about future income. Journal of
Political Economy 89: 974 1009.
Friedman, M. 1957. A theory of the consumption, function.
Princeton: Princeton University Press.
Gollier, C. 1995. The comparative statics of changes in risk
revisited. Journal of Economic Theory 66: 522 536.
Gourinchas, P.-O., and J.A. Parker. 2002. Consumption
over the life cycle. Econometrica 70: 47-89.
Consumer Expenditure (New Developments and the State of Research)
Grossman, S.J., and G. Laroque. 1990. Asset pricing and
optimal portfolio choice in the presence of illiquid
durable consumption goods. Econometrica 58: 25-51.
Gul, F., and W. Pesendorfer. 2001. Temptation and self-
control. Econometrica 9: 1403-1435.
Gul, F., and W. Pesendorfer. 2004. Self-control and the
theory of consumption. Econometrica 72: 119-158.
Hall, R.E. 1978. Stochastic implications of the life cycle-
permanent income hypothesis: Theory and evidence.
Journal of Political Economy 86: 971-987.
Hall, R.E. 1988. Intertemporal substitution in consump-
tion. Journal of Political Economy 96: 339-357.
Hall, R.E., and F.S. Mishkin. 1982. The sensitivity of
consumption to transitory income: Estimates from
panel data on households. Econometrica 50: 461-481.
Hamermesh, D.S. 1984. Consumption during retirement:
The missing link in the life cycle. Review of Economics
and Statistics 66: 1-7.
Hansen, L.P., andK.J. Singleton. 1982. Generalized
instrumental variables estimation of nonlinear
rational expectations models. Econometrica 50: 1269-
1286.
Hansen, L.P., and K.J. Singleton. 1983. Stochastic con-
sumption, risk aversion, and the temporal behavior of
asset returns. Journal of Political Economy 91: 249-
265.
Harris, C., and D. Laibson. 2001. Dynamic choices of
hyperbolic consumers. Econometrica 69: 935-957.
Hayashi, F. 1985. The permanent income hypothesis and
consumption durability: Analysis based on Japanese
panel data. Quarterly Journal of Economics 100: 1083-
1113.
Hayashi, F. 1987. Tests for liquidity constraints: A critical
survey. In Advances in econometrics 11: Fifth world
congress, ed. T. Bewley. Cambridge: Cambridge Uni-
versity Press.
Hayashi, F., J. Altonji, and L. Kotlikoff. 1996. Risk-
sharing between and within Families. Econometrica
64: 261-294.
Heaton, J. 1993. The interaction between time-
nonseparable preferences and time aggregation.
Econometrica 61: 353-385.
Heckman, J.J. 1974. Life cycle consumption and labor
supply: An explanation of the relationship between
income and consumption over the life cycle. American
Economic Review 64: 188-194.
Hubbard, R.G., J. Skinner, and S.P. Zeldes. 1994. The
importance of precautionary motives in explaining
individual and aggregate saving. Carnegie-Rochester
Conference Series on Public Policy 40: 59-125.
Hubbard, R.G., J. Skinner, and S.P. Zeldes. 1995. Precau-
tionary saving and social insurance. Journal of Politi-
cal Economy 103: 360-399.
King, R.G., and C.I. Plosser. 1984. Money, credit, and
prices in a real business cycle. American Economic
Review 14: 363—380.
Laibson, D. 1997. Golden eggs and hyperbolic
discounting. Quarterly Journal of Economics 62: 443—
477.
Lucas, R.E. Jr. 1978. Asset prices in an exchange
2121
MaCurdy, T.E. 1981. An empirical model of labor supply
in a life-cycle setting. Journal of Political Economy 89:
1345-1370.
MaCurdy, T.E. 1999. An essay on the life cycle: Charac-
terizing intertemporal behavior with uncertainty,
taxes, human capital, durables, imperfect capital
markets, and nonseparable preferences. Research in
Economics 53: 5-46.
Mankiw, G.N. 1982. Hall’s consumption hypothesis and
durable goods. Journal of Monetary Economics 10:
417^425?
Meghir, C., and G. Weber. 1996. Intertemporal non-
separability or borrowing restrictions? A disaggregate
analysis using a U.S. consumption panel.
Econometrica 64: 1151-1181.
Miniaci, R., and G. Weber. 1999. The Italian recession of
1993: Aggregate implications of microeconomic evi-
dence. Review of Economics and Statistics 81: 237-249.
Modigliani, F., and R. Brumberg. 1954. Utility analysis
and the consumption function: An interpretation of
cross-section data. In Post Keynesian economics, ed. K.
Kurihara. New Brunswick: Rutgers University Press.
Padula, M. 1999. Euler equations and durable goods.
Working Paper No. 30, CSEF.
Parker, J.A. 1999. The reaction of household consumption
to predictable changes in social security taxes. Ameri-
can Economic Review 89: 959-973.
Pissarides, C.A. 1978. Liquidity considerations in the the-
ory of consumption. Quarterly Journal of Economics
92: 279-296.
Poterba, J. 1988. Are consumers forward looking? Evi-
dence from fiscal experiments. American Economic
Review 78: 413-418.
Sargent, T.J. 1978. Rational expectations, econometric
exogeneity, and consumption. Journal of Political
Economy 86: 673-700.
Shapiro, M.D., and J. Slemrod. 2003. Consumer response
to tax rebates. American Economic Review 93: 381-396.
Sinai, T., and N.S. Souleles. 2005. Owner-occupied hous-
ing as a hedge against rent risk. Quarterly Journal of
Economics 120: 763-789.
Souleles, N.S. 1999. The response of household consump-
tion to income tax refunds. American Economic
Review 89: 947-958.
Summers, L. 1981. Capital taxation and capital accumula-
tion in a life-cycle growth model. American Economic
Review 71: 533-544.
Thurow, L.C. 1969. The optimum lifetime distribution of
consumption expenditures. American Economic
Review 59: 324-330.
Townsend, R.M. 1994. Risk and insurance in village India.
Econometrica 62: 539-591.
West, K.D. 1988. The insensitivity of consumption to news
about income. Journal of Monetary Economics 21: 17-
33.
Zeldes, S.P. 1989. Consumption and liquidity constraints:
An empirical investigation. Journal of Political Econ-
omy 97: 305—346.
2122
Consumer Surplus
Daniel T. Slesnick
Abstract
Over the years, consumer surplus has been used
to measure the welfare effects of price and
income changes. Despite its widespread use, it
provides a measure of well-being that is
ordinally equivalent to the change in utility only Dupuit (
under conditions that are inconsistent with long-
standing empirical evidence. Hicksian surplus
measures, such as the equivalent or
compensating variations, provide exact tory of the debate related to consumer surplus.) We
indicators of the change in utility without such
restrictions. Beginning in the early 1980s,
empirical methods have been developed to p = (pi, p2> - ■ -, p„) ~ a vector of commodity
estimate the equivalent variation that has the
same data requirements as consumer surplus.
Keywords
Aggregation; Compensating variation; Con-
sumer surplus; Equivalent variation; Expendi-
ture function; Indirect utility function; of a change in the price of a single commodity from
Integrability of demand; Intertemporal welfare pi to p\. The change in consumer surplus is given
effects; Linear expenditure system; Marginal by:
utility of income; Representative agent; Roy’s
identity; Social choice; Social expenditure
function; Well-being
JEL Classifications
Dll
How does the market power exercised by firms
influence consumer welfare? What is the effect of
excise taxes on households with different levels of
income? Does governmental regulation increase the
welfare of consumers? Topical issues such as these
indicate that the measurement of welfare is a
fundamental element of public policy analysis.
Indeed, a full consideration of taxes, subsidies,
transfer programmes, health care reform, regulation,
environmental policy, the social security system,
and educational reform must ultimately
Consumer Surplus
address the question of how these policies affect
individual well-being.
While centrally important to many problems of
economic analysis, confusion persists concerning
the relationship between commonly used indicators
of welfare and well-established theoretical
formulations. For more than 150 years, consumer
surplus has been used to measure the welfare effects
of changes in prices and incomes. Its popularity can
be ascribed to its intuitive appeal, the ease with
which it is implemented, and its modest data
requirements. Although it is generally accepted that
) was the originator
of the concept of consumer surplus, it is largely
attributed to Marshall ( ). (Chipman and
Moore ( ), provide a brief survey of the his
begin with the following notation:
prices.
Y k - the income of individual k AK - a vector of
demographic characteristics of individual k.
X,K = X, (P, YH Ak) is the demand for good i by
individual k.
Suppose we are interested in the welfare impact
fp!
A C S * = — xi(t,p2, ...pn, Yk,Ak)dt. (1 )
Jp?
If ACSk is positive (negative), the price change
is judged to have increased (decreased) the welfare
of individual k. Is it ordinally equivalent to the
change in utility? A necessary condition is that the
demand function is generated by a rational
consumer who maximizes utility subject to a budget
constraint. Unless consumers have optimized and
are at the boundaries of their budget sets, it is
impossible to assess the welfare effects of changes
in prices and incomes. (That is, demands must be
‘integrable’ and consistent with a well-behaved
utility function. Hurwicz and Uzawa ( ),
Consumer Surplus 2123

provide a formal statement of the integrability ACS'* = —| y^x;(p, Y k ,A k )dp i

conditions.)
If demands are consistent with rational con- + (Yl-Y°),
sumer behaviour, an indirect utility function V(p, Yh
(4 )
Ak) represents the maximum utility attained at
prices p and income Yh and Roy’s Identity provides where Z is a path between (p°, y^)and (p 1 ,7*).
the link between demands and utility: Chipman and Moore ( ) have demonstrated
that here are no circumstances under which ( ) is
dV($,Yk,&k)ld path independent and ordinally equivalent to the
*1 (p> Y k , A k ) p\ (2)
change in utility of a rational consumer.
av(p,F*,A k )/ar*'

If the marginal utility of income is constant, sub- Hicksian Surplus Measures

stitution of ( ) into ( ) yields an explicit expression
for the change in consumer surplus that is ordinally Given the problems with consumer surplus, how
equivalent to the change in utility: should the welfare effects of price and income
changes be measured? Hicks ( ) developed
an approach that is exactly analogous to ( ) once we
y( P 1 ,r*,A k )-y(p°,r*,A k )
ACS* dV/dYk substitute compensated for uncompensated demand
functions:

While constancy of the marginal utility of MIS,: j 534(p,y,Ak)dA.

income is restrictive, Chipman and Moore ( ,
) have shown that application of con
sumer surplus is more problematical if there are
changes in more than one price. In such circum-
stances, the change in consumer surplus must be
evaluated using a line integral defined over the path
of price changes from p 11 to p 1:
tp1 x
ACS* = Vx,(p,fi,Ak)dpi. (3)
V i
Price paths are not observed so it is essential that ( )
be path independent. This holds if the
uncompensated price effects are symmetric (see, for
example, Angus Taylor and Robert Mann , pp. 500-
4):
Ox, dxj
dpj dpt f o r all i / j.
+ {Yl-Y°k) ( 5)
wherex? (p, V, A k ) is the compensated demand for
the ith good evaluated at utility level V Compen-
sated price effects are symmetric, so the line inte-
gral in ( ) is path independent and the surplus
measure is single-valued.
For simple binary comparisons of policies, the
utility level at which A//S* is evaluated is often
treated as a matter of little consequence. If it is
calculated at the utility attained at prices p 1 and
income Y\ (denoted V1), a generalized version of the
equivalent variation is obtained:
EVk =E( p°,y\A k )
= E (p°, V1, A k ) — E (p°, F°, A k ) (6)

This form of symmetry requires preferences to be

homothetic, which is a restriction that is inconsis-
tent with well-established empirical regularities.
In the most general circumstance of changes in
prices and income, consumer surplus is defined as:
where E{p, V, Ak) is the expenditure function,
defined as the minimum income needed for indi-
vidual k to attain utility Vat prices p. Not only is the
generalized equivalent variation singlevalued, but it
is ordinally equivalent to the change
2124
in utility. That is, EVk is positive if and only if
V1 > V°.
The utility level at which ( ) is evaluated is
important for multiple comparisons of price and
income changes. The generalized equivalent var-
iation will give an ordering of outcomes that is
identical to that based on utility levels. If ( ) is
evaluated at V/0 = V'fp0, Y{)k, Ak), we obtain the
generalized compensating variation:
CVt =E{p°,V°,Ak) -E(p\V°,Ak) + (Y\-Y°)
= E{p\V\Kk)-E(p\V°,Ak).
(7)
Because the utility levels are ‘cardinalized’ using
different prices for each set of binary comparisons,
the ordering of multiple outcomes based on ( ) need
not match the ordering based on utility levels.
Chipman and Moore ( ) have shown
that consistent rankings of outcomes require
restrictions on preferences that are the same as for
consumer surplus.
While the simple static formulation of consumer
surplus is the most frequent application, the
conceptual framework can be extended to analyse
the effects of changes in utility in more general
settings. For example, intertemporal welfare effects
are often represented as the discounted sum of the
within-period equivalent or compensating
variations.
Keen ( ) has shown that this will differ
from the lifetime equivalent variation to the extent
that individuals are able to substitute
intertemporally. As an alternative approach, he
defines Vi to be the maximum level of lifetime
utility of an individual who lives T periods when the
profiles of prices and interest rates are {pt} and \r,}
respectively. If the (optimal) time path of utility
corresponding to VL at these prices and interest rates
is {Vk, (, the lifetime expenditure function can be
represented as:
= * 5 2 g , £(p t ,V*„A kt ),
t
where g, = II'=0(1 + rs)~l. Although conceptually analogous to the equivalent
As in the static framework, the lifetime expen- and compensating variation described previously,
diture function can be used to represent an exact CVk is now random and cannot, in general, be
represented in closed form.
Consumer Surplus
measure of the change in lifetime welfare. Define V{
to be the maximum level of lifetime welfare when
the profile of prices and interest rates are {pj} and
{r)} and denote the corresponding time path of
utility as {V^ } . The reference prices and interest
rates, {pj} and {/f }, yield a lifetime utility level of
V°L and within-period utilities {V®,}. Keen’s exact
measure of the change in lifetime welfare, evaluated
at the reference prices, is exactly analogous to the
generalized equivalent variation:
AW t = Q t ({p?},{i®},V>)
-Q*({p?}.K}.v2).
The concepts of the equivalent and compensat-
ing variation can also be extended to cases in which
the choices made by consumers are discrete rather
than continuous. Dagsvik and Karlstrom ( )
describe the compensating variation in
the context of a random utility model defined as:
(pj> Ak^ + £jk(j — 1 , 2 , . . . , J ) ,
Ujk = V
where Ujk is the utility of individual k in alternative
j, V(.) is a deterministic indirect utility function, and
Ejk are random variables. There are a total of J
choices available to the consumer and, for
simplicity, it is assumed that only prices vary across
alternatives.
Consider the welfare effect of a change in the set
of prices and income facing individual k from
(p?. p°. • • • - p“. Y k ) to ( P 1 - pi • • • - pj’ Y \ )
■ If the
consumer chooses the alternative that maximizes
Ujk, the compensating variation is defined implicitly
as that value CV k that satisfies the following
equality:
max/ y(p?,Tk,Ak) +£jk
= maxjV(p|, Y\ - CVk, A k ) + 8jk.
Consumer Surplus 2125

From Demand Functions to Welfare
Measurement
While it was understood that the equivalent varia-
tion resolved the conceptual problem of welfare
measurement, it had little influence on applied
welfare economics because compensated demand
functions were presumed to be unobservable. Willig
( ) made the first attempt to bridge the
gap between theory and application by showing
that, for a single price change, consumer surplus
can provide an approximation to the equivalent or
compensating variation. However, with multiple
price and income changes, consumer surplus is not
single-valued and is of no use in approximating
changes in economic welfare (McKenzie ).
Shortly after the publication of Willig’s paper,
however, empirical procedures were developed to
estimate the equivalent or compensating variation.
Each method begins with the specification of a
demand function and, under the assumption of
integrability, is used to recover the utility or
expenditure functions. The complexity of this
procedure diminishes if demand functions are lin-
ear, and consideration is restricted to changes in the
price of a single good.
Hausman ( ) provided an analytic solution
to this problem for a demand function given by:
Closed form solutions to the partial differential
equation implied by Roy’s Identity can be obtained
for only a limited class of demand functions. An
alternative approach is to begin with an assumed
form of the indirect utility function and use Roy’s
Identity to obtain a system of demand equations.
Since the form of the utility function is assumed
from the outset, it is unnecessary to solve a
complex system of partial differential equations.
Muellbauer ( ) provided an early example
of this approach. He assumed that demands were
consistent with a Stone-Geary utility function given
by:
np.n) Jn ~ n f. pA) W
where 8 = (8i, 82, .. .,8„) and a = (ai, a2, .. ,,a„) are
unknown parameters. The corresponding
expenditure function is:
1 V U
E ( p , V ) Y,P‘8‘ ( P?)-
The unknown parameters can be estimated by
fitting the linear expenditure system to household
budget data:
ptXi = p^i

x
i = JpPi + JYY/C + YAAk, + a i( Y k - (j = 1,2,
GO)
where yp, yY, and yA are unknown parameters to be
estimated econometrically. Roy’s Identity provides Given estimates of a and S, the expenditure func-
a partial differential equation that can be solved to tion can be used to compute the equivalent or
obtain an expenditure function of the form: compensating variation as in ( ) and ( ).
While this is more general than Hausman’s
E(puV, Ak) =Ve™ approach, it has its own disadvantages. For an
- (1/Yy) [ i p P i + (J p / l r ) + YAAk] • (8) assumed form of the utility function, the functional
forms of the demands are the same for every good,
The expenditure function allows the equivalent which may hinder the ability of the model to fit the
variation to be computed exactly as in ( ) and data. Is it possible to start with an arbitrary demand
Willig-type approximations are unnecessary. system (rather than a utility function) and measure
Hausman’s method has the same data requirements the welfare effects of multiple price changes? Two
as consumer surplus, and only linear regression elegant procedures were proposed that required
methods are needed to estimate the unknown more complicated calculations to recover the
parameters. expenditure function, but did not impose
restrictions on the form of the demand
2126 Consumer Surplus

functions other than the standard integrability 0 < t < 1. As prices and income change, the
conditions. movements of demands along an indifference curve
The first method is based on an approximation can be represented implicitly by the differential
to McKenzie’s ( ) indirect money metric util equation:
ity function defined as:
dYk(i) E*i(p(0,r*(0.Ak)^.
l * ( p , Y k , A k ;p°) = £(p°,V(p,r*,A k ),A k ). dt

McKenzie and Pearce ( ) showed that A/t can Integrating over t yields an expression that can, in
be approximated by a Taylor’s series expansion principle, be solved to obtain £(p(/), V°, A k ) which
about the initial equilibrium: is the centrepiece of the welfare calculations:

A^ = ^Ap(l/2) A
A
' A
£(p(r),y°,A k ) -£(p°,V°,A k )
P 7TTT7AP
t
dpdp'
dp d2p = } Xi
(p(0. E ( P (0> v°, A k ,), A k ) d t .
r 7Ap+ 1/2^4AT)AT
dY dpdY 1 ' dY1 ) 1=1
o
(13)
R
(11)

where R represents higher order terms in the series.
The expression in ( ) can be represented as a
function of uncompensated demand functions when
p is evaluated at the reference prices (this follows
from Roy’s Identity and from the fact that at these
prices the marginal utility of income is equal to one
and all higher income derivatives are zero - see
McKenzie and Pearce , for details):
Ap = —X'Ap
Vartia described several algorithms that can be
used to solve this equation numerically over the
price path p(t) so that, when evaluated at / 1, we
obtain £(p1, V°, Ak). As long as the demands satisfy
the integrability conditions, the solution to ( ) will
be independent of the price path used in the
algorithm. This method is valid for multiple price
and expenditure changes and, because a closed-form
solution is unnecessary, facilitates flexibility in
estimating demand patterns.

Aggregation

The methods described to this point provide esti-

+ (l-J|Ap)Ay + tf. (12)
Given knowledge of the demand functions and the
magnitudes of the price and income effects, one has
all of the information necessary to get as accurate
an estimate of the change in utility as
desired.
Vartia ( ) developed an algorithm that
recovers the expenditure function numerically to
any desired level of accuracy. Let p(t) and Yk(t) be
the paths of price and income changes for
mates of the change in welfare for individuals. In
practice, analysts are more concerned about the
impact of policies on groups. Micro-level estimates
are an essential first step, but, for welfare economics
to be useful to practitioners, a method of
aggregation is essential. The easiest approach is to
assume that market demands are generated by a
representative consumer. Under this condition, the
methods described previously can be applied to
aggregate demands and the utility function of the
representative agent can be recovered.
Consumer Surplus
While frequently applied, this is unsatisfactory
for a number of reasons. Market demands need not
be consistent with a rational representative
consumer. Even if every individual has demands
that are consistent with utility maximization,
aggregate demands need not satisfy any of the
integrabilify conditions other than homogeneity of
degree zero in prices and income (Sonnenschein ).
Moreover, it is unclear what this utility function
actually represents. Kir- man ( ) presents an
example in which the
representative agent prefers (aggregate) market
basket A to B even though all individuals prefer the
reverse. This violation of the most basic principle of
social choice suggests that the utility of the
representative agent should not be used for policy
analysis even in the unlikely event that aggregate
demands are integrable.
An alternative approach is to define aggregate
welfare to be a function of the individual surplus
measures. Such an approach was advocated by
Harberger ( ) in his effort to make consumer
surplus the standard tool for applied welfare anal-
ysis. At a conceptual level, such an indicator of
aggregate welfare appears to be a natural extension
of the positive analysis of welfare measurement at
the micro level. This is obviously not the case
because aggregation necessitates normative
judgements in which the gains to some must be
weighed against the losses to others. Simply sum-
ming the surplus measures, for example, embodies a
version of utilitarianism and ignores distributional
concerns.
Since any method of measuring welfare for
groups of individuals necessarily involves sub-
jective judgements, it seems reasonable to state
explicitly the underlying ethical basis for the
method of ordering outcomes in the aggregate. The
social choice theoretic framework used by Sen (
) provides a reasonable way of pre
senting the normative assumptions related to the
measurability and comparability of individual
welfare levels that facilitate well-behaved social
orderings of outcomes. Under conditions described
by Sen and others, these orderings can be
represented by a social welfare function:
2127
w = w ( v i , v 2, . . . , v k )
where Vk is a welfare indicator of individual k.
A monetary measure of social welfare can be
obtained using Poliak’s ( ) concept of a social
expenditure function:
M(p, W) = min-jY : W ( V i , . . . , V f ) > W , ^ Y k = fj.
This function is exactly analogous to its micro-
level counterpart and is the minimum level of
aggregate income required to attain a specified
social welfare contour. If W° is the social welfare
under policy 0 and W1 is the welfare under policy 1,
the monetary measure of the change is social
welfare is exactly analogous to the generalized
equivalent variation:
AW = M(p,W1)
AW is clearly ordinally equivalent to the
changes in social welfare, and normative judge-
ments are represented explicitly through the spec-
ification of the social welfare function.
See Also
► Coc
-Be
► Cost
Mi
► dicksiar
► adirect
► locialV
Bibliography
Chipman, J.S., and J. Moore. 1976. The scope of con-
sumer’s surplus arguments. In E v o l u t i o n , w e l f a r e
and
time in economics: Essays in honor of
Nicholas
G e o r g e s c u - R o e g e n , ed. A.M. Tang, F.M.
Westfield,
and J.S. Worley. Lexington: Heath-Lexington Books.
Chipman, J.S., and J. Moore. 1980. Compensating varia-
tion, consumer’s surplus, and welfare. A m e r i c a n
Eco-
2128 Consumption and Production

Dupuit, J. 1844. On the measurement of the utility of public
works. Trans R.H. Barback. In R e a d i n g s i n
W e l f a r e E c o n o m i c s , ed. K.J. Arrow and T.
Scitovsky. Home- wood: Richard D. Irwin.
Harberger, A.C. 1971. Three basic postulates for applied
welfare economics. J o u r n a l of Economic
L i t e r a t u r e 9: 785-797.
Hausman, J.A. 1981. Exact consumer’s surplus and dead-
weight loss. A m e r i c a n E c o n o m i c R e v i e w 71:
662-676.
Hicks, J.R. 1942. Consumer’s surplus and index numbers.
R e v i e w o f E c o n o m i c S t u d i e s 9 (2): 126-137.
Hurwicz, L., and H. Uzawa. 1971. On the integrability of
demand functions. In P r e f e r e n c e s , u t i l i t y a n d
d e m a n d , ed. J. Chipman et al. New York: Harcourt,
Brace and Jovanovich.
Keen, M. 1990. Welfare analysis and intertemporal substi-
tution. J o u r n a l o f P u b l i c E c o n o m i c s 42: 47-
66.
Kirman, A.P. 1992. Whom or what does the representative
individual represent? J o u r n a l o f E c o n o m i c
P e r s p e c t i v e s 6 (2): 117-136.
Marshall, A. 1890. P r i n c i p l e s o f e c o n o m i c s .
London: Macmillan.
McKenzie, L. 1957. Demand theory without a utility index.
R e v i e w o f E c o n o m i c S t u d i e s 24 (65): 185-189.
McKenzie, G.W. 1979. Consumer’s surplus without apology:
Comment. A m e r i c a n E c o n o m i c R e v i e w 69:
465-468.
McKenzie, G.W., and I. Pearce. 1976. Exact measures of
welfare and the cost of living. R e v i e w o f E c o n o m i c
S t u d i e s 43: 465^168.
Muellbauer, J. 1974. Prices and inequality: The
U.K. experience. E c o n o m i c J o u r n a l 84 (333): 32-
55.
Poliak, R.A. 1981. The social cost of living index. J o u r n a l
o f P u b l i c E c o n o m i c s 15: 311-336.
Sen, A.K. 1970. C o l l e c t i v e c h o i c e a n d s o c i a l
w e l f a r e . San Francisco: Holden Day.
Sonnenschein, H. 1972. Market excess demand functions.
E c o n o m e t r i c a 40: 549-563.
Taylor, A.E., and R. Mann. 1972. A d v a n c e d c a l c u l u s .
2nd ed. Lexington: Xerox College Publishing.
Vartia, Y.O. 1983. Efficient methods of measuring welfare
change and compensated income in terms of ordinary
goods’ (Sraffa , p. 93). This stands in striking
contrast to the approach of classical political econ-
omy which views the system of consumption and
production as a circular process. This perspective
was first developed by Quesnay ( ) and elab-
orated by Marx ( ) in his
analysis of the econ
omy in general. Marx developed a distinction
between ‘production and productive consumption’
and ‘consumption and consumptive production’ and
related this to the concepts of exchange and
distribution. This distinction fell into diuse with the
rise of neoclassical economics but has been
rehabilitated by Sraffa ( ) in his famous
critique of modem economics. The concept ‘pro-
duction and productive consumption’ provides the
general conceptual framework within which his
particular theory of commodities is elaborated (see
Sraffa , p. 3). Sraffa’s exposition not only advances
our understanding of the theory of commodities, it
also enables us to grasp the essence of Marx’s
important distinction between consumption and
production. Marx expressed himself in rather
obscure Hegelian terms and Sraffa’s simple
numerical examples clarify much of Marx’s
argument.
Following Sraffa, let us suppose that an
extremely simple society is producing just enough
wheat and iron to maintain itself. If 400 quarters of
wheat (hereafter 400 W) were produced using 280
W and 12 tons of iron (121) and 1/2 of the annual
labour supply (1/2 L) as inputs, while 201 were
produced using 120 W, 81 and 1/2 L, then the
methods of production and productive consumption
can be tabulated as follows:
1 /2L + 280W + 121 > 400W
1/2L+120W + 81 —» 201

Consumption and Production In order for the process to be repeated the wheat
industry must exchange 120 W for 121. This
C. A. Gregory restores the original distribution of products and
enables the process to be repeated.
A three-product model takes us from barter to
triangular trade: an n-product model to more com-
Neoclassical economic analysis is carried out within plex forms of exchange and distribution.
a conceptual framework that views the economic The general formulation of the concept ‘pro-
process as a ‘one way avenue’ leading from ‘factors duction and productive consumption’ implicit in
of production’ to ‘consumption Sraffa’s analysis is:
Consumption and Production
labour + things —> things.
In other words, the methods of production and
productive consumption describe the process of the
production of things by means of things and labour.
The production by commodities by means of
commodities is one historically specific form of
these general relations. The emergence of things
and labour as commodities presupposes private
property and the emergence of a class of proletar-
ians (Marx ). This is only one of many social forms
that things and labour can take. In tribal economies,
for example, things and labour assume the social
form of gifts. The social precondition for this to
arise is a relatively egalitarian distribution of land
between clans. Social data of this kind mean that the
principles governing the exchange and distribution
of products will vary greatly from economy to
economy. In a ‘pure’ tribal economy, for example,
profit maximization is not the central organizing
principle of economic life and wages, prices and
profits are not be found (Polanyi ).
A corollary of this general formulation of pro-
duction is that ‘consumption and consumptive
production’ can be described as follows:
things + people —> people.
In other words ‘consumption and consumptive
production’ describes the methods of production of
people by means of people and things.
Neither Marx nor Sraffa analysed these relations
which under capitalism would be called the
‘household economy’ or ‘kinship’. However,
anthropologists who have studied third world tribal
and peasant societies have tended to focus almost
exclusively on these relations, a fact, I would
suggest, which tells something about the relative
importance of production and consumption in
capitalist and tribal/peasant societies respectively.
Some indication of what is involved in this
concept of consumption can be gleaned by elabo-
rating its meaning in the context of Sraffa’s
‘extremely simple economy’. Suppose that the iron
and wheat were produced by two different
2129
households, each household consisting of a father
(M), a mother (F), a boy (m) and girl (f). Repro-
duction of the households, and hence of labour,
requires that the children set up new households
and produce their own children. Incestuous rela-
tions aside, it is clear that the households must
exchange children in a way that is analogous to the
exchange of wheat for iron discussed above. This
can be seen from the following formulation of the
relations of consumption and consumptive
production for this two-household economy:
Mi + F2 —> mi + fi
M2 + Fi —> m2 + fa
where the subscripts represent the respective
households. This particular example is an example
of what anthropologists call ‘cross cousin’ marriage
or ‘sister exchange’. By tracing the relationships
out it will be seen that a man marries his mother’s
brother’s daughter who is also his father’s sister’s
daughter. Take mi for example. His father is M b his
father’s sister is Fi, and the latter’s daughter is f 2
with whom he will set up a household in the next
generation. Tracing the relationships through mi’s
mother (F2) it is obvious that f2 is also his mother’s
brother’s daughter. Relations of this kind are very
important in clan- based societies where a number
of households, usually related either matrilineally or
patrilineally, occupy a common piece of territory
and forbid marriage within the households that
make up the clan. In our own society, where the
clan has no operational significance, and where
marriage is a matter of personal choice rather than a
formal arrangement between groups, the political
and economic significance of kinship and marriage
is relatively unimportant (Gregory ).
Every economic analysis of a particular socio-
economic form such as ‘profits’, ‘prices’ or ‘wages’
involves, either implicitly or explicitly, a general
conceptual framework within which the analysis is
carried out. The general model implicit in
Quesnay’s analysis of 18th-century French agricul-
ture has been elaborated and developed to provide
an extremely useful framework not only for the
development of a 20th-century theory of the value
and distribution of commodities but also for the
2130
analysis of comparative economic systems. By
focusing on the circular process of production and
reproduction, consumption becomes a dynamic pro-
cess rather than the dead end of a one way avenue.
See Also
Bibliography
Gregory, C.A. 1982. G i f t s a n d c o m m o d i t i e s . London:
Academic.
Marx, K. 1859. A contribution to the critique of political
economy. London: Lawrence & Wishart 1971.
Marx, K. 1867. C a p i t a l , vol. I. Moscow: Progress.
Polanyi, K. 1944. T h e g r e a t t r a n s f o r m a t i o n . New
York: Rinehart.
Quesnay, F. 1759. The ‘Tableau Economique’. In T h e
e c o n o m i c s o f p h y s i o c r a c y , ed. R. L. Meek.
London: George Allen, 1962.
Sraffa, P. 1960. P r o d u c t i o n o f c o m m o d i t i e s b y
m e a n s o f c o m m o d i t i e s . Cambridge: Cambridge
University Press.
Consumption Externalities
Robert H. Frank
Abstract
Consumption externalities occur when con-
sumption by some creates costs or benefits for
others. According to Duesenberry’s ‘relative
income hypothesis’, spending is influenced by
the individual’s own standard of living in the
recent past and the living standards of others in
the present. This hypothesis tracks observed
behaviour more closely than Friedman’s ‘per-
manent income hypothesis’, which assumes that
context has no influence on spending. When
context is more important for some goods
(positional goods) than for others (non-positional
goods), positional goods crowd out non-
positional goods, causing welfare losses like
those that occur when bombs crowd out
consumption in military arms races.
Consumption Externalities
Keywords
Bequest motive; Consumption externalities;
Friedman, M; Flirsch, F; Marx, K; Permanent
income hypothesis; Positional goods; Relative
income hypothesis; Revealed preference; Sav-
ings; Smith, A; Veblen, T
JEL Classifications
Dll
Consumption externalities occur when consumption
by some creates external costs or benefits for others.
Their recognition by economists dates at least as far
back as Adam Smith’s discussion of how local
consumption standards influence the goods that
people consider essential (or ‘necessaries’, as Smith
called them). In the following passage, for example,
he described the factors that influence the amount
someone must spend on clothing in order to be able
appear in public ‘without shame’:
By necessaries I understand, not only the commod-
ities which are indispensably necessary for the sup-
port of life, but whatever the custom of the country
renders it indecent for creditable people, even of the
lowest order, to be without. A linen shirt, for exam-
ple, is, strictly speaking, not a necessary of life. The
Greeks and Romans lived, I suppose, very comfort-
ably, though they had no linen. But in the present
times, through the greater part of Europe, a credit-
able day-labourer would be ashamed to appear in
publick without a linen shirt, the want of which
would be supposed to denote that disgraceful degree
of poverty which, it is presumed, no body can well
fall into without extreme bad conduct. (Smith , pp.
869-70)
Consumption externalities received only limited
attention in Smith’s Wealth of Nations and only
occasional mention by economists during the
century that followed its publication. Karl Marx (
), for example, noted that ‘A house may be
large or small; as long as the neighboring houses are
likewise small, it satisfies all social requirement for
a residence. But let there arise next to the little
house a palace, and the little house shrinks to a hut.’
It was not until Thorstein Veblen’s The Theory
of the Leisure Class appeared in 1899 that
Consumption Externalities
consumption externalities received their first seri-
ous, book-length treatment in economics. Veblen’s
thesis was that much of consumption is undertaken
to signal social position. But although his book is
still widely read and cited by scholars in numerous
disciplines, its general theme was largely ignored
by economists during the 50 years following its
publication.
Duesenberry's Relative Income
Hypothesis
Interest in this theme was rekindled with the pub-
lication of James Duesenberry’s Income, Saving,
and the Theory of Consumer Behavior in 1949. In
this volume, Duesenberry offered his ‘relative
income hypothesis’, in which he argued that an
individual’s spending behavioiu is influenced by
two important frames of reference - the individual’s
own standard of living in the recent past and the
living standards of others in the present. Thus, in
Duesenberry’s account, people are subject to both
intrapersonal and interpersonal consumption
externalities.
His theory attempted to explain three important
empirical regularities: (a) long- ran aggregate sav-
ings rates remain roughly constant over time, even
in the face of substantial income growth; (b)
aggregate consumption is much more stable than
aggregate income in the short ran; and (c) indi-
vidual savings rates rise substantially with income
in cross-section data. When Duesenberry’s book
was first published, individual consumption was
generally modelled by economists as a linear
function of income with a positive intercept term.
This model could accommodate rising savings rates
in cross-section data and the stability of
consumption over the business cycle, but not the
long-run stability of aggregate savings rates.
Duesenberry’s hypothesis was hailed as an
advance because of its ability to track all three
stylized fact patterns. The poor save at lower rates,
he argued, because they are more likely to
encounter others with desirable goods that are
difficult to afford. Moreover, since this will be true
no matter how much national income grows,
unfavorable comparisons will always occur more
2131
frequently for the poor - and hence the absence of
any tendency for savings rates to rise with income
in the long ran.
To explain why consumption is more stable than
income in the short ran, Duesenberry argued that
families compare their living standards not only to
those of others around them but also to their own
standards from the past. The high consumption
level once enjoyed by a formerly prosperous family
thus constitutes a frame of reference that makes
cutbacks difficult when income falls.
Despite Duesenberry’s success in fracking the
data, many economists felt uncomfortable with his
relative income hypothesis, which to them seemed
more like sociology or psychology than economics.
The profession was therefore immediately receptive
to alternative theories that purported to explain the
data without reference to softer disciplines. The
most important among these theories was Milton
Friedman’s permanent income hypothesis, variants
of which still dominate today’s research on
spending.
In hindsight, however, there remain grounds for
scepticism about whether Friedman’s theory was a
real step forward. For example, its fundamental
premise - that savings rates are independent of
permanent income - has been refuted by numerous
careful studies (see, for example, Car- roll ). Some
modem consumption theorists have responded by
positing a bequest motive for rich consumers, a
move that begs the question of why leaving
bequests should entail greater satisfaction for the
rich than for the poor.
Another problem is that, contrary to Friedman’s
assertion that the marginal propensity to consume
out of windfall income should be nearly zero,
people actually consume such income at almost the
same rate as permanent income (Bodkin ). To this
observation, Friedman ( ) himself responded
that consumers appear
to have unexpectedly short planning horizons. But
if so, then consumption does not really depend
primarily on permanent income.
Abundant evidence suggests that context influ-
ences evaluations of living standards (see, for
example, Veenhoven ; Easterlin ; Luttmer ). In the
light of this evidence, it seems fair to say that
Duesenberry’s hypothesis
2132
not only has been more successful than Friedman’s
in tracking how people actually spend but also rests
on a more realistic model of human nature. And yet
the relative income hypothesis is no longer even
mentioned in most leading economics textbooks. Its
absence appears to signal the profession’s con-
tinuing reluctance to acknowledge concerns about
relative consumption.
Welfare Implications
In traditional economic models, individual utility
depends only on absolute consumption. These
models lie at the heart of claims that pursuit of
individual selfinterest promotes aggregate welfare.
In contrast, models that include concerns about
relative consumption identify a fundamental
conflict between individual and social welfare. This
conflict stems from the fact that concerns about
relative consumption are stronger in some domains
than in others. The disparity gives rise to
expenditure arms races focused on ‘positional
goods’ - those for which relative position matters
most. The result is to divert resources from ‘non-
positional goods’, causing welfare losses. (The late
Fred Hirsch , coined these terms.)
The nature of the misallocation can be made
clear with the help of two simple thought experi-
ments. In each, you must choose between two
worlds that are identical in every respect except
one. The first choice is between world A, in which
you will live in a 4,000-square-foot house and
others will live in 6,000-square-foot houses; and
world B, in which you will live in a 3,000-square-
foot house, others in 2,000-square-foot houses.
Once you choose, your position on the local hous-
ing scale will persist.
If only absolute consumption mattered, A would
be clearly better. Yet most people say they would
pick B, where their absolute house size is smaller
but their relative house size is larger. Even those
who say they would pick A seem to recognize why
someone might be more satisfied with a 3,000-
square-foot house in B than with a substantially
larger house in A.
In the second thought experiment, your choice is
between world C, in which you would have four
Consumption Externalities
weeks a year of vacation time and others would
have six weeks; and world D, in which you would
have two weeks of vacation, others one week. This
time most people pick C, choosing greater absolute
vacation time at the expense of lower relative
vacation time.
The modal responses in these two thought
experiments suggest that housing is a positional
good and vacation time a non-positional good. The
point is not that absolute house size and relative
vacation time are of no concern. Rather, it is that
positional concerns weigh more heavily in the first
domain than in the second.
When the strength of positional concerns differs
across domains, the resulting conflict between
individual and social welfare is structurally iden-
tical to the one inherent in a military arms race.
When deciding how to apportion available
resources between domestic consumption and
military armaments, each country’s valuations are
typically more context-dependent in the armaments
domain than in the domain of domestic
consumption. After all, being less well armed than a
rival nation could spell the end of political
independence. The familiar result is a mutual
escalation of expenditure on armaments that does
not enhance security for either nation. Because the
extra spending comes at the expense of domestic
consumption, its overall effect is to reduce welfare.
Note, however, that if each country’s valuations
were equally context- sensitive in the two domains,
there would be no arms race, for in that case the
attraction of having more arms than one’s rival
would be exactly offset by the penalties of having
lower relative consumption.
For parallel reasons, the modal responses to the
two thought experiments suggest an equilibrium in
which people consume too much housing and too
little leisure (for a formal demonstration of this
result, see Frank ). In contrast, conventional welfare
theorems, which assume that individual valuations
depend only on absolute consumption, imply
optimal allocations of housing and leisure.
In addition to leisure, goods that have been
classified as non-positional by various authors
include workplace safety, workplace democracy,
savings and insurance. And since public goods
Consumption Function
are, by definition, available in equal quantities to all
consumers, they, too, are inherently non-positional.
The general claim is that unregulated market
exchange will tend to emphasize the production of
positional goods at the expense of these and other
non-positional goods (Frank ). Among the policies
suggested as remedies for this imbalance have been
income and consumption taxes, overtime laws,
hours laws for commercial establishments, legal
holidays, workplace safety and health regulation,
non-waivable workers’ rights, and tax-financed
savings accounts.
Consumption externalities also have implica-
tions for the theory of revealed preference, which
says that, if a well-informed individual chooses a
risky job that pays $600 a week rather than a safer
one that pays only $500, he reveals that the safety
increment is worth less than $100 to him. If safety
is a non-positional good, however, this inference
does not follow, for it ignores the fact that, if all
workers exchange safety for increased income, the
anticipated increase in relative consumption does
not occur. The value that workers assign to safety
may thus be revealed as much in the patterns of
safety regulation they favour as in the nature of the
jobs they choose.
See Also
Bibliography
Bodkin, R. 1959. Windfall income and consumption.
A m e r i c a n E c o n o m i c R e v i e w 49: 602-614.
Carroll, C. 1998. Why do the rich save so much? In D o e s
Atlas Shrug: The economic consequences
o f t a x i n g t h e r i c h , ed. J. Slemrod. New York:
Oxford University Press.
Duesenberry, J. 1949. I n c o m e , s a v i n g , a n d t h e
t h e o r y o f c o n s u m e r b e h a v i o r . Cambridge,
MA: Harvard University Press.
Easterlin, R. 1995. Will raising the incomes of all increase the
happiness of all? J o u r n a l of Economic
B e h a v i o r a n d O r g a n i z a t i o n 27: 35—47.
2133
Frank, R. 1985a. The demand for unobservable and other
nonpositional goods. American Economic
R e v i e w 75: 101-116.
Frank, R. 1985b. C h o o s i n g t h e r i g h t p o n d . New
York: Oxford University Press.
Friedman, M. 1957. A t h e o r y o f t h e c o n s u m p t i o n
f u n c t i o n . Princeton, NJ: Princeton University Press.
Friedman, M. 1963. Windfalls, the horizon, and related
concepts in the permanent income hypothesis. In M e a -
s u r e m e n t i n e c o n o m i c s , ed. C. Christ. Stanford:
Stanford University Press.
Hemenway, D., and S. Solnick. 1998. Is more always better?
Journal of Economic Behavior and
O r g a n i z a t i o n 37: 373-383.
Hirsch, F. 1976. S o c i a l l i m i t s t o g r o w t h . Cambridge,
MA: Harvard University Press.
Luttmer, E. 2005. Neighbors as negatives: Relative earnings
and well-being. Quarterly Journal of
E c o n o m i c s 120: 936-1002.
Marx, K. 1847. Relation of wage-labour to capital. In W a g e
L a b o u r a n d C a p i t a l . Marx/Engels Internet
Archive, 1999. Online. Available at
. Accessed 16 August 2005.
Smith, A. 1776. A n i n q u i r y i n t o t h e n a t u r e a n d
c a u s e s o f t h e w e a l t h o f n a t i o n s , ed. R.
Campbell and A. Skinner. Oxford: Oxford University
Press, 1976.
Veblen, T. 1899. T h e t h e o r y o f t h e l e i s u r e c l a s s .
New York: Modem Library.
Veenhoven, R. 1993. Happiness in nations: Subjective
appreciation of life in 56 nations. Rotterdam: Erasmus
University:
Consumption Function
Michael R. Darby
Keynes ( ) introduced the consumption func
tion as the relationship between consumption and
income. Although Keynes (pp. 95-6) believed this
relationship ‘a fairly stable function’, substantial
shifts in the function were soon observed by
empirical workers. Much work in the post-World
War II era achieved functional forms by the 1970s
which admirers and critics alike could agree were
relatively shiftless. Most recent work has consid-
ered not functional form but whether or not
observed changes in consumption are consistent
with models of efficient markets.
2134
The Keynesian Conception
Keynes conceived of the consumption function as
relating consumption to disposable income as these
are now conventionally measured in the national
income accounts. These concepts were basic to the
model of The General Theory and Keynes was
doubtless pedagogically correct to posit a simple
relationship which could be refined by future
research.
The need for refinement became apparent
shortly. In longer time series, consumption seemed
to vary around a constant fraction of disposable
income. In contrast, consumption functions fitted to
depression-era or cross-section data seemed to
indicate that this ratio (which Keynes called the
average propensity to consume or APC) declined as
disposable income rose. In other words, these
studies estimated that the derivative of consumption
with respect to disposable income (the marginal
propensity to consume or MPC) was less than the
APC.
Alvin Hansen ( ) among others predicted
that a secular stagnation would result unless gov-
ernment spending filled this growing gap between
output and consumption. When the gap failed to
appear, the time was ripe for more sophisticated
theories of the relationship between consumption
and income. These theories were the earliest and
perhaps still most successful resorts to microeco-
nomic foundations for macroeconomics.
Permanent Life-Cycles
In the early 1950s our two dominant models of
consumption developed: the permanent-income and
life-cycle hypotheses. While these models were
once viewed as competing, they can now be seen as
complementary with differences in emphasis which
serve to illuminate different significant problems.
Both models emphasized the distinction between
consumer expenditures measured by the national
income accounts and pure consumption which was
to be explained by optimal allocation of present and
future resources over time. The permanent income
hypothesis (PIH) stressed stochastic variations in
income (and
Consumption Function
consumption) over time and viewed saving in terms
of a bequest motive. The life-cycle hypothesis
(LCH) stressed predictable variations in income
(and consumption) over the life cycle and viewed
saving as resulting from the greater wealth and
numbers of younger savers in comparison to older
dissavers.
The original published references are to
Friedman ( ) for the PIH and Modigliani
and Brumberg ( ) for the LCH. Given the
delays in NBER publication of Friedman’s work
which was widely circulated in manuscript form,
the two hypotheses are generally regarded as
distinct, contemporaneous responses to the
described conflict between earlier studies and
Simon Kuznets’ data on the national income
accounts for the 20th century. From the perspective
of the monumental careers of the two principal
proponents, priority does not seem an issue that
need be resolved here.
The PIH relates (pure) consumption to the per-
petuity stream that could be consumed forever. The
agent is typically regarded as an infinitely lived
individual. This represents the underlying notion of
a family whose generations are linked by operative
transfers from parent to child or vice- versa. Saving
arises to equate the ratio of marginal utility of
present and future consumption to the marginal rate
of transformation implicit in market (real) interest
rates. In this way the PIH is said to emphasize the
bequest motive for saving.
In contrast, the strict LCH had individuals con-
suming their entire endowments over their lifetime.
Saving was supposed to arise because young
workers were more numerous and wealthy (due to
technological progress) than the older generation
who were dissaving to finance retirement. This
provides an avenue by which faster growth can
increase saving. Alternatively, as discussed below,
factors such as social security which change the
extent of mismatch between lifetime consumption
and income patterns are predicted to have profound
effects on aggregate saving.
These approaches - and their synthesis with
inter-generationally linked utility functions - have
led to a rich literature quite apart from the con-
sumption function, but those developments are
beyond the scope of this essay.
Consumption Function
From the point of view of the consumption
function per se, the PIH and LCH imply that pure
consumption is a fraction (variable in principle but
rarely in practice) of wealth or permanent income.
Here wealth is inclusive of human as well as non-
human capital and permanent income is a
(conventionally constant) long-term ex ante real
interest rate times this wealth. (Note that, contrary
to Sargent ( ) and others this wealth is not the
discounted present value of expected future income
to the extent, as in the PIH, that future income is
expected to rise through planned saving.) The
empirical estimation of wealth or permanent
income became a central issue in the specification
of the consumption function.
Friedman proposed a computationally simple
estimator of permanent income as a geometrically
weighted average of past income. Since on this
scheme, permanent income changes - besides
normal growth - by a fraction, say [1, of the dif-
ference between current income and permanent
income, Friedman related this scheme to the
adaptive-expectations approach recently introduced
by his student Phillip Cagan ( ).
Modigliani and his associates proxied normal
labour income by current income and the product of
this variable and the unemployment rate and
attempted to measure non-human wealth by
collecting estimates of the national balance sheet at
market values. In principle, this method seemed
more clearly related to the underlying framework
than Friedman’s permanent-income proxy, but in
practice it suffered several comparative disadvan-
tages: (1) major components of non-human wealth
had no market valuation; (2) the wealth estimates
were not part of the national income accounts and
competing variants were available with substantial
delay and at irregular intervals; (3) for forecasting
purposes, substantial additional equations were
required to forecast (often poorly) future
movements in wealth.
Darby ( ) reconciled these empirical mea
sures of wealth by demonstrating that under the
PIH, Friedman’s geometrically weighted measure
could be derived as the constant real interest rate fj
times a (backward-looking) perpetual inventory of
wealth. This [1 value was estimated as about
1. 10 per annum in contrast to higher values
such
2135
as Friedman’s 0.35 per annum. These higher values
were explained by biases that arise as data deviate
from pure consumption toward expenditures by
consumers.
Empirical work on consumption functions has
frequently floundered on the use of theories of pure
consumption to explain data which are in whole or
part consumer expenditures. Both the PIH and LCH
were theories of pure consumption. Modigliani and
Ando provided one link to consumer expenditures
in their MPC model by modelling household
investment in durable goods analogously to firm’s
investment behaviour. Operating in the PIH
tradition, Darby ( , ) argued that aggregate
transitory
income represented a change in wealth, part of
which change would be invested in consumers’
durable goods. (Darby ( ) in particular argued
that because transitory income is received in non-
human form, a disproportionate effect on durable-
goods purchases may arise during the adjustment
process, a result which explains the results of
Hayashi .) Darby ( , -8)
later combined pure consumption and durable
investment equations to obtain a unified consumer
expenditure function which avoided some of the
inherent difficulties in dividing consumer expen-
ditures into durable and nondurable portions.
The PIH and LCH thus evolved to explain
aggregate consumer expenditures by wealth as a
determinant of pure consumption and by changes in
wealth and other variables which determine
household investment in durable goods. The cor-
relation of the determinants of this household
investment with short-run (transitory) fluctuations
in income explain a MPC which is substantial in
magnitude even though substantially below the
APC.
This brief development has omitted discussion
of alternative views of the consumption function.
Perhaps the most notable of these is the view that
the substantial value of the MPC reflects liquidity
constraints which prevent a substantial share of
consumers (measured by wealth and consumption)
from following their optimal intertemporal
consumption plan. The author of this essay regards
these alternative views as providing qualification of
the dominant wealthbased view.
2136
Efficient-Market Approaches
Hall ( ) proposed to sidestep Friedman’s
backward-looking measure of wealth as well as the
substantial empirical problems involved in
measuring the market value of wealth. Instead, he
posed the question of whether or not changes in
consumption can be modelled empirically as
determined by ‘news’. Specifically, the assertion is
that if wealth estimates and hence consumption are
based on rational expectations, no past information
including past changes in consumption or income
should affect current changes in consumption.
Hall ( ) answered his question affirma-
tively, Flavin ( ) dissented, but Hayashi
( ) showed that excess sensitivity of spending
to changes in wealth appeared to be confined to
consumers’ durable goods purchases. Taken as a
whole, these studies seem to confirm the basic
Friedman-Modigliani conceptions that aggregate
consumption as determined by wealth but that it is
important to distinguish between consumer
expenditures and consumption.
Bequest Versus Life-Cycle Saving
Saying that consumers optimally allocate wealth
leaves several important questions unanswered: Do
consumers have operative linkages in utility
functions across generations? Are consumers able to
see through the veil of government to the ultimate
production possibilities faced by society? If the first
of these questions is answered affirmatively,
transfer programmes such as social security which
change the life-cycle pattern of income receipts will
not affect aggregate consumption and saving. (The
representative infinitely lived individual does not
care whether he or she pays social security taxes
which are refunded as equal benefits.
Intergenerational transfers can be adjusted so that
this representation is acceptable where utility
functions are linked across generations.) If the
second question is also answered affirmatively, then
Ricardian equivalence holds (it is irrelevant whether
government taxes or borrows) and the relevant
income concept for the
Consumption Function
aggregate consumption function is net national
product less government spending for goods and
services.
F eldstein ( ) claimed that aggregate saving
had been significantly reduced by the US social
security programme. As pointed out by Barro ( ),
this effect would not arise with
intergenerationally linked utility functions. Using
different methodology, White ( ) and
Darby ( ) concluded that life-cycle motives
accounted for an at most small fraction of aggregate
saving and wealth. Kotlikoff and Summers ( )
relaxed Darby’s assumptions on smooth
growth of population and labour income without
substantially changing the estimates on the range of
assets attributable to life-cycle motives. These
estimates seem to suggest that intergenerational
linkages are indeed very important, as assumed by
the PIH. The life-cycle effects highlighted in the
LCH would appear more important for analysing
cross-sectional data than as determinants of aggre-
gate consumption.
The Ricardian equivalence idea was urged by
Barro ( ) and Kochin ( ). It requires a
certain suspension of disbelief to assume that bonds
and taxes have equivalent effects on consumer
behaviour, but the data are not very inconsistent
with that notion. Indeed recent studies by S eater (
) and Kormendi ( ) provide some
evidence that Ricardian equivalence is a better
working hypothesis than its denial.
Conclusions
The consumption function suggested by Keynes
provided a useful challenge to theoretical and
empirical economists. The relationship between
changes in consumer expenditures and current
income has been explained generally in a way
which is consistent with microeconomic founda-
tions and which is adequate in a multitude of
specifications for most forecasting purposes.
(Technical differences among empirical specifica-
tions are as large in number as they are
uninteresting to the nonspecialist.) For policy ana-
lytic purposes, two key questions are outstanding:
are life-cycle effects significant in the aggregate,
Consumption Sets
and do individuals effectively see through gov-
ernment? This author’s reading of the evidence
suggests answers of no and maybe, but it is hard to
put much certainty in any answer unless one starts
with dogmatic priors.
The consumption function has faded as a topic
of intense research largely because of the success of
previous work in achieving a workable consensus.
The unsettled issues, however, have crucial policy
implications and there is much value yet to be
added.
See Also
► Consult
► .eyries’
► life Cyc
► .eal
► Relative
► Wealth
Bibliography
Barro, R.J. 1974. Are government bonds net wealth? Jour-
nal of Political Economy 82(6): 1095-1117.
Barro, R.J. 1978. Tile impact of social security on
private saving: Evidence from the U.S. time
series. Washington, DC: American Enterprise Institute.
Cagan, R 1956. The monetary dynamics of hyperinflation. In
Studies in the quantity theory of money, ed. M.
Friedman. Chicago: University of Chicago Press.
Darby, M.R. 1972. The allocation of transitory income among
consumers’ assets. American Economic Review
62(5): 928-941.
Darby, M.R. 1974. The permanent income theory of con-
sumption: A restatement. Quarterly Journal of Eco-
nomics 88(2): 228-250.
Darby, M.R. 1975. Postwar U.S. consumption, consumer
expenditures, and saving. American Economic
Review 65(2): 217-222.
Darby, M.R. 1977-8. The consumer expenditure function.
Explorations in Economic Research 4(5): 645-
674.
Darby, M.R. 1979. The effects of social security on
income and the capital stock. Washington, DC:
American Enterprise Institute.
Feldstein, M. 1974. Social security, induced retirement, and
aggregate capital accumulation. Journal of Political
Economy 82(5): 905-926.
Flavin, M.A. 1981. The adjustment of consumption to
changing expectations about future income. Journal of
Political Economy 89(5): 974-1009.
2137
Friedman, M. 1957. A theory of the consumption
function, NBER general series. Vol. 63. Princeton:
Princeton University Press.
Hall, R.E. 1978. Stochastic implications of the life cycle -
permanent income hypotheses: Theory and evidence.
Journal of Political Economy 86(6): 971-987.
Hansen, A.H. 1939. Economic progress and declining
population growth. American Economic Review 29:
1-15.
Hayashi, F. 1982. The permanent income hypothesis: esti-
mation and testing by instrumental variables. Journal
of Political Economy 90(5): 895-916.
Keynes, J.M. 1936. The general theory of employment,
interest, and money. New York: Harcourt, Brace, and
Co.
Kochin, L.A. 1974. Are future taxes anticipated by con-
sumers? Journal of Money’, Credit, and
Banking 6(3): 385-394.
Komiendi, R.C. 1983. Government debt, government
spending, and private sector behavior. American Eco-
nomic Review 73(5): 994-1010.
Kotlikoff, L.J., and L.H. Summers. 1981. The role of
intergenerational transfers in aggregate capital accumu-
lation. Journal of Political Economy 89(4): 706-
732.
Modigliani, F., and R. Brumberg. 1954. Utility analysis and
the consumption function: An interpretation of cross-
section data. In Post Keynesian economics, ed. K.E.
Kurihara. New Brunswick: Rutgers University Press.
Sargent, T.J. 1978. Rational expectations, econometric
exogeneity, and consumption. Journal of Political
Economy 86(4): 673-670.
Seater, J.J. 1982. Are future taxes discounted? Journal of
Money, Credit, and Banking 14(3): 376-389.
White, B.B. 1978. Empirical tests of the life cycle hypothesis.
American Economic Review 68(4): 647-660.
Consumption Sets
Peter Newman
JEL Classifications
E2
The idea of consumption sets was introduced into
general equilibrium theory in July 1954 in Arrow
and Debreu ( , pp. 268-9) and Debreu ( ,
p.588), the name itself appearing only in the latter
paper. Later expositions were given by Debreu (
) and Arrow and Hahn ( ) and a more
2138
general discussion by Koopmans ( , Essay 1).
Although there have been several articles concerned
with nonconvex consumption sets (e.g. Yamazaki ),
in more recent years their role in general
equilibrium theory has been muted, especially in
approaches that use global analysis (see for
example, Mas-Colell , p. 69). Such sets play no role
in partial equilibrium theories of consumer’s
demand, even in such modem treatments as Deaton
and Muellbauer ( ). Since
general equilibrium theory prides itself on precision
and rigour (e.g. Debreu , p. x), it is odd that on close
examination the meaning of consumption sets
becomes unclear. Indeed, three quite different
meanings can be distinguished within the various
definitions presented in the literature. These are
given below (in each case the containing set is the
commodity space, usually Rn): Ml The consumption
set Cl is that subset on which the individual’s
preferences are defined. M2 The consumption set
C2 is that subset delimited by a natural bound on
the individual’s supply of labour services, i.e. 24
horns a day. M3 The consumption set C3 is the
subset of all those bundles, the consumption of any
one of which would permit the individual to
survive. Each definition in the literature can (but
here will not) be classified according to which of
these meanings it includes. In probably the best
known of them (Debreu , eh. 4), the consumption
set appears to be the intersection of all three subsets
C1-C3. Ml is plain. After all, preferences have to be
defined on some proper subset of the commodity
space, since the whole space includes bundles with
some inadmissibly negative coordinates. M2 is also
reasonable, although a full treatment of
heterogeneous labour services does raise problems
for what is meant by an Arrow-Debreu
‘commodity’ (see for example, that of Arrow-
Hahn , pp. 75-6). It is M3 that gives real difficulty,
both in itself and in relation to the others.
First, there is little reason to expect either Cl or
C3 to be a subset of the other, and so still less to
expect Ml and M3 to define the same set. No
individual would have any problem in preferring
one bundle, the consumption of which would ensure
her survival, to a second bundle, the
Consumption Sets
consumption of which would result in her death by
starvation. However, she might well prefer the
second bundle to a third, whose consumption would
cause her to die from thirst (the representation of
such preferences by a real-valued utility function
might pose problems, but that is another matter). On
the other hand, the same individual might not be
able to rank in order of preference two bundles each
of which contains exotic food and drink, even
though fully assured that the consumption of either
bundle would allow her to survive.
More importantly, M3 implicitly introduces
consumption activities, the actual eating and
drinking and sheltering that are essential to survival.
Such activities constitute what are sometimes
called, by analogy with production, the
consumption technology. Some partial equilibrium
models, such as ‘the new home economics’ and the
theory of characteristics, have treated aspects of
such technologies but so far general equilibrium
theory has not. In particular, Arrow-Debreu theory
has not done so. As a consequence (and unlike some
forms of the classical ‘com model’) it does not give
a coherent account of the birth and death of
individual persons, any more than it does of the
birth and death of individual firms (see general
equilibrium). Hence the third meaning M3, which in
effect presumes that the model contains such an
account when it does not, is hard to interpret. One
major difficulty of interpretation arises with the
Slater-like condition that each individual’s
endowment of goods and services, valued at the
competitive prices p , should be strictly greater than
inf {(x, p ): x € C}, where < .,. > denotes inner
product and C is ‘the’ consumption set (see cost
minirnization and utility maximization). This
condition is important in proofs of existence of
competitive equilibrium, to ensure for example that
the budget correspondence is continuous, or that a
compensated equilibrium is a competitive
equilibrium. It is itself guaranteed by assumptions
(discussed by McKenzie , pp. 821-5) on the
relations between ‘individual’ consumption sets and
the aggregate production set.
If C is taken to contain C3 then the assumptions
just referred to imply that every consumer
Consumption Sets
survives in every competitive equilibrium, not
merely for one period but over the whole (finite)
Arrow-Debreu span. This is a breathtaking assertion
of fact which recalls irresistibly Hicks’s wry
observation: ‘Pure economics has a remarkable way
of producing rabbits out of a hat - apparently a
priori propositions which apparently refer to
reality. It is fascinating to try to discover how the
rabbits got in’( , p. 23).
On the other hand if C is taken to be Cl, then the
assumptions take on a purely technical (and so less
objectionable) aspect, whose role is essentially to
ensure that the system stays within the (relative)
interior of the sets concerned and so displays
appropriate continuity. But then there is no
presumption that individual agents survive in a
competitive equilibrium, even for one period (cf.
Robinson , p. 3). The multi-period versions of the
Arrow-Debreu model are then at risk, since
individuals disappear and take their labour service
endowments with them. This should not come as a
surprise - the problems of time in economics are
really too complicated to be overcome simply by
adding more dimensions to the one-period model.
Some models that include C3 in C attempt to
justify Slater-like conditions directly, on the
grounds that ‘ Not many economies in the present
day are so extremely laissez faire as to permit
people to starve’ (Gale and Mas- Colell , p. 12).
This justification clearly fails as long as the
behaviour of the public agency whose actions
allegedly prevent such starvation is not modelled
explicitly, like that of the private agents.
It is usually assumed that consumption sets are
bounded below, closed and convex. The first two
assumptions are innocuous but the third poses
issues of a conceptual kind, which spring from
difficulties in interpreting the idea of a convex
combination x' = tx 1 + (1 - t)x2 of two bundles A 1
and x2, where fc[0, 1], Consider the example,
sometimes used, in which A 1 is a house in London
and x2 a house in Paris. We cannot take seriously
the claim that A-' is a house in the Channel, so t
cannot refer to distance. An alternative claim that t
refers to the proportion of the period that is spent in
London could arise from many different
2139
finite partitions of the time interval, not all of which
need to be ranked equally by the individual. In
effect, convexity of the consumption set comes
down to the divisibility of consumer goods, an
assumption which in the past has proved not such a
bad approximation if one is interested mainly in
general equilibrium aspects of market demand, and
representative rather than actual consumers.
Indivisibilities of producer goods are of course
much more serious.
See Also
► Arrow-Debreu Model of General Equilibrium
► Cost Minimization and Utility Maximization
► General Equilibrium
► Divisibilities
Bibliography
Arrow, K.J., and G. Debreu. 1954. Existence of an equi-
librium for a competitive economy. Econometrica 22:
265-290.
Arrow, K.J., and F.H. Hahn. 1971. General competitive
analysis. San Francisco: Holden-Day.
Deaton, A., and J. Muellbauer. 1980. Economics and con-
sumer behaviour. Cambridge: Cambridge University
Press.
Debreu, G. 1954. Valuation equilibrium and Pareto optimum.
Proceedings of the National Academy of Sci-
ences 40 (7): 588-592.
Debreu, G. 1959. Theory of value, Cowles commission
monograph no. 17. New York: Wiley.
Gale, D., and A. Mas-Colell. 1975. An equilibrium existence
theorem for a general model without ordered preferences.
Journal of Mathematical Economics 2: 9-15.
Hicks, J.R. 1939. Value and capital. Oxford: Clarendon
Press.
Koopmans, T.C. 1957. Three essays on the state of eco-
nomic science. New York: McGraw-Hill.
McKenzie, L.W. 1981. The classical theorem on existence of
competitive equilibrium. Econometrica 49: 819-841.
Mas-Colell, A. 1985. The theory of general economic
equilibrium. A differentiable approach.
Cambridge: Cambridge University Press.
Robinson, J.V. 1962. The basic theory of normal prices.
Quarterly Journal of Economics 16 (1): 1-20.
Yamazaki, A. 1978. An equilibrium existence theorem without
convexity assumptions. Econometrica 46: 541-555.
2140
Consumption Taxation
James M. Poterba
Abstract
Whether to tax households based on their
income or on their consumption is one of the
central and long-standing questions of tax
design. Most developed nations rely on a com-
bination of income and consumption taxes to
raise revenue. The debate over alternative tax
bases involves both philosophical arguments
about what constitutes a fair measure of ability
to pay and economic arguments about the rel-
ative efficiency of different tax bases. Con-
sumption taxes can be implemented in a variety
of ways, including value added taxes, retail sales
taxes, and savings-exempt income taxes.
Keywords
Capital gains taxation; Consumption taxation;
Distortionary taxation; Flat rate tax; Individual
Retirement Accounts (USA); Progressive and
regressive taxation; Redistribution; Retail sales
tax; Savings-Exempt income tax; Tax compli-
ance; Taxation of income; Value added tax
JEL Classifications
H2
Whether household income or household con-
sumption constitutes a better measure of a house-
hold’s ability to pay taxes, and whether there are
substantial efficiency gains to choosing one tax
base rather than the other, are two of the central
questions of public finance. The debate between
advocates of income taxes and advocates of con-
sumption taxes has spanned several centuries.
While income has often been viewed as the basis
for taxation, and Adam Smith discusses taxation
relative to household incomes, Thomas Hobbes,
John Stuart Mill and Irving Fisher were all strong
proponents of taxing consumption. Consumption
Consumption Taxation
tax supporters argue that the amount that an indi-
vidual draws from the economy’s resource pool
should determine his or her tax burden. They also
point out that an income tax levies a ‘double tax’ on
saving, since saved income is taxed both when it is
earned and when the savings yield a return to
capital. Kaldor ( ) offers a broad review of the
case for consumption taxation. Two notable reports
in the late 1970s, one by the Meade Commission
(Meade ) in the United Kingdom and the other by
the staff of the US Treasury Department ( ),
outlined the modem cases for con
sumption taxation and developed specific proposals.
Proponents of income taxation argue that the
change in an individual’s command over resources
between one period and the next is an appropriate
measure of ‘ability to pay’, even if those resources
are not immediately consumed. This is the measure
of taxable capacity suggested by Robert Murray
Haig and Henry Simons: ‘Haig-Simons’ income.
Moreover, they argue that changes in resources
should be taxed regardless of whether they arise
from labour income or from the returns to past
saving.
Income taxes and consumption taxes exhibit
different time profiles over the course of a lifetime.
When individuals experience a period of retirement
before they die, the time profile of tax payments
under a consumption tax will fall later in the
lifetime than the corresponding payments under an
income tax. This is because individuals continue to
consume after they stop earning labour income.
Retirees under an income tax pay tax only on their
capital income, while retirees under a consumption
tax pay tax on their total outlays, which are likely to
exceed their capital income.
The debate between proponents of consumption
taxation and proponents of income taxation
concerns whether or not capital income should be
taxed. The foregoing philosophical issues not-
withstanding, the efficiency cost of taxing capital
income has been an active subject of economic
research. Chamley ( ) and Judd ( ) argue
that the effective distortions from capital taxes
cumulate over time as the difference between
discounting the future at before-tax and after-
Consumption Taxation 2141

tax interest rates increases with the compounding ^C,/(l+r(l -T))'

horizon. They claim that the optimal steady-state ' =1 T CD
capital income tax rate should be zero. However,
= 5^(1 - T)W,L,/(1 +r(l - x ))'+A 0
they also point out that a one-time capital levy is an i=i
efficient device for raising revenue. A number of
recent studies, described in Auerbach ( ), In this expression, C denotes real consumption
have examined the robust spending, and A0 is the household’s initial wealth
ness of the theoretical claim that the optimal capital endowment.
tax rate is zero. In contrast, the life-cycle budget constraint with
Consumption tax proponents, such as Bradford ( a consumption tax levied at rate 0 is
), claim not only that taxing consump
tion rather than income avoids intertemporal ]T(i + 0)c,/(i + /-)'
distortions, but also that it solves many of the most t=\
T
difficult measurement and accounting problems = w,L,/(1 + r)' + Ao (2)
associated with income taxation. Under a t=\
consumption tax, for example, there would be no
The discount rate in this case is the pre-tax return.
distinction between the tax burden on investment
The consumption tax levied on outlays in each
projects financed with debt and those financed with
period is equivalent to a tax on labour income and
equity, or between realized and unrealized capital
the household’s initial endowment. If (1 — v) =
gains. There would be no need to measure the rate
1/(1 + 0), then Eq. can be rewritten as.
at which long-lived physical assets depreciate, as
one must do under an income tax. Income tax
proponents respond that some components of
^C,/(i +r)' = 5^(1 -v)w,L,/(l + r ) ‘
consumption may be difficult to measure, and that it t= 1 t=\
is more difficult to tailor consumption taxes than + (1 - v)A0 (3)
income taxes to achieve redistributive goals.
The timing of tax payments under the ‘wage-and-
endowment tax’ in ( ) is different from that under
Formalizing Consumption the consumption outlays tax in ( ), but the present
Taxation Vs. Income Taxation value of taxes and the effects on economic incen-
tives are the same under the two systems. The tax
The essential difference between a consumption tax
on initial endowment is an essential component of
and an income tax can be illustrated by comparing
this equivalence: a wage tax alone is not equivalent
the lifetime budget constraints that consumers
to a consumption tax because initial assets escape
would face under each tax system. An income tax is
taxation when only wages are taxed.
levied on both labour and capital income. When a
The current tax system in most developed
household has assets of At _ i at the beginning of
nations is a hybrid structure, reflecting some ele-
period t, these assets earn a pretax return r and the
ments of income taxation but also embodying
household earns labour income of wL where w
components of a consumption tax. This is most
equals the real wage and L denotes labour supply,
apparent in nations that rely on both an income tax
the income tax base is wL + rAt _ i. The income tax
and a consumption tax, such as a value added tax,
not only reduces the after-tax real wage but also
for a substantial share of government revenue. Even
lowers the after-tax return to saving. In a life-cycle
within many income tax systems, however, there
model in which a household lives for Tperiods and
are provisions that move toward an income
in which there is no inflation, the life-cycle budget
constraint with an income tax is
2142
tax-consumption tax hybrid. In the United States,
for example, capital income that accrues in
employer-provided pension plans and in a variety of
taxpayer-directed retirement saving accounts, such
as Individual Retirement Accounts (IRAs), is
excluded from income taxation. Some types of
capital income are taxed at rates below the top
statutory tax rates on wage income. Realized capital
gains have often been taxed at preferential rates,
and in some cases dividend income to households is
also subject to reduced rates of tax. There is
substantial variation in tax structures across nations,
but the principle of allowing some tax reduction on
capital income is widespread. This makes it difficult
to assess where any particular nation’s tax system
falls on the spectrum between an income tax and a
consumption tax.
Types of Consumption Taxes
In practice, there are many ways to implement a
consumption tax. Two, the retail sales tax and the
value added tax, are widely used in practice. Both
are examples of indirect consumption taxes,
because they are levied without any reference to the
consumer’s identity. Direct consumption taxes, in
contrast, are levied on households by computing
their total consumption. In contrast to indirect
consumption taxes, direct consumption taxes can be
levied at progressive rates. While direct
consumption taxes have never been used as the
primary revenue source in any nation, they have
been actively debated in the policy reform literature.
Tax structures that closely resemble direct
consumption taxes have been adopted as
components of existing tax systems. The two most
widely discussed direct consumption tax options are
the savings-exempt income tax and the ‘X-tax,’ a
combination of a cash-flow tax on business income
and a household wage tax.
A retail sales tax (RST) is the simplest con-
sumption tax. It is collected by retailers at the point
of final sale, and it corresponds directly to the tax
on consumption spending described in Eq. above. In
2006,44 of the 50 US states levied some form of
sales tax, with rates typically between four and
seven per cent. There is little
Consumption Taxation
experience with RSTs above ten per cent. One
unresolved question with regard to proposals that
call for significantly higher RSTs is whether the
difficulty of monitoring all points of purchase
would lead to substantial problems of tax evasion.
A value added tax (VAT) is a very common
form of consumption tax. Virtually all developed
nations with the exception of the United States levy
some form of VAT, with rates ranging up to 25 per
cent in Denmark, Norway and Sweden. The VAT is
collected from businesses on the difference between
the gross value of their sales and the cost of any
inputs that they purchase from other entities that
have already paid VAT.
To illustrate the operation of VAT, consider a
bakery that produces and sells bread for $ 100. The
baker’s input costs are $30 for flour and $65 for an
employee. The bakery earns a $5 profit. If flour is
purchased from another firm that has already paid
VAT, then the bakery’s VAT liability equals $70
times the VAT tax rate, since its value added equals
its sales of $100 minus input purchases that have
already paid VAT, or $30. Wages are not deducted
from sales when computing value added. Although
the VAT is collected in stages from all firms in a
production chain, it is equivalent to an RST at the
same rate. One attractive feature of the VAT is that
downstream firms, such as the baker in this
example, help ensure VAT compliance by upstream
firms that supply intermediate goods. In this
example if the flour seller cannot provide
documentation for its VAT payment, the baker will
face tax on value added of $100. Thus the baker has
an incentive, all else equal, to purchase inputs from
suppliers who pay VAT.
Ebrill et al. ( ) offer a comprehensive dis
cussion of VAT implementation issues and sum-
marize experience with the VAT in both developed
and developing nations. The VAT accounts for a
substantial share of revenue in most industrialized
nations. The treatment of international transactions
has proven a source of difficulty in some nations,
since exporting firms are typically granted a rebate
for their VAT payments. Some tax evasion schemes
involve exporting goods to qualify for the rebate
and re-importing the same goods without paying
VAT on the import. The
Consumption Taxation
taxation of financial services also proves challeng-
ing under the VAT.
A savings-exempt income tax (SEIT) is a con-
sumption tax that is built on an income tax model.
For those who are familiar with an income tax
system, it provides a way of shifting to a con-
sumption tax without drastic administrative changes
in the tax system. The Nunn-Domenici ‘USA Tax’,
introduced in the US Senate in the mid-1990s and
analysed in Ginsburg ( ), was
a based on this type of consumption tax.
Under the SEIT, the tax base is income less
saving. To prevent taxpayers from simply claiming
high levels of saving and thereby avoiding tax
liability, saving must be documented in the form of
a contribution to a ‘qualified account’. Income
earned on assets held in the qualified account is not
taxed, but withdrawals from the qualified account
are included in the tax base. Thus a taxpayer who
earns $50,000 and contributes $5,000 would be
taxed on $45,000 in the contribution period. If,
some years later, when earnings equal $25,000, the
taxpayer withdraws $10,000 from the qualified
account, she would be taxed on $35,000.
Even though the SEIT taxes the earnings that
have accrued on the contributions to the qualified
account when the fluids are withdrawn from this
account, the return on capital is untaxed in this
setting. Taxing accumulated capital income when
the proceeds are withdrawn is not equivalent to
taxing capital income as it accrues: this is the reason
Individual Retirement Accounts, 401(A) plans and
other tax-deferred saving programmes provide an
incentive for personal saving. When capital income
is taxed as it accrues, the value of earning one
dollar, paying tax on it at rate T, and then investing
it for T periods at a pretax rate of return r but with
an accrual tax rate T, is (1 — T) (1 + (1 - T)rf. In
contrast, if the initial earnings are excluded from
taxation, there is no taxation of accruing capital
income, and withdrawals are taxed at 100t per cent,
then the value after T periods is (1 - T)(1 + r)T. The
qualified account approach eliminates the tax
burden on the ‘inside build up’ of capital assets.
One of the key challenges in implementing a SEIT
is avoiding the wholesale reallocation of
2143
existing wealth into ‘qualified accounts’ at the time
the SEIT is adopted. Such transfers could sharply
reduce tax collections, but, since they involve
previously accumulated assets, they would not
translate into marginal incentives for new saving. If
it were possible to inventory the assets of each
taxpayer when the SEIT was implemented, this
would make it possible to design regulations to limit
the transfer problem. Absent such information on
previously accumulated wealth, however, transfers
of pre-existing wealth into qualified accounts are
likely to prove a difficult implementation issue for
the savings-exempt income tax.
An X-tax combines a cash flow tax on busi-
nesses, much like a VAT with a deduction for
wages, with a household-level tax on wage income.
The X-tax and its relatives are descended from
proposals in the US Treasury Department’s ( )
report on fundamental tax reform. Bradford ( )
discusses several plans of this type,
and one widely discussed variant was developed by
Hall and Rabushka ( ). The X-tax has
greater flexibility than a VAT for achieving distri-
butional goals, since the household level tax can
include progressive rates or transfers to low-
earning households. This illustrates the distribu-
tional flexibility of direct rather than indirect con-
sumption taxes. If the household tax is a flat rate tax
on wages at the same rate as the corporate cash flow
tax, then the X-tax is equivalent to a VAT or an
RST. When the rates are different, then the X-tax
becomes a combination of a VAT and an additional
tax or subsidy on labour income. The cash flow
nature of the business tax eliminates the need to
measure depreciation, since firms can claim an
immediate deduction - expensing - for purchases of
capital goods.
In practice, neither the RST nor the VAT is
implemented strictly along the principles described
above. Proposals for both the SEIT and the X-tax
also include additional features that often introduce
efficiency costs that would not arise in ‘textbook’
versions of these taxes. The RST, for example,
typically exempts some goods and services.
Expenditures on food, medical care and clothing are
often excluded from the tax base, thereby achieving
a more progressive distribution of tax burdens while
creating
2144
distortions between various classes of consumption
goods. The VAT is often implemented at different
rates on different goods, with exemptions for some
goods, creating the same distortionary effects.
Because both the savings- exempt income tax and
the X-tax require households to file tax returns, they
are prone to modification to allow deductions for
some expenditure categories, such as mortgage
interest or health insurance premiums. While
neither of these consumption tax plans has been
tried in practice, they probably would be influenced
by the same political pressures that have generated a
wide array of tax expenditures in the current income
tax code.
Efficiency Gains from Replacing an
Income Tax with a Consumption Tax
Income taxes create two distortions: one between
the before-tax and the after-tax real product wage,
which distorts the labour-leisure margin, and one
between the beforetax and the after-tax real rate of
return to saving. The latter distorts the lifetime
allocation of consumption relative to the pattern that
would be chosen if the return to delaying
consumption equalled the economy’s pre-tax mar-
ginal product of capital. Shifting from an income
tax to a consumption tax eliminates the second
distortion. The key analytical issue in evaluating the
welfare consequences of replacing an income tax
with a consumption tax is therefore measuring the
efficiency costs associated with the taxation of
saving and investment. This efficiency cost depends
on the underlying structure of consumer
preferences. The interest elasticity of saving is often
invoked as a summary measure of the key
preference parameters. When changes in after-tax
returns induce only modest changes in household
saving, the efficiency gain from switching from an
income tax to a consumption tax will be smaller
than when the interest elasticity of saving is large.
Auerbach and Kotlikoff ( ) use a dynamic
general equilibrium model, including a realistic
treatment of household life-cycle income and con-
sumption streams, to evaluate the efficiency gains
Consumption Taxation
from replacing an income tax with a consumption
tax. Their results suggest that for a given revenue
requirement, the steady-state capital stock is larger
with a consumption tax than with an income tax.
This translates into higher steady-state per capita
utility under the consumption tax than the income
tax.
The steady-state comparison is not the only
consideration when evaluating two alternative tax
systems, however. It is possible to design tax
reforms that raise steady-state welfare but cause
welfare losses in the transition from an initial
equilibrium to the new steady state. The tradeoff
between short-run and long-run policy effects
depends on the policymaker’s discount rate and in
calibrated general equilibrium models it is possible
to compute the present discounted value of the gains
and losses to the cohorts alive at different dates.
Transition from One Tax Regime to
Another
Focusing on the present value of welfare gains and
losses draws attention to the transitional rules that
govern the switch from one tax system, say an
income tax, to another, such as a consumption tax.
These transition rules can determine whether a
policy reform represents a net gain or a net loss
relative to continuation of the initial income tax
regime. Altig et al. ( ) illustrate this important
point using a more elaborate version of the model
developed in Auerbach and Kotlikoff ( ).
They find that if the tax basis of existing assets is
extinguished when the income tax is replaced by a
consumption tax, so that depreciation allowances
are no longer claimed after the reform, and if
investors who accumulated savings under the
income tax regime do not receive any relief from
the consumption tax burden they will face when
they draw down their assets, then the efficiency
gains from adopting a consumption tax may be as
large as five per cent of national income.
‘Grandfathering’ existing assets sharply reduces
these efficiency gains, because it reduces the base
of the consumption tax and requires
Consumption-Based Asset Pricing Models (Empirical Performance) 2145

higher tax rates to satisfy a given revenue con-

straint. This results in greater distortions on the Consumption-Based Asset Pricing
labour-leisure margin. Designing transition relief Models (Empirical Performance)
that participants in the political process will view as
fair, without forgoing most of the efficiency gains Fatih Guvenen and Hanno Lustig
from a stark consumption tax transition, is likely to
be one of the greatest challenges in any
consumption-oriented tax reform.
Abstract
Asset pricing is a branch of financial economics
that is rich in puzzles and anomalies - that is,
See Also
stylized empirical facts not easily explained by
► ax Expen the canonical asset pricing models. These range
► axation o from the equity premium puzzle and the risk-
► ;alu( Add free rate puzzle to the fact that stock returns are
highly predictable. This article discusses
different consumption-based asset pricing
Bibliography models that have been developed to resolve
Altig, D., A.J. Auerbach, L.J. KotUkofif, K.A. Smetters, and
these puzzles, and it evaluates their empirical
J. Walliser. 2001. Simulating fundamental tax reform in performance.
the United States. American Economic Review 91: 574-
595.
Keywords
Auerbach, A.J. 2006. The choice between income and
consumption taxes: A primer. Working Paper Capital asset pricing model; Consumption-
No. 12307. Cambridge, MA: NBER. based asset pricing models; Elasticity of
Auerbach, A.J., and L.J. Kotlikoff. 1987. Dynamic fiscal intertemporal substitution; Equity premium
policy. Cambridge: Cambridge University Press.
Bradford, D. 1980. The case for a personal consumption tax.
puzzle; External habit; Habit formation;
In What should be taxed? Income or expenditure, ed. Heteroskedasticity; Imperfect risk sharing;
J. Pechman. Washington, DC: Brookings Institution. Incomplete markets; Precautionary savings;
Bradford, D. 1986. Untangling the income tax. Cambridge, Real business cycles; Recursive preferences;
MA: Harvard University Press.
Chamley, C. 1986. Optimal taxation of capital income in
Representative agent; Risk aversion; Riskfree
general equilibrium with infinite lives. Econometrica 54: rate puzzle; Stochastic discount factor
607-622.
Ebrill, L., M. Keen, J.-P. Bodin, and V. Summers. 2001. The
modern VAT. Washington, DC: International Monetary
Fund. JEL Classifications
Ginsburg, M. 1995. Some thoughts on working, saving, and D4; D10; G12
consuming in Nunn-Domenici’s tax world. National Tax
Journal 48: 585-602. The aim of consumption-based asset pricing models
Hall, R., and A. Rabushka. 1995. The flat tax. 2nd ed.
is to explain a number of important and puzzling
Stanford: Hoover Institution Press.
Judd, K.L. 1985. Redistributive taxation in a simple perfect features of asset returns using standard economic
foresight model. Journal of Public Economics 29: 59-83. theory. Perhaps the best-known challenge for these
Kaldor, N. 1955. An expenditure tax. London: George Allen models is the equity premium puzzle. Let us start
& Unwin. from the Euler equations for stock and bond choice,
Meade, J.E. 1978. The structure and reform of direct taxation.
London: Allen & Unwin.
and let us assume that both of these Euler equations
US Treasury Department. 1977. Blueprints for basic tax hold with equality. If agents have constant relative
reform. Washington, DC: US Treasury Department. risk aversion (CRRA ) preferences and if returns
and consumption growth are jointly log-normal,
then the
2146 Consumption-Based Asset Pricing Models (Empirical Performance)

Sharpe ratio (that is, the equity premium per unit of because these households are extremely unwilling
risk) can be decomposed as: to substitute consumption over time. As a result,
they desire a flat consumption profile and, therefore,
would like to transfer resources from the future to
\ max std(Ac) x corr(Ac,R ), (1)
e
e
std(R ) y y
’ today. But since this is not feasible in an
endowment economy, the equilibrium risk-free rate
where R is the excess return on stocks over bonds, needs to be very high to discourage this type of
e

a is the relative risk aversion (RRA) parameter, and consumption smoothing and make individuals
Ac denotes log consumption growth. The equity willing to consume their endowment every period.
premium is about 6 per cent per year in the US data The last term in ( ) captures the precautionary
with a standard deviation of 15 per cent, producing savings motive, which becomes active in the pres-
e e
a Sharpe ratio (E(R )/ std(R )) of 0.4. Mehra and ence of uncertainty. For very high levels of risk
Prescott ( ) used aversion, this effect dominates the intertemporal
the construct of a representative agent who substitution effect, and an increase in the RRA
consumes the aggregate endowment stream. coefficient reduces the risk-free rate. Epstein and
Constantinides ( ), Rubinstein ( ) and Zin ( ) developed a class of reclusive prefer
Wilson ( ) derived aggregation results that ences that disentangles the inverse of the elasticity
rely on either complete markets or the absence of of intertemporal substitution from the coefficient of
idiosyncratic income risk. By appealing to these risk aversion. As discussed below, these preferences
aggregation results, Mehra and Prescott could allow one to make progress on the equity premium
substitute per-capita consumption growth into ( ). puzzle without running into the risk-free rate
This series has a standard deviation of less than 2 puzzle.
per cent in the post-war US data, and a low Against the backdrop of Mehra and Prescott’s
correlation with stock returns - less than 0.25 by benchmark model, subsequent papers that attempt to
most estimates. Substituting these values into the resolve these puzzles can be categorized according
expression above implies a lower bound for the to whether they modify (i) the preferences, (ii) the
relative risk aversion coefficient of 80, which is endowment process, or (iii) the market and asset
implausibly high judging by its implications for an structure. We discuss each of these approaches in
individual’s choices in other settings. In other turn.
words, we need extremely high risk aversion to
rationalize the observed equity premium, and that is
the puzzle. Furthermore, even if one is willing to
The Utility Function
accept such a high coefficient of risk aversion, this
Recursive Preferences
choice creates different puzzles itself - a point first
In the case of CRRA utility, the stochastic discount
noted by Weil ( ).
factor (SDF) has the following form: M, ,+i =
To understand Weil’s ‘risk-free rate puzzle’,
P(Ct+ilCt)~a, where C denotes the level of
first note that the Euler equation for the risk-free
consumption. A drawback of this specification is
asset choice can be linearized to obtain:
that it restricts the elasticity of intertemporal
substitution (EIS) to be the reciprocal of the RRA
a-
E[R^] m —In [i + aE(Ac) — —var(Ac). (2 ) parameter when in fact these two parameters
capture conceptually distinct aspects of individuals’
preferences. Building on work by Kreps and Porteus
Let us assume a positive time discount rate (ft < 1 ),
( ), Epstein and
and an average consumption growth rate of 1.5 per
Zin ( ) and Weil ( ) introduced ‘recur
cent per year. Let us also abstract from uncertainty
sive preferences’ (also called ‘non-expected
for the moment. Then a risk aversion of 40 would
utility’):
imply an implausibly high interest rate of nearly 60
per cent per year simply
Consumption-Based Asset Pricing Models (Empirical Performance)
(1 - P)C? + faiuttY11-*' ih
u, (3)
where a is still the RRA parameter, but now the EIS
is captured by a separate parameter: 1/(1 —p). In
this case, the SDF is given by:
where y = dp, and Rf is the total return on the
investors’ wealth portfolio (including human capital
which must be tradable for this representation to be
derived; see Epstein and Zin and Weil ). An
appealing feature of this SDF is that it combines
two components that are each central to separate
asset pricing theories: in particular, the SDF is a
geometric average of consumption growth and the
market return, where the latter is the relevant SDF
in the standard capital asset pricing model (CAPM).
Moreover, when a = 0 (logarithmic risk
preferences), then the CAPM emerges as a special
case whereas a p reduces it to the standard case of
expected utility (see Epstein and Zin ; Campbell ).
In addition, this preference specification is
flexible enough to allow a choice of a coefficient of
relative risk aversion that is high enough to match
the equity premium without being forced to accept
a very low EIS. The low EIS is responsible for the
risk-free rate puzzle, as explained above. Bansal
and Yaron ( ) exploit this
agent’s concern for long-run consumption risk by
introducing a small predictable component in
consumption growth.
Habit Formation and Catching-Up
with the Joneses
Another approach, pioneered by Simdaresan ( ), economic environments that give rise to SDFs
Abel ( ) and Constantinides ( ), similar to those resulting from external habit
starts from the following specification of the inves- preferences (such as the one used in CC). Examples
tor’s preferences over consumption streams G:
‘ 1—a
where Xt is some function of either (i) the individ-
ual’s own past consumption or (ii) the past
2147
consumption of a reference group, such as an
individual’s peers, neighbours, or the population as
a whole. Abel’s specification features the ratio of G
to Xt instead of the level difference. The first
approach allows an individual’s marginal utility to
depend on her own past consumption history. This
is commonly referred to as habit formation,
endogenous habit, or internal habit. The second
interpretation allows an individual’s utility to
depend on her status relative to her peers, neigh-
bours or the population as a whole. This is referred
to as catching-up with the Joneses or as external
habit. These preference specifications amplify the
effect of consumption growth shocks on the mar-
ginal utility growth of investors, in turn generating a
high equity premium.
A particularly successful version of the
catching-up-with-the-Joneses specification was
developed by Campbell and Cochrane ( )
(henceforth CC) who choose the sensitivity of X to
consumption growth shocks to match the
conditional and unconditional moments of returns.
In the baseline CC model, aggregate consumption
and dividend growth are i.i.d. over time. Menzly et
al. ( ) introduce additional cash
flow dynamics to explain the time series and cross-
section of stock returns, while Santos and Veronesi
( ) emphasize the importance of
labour income share variation to understand time
variation in risk premia. Wachter ( ) applies a
version of the CC model to the term structure, while
Verdelhan ( ) uses the same model to
explain the forward premium puzzle.
Looks Like Habit
Several recent papers have proposed models with
standard preferences (such as CRRA) but consider
include work by Piazzesi et al. ( )
who introduce housing services consumption into
this framework, and by Yogo ( ) who con
siders durable consumption broadly defined,
building on earlier work by Dunn and Singleton ( )
and Eichenbaum and Hansen ( ).
Finally, Guvenen ( ) studies a model with
limited stock market participation and shows that
2148 Consumption-Based Asset Pricing Models (Empirical Performance)
while the asset pricing implications of his model are
similar to those in CC, the implications for
macroeconomic questions (such as policy analysis,
and so on) are quite different.
Additional Arguments in the Utility Function
The models discussed so far assume that investors
only derive utility from non-durable consumption.
In exchange economy models (in which the con-
sumption process is exogenous) this is equivalent to
assuming that non-durable consumption enters the
utility function in a separable manner. Some recent
papers explicitly model the utility flow from
housing consumption (in a non-separable manner),
and find that such an extension improves the asset
pricing performance (see Grossman and Laroque ;
Piazzesi et al. ; Flavin and Yamashita ). Similarly, a
labour-leisure choice was introduced by Boldrin et
al. ( ) These asset pricing puzzles have also attracted a lot
and Danthine and Donaldson ( ), in a repre-
sentative agent framework, and by Uhlig ( ) basic framework used in Mehra and Prescott ( )
in an incomplete markets framework. However,
these authors find that this extension negatively
affects the performance of asset pricing models,
because it allows households to smooth their mar-
ginal utility by adjusting on the labour-leisure
margin. As a result, one needs to introduce
additional - typically labour market - frictions to
counteract this new smoothing opportunity'.
Consumption Dynamics
In consumption-based asset pricing models, it is
common to assume that aggregate consumption
growth is i.i.d. overtime, because the evidence for
consumption growth predictability in the data is
weak. In the i.i.d. case, the conditional market price
of risk, which can be approximated by the
conditional standard deviation of the log SDF,
at(logMt t-i) = a x <7,(Ac), is constant. Therefore,
these models cannot generate any time variation in
risk premia on equity or any other asset.
In the context of a standard representative agent
model, Kandel and Stambaugh ( ) gen
erate time-variation in risk premia by introducing
heteroskedasticity in aggregate consumption
growth. Bansal and Yaron ( ) deviate from
the i.i.d. assumption by introducing a small pre-
dictable component in consumption growth that is
statistically hard to detect. This long-run component
increases the market price of consumption risk. In
addition, they add some time variation in the size of
the long-run risk component. Colacito and Croce (
) show these long-run risk
models can reconcile the low volatility of exchange
rate changes with the large market price of risk.
Finally, Longstaff and Piazzesi ( ) argue that
corporate earnings are much
more risky than aggregate consumption growth, and
that this can account for a large share of the equity
premium puzzle.
Production Economy Models
attention from macroeconomists because the same
also forms the backbone of the Kydland
and Prescott ( ) model and the subsequent real
business cycle literature. Therefore, understanding
why individuals dislike risk in financial markets
could help shed light on individuals’ perceptions of
macro risk and consumption fluctuations, which are
key issues for macroeconomic policy. However,
macroeconomists are also interested in the
determination of quantities, such as output,
investment and consumption, making the exchange
economy framework unsuitable for their purposes.
Therefore, macroeconomists replace the exogenous
endowment stream with the endogenous
equilibrium consumption process generated by a
standard neoclassical production economy that
faces technology shocks. One of the first findings of
this approach, summarized in Rouwenhorst ( ), is
that resolving the equity
premium puzzle in a production economy is far
more challenging than in an exchange economy,
because this endogenous consumption process
becomes too smooth if one increases risk aversion.
As a result, one needs to resort to real frictions such
as large adjustment costs in Jermann’s ( )
model. Furthermore, and as noted above,
allowing for an endogenous labour supply choice,
as is common in macroeconomic analysis, gives
Consumption-Based Asset Pricing Models (Empirical Performance)
consumers another margin to smooth marginal
utility and further reduces the equity premium.
Boldrin et al. ( ) and Uhlig ( ) have suc
cessfully introduced labour market frictions to
effectively shut down this channel.
Market and Asset Structure
The aggregation results we appeal to in order to use
a representative agent in asset pricing depend on
market completeness. A natural question is to ask
what happens if some of these markets are shut
down.
Incomplete Markets
In an attempt to resolve the equity premium puzzle,
uninsurable idiosyncratic income risk has been
introduced into consumption-based asset pricing
models by Aiyagari and Gertler (
Telmer ( ), Lucas ( ), Heaton and Lucas
( ), Krusell and Smith ( ) and Marcet and
Singleton ( ), among others. Their main
results, obtained numerically for a range of
parameter values, suggest that the impact of
uninsurable labour income risk on the equity pre-
mium is small, because agents manage to smooth
consumption quite well by trading a risk-free bond.
In fact, Levine and Zame ( ) show that
under general conditions the equilibrium allocations
and prices in incomplete market economies
converge to the complete market counterparts as
households become more patient, rendering the
incompleteness moot.
So when does imperfect risk sharing matter?
Mankiw ( ) derives a sufficient condition for
imperfect risk sharing to increase the equity risk
premium: the cross-sectional variance of con-
sumption growth needs to increase when returns are
low (that is, in recessions). Constantinides and
Duffie ( ) embed this counter-cyclical
cross-sectional variance mechanism in a general
equilibrium model. Grossman and Shiller (
show that the Mankiw-Constantinides-Duffie
(MCD) mechanism breaks down in continuoustime
diffusion models, because the cross- sectional
variance of consumption growth is deterministic.
2149
Discussion of Other Models
Rietz ( ) was the first to argue that countries
like the United States may simply have been very
lucky. Hence, the observed history of the US econ-
omy may understate the actual probability of eco-
nomic disasters, such as the Great Depression (at
least as perceived by investors). In this case, the
volatility of the SDF may be significantly higher
than the one estimated from historical time series.
As a result, investors will shun stocks and demand a
much higher equity premium to hold them. One
difficulty with this explanation is that many
economic disasters also result in governments
reneging on their debt obligations. Barro ( )
extends Rietz’s framework by
distinguishing between two types of disasters -
those that only affect the stock market and those
that affect all asset markets - and explores the
empirical implications of this mechanism in recent
),
work.
See Also
►
► Incomplete Markets
► Recursive Preferences
Bibliography
Abel, A.B.. 1990. Asset prices under habit formation and
catching up with the Jones. American Economic Review
80: 38M2.
Aiyagari, S.R., and M. Gertler. 1991. Asset returns with
transaction costs and uninsured individual risk. Journal
of Monetary Economics 27:311-331.
Bansal, R., and A. Yaron. 2004. Risks for the long run: A
potential resolution of asset prizing puzzles. Journal of
Finance 59: 1481-1509.
) Barro, R. 2006. Rare disasters and asset markets in the
twentieth century. Quarterly Journal of Economics
121:823-866.
Boldrin, M., L. Christiano, and J. Fisher. 2001. Habit
persistence, asset returns, and the business cycle. Amer-
ican Economic Review 91: 149-166.
Campbell, J.Y. 2000. Asset pricing at the millennium. Journal
of Finance 55: 1515-1567.
2150 Consumption-Based Asset Pricing Models (Empirical Performance)
Campbell, J.Y.. and J.H. Cochrane. 1999. By force of habit: A
consumption-based explanation of aggregate stock market
behavior. J o u r n a l o f P o l i t i c a l E c o n o m y 107:
205-251.
Colacito. R.. and M. Croce. 2005. Risks for the long-run and
the real exchange rate. Working paper, New York
University.
Constantinides. G.M. 1982. Intertemporal asset pricing with
heterogeneous consumers and without demand
aggregation. J o u r n a l o f B u s i n e s s 55: 253 267.
Constantinides. G.M. 1990. Habit-formation: A resolution of
the equity premium puzzle. J o u r n a l o f P o l i t i c a l
E c o n o m y 98: 519-543.
Constantinides. G.M., and D. Duffie. 1996. Asset pricing with
heterogeneous consumers. J o u r n a l o f P o l i t i c a l
E c o n o m y 104: 219-240.
Danthine, J.-P, and J.B. Donaldson. 2002. Labour relations
and asset returns. Review of Economic Studies 69: 41-64.
Dunn, K.., and K. Singleton. 1986. Modeling the term
structure of interest rates under nonseparable utility and
durability of goods. Journal of Financial Economics 17:
769-799/
Eichenbaum, M., and L.P. Hansen. 1990. Estimating models
with intertemporal substitution using aggregate time
series data. Journal of Business and Economic Statistics 8:
53-69.
Epstein, L.G., and S. Zin. 1989. Substitution, risk aversion
and the temporal behavior of consumption and asset
returns: A theoretical framework. Econometrica 57: 937
969.
Flavin, M., and T. Yamashita. 2002. Owner-occupied housing
and the composition of the house-hold portfolio.
A m e r i c a n E c o n o m i c R e v i e w 79: 345 362.
Grossman, S.. and G. Laroque. 1990. Asset pricing and
optimal portfolio choice in the presence of illiquid durable
consumption goods. E c o n o m e t r i c a 58:25 51.
Grossman, S.. and R. Shiller. 1982. Consumption
correlatedness and risk measurement in economies with
non-traded assets and heterogeneous information.
J o u r n a l o f F i n a n c i a l E c o n o m i c s 10: 195 210.
Guvenen, F. 2005. A parsimonious macroeconomic model for
asset pricing: Habit formation or cross-sectional
heterogeneity? Working paper, University of Texas at
Austin.
1 lansen. L.P.. and K. Singleton. 1983. Stochastic consump-
tion. risk aversion, and the temporal behavior of asset
returns. J o u r n a l o f P o l i t i c a l E c o n o m y 91: 249
265.
Heaton, .1.. and D. Lucas. 1996. Evaluating the effects of
incomplete markets on risk sharing and asset pricing.
J o u r n a l o f P o l i t i c a l E c o n o m y 104: 668 712.
Jermann, U. 1998. Asset pricing in production economies.
J o u r n a l o f M o n e t a r y E c o n o m i c s ' . 257 275.
Kandel. S.. and R.F. Stambaugh. 1990. Expectations and
volatility of consumption and asset returns. R e v i e w o f
F i n a n c i a l S t u d i e s 3: 207-232.
Kreps, D.. and E.L. Porteus. 1978. Temporal resolution of
uncertainty and dynamic choice theory.
E c o n o m e t r i c a 46: 185-200.
Krusell. P. and J.A. Smith. 1997. Income and wealth
heterogeneity, portfolio selection, and equilibrium asset
returns. M a c r o e c o n o m i c D y n a m i c s 1: 387 422.
Kydland, Finn E.. and Edward C. Prescott. 1982. Time to
build and aggregate fluctuations. E c o n o m e t r i c a 50:
1350-1370.
Levine, D.. and W. Zame. 2002. Does market incompleteness
matter1? E c o n o m e t r i c a 70: 1805 1839.
Longstaff. F.. and M. Piazzesi. 2002. Corporate earnings and
the equity premium. Working paper, UCLA Anderson
School.
Lucas, D. 1994. Asset pricing with unidiversifiable income
risk and short sales constraints: Deepening the equity
premium puzzle. J o u r n a l of Monetary
E c o n o m i c s 34: 325 341.
Mankiw, G.N. 1986. The equity premium and the concen-
tration of aggregate shocks. J o u r n a l o f F i n a n c i a l
E c o n o m i c s 17: 2 11 219.
Marcel, A., and K.. Singleton. 1999. Equilibrium asset prices
and savings of heterogeneous agents in the presence of
incomplete markets and portfolio constraints.
M a c r o e c o n o m i c D y n a m i c s 3: 243-277.
Mehra, R., and E. Prescott. 1985. The equity premium: A
puzzle. J o u r n a l o f M o n e t a r y E c o n o m i c s 15:
145 161.
Mcnzly, 1.., T. Santos, and P. Veronesi. 2004. Understanding
predictability. J o u r n a l o f P o l i t i c a l E c o n o m y
112: 1 47.
Piazzesi, M., M. Schneider, and S. Tuzel. 2007. Housing,
consumption, and asset pricing. J o u r n a l of
F i n a n c i a l E c o n o m i c s 83: 531-569.
Rietz. T.A. 1988. The equity risk premium: A solution 1?
J o u r n a l o f M o n e t a r y E c o n o m i c s 22: 117-
131.
Rouwenhorst, G. 1995. Asset pricing implications of equi-
librium business cycle models. In F r o n t i e r s o f
b u s i n e s s c y c l e r e s e a r c h , ed. T.F. Cooley.
Princeton: Princeton University Press.
Rubinstein, M. 1974. An aggregation theorem for security
markets. J o u r n a l o f F i n a n c i a l E c o n o m i c s 1:
225-244.
Santos, .1., and P. Veronesi. 2005. Labor income and pre-
dictable stock returns. R e v i e w o f F i n a n c i a l
S t u d i e s 19: 1 43.
Sundaresan, S. 1989. Intcrtemporally dependent preferences
and the volatility of consumption and wealth. R e v i e w
o f F i n a n c i a l S t u d i e s 2 : 73 89.
Telmer, C. 1993. Asset-pricing puzzles and incomplete
markets. J o u r n a l o f F i n a n c e 48: 1803 1832.
Uhlig, II. 2006. Macroeconomics and asset prices: Some
mutual implications. Working paper, Humboldt
University.
Verdelhan, A. 2004. Habit-based explanation of the exchange
rate risk premium. Working paper, Boston University.
Wachter, .1. 2002. 1 labit formation and returns on bonds and
stocks. Unpublished paper, Stern School of Business,
New- York University.
Weil, P. 1989. The equity premium puzzle and the risk-free
rate puzzle. J o u r n a l o f M o n e t a r y E c o n o m i c s
24: 401 424.
Consumption-Based Asset Pricing Models (Theory)
Wilson, R. 1968. The theory of syndicates. E c o n o m e t r i c a
36: 119-132.
Yogo, M. 2006. A consumption-based explanation of the cross-
section of expected stock returns. J o u r n a l o f
F i n a n c e 61: 539-580.
Consumption-Based Asset Pricing
Models (Theory)
Fatih Guvenen and Hanno Lustig
Abstract
The essential element in modem asset pricing
theory is a positive random variable called ‘the
stochastic discount factor’ (SDF). This object
allows one to price any payoff stream. Its exis-
tence is implied by the absence of arbitrage
opportunities. Consumption-based asset pricing
models link the SDF to the marginal utility
growth of investors - and in turn to observable
economic variables - and in doing so they
provide empirical content to asset pricing theory.
This article discusses this class of models.
Keywords
Consumption-based asset pricing models; Equity
premium puzzle; Euler equations; Sharpe ratio;
Stochastic discount factor
JEL Classifications
D4; DIO; G12
Consumption-based asset pricing models study the
pricing of payoff streams using the covariance of
these payoffs with the marginal utility growth of
investors.
The central component of a consumption-based
asset pricing model is die Euler equation, which
imposes restrictions on die covariance between asset
returns and the marginal utility growth of investors.
An easy and intuitive way to derive this equation is
by using a variational argument. Suppose that the
optimal consumption path of investor i is given
2151
by {C'}^ () where T is possibly infinite. Suppose
further that an asset j is available with a return R\ t+l
between periods / and t + 1 , and the investor is not
facing a binding portfolio constraint with respect to
this asset Then a feasible strategy is to reduce con-
sumption at time l by a small amount e, invest it in
asset j, and consume the proceeds, C'f j + , in
the next period. Assuming a time-separable utility
function, with the one-period felicity function
denoted by U and a time discount factor of ff this
strategy changes the investors’ expected lifetime
Uc (CrX,)e + E, \pUc ( C ' + 1 £^ + 1 )] ,
utility by
where Et is the mathematical conditional expectation
operator; X represents the arguments of the utility
function other than consumption; and XJC denotes
the partial derivative with respect to consumption.
The optimality of the original sequence implies that
this strategy cannot be profitable for any amount e
and any asset available. Setting this gain to zero and
rearranging yields the Euler equation:
M/,t+1^,,+! where
Uc(Cl+1,X'+1) (1)
Mm + 1 = p
UciCrX,)
This Euler equation was first derived by Rubin-
stein ( ) and Lucas ( ) in discrete time,
and by Breeden ( ) in continuous time. While
this class of models can in principle be used to study
a broad variety of assets, this article will focus on
stocks and short-term bonds, which have received
the greatest attention in the consumption- based
asset pricing literature.
In the case of a one-period discount bond with
gross return pft , + 1 = 1 //f — a bond that costs P1,
dollars today and pays off 1 dollar tomorrow - the
Euler equation can be rewritten as
Pf, =E,[MI,I+1\. (2 )
Similarly, when the asset is a stock with ex-dividend
price Pst and dividend payment Dt, the Euler
equation can be rearranged to read P\ = E, [/W,., I
(F) , + D,+i)]. By forward substitution this equation
yields:
2152
P ts E, } ',Mt^l+sDt+s > (3)
\i
which determines the price of a share of equity as
the value of all future dividends it entitles
discounted by the SDF.
Lucas ( ) and Mehra and Prescott ( ) (MM+1,^+1 <),
used a representative-agent endowment economy
structure in which the dividend stream, {£>,}“,, is
exogenously produced by a ‘tree’. Furthermore,
these dividends are assumed to be perishable
(‘fruit’), so in equilibrium the price of equity (in the
tree) adjusts to the point where the representative
agent is willing to consume all available dividends:
Ct = Dt. Substituting this condition into the
expression for M in Eq. ( ), and then using M in Eqs.
( ) and ( ) shows that the price of this stock and that
of the one-period bond are entirely determined by
the stochastic process for Dt together with the
functional form for U (we ignore X, for now).
Hansen and Singleton ( ) tested the repre
sentative agent’s Euler equation on US consumption
data, and found that the model was rejected. In a
famous paper, Mehra and Prescott ( ) an SDF with a coefficient of variation (standard
showed that when one chooses the properties of C,
to match the moments of aggregate consumption in
the data (‘calibrate the model to data’), the equity
premium E (R!' , K, J generated by the
model was about 60 times smaller than that
observed in the historical US data. This ‘equity
premium puzzle’ has generated enormous interest
and led to the development of a wide range of
consumption-based asset pricing models in an
attempt to resolve it. For further discussion of the
empirical performance of these models, see
consumption-based asset pricing models (empirical
performance).
An alternative way to explain the hurdles these
models face is by deriving an empirical lower
bound on the volatility of the stochastic discount
factor (SDF). Subtracting the Euler equation for
bond returns from the one for stock returns yields:
M, t, t-\-\ 0. Noting that the left-
hand side of this condition can be rewritten as
COV(M,',+1(^Rs1+1 — + E(Mltl+i')E(R,+l .s —/?f),
Consumption-Based Asset Pricing Models (Theory)
some simple manipulations yield the following key
decomposition:
( h\ /^) a {M tt+l )
E R
--------------------------------------— =------------------------------------------- — CO YY
£ M
( M+ 0 (4)
where cr(-) denotes the standard deviation.
Observing that the correlation term is bounded from
above in absolute value by 1 , we get
E RS
( t+1 ^) < 1)
cr(Rst+1 ~E(Mut+1)
The left-hand side of this inequality is the
‘Sharpe ratio’ - the (expected) excess return
demanded by investors per unit (standard deviation)
of risk they bear - which averages about 0.40 in
annual US data. The right-hand side is called the
‘market price of risk’ or the ‘maximum Sharpe
ratio’. This inequality bound implies that a
consumption-based model must be able to generate
deviation normalized by mean) of at least 40 per
cent to be consistent with the Sharpe ratio observed
in the data. This observation - developed by Shifter
( ) and further general-
ized by Hansen and Jagannathan ( )—
provides a ‘volatility bounds’ test for potential
candidate models. As discussed in consumption-
based asset pricing models (empirical performance),
the majority of plausibly calibrated asset pricing
models fail this test.
When the investor faces a binding borrowing
constraint, she cannot increase her consumption
today by reducing the holdings of asset j. As a
result, her marginal utility today will remain higher
than the value implied by the equality condition in
( ), and the Euler condition for that asset will
instead be an inequality:
Mt,t+ \r1i,i+i < 1. This relaxes the lower
bound on the volatility of the SDF derived in Eq. ( )
(cf. Luttmer ).
To develop further implications of consumption-
based models it is necessary to
Consumption-Based Asset Pricing Models (Theory)
impose additional structure on Mt t, |, which requires
being more specific about (1 ) the functional form
and the arguments of the utility function; (2 ) the
stochastic properties of variables affecting marginal
utility (that is, consumption, leisure, and so on); and
(3) the market structure. The latter determines
whether an appropriate aggregation theorem holds
(which happens for example when markets are
complete), in which case C, can be replaced with
aggregate consumption. Therefore, consumption-
based models can be broadly categorized based on
the assumptions they make along these three
dimensions. These different models are discussed in
consumption-based asset pricing models (empirical
performance).
Another feature of asset markets that has
received much attention in the literature concerns
the high volatility of stock prices. For example, the
standard deviation of the log price/dividend (P/D)
ratio of stocks is about 40 per cent per annum in the
US data. In a world with a constant SDF (as would
be the case with risk-neutral investors), it is
impossible to rationalize this high volatility with
the relatively low variability of the underlying
dividend stream (LeRoy and Porter ; Shiller ). Let
p, denote the log price, d, denote the log dividend,
and r, denote the log stock return. Using a first-
order approximation, Campbell and Shiller ( )
show that the log
P/D ratio can be decomposed as follows:
OO
p, — d, = constant + El'^pi~1 [ A — rt+j\
dl+j
j=i
with p = exp(prf)/l (l + exp(pd)) and pd denotes the
average log P/D ratio. The first term in the square
brackets is referred to as the cash flow component,
and the second part is referred to as the discount
rate component. This decomposition implies that
the variance of the log P/D ratio can be stated as:
var(p, - d,)cov ^p, - d„ ^ Ad,+;^
- cov ^p, - dh p'~1rl+^j .
2153
This expression shows that the P/D ratio moves
only because it predicts future returns on stocks or
because it predicts future dividend growth. In the
data, most of the volatility in P/D ratio is due to
news about future expected returns (‘discount
rates’), not due to future dividend growth (‘cash
flows’) (Campbell ; Cochrane ). There is a large
literature that documents the predictability of stock
returns over longer holding periods, starting with
work by Campbell and Shiller ( ,
), Poterba and Summers ( ) and Fama
and French ( , ). Other variables that
predict returns include the spread between long and
short bonds (Fama and French ) and the T-bill rate
(Lamont ). More recently, more attention has been
paid to macroeconomic variables that predict
returns, most notably in the work by Lettau and
Ludvigson ( ) who doc
ument that the consumption/wealth ratio is a pow-
erful predictor of stock returns.
So, the volatility of P/D ratio implies that excess
returns on stocks are highly predictable. In other
words, expected excess returns change a lot over
time, even per unit of risk. We use the conditional
version of the expression in ( ) to understand the
implications of this finding:
e,(rsi+i -#)
*(*;+!-*()
I) (6)
= —7
--------------; rCorr (M,,,+URt-\-1
s
t
where a, denotes the conditional standard deviation.
Good models need to produce a lot of time variation
in the right-hand side of ( ) and this happens mostly
through variation in the conditional market price of
risk (first term). This is an upper bound on the
conditional Sharpe ratio. (See also Lettau and
Ludvigson , on how to measure variation in the
conditional Sharpe ratio.) Another test of
consumptionbased asset pricing models is whether
they are able to generate as much predictability as
found in the data. Examples of early models that
match the variation in the conditional market price
of risk include Kandel and Stambaugh ( ),
Campbell and
2154
Cochrane ( ) and Barberis et al. ( ). More
recent work includes the work by Santos and
Veronesi ( ), Menzly et al. ( ), Piazzesi
et al. ( ), Guvenen ( ), Lustig and Van
Nieuwerburgh ( , ) and Bansal and
Yaron ( ). These models are discussed in
detail in consumption-based asset pricing models
(empirical performance).
See Also
Bibliography
Bansal, R., and A. Yaron. 2004. Risks for the long-run: A
potential resolution of asset pricing puzzles. J o u r n a l
o f F i n a n c e 59: 1481-1509.
Barberis, N., M. Huang, and T. Santos. 2001. Prospect theory
and asset prices. Q u a r t e r l y Journal of
E c o n o m i c s 116: 1-53.
Breeden, D.T. 1979. An intertemporal asset pricing model
with stochastic consumption and investment opportu-
nities. J o u r n a l o f F i n a n c i a l E c o n o m i c s 7:
265-296.
Campbell, J.Y. 1991. A variance decomposition for stock
returns. E c o n o m i c J o u r n a l 101: 157-179.
Campbell, J.Y., and J.H. Cochrane. 2000. Explaining the poor
performance of consumption-based asset pricing models.
J o u r n a l o f F i n a n c e 55: 2863-2878.
Campbell, J.Y., and R.J. Shiller. 1988. The dividend-price
ratio and expectations of future dividends and discount
factors. R e v i e w o f F i n a n c i a l S t u d i e s 1: 195-
227.
Campbell, J.Y., and R.J. Shiller. 1998. Stock prices, earnings
and expected dividends. J o u r n a l o f F i n a n c e 43:
661-676.
Cochrane, J.H. 1991. Explaining the variance of price-
dividend ratios. R e v i e w o f F i n a n c i a l S t u d i e s 5:
243-280.
Fama, E.F., and K.R. French. 1988. Dividend yields and
expected stock returns. J o u r n a l o f F i n a n c i a l
E c o n o m i c s 22: 3-27.
Fama, E.F., and K.R. French. 1989. Business conditions and
expected returns on stocks and bonds. J o u r n a l o f
F i n a n c i a l E c o n o m i c s 25: 23^49.
Guvenen, F. 2005. A parsimonious macroeconomic model for
asset pricing: Habit formation or cross-sectional
heterogeneity? Working paper, University of Texas at
Consumption-Based Asset Pricing Models (Theory)
Hansen, L.P., and R. Jagannathan. 1991. Implications of
security markets data for models of dynamic economies.
J o u r n a l o f P o l i t i c a l E c o n o m y 99: 252-262.
Hansen, L.R, and K. Singleton. 1983. Stochastic consumption,
risk aversion, and the temporal behavior of asset returns.
J o u r n a l o f P o l i t i c a l E c o n o m y 91: 249-265.
Kandel, S., and R.F. Stambaugh. 1990. Expectations and
volatility of consumption and asset returns. R e v i e w o f
F i n a n c i a l S t u d i e s 3: 207-232.
Lamont, O. 1998. Earnings and expected returns. J o u r n a l
o f F i n a n c e 53: 1563-1587.
LeRoy, S.F., and R.D. Porter. 1981. The present-value
relation: Tests based on implied variance bounds.
E c o n o m e t r i c a 49: 555-574.
Lettau, M., and S.C. Ludvigson. 2001a. Consumption,
aggregate wealth and expected stock returns. J o u r n a l
o f F i n a n c e 56: 815-849.
Lettau, M., and S.C. Ludvigson. 2001b. Measuring and
modeling variation in the risk-return tradeoff. In H a n d -
b o o k o f F i n a n c i a l E c o n o m e t r i c s , ed. Y. Ait-
Sahalia and L.P Hansen. Amsterdam: North-Holland.
Lucas, R. 1978. Asset prices in an exchange economy.
E c o n o m e t r i c a 46: 1429-1454.
Lustig, H. and S.V. Nieuwerburgh. 2006. C a n h o u s i n g
collateral explain long-run swings in
a s s e t r e t u r n s ? Working Paper NYU Stem and
UCLA.
Lustig, H., and S. van Nieuwerburgh. 2005. Housing col-
lateral, consumption insurance and risk premia: An
empirical perspective. J o u r n a l o f F i n a n c e 60:
1167-1219.
Luttmer, E. 1996. Asset pricing in economies with frictions.
E c o n o m e t r i c a 64: 1439-1467.
Mehra, R., and E. Prescott. 1985. The equity premium: A
puzzle. J o u r n a l o f M o n e t a r y E c o n o m i c s 15:
145-161.
Menzly, L., T. Santos, and P Veronesi. 2004. Understanding
predictability. J o u r n a l o f P o l i t i c a l E c o n o m y
112: 1-47.
Piazzesi, M., M. Schneider, and S. Tuzel. 2007. Housing,
consumption, and asset pricing. J o u r n a l of
Financial
E c o n o m i c s 83: 531-569.
Poterba, J.M., and L.H. Summers. 1986. The persistence of
volatility and stock market fluctuations. A m e r i c a n
E c o n o m i c R e v i e w 76: 1142-1151.
Rubinstein, M. 1976. The valuation of uncertain income
streams and the pricing of options. B e l l J o u r n a l o f
E c o n o m i c s 7: 407-425.
Santos, J., and P. Veronesi. 2005. Labor income and pre-
dictable stock returns. R e v i e w o f F i n a n c i a l
S t u d i e s 19: 1-43.
Shiller, R.J. 1981. The use of volatility measures in assessing
market efficiency. J o u r n a l o f F i n a n c e 36: 291-
304.
Shiller, R.J. 1982. Consumption, assest markets and mac-
roeconomic fluctuations. C a r n e g i e - R o c h e s t e r
S e r i e s o n P u b l i c P o l i c y 17: 203 238. North-
Contemporary Capitalism 2155

volatility; Stock repurchases; Strategic control;

Contemporary Capitalism Technology; Venture capital

William Lazonick
JEL Classifications
PI

Abstract
The key to understanding ‘capitalism’ as a mode
of resource allocation that generates economic What Is 'Capitalism'?
growth is the organization and performance of
its most innovative business enterprises. The At the beginning of the 21st century, ‘capitalism’
‘Old Economy business model’ that made the has triumphed as the dominant system for allocat-
United States the world’s most powerful nation ing a society’s economic resources. The last time in
in the post-Second World War decades came history in which the persistence of capitalism in the
under challenge in the 1970s and 1980s, and the world’s most advanced economies was seriously
ideology of ‘maximizing shareholder value’ called into question was the Great Depression of the
arose to legitimize a redistribution of income 1930s - a decade during which the unemployment
from labour interests to financial interests. The rate in the United States remained at 15 per cent or
‘New Economy business model’ emerged in the higher, notwithstanding unprecedented state
1980s and 1990s to drive the innovation process, intervention under the New Deal. It took the Second
contributing, however, to unstable and World War to pull the United States and the world
inequitable economic growth. economy out of depression, and in the subsequent
decades it took substantial and sustained
government spending in the rich economies of
North America and western Europe to hold
Keywords
unemployment to acceptable levels.
Acquisitions; Business enterprises; Capitalism;
In the post-war era, the Soviet Union’s highly
Collective capitalism; Competitive advantage;
planned economy posed as a possible alternative to
Conglomerate movement; Contemporary
capitalism. The purported strength of the Soviet
capitalism; Corporations; Creative destruction;
challenge, however, turned out to be based at least
Defined-benefit pensions; Developmental state;
as much on Cold War ideology emanating from the
Dividend yield.; Division of labour; Economic
United States as on the actual productive power of
development; Entrepreneurship; Financial
the Soviet Union and its satellites. By the 1990s the
commitment; Foreign direct investment;
Soviet model had virtually vanished, as Russia
Globalization; Great Depression; Hostile take-
itself sought to make the transition to a ‘market
overs; Information and communications tech-
economy’, guided, tragically, by a mythical
nology; Innovation; Japan, economics in; Junk
ideology of how capitalism is supposed to operate,
bonds; Leveraged buyouts; Lifelong employ-
imported from the United States.
ment; Mergers; Milken, M.; NASDAQ; New
Over the same period capitalism entrenched
Deal; New Economy business model; New York
itself in East Asia. During the 1970s and 1980s
Stock Exchange; Old Economy business model;
Japan became a rich economy on the basis of a
Organization man; Organizational integration;
distinctive model of ‘collective capitalism’, and in
Outsourcing; Patents; Research and
the 1980s and 1990s the East Asian ‘Tigers’ - Hong
development; Schumpeter, J. A.; Separation of
Kong, Singapore, South Korea
ownership and control; Shareholder value;
Social inclusion; Stock options; Stock price
2156
and Taiwan - closed the gap, each with its own
variant of the Japanese model. More recently China
and India, with one-third of the world’s population,
have experienced rapid economic growth, driven by
what many would call ‘capitalist’ institutions. Yet,
even as firms cross the globe to access Indian
software engineers, and vice versa, India remains a
nation with one-third of the world’s illiterates.
Meanwhile the fact that China, the world’s second
largest economy since the early 1990s, continues to
be guided by an avowedly Communist government
raises the question of what ‘capitalism’ really is.
Defining contemporary capitalism is not merely
a question of semantics. If, as has been
demonstrated since the mid-20 th century, ‘capi-
talism’ is a powerful engine of economic growth,
we want to know how it functions as a mode of
resource allocation and the social conditions under
which capitalist growth is not only strong but also
stable, and equitable. We also want to know how
the institutions of contemporary capitalism that
generate growth might be transferred to those parts
of the world - first and foremost Africa but also
eastern Europe and Latin America as well as parts
of the Middle East - that have economically been
left behind. Given its pervasiveness and dominance,
a depiction of the institutions that define
contemporary capitalism is tantamount to a
description of the economic world in which perhaps
one-half of the world’s population now lives and to
which much of the other half now aspires.
There is no consensus among economists on the
definition of contemporary capitalism. The
dominant approach to analysing resource allocation
and the economic performance of an advanced
economy rests on the notion that a capitalist
economy is essentially a market economy that
allocates resources to their most productive uses.
But what at any time and in any place, the student
of economic development asks, explains how those
most productive uses come to exist? And why in
certain times and places? Fundamental to capitalist
growth is ‘innovation’, the process that generates
goods and services that, even with factor prices held
constant, are of higher quality and lower cost than
those previously available
Contemporary Capitalism
(Lazonick ). Can a theory of capitalism as a market
economy comprehend the innovation process?
In the early 20th century a young Joseph
Schumpeter asked this question. As a Viennese
economics student, Schumpeter was versed in the
relatively recent, and increasingly influential,
Austrian and Walrasian theories of how, through the
equilibrating mechanism of the market, the
economy could achieve an ‘optimal’allocation of
resources across productive uses. Schumpeter’s
insight was to recognize that such a view of the
economic world could not explain economic
development. In 1911 Schumpeter wrote The The-
ory of Economic Development (first translated into
English in 1934) to argue that entrepreneurial
activity that results in innovation - what he called
the ‘Fundamental Phenomenon of Economic
Development’ - can disrupt the ‘Circular Flow of
Economic Life as Conditioned by Given Cir-
cumstances’ to change the ways in which the
economy operates and performs. Without such
disruption of equilibrium conditions, the economy
would not develop. Over the next four decades
Schumpeter sought to elaborate a theory of
economic development informed by his own,
evolving, understanding of the changing reality of
the most advanced capitalist economies.
In particular, Schumpeter sought to understand
the role of the business enterprise in advanced
capitalist development. By the 1940s he had taken
definitive leave of his youthful conceptions of the
innovative entrepreneur as an individual actor and
innovation as simply ‘new combinations’ of
existing resources. Rather, he saw that powerful
business organizations both developed and utilized
productive resources to create new technologies and
access new markets. The creation of new tech-
nologies, moreover, destroyed the commercial via-
bility of old technologies. In Capitalism, Socialism,
and Democracy, first published in 1942,
Schumpeter argued that the process of ‘creative
destruction’ had become embodied in established
corporations as ‘technological “progress” tends,
through systematization and rationalization of
research and of management, to become more
effective and surefooted’, being ‘the business of
teams of trained specialists who turn out what is
Contemporary Capitalism
required and make it work in predictable ways’
(Schumpeter , pp. 118, 132).
This article takes as its point of departure the
proposition, suggested by Schumpeter, that the key
to understanding ‘capitalism’ as a mode of resource
allocation that generates economic growth is the
organization and performance of its most innovative
business enterprises. That is not to say that markets
and states are unimportant to the operation, and
hence definition, of capitalism. Historically,
however, well-functioning markets are outcomes of
successful capitalist development. For the
individual, markets create the possibility of
choosing what to consume and for whom to work,
including the prospect of working for oneself. But
markets cannot explain the development of the new
products and processes that drive the growth of the
capitalist economy. The innovation process is
uncertain, collective and cumulative (see O’sullivan
). The uncertain character of innovation means that
investments in innovation require strategic control
over resource allocation by individuals who have
intimate knowledge of the technologies, markets
and competitors that an innovative strategy must
confront. The collective character of innovation
means that the implementation of an innovation
strategy requires the organizational integration of a
hierarchical and functional division of labour into a
process of organizational learning. The cumulative
character of innovation means that the process
requires financial commitment until it can generate
financial returns. Enterprises, not markets, engage
in strategic control, organizational integration and
financial commitment (Lazonick ).
Nor can one explain innovation by appealing to
the notion of the developmental state as its driving
force, as has often been done for the East Asian
economies. Implicit, and at times explicit, in this
view is an acceptance of the ideology that the
economic development of the United States is an
exemplar of the workings of the market economy.
Yet from gun manufacture and interchangeable
parts in the first half of the 19th century to the
computer revolution and Internet in the late 20th
century, as well as railroads, aviation and the life
sciences in between, the history of US capitalism
2157
is replete with examples of the critical role of the
developmental state in allocating resources to the
processes of knowledge creation that then provided
the foundations for US industrial leadership. Yet, as
important as the developmental state has been even
in a so-called ‘market economy’ such as the United
States, the allocation of resources to knowledge
creation would have been wasted, and would
probably never have been made, had it not been for
the presence and influence of innovative enterprises
that have made use of this knowledge to generate
higherquality, lower-cost products than had
previously been available.
In this article I focus on the changing role of
innovative enterprise in determining resource
allocation and economic performance in contem-
porary capitalism. Space constraints dictate that I
confine the analysis of contemporary capitalism to
the case of the United States, with the caveat that,
even in a highly globalized economy in which one
might expect convergence to a common business
model, there are almost as many distinctive
‘varieties of capitalism’ in terms of governance,
employment and investment institutions, as there
are advanced capitalist nations. The US economy is,
however, the world’s largest and richest economy.
It is also the one in which market ideology is most
virulent and the actual mode of resource allocation
most misunderstood. Section “ ” of
this article provides historical background to
understanding contemporary US capitalism by
describing the key characteristics of the ‘Old
Economy business model’ (OEBM) that made the
United States the world’s most powerful nation in
the decades after the Second World War. Section “
analyses the challenges that confronted OEBM in
the 1970s and 1980s, and how the ideology of
‘maximizing shareholder value’ arose to legitimize
a redistribution of income from labour interests to
financial interests. Section “
” shows how the ‘New
Economy business model’ (NEBM) emerged in the
1980s and 1990s to drive the innovation process,
but in ways that have contributed to unstable and
inequitable economic growth. Section “
2158 Contemporary Capitalism

concludes with some The separation of ownership from control has

questions about the future of the US model in a worked effectively to generate innovation when the
global economy in which many distinctive business interests of salaried executives who exercise
models still compete. strategic control have been aligned with those of
employees who engage in the development and
ensure the utilization of the company’s productive
The Old Economy Business Model resources. In the post-Second World War decades
the organizational integration of the capabilities of
The United States emerged from the Second World administrative and technical specialists enabled US
War as the undisputed world leader in GDP per firms to develop the world’s most competitive
capita, a position that it still retains. With western systems of mass production. These personnel were
Europe and Japan still in recovery from the war, the products of the US system of higher education,
United States was at its peak of dominance in the which since the early decades of the century had
1950s on the basis of a highly collective model of prepared the labour force to enter employment in
capitalism embodied in the managerial corporation, bureaucratic organizations.
and personified in the concept of the ‘organization A distinctive feature of the US business model
man’ (Whyte ). The stereotypical ‘organization was the organizational segmentation between these
man’ was white, Anglo-Saxon and Protestant, salaried managers, in whose training and experience
obtained a college education, got a well-paying job the corporation made substantial investments, and
with an established company early in his career, and so-called ‘hourly’ workers. (Nonsalaried employees
then worked his way up and around the corporate were classified as ‘hourly’, or ‘non-exempt’,
hierarchy over decades of employment, with a sub- workers because of the stipulation of the National
stantial ‘defined benefit’ pension, complete with Labor Relations Act that emerged from the New
highly subsidized medical coverage, awaiting him Deal era that required employees who were paid an
on retirement. The employment stability offered by hourly wage receive 150 per cent of that wage if
an established corporation was highly valued, while they worked longer than the normal working hours.
inter-firm labour mobility was shunned. The overtime work of salaried personnel is exempt
‘Organization men’ rose to top executive posi- from this provision.) The corporation viewed these
tions where, as salaried managers rather than operatives, who were typically high-school
owners, they exercised strategic control. This sep- graduates, as interchangeable commodities in whose
aration of share ownership and managerial control, capabilities the company had no need to invest,
which continues to characterize the US industrial notwithstanding the fact that they often spent their
corporation, resulted from the widespread entire working fives with one company. At the same
distribution among shareholders of the time, these industrial corporations needed reliable
corporation’s publicly traded stock. In principle, even if lowskill workers to tend mass production
boards of directors representing the interests of processes. The combination of dominant product-
shareholders monitor the decisions of these man- market positions and union power, which advanced
agers. In practice, incumbent top executives choose the pay and protected the employment of senior
the outside directors and are themselves members of workers, enabled the hourly worker to receive good
the board. Shareholders can challenge management pay and benefits, including a defined-benefit
through proposals to the annual general meeting, pension that assumed long-term employment with a
but over the course of the 20 th century a body of single company.
law evolved that enables management to exclude The developmental state played an indispens-
shareholder proposals that deal with normal able role in the innovation process by partially
business matters (for example, downsizings) as funding the system of higher education as well as,
distinct from social issues (for example, sex in the forms of research labs, subsidies and
discrimination).
Contemporary Capitalism
contracts, programmes for technology development
in sectors such as aerospace, computers and life
sciences. The development of the productive
potential of these government investments relied in
turn on corporate research capabilities. Retained
earnings formed the foundation of committed
finance for new corporate investments in
innovation. When corporations needed additional
investment financing, they issued corporate bonds
at favourable rates that reflected the established
position of the company as well as its conservative
debt-equity ratios. Companies used bank loans
almost exclusively for working capital, and made
only limited use of the stock market as a source of
investment funds.
These social conditions enabled US corporations
to grow very large in the post-war decades. The 50
largest US industrial corporations by revenues on
the Fortune 500 list averaged 87,070 employees in
1957, 117,393 in 1967, and 119,093 in 1977. These
figures do not include employment at AT&T, the
regulated telephone monopoly, which in 1971
employed 1,015,000 people, of whom 700,000 were
union members with good wages, stable
employment and excellent benefits. By the late
1960s and early 1970s increasing numbers of blacks
were moving into union jobs in the steel,
automobile, electrical equipment, consumer durable
and telecommunications industries. The growth of
established corporations in these industries in the
three decades after the Second World War
contributed to a more equal distribution of family
income in the US economy.
'Maximizing Shareholder Value'
During the 1970s the US model faltered in the face
of Japanese competition. Building on innovative
capabilities developed for their home markets
during the 1950s and 1960s, Japanese companies
gained competitive advantage over US companies
in industries such as steel, memory chips, machine
tools, electrical machinery, consumer electronics
and automobiles. US companies had entered the
1970s as world leaders in these industries. Many
US observers attributed
2159
the rapid increase in Japanese exports to the United
States in the 1970s to Japan’s lower wages and
longer working hours. By the early 1980s, however,
with real wages in Japan continuing to rise, it
became clear that Japanese advantage was based on
the superior organization of their enterprises, and in
particular on a more thoroughgoing integration of
participants in the functional and hierarchical
divisions of labour for the dual purposes of
transforming technologies and accessing new
markets. Indeed, during the 1980s Japan exported
management practices as well as material goods to
the West. From the second half of the 1980s, with
the yen strengthening and trade surpluses generating
political backlash, Japanese companies made a
transition to direct investment in the United States
and other advanced economies.
A growing financial orientation of US business
that had surfaced in the conglomerate movement of
the 1960s undermined the abilities and incentives of
established US corporations to respond to the
Japanese challenge. To some extent the growth of
the US industrial corporation in the post-war
decades had been based on strategic investments in
new product lines and geographic areas that built on
the corporation’s existing productive capabilities,
and yielded economies of scale and scope. The
conglomerate movement, however, saw major
corporations invest in scores of unrelated
businesses, often through mergers and acquisitions,
based on the prevailing, but erroneous, ideology
that a good corporate executive could manage any
type of business, and that conglomeration offered
the synergies of superior corporate management.
The conglomerate movement failed because it
segmented top executives, in positions of strategic
control, from the rest of the managerial organization
that had to develop and utilize productive resources
to sustain the firm’s competitive advantage
(Lazonick ).
In the late 1970s and early 1980s the conglom-
erates unraveled. In the mid-1970s Michael Milken,
a Drexel Burnham investment banker, had created
the junk bond market by convincing institutional
investors, in search of higher yields in an
inflationary era, to hold downgraded corporate
securities, many of them ‘fallen angels’ from
 1960-9 1970-9
Real stock yield 6.63 -1.66
Price yield 5.80 1.35
Dividend yield 3.19 4.08
Change in CPI 2.36 7.09
Real bond yield 2.65 1.14
Source'. Council of Economic Advisers ( , Tables B-62, B-73, B-95 and B-96)
Notes: Stock yields are for Standard and Poor’s composite index of 500 US corporate stocks (424 of which are, as of 28 March
2006, NYSE). Bond yields are for Moody’s Aaa-rated US corporate bonds
unsuccessful conglomeration. By the late 1970s,
with the junk-bond market well developed, it
became possible to issue new junk bonds to finance
leveraged buyouts (LBOs) in which the top
managers of a conglomerate division turned it into
an independent company to recapture strategic
control over resource allocation. By the late 1980s,
however, the junk bond had become an instrument
for the hostile takeover of entire companies, with
KKR’s 1989 LBO of RJR Nabisco for $24.5 billion
marking the height of what became known as ‘the
deal decade’.
The ideology that justified hostile takeovers was
that the corporation should be run to ‘maximize
shareholder value’ (see Lazonick and O’sullivan ).
Proponents of shareholder value charged that, either
because of opportunism or incompetence, many
incumbent corporate managers were making poor
allocative decisions. By exercising their influence
through the market for corporate control,
shareholders could force incumbents to alter their
allocative decisions, replace them with those who
would maximize shareholder value, or distribute
cash to shareholders in the forms of dividends and
stock repurchases so that shareholders themselves
could, so the argument goes, reallocate the
economy’s resources to their best alternative uses.
While the hostile takeover movement did not
directly threaten high-tech companies (in which the
most valuable assets could walk out the door), by
the end of the 1980s the top executives of virtually
all US industrial corporations had embraced the
ideology of maximizing shareholder value and
made it their own. By the 1980s executive stock
option compensation was a well-established
practice. Since in the United
1980-9 1990-9 2000-5
11.67 15.01 - 1.87
12.91 15.54 -0.76
4.32 2.47 1.58
5.55 3.00 2.67
5.79 4.72 3.60
States option awards did not require that the
company’s stock price outperform the stock market
or even the stock prices of a group of competitors,
those who received these awards could only gain
from what, from July 1982 to August 2000, turned
out to be the longest stock market boom in US
history, with the Dow Jones Industrial Average and
the S&P500 Index both rising about 1,300 per cent.
As Table shows, stock-price appreciation drove
the extraordinary real stock yields that were
sustained over the 1980s and 1990s. The relatively
low dividend yields in the 1990s did not reflect
stinginess on the part of US corporations; the US
corporate payout ratio - the amount of dividends as
a percentage of after-tax corporate profits (with
inventory evaluation and capital consumption
adjustments) - averaged 48 per cent in the 1980s
and 57 per cent in the 1990s compared with 3 9 per
cent in the 1960s and 41 per cent in the 1970s. It
was just that the rate of increase of stock prices
outstripped the rate of increase of dividend
payments, thus depressing the dividend yield. The
form that the stock yield takes is of significance
because investors can capture the dividend yield by
holding stocks, whereas they can capture the price
yield only by buying and selling stocks. Inherent in
high-price yields, therefore, is a volatile stock
market.
A volatile stock market benefits those who are
compensated in stock options on an annual basis,
especially when, as is the case in the United States,
options vest as quickly as one year from the date of
grant and can be exercised for up to ten years. It has
been estimated that, largely because of the gains
from exercising stock options, on average the ratio
of CEO pay of an S&P500
Contemporary Capitalism
company to that of a production worker was 42 in
1985, 107 in 1990, 525 in 2000, and 411 in 2005.
Top executives took a keen interest in their
company’s stock price, and in the 1980s and 1990s,
in the name of ‘maximizing shareholder value’, they
found ways in which they could use their positions
of strategic control over corporate resource
allocation to influence it. They could cook the
corporate books to boost current earnings, a practice
that became widespread in these decades and one
for which a few executives have been fined or even
jailed. The American Competitiveness and
Corporate Accountability Act of 2002, better known
as Sarbanes-Oxley, has sought to stem this practice.
But quite apart from artificially inflating corporate
earnings, top corporate executives also found that
downsizing the labour force and repurchasing
corporate stock helped to boost a company’s stock
price, even though these resource allocations did not
necessarily improve the company’s competitive
performance.
The era of corporate downsizing took hold in the
recession of 1980-2 when hundreds of thousands of
stable, well-paid blue-collar jobs were lost that were
never subsequently restored (see Lazonick ). It
would appear that the blacks who had relatively
recently moved into these types of jobs were
particularly hard hit; last hired, they tended to be the
first fired. The subsequent ‘boom’ years of the mid-
1980s witnessed hundreds of plant closings. In the
‘white-collar’ recession of the early 1990s tens of
thousands of professional, administrative and
technical employees found that their jobs had been
eliminated, although once again it was blue-collar
workers who bore the brunt of the downturn. In
1980 manufacturing employment was 22 per cent of
the labour force; by 1990 it had fallen to 17 per cent
and by 2001 to 14 per cent. While the employment
picture generally became much better during the
Internet boom of the last half of the 1990s, job
cutting remained a way of life for many major US
corporations. According to data on layoff
announcements by companies in the United States
collected by the recruitment firm, Challenger, Gray
and Christmas, announced job cuts averaged just
under 550,000 per year for the
2161
period 1991-4, 450,000 per year in 1995-7, and
656,000 per year during the boom years 1998-2000.
Meanwhile, from the mid-1980s US corpora-
tions began to actively support their stock prices
through large-scale stock repurchases. Companies
included in the S&P500 in March 2006 distributed
more cash to shareholders in repurchases than in
dividends in 1997 through 2000 and again in 2004,
and just slightly less in 2001 through 2003. Since
1978 net equity issues by US non-financial
corporations has been positive in only six of 28
years (1980, 1982, 1983, 1991, 1992, 1993); since
the early 1980s US industrial corporations have in
aggregate been supplying capital to the stock market
rather than vice versa. In 2005 the net flow of cash
from non-financial corporations to the stock market
was a record $366 billion, 1.42 times in real dollars
the previous high in 1998 (Lazonick ).
The New Economy Business Model
On 29 December 1995, AT&T announced that, as
part of the process of breaking itself up into three
separate companies, it would be cutting 40,000 jobs.
AT&T was a company that could trace its origins
back to the 1870s, had created the world’s most
advanced telephone system, was the home of the
famous Bell Labs that among many other
accomplishments invented the transistor in 1947,
and, despite having lost its status as a regulated
monopoly in 1984, still employed 308,700 people.
Now, however, AT&T became emblematic of the
failure of US Old Economy corporations to continue
to provide employment opportunities. With
campaigning for the 1996 presidential election
picking up steam, Patrick Buchanan, a right- wing
politician, caught the attention of the media by
denouncing the highly paid executives of AT&T
and other downsizing corporations as ‘corporate hit
men’. Fuel was added to the fire by the revelation
that, in the name of ‘creating shareholder value’, A1
Dunlap, whom the American public came to know
as ‘Chainsaw AT, had in 20 months as CEO of Scott
Paper devastated the 115-year old company while
putting an estimated
 Old economy business model (OEBM)
Strategy, Firm growth based on multidivisional structure: multi-
product product firm
Strategy, Vertical integration of the value chain; in-house standards Vertical specialization of the value chain; industry
process and proprietary R&D
Finance Venture finance from savings, family and business
associates; NYSE listing, growth finance from retentions, Organized venture capital; early IPO on NASDAQ;
after dividends, and bond issues
Organization
Secure employment; ‘organization man’ (career with one broad-based stock options, nonunion, defined
company), industrial union; defined- benefit pension, contribution pension, employees bear more burden of
good medical coverage in employment and retirement
$100 million in his own pocket. In March 1996, the
New York Times ran a seven-part series, later
released as a paperback, on ‘the downsizing of
America’ (Lazonick ).
By the spring of 1996, however, the fbror over
corporate downsizing had disappeared. In its place,
Americans became enthralled by the prosperity
promised by what in the second half of the 1990s
came to be called the ‘New Economy’. In the
United States the previous half-century had seen a
massive accumulation of information and
communications technology (ICT) capabilities. The
development of computer chips from the late 1950s
had provided the technological foundation for the
microcomputer revolution from the late 1970s,
which in turn had provided the technological
infrastructure for the Internet boom of the second
half of the 1990s. The research funding for this
accumulation of ICT capabilities had come mainly
from the US government and the research
laboratories of established Old Economy high-
technology corporations. Each wave of techno-
logical innovation, however, created opportunities
for the emergence of start-up companies that were
to become central to the commercialization of the
new technologies.
Although by the mid-1980s the Japanese had
outcompeted even the leading US semiconductor
firms in the memory chip market, US companies
such as Intel, Motorola and Texas Instmments
New economy business model (NEBM)
New firm entry into specialized ICT markets; accumulate
new capabilities by acquiring (other) young technology
firms
technology standards; R&D for cross-licensing and
alliances; outsourcing routine work to specialist contract
manufacturers and/or offshoring routine work to low-
wage nations
retentions with zero dividends; use of own stock as a
compensation and combination currency; systematic
stock repurchases to support stock price
Insecure employment; interftrm mobility of labour,
medical insurance
continued to dominate the microprocessor and logic
chip markets that drove product innovation in the
microelectronics industry (Lazonick ). While
Silicon Valley was not the only US location for
innovation in this industry, the concentration of
semiconductor start-ups in the region from the late
1950s resulted in the emergence by the 1980s of a
distinctive mode of combining strategy, finance and
organization: the ‘New Economy business model’
(NEBM) (see Table ). During the 1990s NEBM
spread beyond Silicon Valley start-ups and was
adopted successfully by leading Old Economy ICT
companies such as Hewlett-Packard and IBM. In
the Internet boom of the late 1990s elements of
NEBM diffused to other ICT companies, including
an Old Economy company such as Lucent
Technologies, spun off from AT&T in its 1996
trivestiture, which almost destroyed itself in
attempting to adopt the business model. In the
2000s NEBM characterizes the most innovative
sectors of the US economy (for the case of
biotechnology, see Pisano )•
The founders of New Economy firms have
typically been scientists and engineers who have
gained specialized experience in existing firms,
although in some cases they have been university
faculty members intent on commercializing their
academic knowledge. Some of these entrepreneurs
have come from existing Old Economy
Contemporary Capitalism
companies, where it was often difficult for their
new ideas to get internal backing. But New Econ-
omy companies themselves have become increas-
ingly important as sources of new entrepreneurs
who left their current employers to start new firms.
Large numbers of high-tech entrepreneurs in the
United States have been foreign-bom, coming
mainly from India and China (Saxenian ).
Typically, the founding entrepreneurs of a New
Economy start-up seek committed finance from
venture capitalists with whom they share not only
ownership of the company but also strategic control.
In the 2000s Silicon Valley remains by far the
leading location in the United States and the world
for venture-backed high-tech start-ups. The region
acquired this position from the 1960s as a
distinctive venture capital industry emerged out of
the opportunities for start-ups created by the
microelectronics revolution. Besides sitting on the
board of directors of the new company, the venture
capitalist generally recruits professional managers,
who are given company stock along with stock
options, to lead the transformation of the firm from
a new venture to a going concern. This stock-based
compensation gives these managers a powerful
financial incentive to develop the innovative
capabilities of the company to the point where it can
do an initial public offering (IPO) or private sale to
a listed company, thus enabling the start-up’s
privately held shares to be transformed into publicly
traded shares. Both before and after making this
transition, their tenure with, and value to, the
company depends on their managerial capabilities,
not their fractional ownership stakes (Lazonick ).
The stock market speculation of the ‘dotcom’
era made it all too easy to cash out of a start-up, as
many high-tech firms that had not engaged in
innovation did IPOs or were sold to established
companies. When start-ups do innovate, the key to
making the transition from new venture to going
concern has been the organizational integration of
an expanding body of technical and administrative
‘talent’. As Silicon Valley developed from the
1960s, this educated and experienced labour had to
be induced to trade secure employment with an Old
Economy company for insecure employment with a
start-up. To attract
2163
these highly mobile people and retain their services,
Silicon Valley firms increasingly adopted ‘broad-
based’ employee stock option plans that extended
this form of compensation to a large proportion,
sometimes all, of the firm’s non-executive
employees rather than just to top executives. In
start-ups, stock options usually served as a partial
substitute for cash salaries, and the eventual gains
from exercising options were viewed as a substitute
for a company-funded pension (Lazonick ).
Again, the underlying stock would become
valuable if and when the start-up did an IPO or a
private sale to a publicly listed company. Short-
ening the expected period between the launch of a
company and an IPO was the practice of most
venture-backed high-tech start-ups of going public
on NASDAQ, created in 1971 as an electronic
exchange for the over-the-counter markets with less
stringent listing requirements than the ‘Old
Economy’ New York Stock Exchange (NYSE). The
1978 cut in the capital gains tax rate to 28 per cent,
after it had been raised to 49 per cent just two years
before, provided further encouragement to
entrepreneurs and venture capitalists to found new
companies, and for employees of these companies,
rewarded with stock options, to provide the skills
and efforts needed to transform new ventures into
going concerns. In 1979 the clarification of the
‘prudent man’ rule as applied to the Employee
Retirement Income Security Act (ERISA) of 1974
gave asset managers the green light to allocate a
portion of their portfolios to riskier stocks and
venture capital funds, and resulted in a flood of new
money, especially from pension funds, into the
venture capital industry. The American Electronics
Association and the National Venture Capital
Association, with their strongest and deepest roots
in Silicon Valley, were the frontline Washington
lobbyists for these regulatory changes.
While institutional money provided capital to
NEBM, high-tech labour became more mobile from
one firm to another than it had been in the Old
Economy. Employee stock options induced this
mobility, but what made it possible in terms of the
knowledge bases that managers and engineers
possessed were industry standards, as distinct
2164
from in-house standards, that emerged in the var-
ious sectors of ICT. In the Old Economy in-house
standards promoted the growth of large vertically
integrated firms on the basis of proprietary tech-
nologies, whereas in the New Economy industry
standards encouraged new entry. Nevertheless, as
demonstrated by the important cases of Intel and
Microsoft in the development of the microcomputer
industry, those New Economy firms that dominated
in the setting of the industry standards could also
grow very large (at the end of fiscal 2005 Intel
employed 99,900 and Microsoft 61,000). By
establishing industry standards, their growth
encouraged rather than discouraged startups, which
in turn depended on the availability of not only
venture capital (which came from many sources
besides the formal venture capital industry) but also
mobile labour whose knowledge and experience
could be easily integrated into the start-up’s
learning processes.
Of critical importance in setting industry stan-
dards in microelectronics was IBM’s decision in
1980 to enter what became known as the personal
computer (PC) industry with Intel supplying the
microprocessor and Microsoft the operating system.
At the time IBM controlled about 80 per cent of the
computer market, had over 341,000 employees, and,
with an explicit system of ‘lifelong employment’,
trumpeted the fact that since 1921 it had not
terminated an employee involuntarily. Yet between
1990 and 1994 IBM slashed its employment from
374,000 to 220,000. In 1991-3, the company had
losses of $16 billion (including more than $8 billion
in 1993, at the time the largest annual loss in US
corporate history) on total revenues of $192 billion,
and encouraged the media to believe that the mass
layoffs were necessary to avoid bankruptcy. Yet
virtually all of the losses came from ‘restructuring’
charges, that is, the cost of terminating employees
(Lazonick )•
In retrospect, it is clear that these charges were
the cost of ridding the company of its 70-year-old
system of lifelong employment. The industry
standards in ICT, which IBM had played a leading
role in establishing, served to reduce the value to
the company of older employees with experience
accumulated at IBM over the course of their
Contemporary Capitalism
careers and to increase the value of younger
employees who may have had experience working
for other ICT companies. Explicitly reflecting this
change in employment policy, in 1999 IBM
announced that it would replace its traditional
defined-benefit pension plan, which favoured long-
term employees, with a portable ‘cash- balances’
plan that would be much more attractive to younger
employees who did not envisage a lifelong career
with IBM. In December 2004, as its employment
reached 329,000, IBM announced that new
employees would no longer be eligible for the cash-
balances pension fund. Instead the company would
offer them a defined-contribution pension, with the
company matching the employee contribution up to
six per cent of his or her compensation.
From the mid-1990s, with the Old Economy
commitment to its employees out of the way, IBM
adopted all of the elements of NEBM. It shifted out
of hardware into services, and outsourced its
manufacturing. It became by far the leading
patenter in the United States, even as it cut R&D
from the ten per cent of sales that prevailed in the
1980s to six per cent of sales since the mid-1990s,
this change reflecting an expressed shift to product
development and away from basic research. Since
the early 1990s IBM has engaged in patenting much
less to control proprietary technologies, as had been
the case in the past, and much more to gain access
through cross-licensing to technologies controlled
by other companies and to generate intellectual
property revenue ($1.3 billion per year in the
2000 s).
As it rid itself of lifelong employment in the
early 1990s IBM began to extend stock options,
previously reserved for top executives, to a broad
base of employees. In 1990 options outstanding
were only four per cent of all shares outstanding; in
2005, 15.2 per cent. As for distributions to
shareholders, in New Economy fashion, subsequent
to its early 1990s restructuring IBM has favoured
repurchases over dividends. In 1981-90 IBM’s
dividends were 48 per cent and repurchases 12 per
cent of net income; in 1993-2005 dividends were 15
per cent and repurchases 91 per cent. In an effort to
offset dilution of shareholdings as employees
exercise
Contemporary Capitalism
stock options, and more generally to boost its stock
price, in 1995-2005 IBM has spent $62.6 billion on
stock repurchases. Over the same period the
company has spent $56.6 billion on R&D.
As for a New Economy company that, unlike
IBM, started out that way, Cisco Systems, which
since the late 1990s has controlled about 75 per cent
of the Internet router market, is a prime example of
the importance, and implications, of broad-based
stock options in NEBM compensation. Founded in
Silicon Valley in 1984, Cisco grew from about 200
employees at the time of its IPO in 1990 to 40,000
employees during 2000. Throughout its history
Cisco has awarded stock options to virtually all of
its employees. By the end of fiscal 2000 stock
options outstanding accounted for 14 per cent of the
company’s total stock outstanding; by 2005 that
number was 23 per cent. In March 2000, at the peak
of the Internet boom, Cisco had the highest market
capitalization of any company in the world. Under
such conditions its stock options were very lucra-
tive. I have estimated that over the 11 years 1995-
2005 (all years for which data are reported refer to
fiscal year’s end, the last week in July), Cisco
employees, totaling about 256,000 employee-years,
shared $21.5 billion in gains from exercising stock
options, for an average of $84,000 per employee-
year. The annual averages per employee ranged
from less then $9,000 in 2003 to more than
$281,000 in 2000. Of the total amount, the highest
paid executives, totaling 57 executive-years, shared
$893 million, for an average of $15.7 million per
executive-year, with annual averages ranging from
$1.3 million in 2003 to $51.3 million in 2000. The
annual ratios of average top-executive to average
employee gains from exercising stock options
ranged from 36:1 in 1997 to 594:1 in 2005
(Lazonick ). Cisco employees have a clear financial
interest in the company’s stock price, and the
company’s top executives even more so.
Besides using their own stock as a compensation
currency, during the 1990s some New Economy
companies grew large by using their stock, instead
of cash, to acquire other, smaller and typically
younger, New Economy firms in order to
2165
gain access to new technologies and markets. Cisco
mastered this growth-through-acquisition strategy.
From 1993 through 2005 Cisco made 106
acquisitions valued in nominal terms at $46.9
billion, over 80 per cent of which was paid in the
company’s stock rather than cash. In 1999 and 2000
alone, Cisco did 41 acquisitions at a cost of $26.7
billion with over 99 per cent paid in stock
(Carpenter et al. ).
At the same time, like many if not most New
Economy companies, Cisco conserved cash by
paying no dividends. Along with its use of stock as
a combination currency, this payout policy enabled
Cisco to become a giant company in the 1990s
without taking on any long-term debt. Since the
bursting of the Internet bubble from mid-2000
through 2005, however, Cisco has spent $27.2
billion repurchasing its own stock to support its
sagging stock price. In 2004-5, as it spent $19.3
billion on stock repurchases, Cisco used $8.3 billion
in cash - including $6.5 billion of it raised through
its first-ever bond issue - to do 24 acquisitions
rather than continue to use its stock as an
acquisition currency that it would then feel
compelled to offset with repurchases. (Cisco’s
decision to use cash rather than stock for
acquisitions was helped by the Financial
Accounting Standards Board’s 2001 closing of the
‘pooling-of-interests’ loophole that enabled
companies like Cisco that did all-stock acquisitions
to record them on their balance sheets at book
values, which were generally a small fraction of
market values, and thus inflated future earnings.
Nevertheless, in 2002 and 2003, with pooling-of-
interests accounting outlawed, Cisco still used stock
for payment of over 97 per cent of the price of its
nine acquisitions.)
The corporate obsession with supporting its
stock price through massive stock repurchases has
therefore taken hold of companies in the most
innovative sectors of the US economy. As further
notable examples, for the years 1995-2005 Intel
distributed $51.3 billion in repurchases along with
$6.0 billion in dividends compared with R&D
spending of $38.0 billion, while Microsoft
distributed $45.4 billion in repurchases and $38.7
billion in dividends compared with R&D spending
of $40.8 billion.
2166
Microsoft’s dividends included a one-time payment
of $36.1 billion in November 2004.
These companies would argue that R&D
spending and stock repurchases are both working
toward the same end: to enhance the company’s
innovative capabilities by, in the case of R&D,
generating new knowledge, and in the case of
repurchases, competing for high-tech labour capable
of transforming that knowledge into innovative
products and processes. By boosting stock prices, it
is argued, repurchases help to attract, retain and
motivate people who choose to work for companies
in which they are partially compensated with stock
options. In the case of Microsoft the argument has
had less weight since July 2003 when the company
ended its option programme (although many
Microsoft employees still have unexercised options
awarded prior to that date). In the 2000s, moreover,
the extent and location of the talented labour
supplies for which companies like Cisco, IBM, Intel
and Microsoft compete have changed dramatically
with the rise of India and China (Lazonick ). These
dramatically changed labour market conditions for
high-tech labour raise serious questions concerning
which employees benefit from a company’s stock
price performance and for how long, and indeed
whether established high-tech companies even need
to use employee stock options to compete
successfully for high-tech labour.
The offshoring to India and China in the 2000s
of high value-added jobs of software engineers and
computer programmers that it was previously
thought could not go abroad represents the latest
stage in four decades of the globalization of NEBM.
Beginning in the early 1960s Silicon Valley
semiconductor companies were among the first to
offshore assembly to East Asia, and by the early
1970s virtually every US semiconductor
manufacturer had followed suit. When these com-
panies set up plants in places like South Korea,
Taiwan, Hong Kong, Singapore and Malaysia, they
employed, alongside unskilled and predominantly
female assembly labour, indigenous university
graduates as managers and engineers. Over time the
US companies upgraded their facilities in these
locations, and offered more and
Contemporary Capitalism
better employment opportunities for the indigenous
well-educated labour force. As a striking example,
in 1984 Intel claimed that, of its 8,500 employees
outside of the United States (of26,000 employees
worldwide), only 60 were US citizens. This
indigenous employment through foreign direct
investment encouraged the national governments to
increase the level of investment in their already
well-developed systems of higher education, thus
augmenting the future high-tech labour supply
(Lazonick ).
In the 1990s established US ICT companies, led
by IBM and Hewlett-Packard, dramatically reduced
their employment of production workers by
outsourcing manufacturing operations to electronic
manufacturing service providers, also known as
contract manufacturers (Lazonick ). Indeed,
younger companies like Cisco grew rapidly without
doing any in-house manufacturing. Initially the
contract manufacturers would set up operations or
take over existing plants of their customers in the
United States. But a key capability of the leading
contract manufacturers is to shift production that
has become more routine and cost-sensitive to
lower wage areas of the world. In the late 1990s and
early 2000 s the leading contract manufacturers grew
at a rapid pace; at the end of 2005 employment at
the five largest - Flextronics, Solectron, Sanmina-
SCI, Celestica and Jabil Circuit - totalled 260,000.
While we do not know the global distribution of
this labour force, North America accounts for only
an estimated 25 per cent of the sales of these five
companies.
Meanwhile, in the 1990s and 2000s hundreds of
thousands of foreigners, especially Indians, with
college degrees in science and engineering have
migrated to the United States for graduate education
and work experience (Lazonick ). Many acquired
permanent resident (immigrant) status in the United
States, as the US government expanded
employment-based preferences in the issuance of
immigrant visas. For access to US work experience,
however, the most important mode of entry for
high-tech employees has been on non-immigrant H-
1B and L-l visas. The H-1B programme enables
Contemporary Capitalism
non-immigrants, the vast majority of whom have at
least a bachelor’s degree and whose skills are
purportedly unavailable in the United States, to
work in the United States for up to six years. In the
first half of the 2000s about 70 per cent of H-1B
visa holders had science or technology degrees, and
40-50 per cent came from India (the next largest
national group is from China, at less than ten per
cent). The L-l visa programme permits a company
with operations in the United States to transfer
foreign employees to the United States to acquire
work experience, with no limitation of time. In
2001, there were an estimated 810,000 people on H-
1B visas in the United States, and possibly as many
on L-l visas.
Many of these non-immigrant visa holders have
continued to work in the United States by obtaining
permanent resident status. But most have returned
to their native countries with valuable industrial
experience that can be used to start new firms and,
more typically, to work as technical specialists for
indigenous or foreign companies. As a result of
both the migration of US companies abroad in
search of high-tech labour as well as the migration
of foreign high-tech labour to the United States, and
then back to their home countries, in the 2000 s, to
an extent never before imagined, even the best-
educated US high-tech employees compete with a
truly global labour supply for jobs.
Stable and Equitable Growth?
On 16 March 2005 the Semiconductor Industry
Association (SIA) organized a Washington, DC
press conference in which it exhorted the US
government to step up support for research in the
physical sciences, including nanotechnology, to
assure the continued technological leadership of the
United States. Intel CEO Craig Barrett was there as
a SIA spokesperson to warn: ‘U.S. leadership in
technology is under assault’
(Electronic News ):
The challenge we face is global in nature and
broader in scope than any we have faced in the
past. The initial step in responding to this challenge
2167
is that America must decide to compete. If we don’t
compete and win, there will be very serious conse-
quences for our standard of living and national
security in the future.. .U.S. leadership in the nano-
electronics era is not guaranteed. It will take a
massive, coordinated U.S. research effort involving
academia, industry, and state and federal govern-
ments to ensure that America continues to be the
world leader in information technology.
At the time Barrett was a member of the US
National Academy of Sciences Committee on
Prospering in the Global Economy in the 21st
Century, which delved into deficiencies in the
development of science and engineering capabilities
in the United States. Notwithstanding his obvious
concern about these problems from a public policy
perspective, on a radio talk show in February 2006
Barrett (by this time Chairman of Intel) remarked:
‘Companies like Intel can do perfectly well in the
global marketplace without hiring a single US
employee’ (wbur.org ).
The problem with this statement is not that US
workers should have privileged access to jobs at a
US-based company like Intel (which still employs
half of its almost 100,000 employees in the United
States). The problem is that, if a powerful company
like Intel is not dependent on US high-tech
employees for its future labour force, why should it
be concerned about supporting the mass educational
infrastructure in the United States needed to
develop this future labour force? And what does it
mean to say that ‘America must decide to compete’
if, as I would argue is the case (Lazonick ), the most
innovative US corporations have more of an interest
in the Malaysian or Indian system of mass
education than in the US system?
Since the mid-1970s the US mass education
system has been performing poorly in science and
mathematics by the standards of both the advanced
and many developing economies. Such was much
less the case in the three decades or so after the
Second World War, when the Old Economy
corporation was more dependent upon a labour
force that was well-educated at the primary and
secondary levels in the United States. This shift in
the performance of the mass education
2168
parallels the reversal of post-war progress towards a
more equal distribution of income that began about
three decades ago. The much less secure
employment of most US corporate employees in the
shift from OEBM to NEBM would seem to have
contributed to this reversal.
Meanwhile in the 2000s the compensation of the
CEOs of US corporations has long since passed
levels that are at a minimum unseemly and some
would say obscene. The ‘explosion in CEO pay’,
which has been discussed in the United States since
the mid-1980s, seems to have no limits, especially
if, when the corporate stock price falls, it can be
once again pumped up or boards of directors can
replace the ‘lost’ stock option income by other
forms of remuneration such as salaries, bonuses or
restricted stock. The seemingly endless explosions
in top-executive pay reflect the obsession of US
corporate executives with ‘maximizing shareholder
value’ and, cash flow permitting, disgorging
billions upon billions of corporate cash to
shareholders in the forms of repurchases and
dividends to try to make it happen.
In terms of public policy initiatives, virtually
nothing has been done to control top executive pay
in the United States. One well-known attempt was
misguided. In 1993 President Clinton carried out a
campaign promise to control CEO pay by
legislating a cap of $1 million on the amount of
‘non-performance-related’ top executive
compensation - salary and bonus - that a
corporation could claim as a tax deduction. One
perverse result of this law was that companies that
were paying CEOs less than $1 million in salary
and bonus raised these components of CEO pay
towards $1 million, which executives now viewed as
the government-approved CEO ‘minimum wage’.
The other perverse result was that companies
increased CEO option compensation, for which tax
deductions were not in any case being claimed, as
an alternative to exceeding the $1 million salary-
and-bonus cap.
That having been said, the limits to the gains
from stock options, not just for top executives but
also for broad bases of the employees of US high-
tech corporations, would long ago have been
reached if not for the fact that many of these
Contemporary Capitalism
corporations have been in the forefront of innova-
tion. Given the unchallenged sway that the ideology
of ‘maximizing shareholder value’ has over the
governance of these corporations, I have no doubt
that instability, as reflected in the boom and bust of
the stock market in the late 1990s and early 2000 s,
and inequity, as reflected in the worsening of the
distribution of income, will continue to beset the US
economy.
Whether US corporations will remain in the
forefront of innovation that, by necessity, must
underpin long-term economic growth is another
matter. Notwithstanding globalization, the US
model of contemporary capitalism is not a global
model. No other contemporary capitalist economy
has made the commitment to ‘shareholder value’
that is the most distinctive feature of the US model.
Japan has come through the stagnation of the 1990s
as a highly innovative economy, while eschewing
shareholder value ideology and practices
(Lazonick ). In western European nations the
ideology has been tempered by a commitment to
‘social inclusion’; the question is whether the equity
and stability that social inclusion brings can be
harnessed to support innovative enterprise. In the
emerging giants, India and China, the stock market
has come to play a more important, and possibly
dangerous, role. In all of these economies, the
success of innovative companies has been based,
however, not on the stock market, but on the
principles of strategic control, organizational
integration, and financial commitment. Historically
these principles also underpinned innovative
enterprise in the United States. Many corporate
executives who exercise control over resource
allocation in the US economy may, however, have
forgotten these principles, or worse yet, while they
have been busy enriching themselves, they may
have never bothered to leam them.
Bibliography
To conserve both the word-count and flow of this essay, I have
kept bibliographic references to a minimum, indicating
instead works of mine in which these references can be
found.
Contestable Markets 2169

Carpenter, M., W. Lazonick, andM. O’sullivan. 2003. The

stock market and innovative capability in the New Contestable Markets
Economy: The optical networking industry. Industrial
and Corporate Change 12: 963-1034.
Council of Economic Advisers. 2006. Economic report of the
Robert D. Willig
president, 2006. Washington, DC: Executive Office of the
President.
Electronic News. 2005. US could lose race for nanotech
leadership, SIA panel says. 16 March. Online. Available at
http //w Keywords
= 1344283927. Antitrust enforcement; Barriers to entry; Con-
Accessed 5 Sept 2006.
testable markets; Increasing returns; Integer
Lazonick, W. 2003. The theory of the market economy and the
social foundations of innovative enterprise. Economic and problem; Long-run competitive equilibrium;
Industrial Democracy 24: 9^14. Marginal and average cost pricing; Natural
Lazonick, W. 2004. Corporate restructuring. In The Oxford monopoly; Oligopoly; Partial equilibrium;
handbook of work and organization, ed. S. Ackroyd et al.
Perfect competition; Perfectly contestable mar-
Oxford: Oxford University Press.
Lazonick, W. 2005. The institutional triad and Japanese kets; Ramsey optimum; Scale economies; Weak
development [translated into Japanese], In The contem- invisible hand theorem of natural monopoly
porary Japanese enterprise, vol. 1, ed. G. Hook and A.
Kudo. Tokyo: Yukikaku Publishing.
Lazonick, W. 2006a. Evolution of the new economy business
model. In Internet and digital economics, ed. E.
Brousseau and N. Curien. Cambridge: Cambridge JEL Classifications
University Press. D4
Lazonick, W. 2006b. Globalization of the ICT labor force. In
The Oxford handbook on ICTs, ed. R. Mansell et al.
Oxford: Oxford University Press. Contestable markets are those in which competitive
Lazonick, W. 2006c. Innovative enterprise and economic pressures from potential entrants exercise strong
development. In Business performance in
constraints on the behaviour of incumbent suppliers.
twentieth century: a comparative perspective,
ed. Y. Cassis and A. Colli. Cambridge: Cambridge For a market to be contestable, there must be no
University Press. significant entry barriers. Then, in order to offer no
Lazonick, W. 2006d. The US stock market and the profitable opportunities for additional entry, an
governance of innovative enterprise. INSEAD: equilibrium configuration of the industry must
Working paper.
Lazonick, W., and M. O’sullivan. 2000. Maximizing share- entail no significant excess profits, and must be
holder value: a new ideology for corporate governance. efficient in its pricing and in its allocation of
Economy and Society 29: 13-35. production among incumbent suppliers. This is so
New York Times. 1996. The downsizing of America. of a contestable market whether it is populated with
New York: Times Books.
O’sullivan, M. 2000. The innovative enterprise and corporate
only a monopolist or with a large number of
governance. Cambridge Journal of Economics 24: actively competing firms, because it is potential
393-416. competition from potential entrants rather than
Pisano, G. 2006. The science bsiness: Strategy’, competition among active suppliers that effectively
organization, and leadership in biotechnology.
Cambridge, MA: Harvard Business School Press.
constrains the equilibrium behaviour of the
Saxenian, A. 2006. The new argonauts: Regional incumbents.
advantage in a global economy. Cambridge, MA: Perfectly contestable markets (PCMs) are a
Harvard University Press. benchmark for the analysis of industry structure - a
Schumpeter, J. 1934. The theory of economic
development. Oxford: Oxford University Press. benchmark based on an idealized limiting case.
Schumpeter, J. 1950. Capitalism, socialism and Perfectly contestable markets are open to entry by
democracy. 3rd ed. New York: Harper. entrepreneurs who face no disadvantages vis-a-vis
Wbur.org. 2006. Sharpening the cutting edge. On Point. 9 incumbent firms and who can exit without loss of
February. Online. Available at
any costs that entry required to be sunk. The
Accessed 5 Sept 2006. potential entrants have available the same
Whyte, W. 1956. The organization man. New York:
Simon and Schuster.
2170
best-practice production technology, the same input
markets and the same input prices as those available
to the incumbents. There are no legal restrictions on
market entry and exit, and there are no special costs
that must be borne by an entrant that do not fall on
incumbent firms as well. Consumers have no
preferences among firms except those arising
directly from price or quality differences in firms’
offerings.
Potential entrants into perfectly contestable
markets are profit-seekers who respond with pro-
duction to profitable opportunities for entiy. They
assess the profitability of their marketing plans by
making use of the current prices of incumbent
firms. Thus, for example, an entrepreneur will enter
a market if he anticipates positive profit from
undercutting the incumbent's price and solving the
entire market demand at the new lower price.
Potential entrants are undeterred by prospects of
retaliatory price cuts by incumbents and, instead,
are deterred only when the existing market prices
leave them no room for profitable entry.
These features of the behaviour of potential
entrants are key to the workings of perfectly con-
testable markets, and they are fully rational only
where entry faces no disadvantages and is costlessly
reversible. Hence, the benchmark case of perfect
contestability excludes the sunk costs, pre-
commitments, asymmetric information and strategic
behaviour that characterize many real markets and
that are the focus of much current research attention
in the field of industrial organization. With
irreversibilities and the inducements for strategic
behaviour absent, industry structure in PCMs is
determined by the fundamental forces of demand
and production technology.
Of course, this is also true of perfectly compet-
itive markets. However, this most familiar idealized
limiting case is not a satisfactory benchmark for the
study of industry structure in general, because it is
intrinsically inapplicable to a variety of significant
cases. In particular, where increasing returns to
scale are present, perfectly competitive behaviour is
logically inconsistent with the long-run financial
viability of unsubsidized firms.
Perfectly contestable markets can serve in place
of perfectly competitive markets as the general
standard of comparison for the organization
Contestable Markets
of industry whether or not scale economies are
prevalent. Where they are not, perfectly competitive
behaviour is necessary for equilibrium in PCMs,
and, where scale economics do prevail, equilibrium
in PCMs entails behaviour different from that found
in perfectly competitive markets but which none the
less tends to exhibit desirable welfare properties. In
other words, perfect contestability is a
generalization of perfect competition that has strong
implications in significant circumstances where the
latter is inapplicable.
In order to clarify and expand on these ideas,
subsequent sections offer analytic outlines of the
theory of perfectly contestable markets and appli-
cations of the theory to single-product and multi-
product industries. Finally, observations are offered
on the implications of this theory for the
formulation of government policy towards industry.
Perfectly Contestable Markets: Definitions and
Basic Properties
The theory presented here lies in the realm of partial
equilibrium. It deals with the provision of
the set of products N = {1.........n}, some of which
may not actually be produced, and which is a proper
subset of all the goods in the economy. The prices
of these products are represented by vectors p t R+
+, and other prices are assumed to be exogenous and
are suppressed in the notation. Qip) c R"+ is the
vector-valued market demand function for the
products in N, and it suppresses consumers’
incomes which are assumed to be exogenous. For
any output vectory e R", C(v) is the cost at
exogenously fixed factor prices when production is
efficient, the underlying technology is assumed to
be freely available to all incumbent firms and all
potential entrants. Where necessary, C(y) and Q(p)
will be assumed to be differentiable.
Definition 1 A feasible industry configuration is
composed of in firms producing output vectors i
c Rfi, at prices p c R" such that the
markets clear, W™ , y‘ = Q(p), and that each firm at
least breaks even,/) . / - C(y') > 0,i = 1, ..., in.
Contestable Markets
Thus, the industry configuration is taken to be
comprised of m firms, where m can be any positive
integer, so that the industry structure is monopo-
listic if m = 1, competitive if m is sufficiently large,
or oligopolistic for intermediate values of m. The
term ‘feasibility’ refers to the requirements that
each of the firms involved selects a nonnegative
output vector that permits its production costs, C(y')
to be covered at the market prices, p, and that the
sum of the outputs of the m firms satisfies market
demands at those prices.
Definition 2 A feasible industry configuration
over N, with prices p and firms’ outputs y\.. is
sustainable if//. y’ < C(ye), for all pe e R"++, ye <5
R"+,pe <p, and/ < Q(p ).
The interpretation of this definition is that a
sustainable configuration affords no profitable
opportunities for entry by potential entrants who
regard incumbents’ prices as fixed (for a period
sufficiently long to make C( ) the relevant flow cost
function for an entrant). Here, a feasible marketing
plan of a potential entrant is comprised of prices, pe,
that do not exceed the incumbents’ quoted prices, p,
and a quantity vector, /, that does not exceed market
demand at the entrant’s prices, Q(pe). The
configuration is sustainable if no such marketing
plan for an entrant offers a flow of profit, pe . / —
C(/ ), that is positive.
Definition 3 A perfectly contestable market (PCM)
is one in which a necessary condition for an
industry configuration to be in equilibrium is that it
be sustainable.
A PCM so defined may be interpreted, heuris-
tically, as a market subject to potential entry by
firms that have no disadvantage relative to incum-
bents, and that assess the profitability of entry on
the supposition that incumbents’ prices are fixed for
a sufficiently long period of time. Then, since one
requirement for equilibrium is the absence of new
entry, an equilibrium configuration in a PCM must
offer no inducement for entry; that is, it must be
sustainable.
Definition 4 A feasible industry configuration
over N, /?; y1,..., ym, is a long-run competitive
equilibrium ifp . y < C(y) Vy eRn+.
2171
So defined, a long-run competitive equilibrium
has precisely the characteristics usually ascribed to
it. Together, p . / > C(V) and p.y < C{y),MyeRn+,
imply that p . / = C(y') and that they' e arg max v,
[p.y— C(y)]. Thus, each firm in the configuration
takes prices as parametric, chooses output to
maximize profits, earns zero profit, and equates
marginal costs to prices of produced outputs. It is
now easy to show
Proposition 1 A long-mn competitive equilibrium
is a sustainable configuration, so that a perfectly
competitive market is a PCM.
Proposition 2 Sustainable configurations need not
be long-mn competitive equilibria, and a PCM need
not be perfectly competitive.
The simplest example sufficient to prove this
second proposition is an industry producing a single
product with increasing returns to scale over the
relevant range of output. Here, the only feasible
configuration that is sustainable entails one firm
producing the maximal output level y* given by the
intersection of the declining average cost curve with
the industry demand curve, and selling at the price
p* given by the corresponding level of average cost.
This configuration is sustainable because, at a price
equal to or less than p*, sale of any quantity on or
inside the demand curve yields revenue no greater
than production cost; in this range, price does not
exceed average cost. Yet this configuration is not a
long-run competitive equilibrium, as defined above,
because sale of a quantity greater than y* would
earn positive profit if the price could remain at p*,
and because at y* price exceeds marginal cost
which is less than average cost. In fact, in this
example there is no possible long-mn competitive
equilibrium since marginal cost lies below average
cost throughout the relevant range of output levels
given by demand. In contrast, there is a sustainable
configuration.
Hence, Propositions and show that the
sustainable industry configuration is a substantive
generalization of the long-run competitive equi-
librium, and that the PCM is a substantive gener-
alization of the perfectly competitive market. The
2172 Contestable Markets

following propositions summarize some charac- be dispensed with, almost entirely, with exclusive
teristics of equilibria in PCMs. reliance on the freedom of entry that characterizes
PCMs.
Proposition 3 Let /?; y1,..ym be a sustainable
industry configuration. Then each firm must (i) earn Proposition 5 Let /?; y 1 ,..., ym be a sustainable
zero profit by operating efficiently, p . y C(y‘) = 0; configuration. Then, for any y1, ... ,yk with
(ii) avoid cross-subsidization, Ps ■y‘s > C(y‘) — k mk m

C(y‘N s),MSCN (where the vector xT agrees with the E ^' = Ey> Ec (^)> Ec (y).
i= i i= i i= i i i
vector x in components j € T and has zeros for its
other components); (iii) price at or above marginal That is, a sustainable configuration minimizes
cost,/?, > dC(y')ldyj. the total cost to the industry of producing the total
The interpretation of condition (ii) is that the industry output.
revenues earned from the sales of any subset of the This proposition is a generalization to PCMs of
goods must not fall short of the incremental costs of a well-known result for perfect competition. It can
producing that subset. Otherwise, in view of the be interpreted as a manifestation of the power of
equality of total revenues and costs, the revenues unimpeded potential entry to impose efficiency
collected from the sales of the other goods must upon the industry. For example, the proposition
exceed their total stand-alone production cost. In implies that if a monopoly occupies a PCM it must
be a natural monopoly - production by a single firm
PCMs, such pricing invites entry into the markets
must minimize industry cost for the given output
for the goods providing the subsidy.
vector. Thus, Propositions , and are powerful tools
Proposition 4 Let p; y1,..., ym be a sustainable for the analysis of industry structure in PCMs.
configuration with yj < ELi yj ■ Thcn Pj = dCiy^/dyj. Proposition permits information on the properties of
That is, if two or more firms produce a given good production costs to be used to assess the scale and
in a PCM, they must select input-output vectors at scope of firms’ activities in PCMs. Then,
which their marginal costs of producing it are equal Propositions and permit inferences to be drawn
to the good’s market price. about the corresponding equilibrium prices.
The implications of this result are surprisingly
strong. The discipline of sustainability in perfectly PCMS with a Single Product
contestable markets forces firms to adopt prices just
equal to marginal costs, provided only that they are This analytic approach leads to very strong results
not monopolists of the products in qst. Conventional in the single-product case. Propositions - show that
wisdom implies that, generally, only perfect there are only two possible types of sustainable
competition involving a multitude of firms, each configurations in single-product industries. The first
small in its output markets, can be relied upon to type involves a single firm which charges the lowest
provide marginal-cost prices. Here we see that price that is consistent with nonnegative profit. The
potential competition by prospective entrants, rather firm must be a natural monopoly when it produces
than rivalry among incumbent firms, suffices to the quantity that is demanded at this price. And, in
make marginal-cost pricing a requirement of this circumstance, the result maximizes welfare
equilibrium in PCMs, even those containing as few subject to the constraint that all firms in question be
as two active producers of each product. The viable financially without subsidies. Such a second-
conventional view holds that the enforcement best maximum is referred to as a ‘Ramsey
mechanism of full competitive equilibrium requires optimum’.
the smallness of each active firm in its product The second type of sustainable configuration
market, in addition to freedom of entry. We see that involves production by one or more firms of
the smallness requirement can
Contestable Markets
outputs at which both marginal cost and average
cost are equal to price. Here, in the long run, all
active firms exhibit the behaviour that characterizes
perfectly competitive equilibrium. And, of course,
the result involves both (first-best) welfare
optimality and financial viability. Hence, in this
case, Ramsey optimality and the first-best coincide.
This establishes the result that in a singleproduct
industry any sustainable configuration is Ramsey
optimal.
However, in general, because of the ‘integer
problem’, sustainable configurations may generally
not exist. This problem arises, for example, where
there is only one output at which a firm’s marginal
and average costs coincide, and where the quantity
of output demanded by the market at the
competitive price is greater than this, but is not an
integer multiple of that amount. Then, no sus-
tainable configurations exist.
There is, however, a plausible assumption,
supported by empirical evidence, at least to some
degree, that eliminates the integer problem.
Suppose that a firm’s average cost curve has a flat-
bottom rather than being ‘U’-shaped. In particular,
suppose that the minimum level of average cost is
attained not only at one output, but (at least) at all
outputs between the minimum efficient scale, ym,
and twice the minimum efficient scale. Then any
industry output, y1, that is at least equal to ym can be
apportioned among an integer number of firms,
each of which achieves minimum average cost.
Specifically, y1 can be divided evenly among
L);1/y"J firms (where |vj is the largest integer not
greater than x) and each firm’s output, y1/ L);1/y'"J
must lie in the (half-open) interval between y,„ and
2y,„. Hence, in this case, the Ramsey optimum can
either be a sustainable configuration of two or more
firms performing competitively, or a sustainable
natural monopoly. Such a monopoly may either
produce an output at which there are increasing
returns to scale and it will then price at average
cost, or it may produce an output between y,„ and
2y,„ with locally constant returns to scale and adopt
a price equal both to average and marginal cost.
This, together with the preceding argument,
establishes the following result.
2173
Proposition 6 In a single-product industry in
which the firm’s average cost curve has a flat-
bottom between minimum efficient scale and twice
minimum efficient scale, a configuration is
sustainable if and only if it is Ramsey optimal.
This result shows that, under the conditions
described, there is equivalence between welfare
optimality and equilibrium in PCMs. This extends
the corresponding result for perfectly competitive
equilibria to cases of increasing returns to scale.
Moreover, since the behavioural assumptions
required for a PCM are weaker than those under-
lying perfectly competitive markets, the equiva-
lence result is more sweeping. In particular,
Proposition implies that PCMs can be expected to
perform well, whatever the number of firms
participating in equilibrium. It is the potential
competition of potential entrants, rather than the
active competition of existing rivals, that drives
equilibrium in PCMs with a single product to
welfare optimality.
Multi-Product Perfectly Contestable
Markets
In industries that produce two or more goods, a rich
variety of industry structures become possible, even
in PCMs. Here, while the constraints imposed upon
incumbents by perfect contestability are not nearly
as effective in limiting the range of possible
outcomes as they are in single product industries,
they nevertheless provide a helpful basis for
analysis. In particular, Propositions - indicate
connections among various qualitative properties of
multi-product cost functions and various elements
of industry structure in PCMs. These connections
constitute one theme of this section. The other
theme is the normative evaluation of the industry
structures that arise in multi-product PCMs.
Before proceeding, it may be useful to provide
definitions of some of the multiproduct cost prop-
erties that are used in the analysis.
Definition 5 Let P = [ 7',, ... , Tk) be a nontrivial
partition of S C N. There are (weak)
2174 Contestable Markets

economies of scope at ys with respect to the partition over the set of goods it produces. When strict
P if ffki=i C(yTi) > (>)C'(js). If no partition is economies of scope are present, there must be at
mentioned explicitly, then it is presumed that least one multi-product firm in any PCM that
T, = {/}• supplies more than one good.
The second basic question that arises is whether
Definition 6 The degree of scale economies there can be two or more firms actively producing a
defined over the entire product set, N {1,.. n), at}’, particular good in a PCM. If there are, then, by
is given by = C(y)/y.VC(y). Proposition , marginal cost pricing must result. The
Returns to scale are said to be increasing, con- answer depends upon the availability of product-
specific scale economies.
stant or decreasing as SA- is greater than, equal to or
less than unity. This occurs as the elasticity of ray Proposition 8 Any product with average incre-
average cost with respect to t is negative, positive or mental costs that decline throughout the relevant
zero; where ray average cost is RAC (O'0) - range (that is, that offers product-specific increasing
C{ty°)/t. returns to scale) must be produced by only a single
Definition 7 The incremental cost of file product firm (if it is produced at all) in a PCM. Further,
set T C Naty is given by IC j{y)= Ciy) C (yy _ f). such a product must be priced above marginal cost,
The average incremented cost of T is AICr (y) unless the degree of product- specific scale
=ICifyVEy e 7}’j. economies is exactly one.
The average incremental cost of T is decreasing, Thus, regardless of file presence or absence of
increasing or constant at y if MCftyT-yN T) is a economies of scope, globally declining average
decreasing, increasing or locally constant function incremental costs imply that a product must be
of t at t = 1. These cases are labelled respectively, monopolized in a PCM. It is an immediate corollary
increasing, decreasing or constant returns to the that if all goods in the set N exhibit product-
scale of the product line T. The degree of scale specific scale economies, and if there are economies
economics specific to T is of scope among them all, then the industry is a
natural monopoly that must be monopolized in a
PCM.
dC(y)
ICr(y)/$>f Another route to this result is provided by the
it T c l - ‘weak invisible hand theorem of natural monopoly’.

Definition 8 A cost function C(y) is trans-rayProposition 9 Trans-ray convexity of costs

together with global economies of scale imply nat-
convex through some point y* = (y*v-yn) if there
ural monopoly. If, in addition certain other technical
exists at least one vector of positive constants w! , ...
conditions are met, a monopoly charging Ramsey-
, w„ such that for every two output vectors ya = (yf,
optimal prices is a sustainable configuration.
■■■yar) andy6 = (y\, .. ,yhn) that lie on the hyperplane
Yyvy, = w0 through point y*, C[kya + (1 - k)yb] < In general, there may exist natural monopoly
kC(y>a) + (1 - k)C(yh for k e (0, 1). situations in which no sustainable prices are pos-
sible for the Ramsey optimal product set. Further,
In view of the general result that sustainable
configurations minimize industry-wide costs even where sustainable prices exist, the Ramsey
optimal prices may not be among them. However,
(Proposition ), these cost properties permit infer-
under the conditions of the weak invisible hand
ences to be drawn about industry structure in multi-
theorem, the Ramsey optimal prices for the Ramsey
product PCMs. The first issue that arises is when
optimal product set are guaranteed to be sustainable,
multicommodity production is characteristic of
so that PCMs are consistent with (second-best)
equilibrium in a PCM.
welfare optimal performance by a natural
Proposition 7 A multi-product firm in a PCM monopoly.
must enjoy (at least weak) economies of scope PCMs will yield first-best welfare optimality if
there exist sustainable configurations with at least
Contestable Markets
two firms actively producing each good. For in this
case Propositions and guarantee industrywide cost
efficiency and marginal-cost pricing of all products.
Here, two issues must be resolved: Does industry-
wide cost minimization require at least two
producers of each good? And if so, do sustainable
configurations exist?
The existence problem can be solved in a man-
ner analogous to its solution in the case of single-
product industries: by assuming that ray average
costs remain at their minimum levels for output
vectors that lie (on each ray) between minimum
efficient scale and twice minimum efficient scale.
And the presence of at least two producers (or one
operating in the region where constant returns
prevail) of each good is assured if the quantities
demanded by the market at the relevant marginal-
cost prices are no smaller than minimum efficient
scale (along the relevant ray) and if the cost func-
tion exhibits trans-ray convexity.
Policy Implications of PCMS
One of the principal lessons of the analysis of PCMs
is that monopoly does not necessarily entai I
welfare losses. Rather, the ‘weak invisible hand th’
shows that under certain conditions sustainability
and Ramsey optimality are consistent, so that the
total of consumers’ and producers’ surpluses may
well be maximized (subject to the constraint that
firms be self-supporting) in the equilibrium of a
monopoly which operates in contestable markets.
Even stronger results follow from the discussed
results that under certain conditions sustainability
and a first-best solution are consistent in an
oligopoly with a small number of firms. When
minimization of industry cost requires that each
good be produced by at least two firms,
sustainability requires any equilibrium to satisfy the
necessary conditions for a first-best allocation of
resources. Thus, in these cases, the invisible hand
has the same power over oligopoly in perfectly
contestable markets that it exercises over a perfectly
competitive industry.
This theory suggests that in a market that
approximates perfect contestability, the general
public interest is well-served by a policy of
2175
laissez-faire rather than active regulation by
administrative or antitrust means. Small numbers of
large firms, vertical and even horizontal mergers
and other arrangements which have traditionally
been objects of suspicion of monopolistic power,
are rendered harmless and perhaps even beneficial
by the presence of contestability.
On the other hand, contestability theory does not
lend support to the proposition that the unrestrained
market automatically solves all economic problems
and that virtually all regulation and antitrust activity
entails unwarranted and costly intervention. The
economy of reality is composed of industries which
vary widely in the degree to which they
approximate the attributes of perfect contestability.
Before the theory of contestability can be
legitimately applied to reach a conclusion that
intervention is unwarranted in a specific sector, it
must first be shown that the sector ties unprotected
by entry barriers and that the force of potential entry
therefore actively constrains the behaviour of
incumbent firms. This then becomes the appropriate
first stage in an analysis of efficient government
policy towards an industry. Only where the
conditions of contestability are found to
characterize the reality of an industry can there be
validity in applying the normative conclusions of
contestability theory concerning the power of
potential competition actually to enforce efficient
behaviour on incumbents.
Even where contestability is absent in reality,
the formulation on efficient regulation can be use-
fully guided by the theory of contestable markets
instead of the theory of perfectly competitive mar-
kets. The first-best lesson of the perfect competition
model, calling for prices to be set equal to marginal
costs, has no doubt contributed to the common
regulatory ethos which equates price to some
measure of cost. This doctrine has been used
frequently where it is completely inappropriate and
without logical foundation, that is, in cases where
prices should be based on demand as well as cost
considerations, because of the presence of
economies of scale and scope. Such arbitrary
measures as fully distributed costs cannot substitute
for marginal cost measures as decision rales for
proper pricing, and the search for a substitute is a
remnant of inappropriate reliance on the
2176
model of perfect competition for guidance in
regulation.
In contrast, contestability theory suggests cost
measures that are appropriate guideposts for reg-
ulated pricing - incremental and stand-alone costs.
The incremental cost of a given service is, of
course, the increment in the total costs of the
supplying firm when that service is added to its
product line. In perfectly contestable markets, the
price of a product will lie somewhere between its
incremental and its stand-alone cost, just where it
falls in that range depending on the state of demand.
One cannot legitimately infer that monopoly power
is exercised from data showing that prices do not
exceed stand-alone costs, and stand-alone costs
constitute the proper cost- based ceilings upon
prices, preventing both cross-subsidization and the
exercise of monopoly power. A simple example
will show why this is so.
First, suppose that a firm supplies two services,
A and B, which share no costs and that each costs
10 units a year to supply. The availability of
effective potential competition would force reve-
nues from each service to equal 10 units a year. For
higher earnings would attract (profitable) entry, and
lower revenues would drive the supplier out of
business. In this case, in which common costs are
absent, incremental and stand-alone costs are equal
to each other and to revenues, and the competitive
and contestable benchmarks yield the same results.
Next, suppose instead that of the 20 unit total
cost 4 are fixed and common to A and B, while 16
are variable, 8 of the 16 being attributable to A and
8 to B. If, because of demand conditions, at most
only a bit more than 8 can be generated from
consumers of A, then a firm operating and surviving
in contestable markets will earn a bit less than 12
from B. These prices lie between incremental costs
(8) and standalone costs (12), are mutually
advantageous to consumers of both services, and
will attract no entrants, even in the absence of any
entry barriers. In contrast, should the firm attempt
to raise the revenues obtained from B above the 12
unit stand-alone cost, it would lose its business to
competitors willing to charge less. Similarly, the
same fate would befall it in contestable markets if it
priced B in a way that earned more than
Contingent Commodities
8 plus the common cost of 4, less the contribution
towards that common cost from service A.
Thus, the forces of idealized potential competi-
tion in perfectly contestable markets enforce cost
constraints on prices, but prices remain sensitive to
demands as well. Actual competition and potential
competition are effective if they constrain rates in
this way, and in such circumstances regulatory
intervention is completely unwarranted. But if, in
fact, market forces are not sufficiently strong, then
there may be a proper role for regulation of natural
monopoly, and the theoretical guidelines derived
from the workings of contestable markets are the
appropriate ones to apply. That is, prices must be
constrained to lie between incremental and stand-
alone costs. (This is the approach recently adopted
by the Interstate Commerce Commission to
determine maximum rates for US railroad services,
and the method has already withstood appeals to the
federal courts.)
See Also
► F....
Bibliography
Baumol, W..L and R.D. Willig. 1986. Contestability:
Developments since the book. Oxford Economic Papers 38:
9-36.
Baumol, W.J., J.C. Panzar, and R.D. Willig. 1982. Contestable
markets and the theory of industry structure. New York:
Harcourt Brace Jovanovich.
Baumol, W.J., J.C. Panzar, and R.D. Willig. 1985. On the theory
of perfectly contestable markets. In New developments in
the analysis of market structure, ed. J. Stightz and F.
Mathewson. New York: Harcourt Brace.
Contingent Commodities
Zvi Safra
Keywords
Contingent commodities; Convexity; General
equilibrium; Risk; Risk aversion; Uncertainty
Contingent Commodities
JEL Classification
D5
The theory of general competitive equilibrium was
originally developed for environments where no
uncertainty prevailed. Everything was certain and
phrases like ‘it might rain’ or ‘the weather might be
hot’ were outside the scope of the theory. The idea
of contingent commodity, that was introduced by
Arrow ( ) and further
developed by Debreu ( ), was an ingenious
device that enabled the theory to be interpreted to
cover the case of uncertainty about the availability
of resources and about consumption and production
possibilities. Basically, the idea of contingent
commodity is to add the environmental event in
which the commodity is made available to the other
specifications of the commodity. With no
uncertainty every commodity is specified by its
physical characteristics and by the location and date
of its availability. It is fairly clear, however, that
such a commodity can be considered to be quite
different where two different environmental events
have been realized. The following examples clarify
this: an umbrella at a particular location and at a
given date in case of rain is clearly different from
the same umbrella at the same location and date
when there is no rain; some ice cream when the
weather is hot is clearly different from the same ice
cream (and at the same location and date) when the
weather is cold; finally, the economic role of wheat
with specified physical characteristics available at
some location and date clearly depends on the
precipitation during its growing season. Thus,
specifying commodities by both the standard
characteristics and the environmental events seems
very natural, whereas the role of the adjective in
‘contingent commodities’ is simply to make it clear
that one is dealing with commodities the availability
of which is contingent on the occurrence of some
environmental event. With this specification the
model with contingent commodities is very similar
to the classical model of general competitive
equilibrium and thus questions like the existence of
equilibrium and its optimality (with the additional
aspect of efficient allocation of risk bearing) are
answered in a
2177
similar way. Note that, although this model deals
with uncertainty, no concept of probabilities is
needed for its formal description.
To make things more explicit we look at a
simple model with contingent commodities.
Assume that, without referring to uncertain events,
there arc k > 1 commodities, indexed by i, and that
there are n > 1 mutually exclusive and jointly
exhaustive events (or states of nature ), indexed by
s, where k and n are finite. Thus a contingent
commodity is denoted by x i s and the total number of
these commodities is hi, which is greater than k but
still finite. Consumption and production sets are
thus defined as subsets of the foi-dimensional
Euclidean space, and the economic behaviour of
firms and consumers naturally follows from profit
maximization (by firms) and utility maximization
(by consumers). The price pIS of the contingent
commodity xis is the number of units of account that
have to be paid in order to have the zth commodity
being delivered at the vth event. It is assumed that
the market is organized before the realization of the
possible events. Thus payment for the contingent
commodity xiy is done at the beginning while
delivery takes place after the realization of events
and only in case event s has occurred. Note that the
price of the (certain) rth commodity, that is, the
number of units of account that have to be paid in
order to have the zth commodity for sure, is the sum
over 5 of the prices pis. For example, assume that
the price of one quart of ice cream if the weather is
hot is $2.00, the price of one quart of the same ice
cream if the weather is cold is $1.00 and that n = 2
(either it is hot or cold). Thus the price of having
one quart of that ice cream for sure is $2.00 + $1.00
= $3.00.
It should be noted that, although the probabili-
ties of the possible events do not explicitly enter the
model, the attitude towards risk of both consumers
and producers is of interest and does play a signif-
icant role in this framework. The preference rela-
tions of consumers defined on subsets of the An-
dimensional Euclidean space reflect not only their
‘tastes’ but also their subjective beliefs about the
likelihoods of different events as well as their
attitude towards risk. Convexity of consumers’
preferences, for example, is interpreted as risk aver-
sion while, in the same spirit, profit maximization
2178 Contingent Valuation

of firms is interpreted as risk neutrality. It should be reliable revealed preference data about demands
mentioned that both Arrow and Debreu basically for certain types of goods. A number of
assume expected utility maximizing behaviour, in significant lawsuits have promoted their use in
the sense of the Savage ( ) framework. A more estimating demand for environmental goods.
general approach to such preference relations can They are also used by transportation economists,
be found in Yaari ( ), where, again, convexity is health economists and market researchers.
taken to mean risk aversion. Although the degree of acceptance of these
A unified and more formal treatment of time methods varies, many economists agree that a
and uncertainty using contingent commodities can value based on stated preferences derived from
be found in Debreu ( , ch. 7). Radner carefully conceived and executed research is
( ) presents an extension of the above model almost certainly preferable to no number at all.
to the case in which different economic agents have
different information.
Keywords
See Also Conjoint analysis; Construct validity assess-
ments; Contingent valuation; Discrete- choice
► Arrow-Debreu Model of General Equ
►J models; Dummy variables; Environmental
economics; Expectations; Health economics;
Household surveys; Indirect utility functions;
► Uncertainty and General Equ
Logit models; Marginal utility of income;
Maximum likelihood; Neuroeconomics; Probit
Bibliography models; Random utility models; Revealed
Arrow, K.J. 1953. Le role de valeurs boursieres pour la repartition preference; Scope tests; Stated preference;
la meilleure des risques. Econometrie. Paris: CNRS. English Willingness to pay
translation ‘The role of securities in the optimal allocation of risk-
bearing’ in Review of Economic Studies (1964); reprinted in
K.J. Arrow, Essays in the theory of risk-bearing. Chicago:
Markham, 1971. Debreu, G. 1953. Une economic de Tincertain. JEL Classifications
Mimeo, Paris: Electricite de France. Q51
Debreu, G. 1959. Theory of value. New York: Wiley. Radner, R.
Most economists would agree that no researcher
1968. Competitive equilibrium under uncertainty. Econometrica
36: 31-58.
should prefer demand data from hypothetical mar-
Savage, L. J. 1954. Foundations of statistics. New York: Wiley.
kets if data concerning the identical goods or
Yaari, M.E. 1969. Some remarks on measures of risk aversion
services, based on real markets, are readily avail-
and their uses. Journal of Economic Theory 1:315-329.
able. However, there are many situations when even
the cleverest empirical economist cannot come up
with revealed preference data from actual markets
that can be relied upon for information about
household demands for some types of goods.
Environmental goods are one class of goods where
Contingent Valuation real-market demands sometimes cannot be
measured adequately. In the 1980s, environmental
Trudy Ann Cameron
economists began in earnest to exploit stated-
preference demand information, usually collected
using household surveys. This demand information
is used primarily to produce utility-theoretic
Abstract measures of the social benefits of environmental
‘Contingent valuation’ methods are used to protection measures for benefit-cost analyses.
generate demand data, usually from household
surveys, when real markets do not supply
Contingent Valuation
Environmental economists called these methods
‘contingent valuation methods’ (CVM) because the
valuations were elicited ‘contingent upon the
conditions described in the survey’. Research that
focused on the development and assessment of
CVM in environmental economics was well under
way by the mid-1980s. However, two events in
1989 thrust the method to the forefront of the field.
First, the Exxon Valdez, an ocean tanker, ran
aground in Prince William Sound in Alaska, spilling
11 million gallons of oil in an environmental
disaster that attracted a huge amount of media
attention worldwide. Second, just a few months
later, the US Court of Appeals held that the
economic damages assessed for spills of oil or other
hazardous substances could include ‘lost passive use
values’, and that these values could legally be
measured via CVM.
Plaintiffs and defendants in the Exxon Valdez
case thus had a big incentive to advocate and
derogate CVM, respectively. For at least a time, the
discussion in the literature teetered on the brink of
losing its polite academic tone. Given the escalation
of the controversy over CVM, the US National
Oceanic and Atmospheric Administration (NOAA)
convened a panel of experts (untainted by any active
role in the Exxon Valdez litigation) to assess CVM.
This exercise, by Arrow et al. ( ), produced a
set of pronounce
ments concerning best practices for the conduct of
CVM studies. While the 1993 NOAA Panel report
cannot be considered the last word on CVM, it was
very influential, and there has since been strong
pressure on researchers either to conform to the
NOAA best practices or to fully justify any
departures from them.
As a result of the Exxon Valdez case, much
doubt about the reliability of stated preference data
led to numerous comparisons of the implications of
stated and revealed preference data (for example,
Carson et al. ). CVM works best when respondents
have a clear sense of the consequences of their
choices - in terms of both their own budgets and the
exact nature of the good that they are being asked to
consider paying for - and when they are reasonably
familiar with market transactions involving that
good. This means that CVM is, unfortunately, most
successful when it is
2179
least needed. The challenge for researchers is to
ensure that demand information gathered using
CVM, in less-than-ideal contexts, is as valid and
useful as possible.
Myriad biases and qualifications may afflict
poorly executed CVM studies. A partial list
includes incentive compatibility, hypothetical bias
(if the choice is perceived to have absolutely no real
consequences), strategic bias (when people try to
manipulate the outcome by misrepresenting their
preferences), non-response bias (since people
cannot be compelled to participate), starting-point
bias (for surveys with iterative bids), interviewer
bias (for in-person surveys), and information bias
(when some portion of the value is constructed
during the survey where it did not exist before).
Other problems include yea-saying, part-whole bias
or embedding, scenario rejection, and the potential
for respondents not to pay sufficient attention to
their real budget constraints.
Choice formats have been an important issue in
the development of CVM. For example, in some
early applications of CVM survey respondents were
asked directly to identify the single highest dollar
amount that they would pay to obtain some change
in conditions. These were called open- ended CV
questions. Researchers quickly realized that such a
task was difficult for consumers who were
unfamiliar with naming their own price, especially
for goods they may never before have thought much
about having to pay for. CVM elicitation techniques
evolved fairly quickly to a dichotomous-choice
format, where respondents are given a choice
between two states of the world. One state is
typically the status quo, while the other involves a
specified change or set of changes (such as an
improvement in environmental quality, or some
other rationed public good) that come at a price
(typically a lump-sum payment). This binary choice
format was found to fit naturally into a random
utility model (RUM) framework that had also
become a popular approach to consumer choice
problems, both real and hypothetical, in the
transportation mode- choice literature and
elsewhere in economics.
Respondents’ preferences, based on then-
answers to dichotomous-choice CV questions,
2180
can be characterized either in an ad hoc fashion or
in a more formal utility-theoretic framework. One
standard approach is to specify an indirect utility
function shared by all respondents. In its simplest
form, the level of indirect utility' is assumed to
depend on the individual’s net income under each of
the two alternatives, and upon a discrete indicator of
whether there is a change in the rationed public
good, or no change, under each alternative.
Respondents can choose the environmental
improvement programme along with its associated
cost (implying lower net income), or decline the
environmental improvement programme in order to
avoid the cost (preserving their net income). If a
respondent prefers the programme with its
associated cost, the researcher assumes that the
respondent's utility' level is higher under that
alternative. Equivalently, this means that the net
indirect utility' associated with the programme
alternative is positive.
A discrete-choice econometric model, typically
involving a binary logit estimator, is used to
estimate the sample average marginal utilities of
(</) net income and (b) the discrete bundle of
changes represented by the programme in question.
It is of course possible to allow for heterogeneity
across the sample in these marginal utilities. Most
often, heterogeneity is introduced by allowing the
otherwise scalar marginal utility associated with
going from ‘no programme’ to ‘programme’ to
become a systematically varying parameter. Of
course, if the identical programme is offered to all
respondents, it is not possible to allow this marginal
utility to vary with attributes of the programme.
However, it can easily be allowed to vary with
characteristics of the respondent.
Less commonly, the marginal utility of income
is also allowed to vary across respondents, either
with the respondent’s income (to allow' for
diminishing marginal utility' of income) or with
other respondent characteristics. However, there is a
premium on simplicity for the marginal utility of
income, stemming from the need to use the
estimated marginal utility' of income parameter (s)
to recover demand information. For this reason,
many researchers will, if it is justified by the data,
prefer a choice model that is linear and
Contingent Valuation
additively separable in net income under each
alternative.
Linearity and additive separability' in income is
convenient (w'hen warranted) because the willing-
ness to pay (WTP) function associated with the
fitted model is given by the marginal rate of sub-
stitution (MRS) between the programme and
income. This MRS is given by the ratio of the
marginal utility of the programme to the marginal
utility of income, producing a result that can be
expressed in dollars per 'unit' of the programme,
where the program indicator is either zero or 1 in
the simple binary case. In the non-stochastic case,
for a simple dichotomous choice CVM model, this
is a single number - 'WTP for the program’ - if the
researcher has assumed homogeneous preferences
throughout the sample.
Some extra empirical housekeeping is necessary
w'hen it is acknowledged that this point estimate is
constructed as the ratio of two estimated quantities,
each of which (due to the use of maximum
likelihood estimation methods for the logit or probit
model) is an asymptotically normally distributed
random variable. In theoretical terns, the ratio of
two normally distributed random variables has an
undefined mean, because zero is a possible value
for the denominator. As a practical matter, some
researchers use simulation methods to build up a
sampling distribution for the estimated WTP. It is
possible to use packaged soft- ware to make a large
number of random draws from the joint distribution
of the logit or probit parameters (based on the
estimated parameter point estimates and the
parameter variance- covariance matrix). One can
then build up a sampling distribution for the needed
ratio. Other strategies for dealing with this
inconvenience involve estimating the model in
‘WTP-space’ rather than ‘utility-space’ or
employing the new'er mixed logit (random-
parameters logit models) and stipulating that the
marginal utility of income parameter be distributed
lognormal (since it should be strictly positive, on
average), rather than normal, so that the potential
divide-by-zero problem goes away.
Over the 1990s contingent valuation researchers
in environmental economics gradually made better
contact with their counterparts working in other
literatures who w'ere confronted
Contingent Valuation
by similar problems where there is a lack of market
data for products or public goods that need to be
valued. In the transportation literature, researchers
had grappled early on with the problem of
forecasting demand for public transportation
projects, or new types of vehicles, that did not yet
exist. Researchers began to introduce hypothetical
new transportation options which could be
characterized in the same terms as existing options
(in ‘attribute space’) but which had some attributes
that lay well outside the set of existing options on
some dimensions, or which involved attributes that
were not relevant for existing options (such as travel
range or recharge time for prospective electric
vehicles). One key difference from contingent
valuation was the practice of asking survey
respondents to consider more than just ‘the status
quo versus a single alternative’. Furthermore, the
alternatives were more richly specified. Instead of
using simply a dummy variable to indicate whether
the policy, programme or public good was present
or absent, each alternative was characterized in
terms of an array of attributes.
Similar problems were also being addressed in
the marketing literature, particularly in the context
of ‘pre-test’ marketing. Companies considering
whether to develop and introduce new products
needed to know in advance about the likely demand
for these products, perhaps as a function of
alternative possible product configurations. Market
researchers developed a set of techniques they
called ‘conjoint analysis’. In the marketing
literature, the specifications used for the choice
models were initially very ad hoc. Little attention
was paid to the interpretation of the estimated
coefficients as marginal utilities, and simple linear
and additively separable specifications were very
common. The slope coefficients were known as
‘part-worths’ rather than marginal utilities. How-
ever, much was learned about the degree of con-
sistency between planned purchase behaviour and
actual purchase behaviour.
CVM has also recently grown in popularity in
other sub-disciplines, notably health economics.
However, Smith ( ) surveys that literature and
suggests that researchers in that field have not yet
developed a set of best practices for the use of
2181
CVM with the types of choices that are most
common in health economics contexts.
In the transportation and marketing literatures,
the desired demand information often spanned a
number of possible alternative products or services.
Stated preference studies were often conducted not
just to determine respondents’ willingness to pay
for a single well-defined good but to understand
how willingness to pay might be affected by
variations in the mix of attributes making up a
prospective good. It was often necessary to
anticipate demands for differentiated products,
where each product could be characterized as a
bundle of attributes and the levels of these attributes
differed across alternatives.
In contrast, more of the impetus for CVM non-
market valuation research in environmental
economics stemmed from a number of significant
lawsuits. In the legal context, there is a premium on
simplicity in economic modelling so as not to
confuse the jury or the judge. It is often best to
produce one value for one clearly defined com-
modity. (Providing a judge or jury with a function
that describes demand, where WTP depends upon a
wide array of attributes, conditions or respondent
attributes, can actually be a liability when attorneys
are trying to make a simple, clear and persuasive
case. In a legal context, it is most incisive to value
one thing, and to value it as precisely as possible.)
Eventually, however, environmental economists
began to acknowledge the value of understanding
the heterogeneity in demands for environmental
goods, since this knowledge can be very helpful to
policymakers who wish to consider how different
versions of a policy might affect different
constituencies.
There are many commonalities between the
tasks faced by environmental economists and those
faced by transportation economists and market
researchers, but there is one key difference. In
transportation economics and market research, it is
often the case that the public transit system in
question will actually be built, or the new product
will actually be developed and put on the market.
There is an opportunity to go back and see whether
the level of demand predicted by the stated
preference study actually materializes when there is
a real market. In the environmental
2182
economics literature, there are typically fewer
opportunities to ‘validate’ the stated preference
demand information with revealed preferences for
the same product.
One common expectation for a good CVM study
is now a demonstration that the demand function
that has been estimated should ‘walk and talk’ like a
demand function. For example, is willingness to pay
to preserve big-game hunting opportunities lower,
on average, for elderly women than for middle-aged
males? Is willingness to pay to preserve air quality
higher for people with lung disease or asthma, or
for people who have family members with these
illnesses? These tests are commonly called
‘construct validity’ assessments. Contingent
valuation studies that pass a battery of plausibility
tests such as these can generally be viewed as more
reliable.
Another common test of contingent valuation
estimates that these stated preference demand
functions are typically expected to satisfy is some-
thing called a ‘scope test’. This means that, on
average, respondents’ willingness to pay for an
alternative that involves more of the ‘good’ in
question should be greater than that for an alter-
native that involves less of the ‘good’. It is of
course possible that marginal utility may be positive
(as the scope test implies) at low levels of the good,
but also that it may go to zero if the quantity of the
good is high enough for satiation to set in, and there
is no theoretical basis for expecting willingness to
pay to be proportional to the amount of the good in
question.
CVM data can also sometimes be pooled with
actual choice data. This can allow portions of the
underlying indirect utility function to be anchored
upon real choices, even though the variability in
attributes in the real alternatives may not span the
full domain relevant to pending policy decisions.
The stated choice questions can be used to extend
the domain of the estimated demand function.
While economists will remain uncomfortable
about reliance upon stated preference information,
many now acknowledge that there are cir-
cumstances where stated preference data are all
Contingent Valuation
that can be collected. In fact, the need for econo-
mists to rely upon survey data (what people say as
opposed to what they actually do in markets) is now
being acknowledged in the other contexts in
economics. For example, expectations about future
income or life expectancy figure prominently in a
number of economic theories. These expectations
typically cannot be measured directly, but can
sometimes be elicited using surveys and put to good
use empirically (see Manski )•
It is worth noting that not just stated preference
data but also revealed preference data can be highly
variable in its quality. Much revealed preference
demand data is also drawn from consumer surveys.
It is not always clear that the individual respondent
sees the need for accuracy and completeness to be
as critical as researchers using the data might hope.
In consumer expenditure surveys, for example,
interviewers prompt subjects for different types of
expenditures, but the enthusiasm and engagement of
the survey subject often determines the acciuacy
and completeness of the data. Rather than viewing
revealed preference data as of unambiguously high
quality and stated preference data (such as that
produced by CVM) as of unambiguously low
quality, it may be prudent simply to acknowledge
that both types of data can have their problems.
A partial list of current frontiers in CVM-related
research is possible. These frontiers include
continuing assessment of (a) alternative elicitation
formats (there are many candidates beyond the
simple NOAA-recommended binary choice format),
(b) the choice contexts presented to subjects, (c) the
effects of allowing subjects to express uncertainty
about their choices, (d) the effects of practice and
fatigue when several CVM questions are presented
to each respondent, (<?) integrating stated choices
with additional types of real market information, (/)
how the degree of complexity of the CVM choice
scenarios interacts with the cognitive capacity of the
subject and/or the subject’s inclination to pay
attention, and (g) the neuroeconomics of real as
opposed to stated choices.
Contingent Valuation
Two of the classic books on CVM are Cum-
mings et al. ( ) and Mitchell and Carson
( ). Following the Exxon Valdez case, a pro
vocative debate was featured in the Journal of
Economic Perspectives (Diamond and Hausman ;
Hanemann ; Portney ). McFad- den ( ) raised some
specific concerns about
the reliability of CVM data in the context of an
empirical application to the existence value of
wilderness areas in the western United States. In the
intervening years, however, research concerning
CVM has continued apace. Helpfiil expositions and
discussions of recent innovations have made their
way into textbook form, with one particularly useful
summary being provided in Chapter 6 of Freeman (
). A brief, accessible
and very helpful introduction to CVM for non-
specialists is contained in Carson (
Louviere et al. ( ) offer a comprehensive dis
cussion of stated choice methods broadly defined,
including experimental design, modelling, esti-
mation and combining revealed and stated prefer-
ence data, with illustrations in marketing,
transportation and environmental economics. An
inventory of the wide range of practical issues to
consider in actually implementing a C VM study is
provided by Boyle ( ), while Holmes and
Adamowicz ( ) update the state of the art for
attribute-based (conjoint choice) methods.
There is still considerable variation in individual
researchers’ levels of comfort with CVM and stated
preference data more generally. We might
reconsider the question posed by Diamond and
Hausman ( ): ‘ Contingent valuation - is some
number better than no number?’ There are now
many economists who would agree that a value
based on stated preferences - from a study that is
carefully conceived and executed, based on a suf-
ficiently large sample that is representative of its
intended population, that has been put through a
battery of consistency and validity assessments, and
that produces an implied demand function that
behaves the way we would expect a ‘real’ demand
function to behave - is almost certainly better than
no number. This is especially true when ‘no
number’ creates the risk that a value of
2183
zero would otherwise be imputed, by default, for
use in policy decisions.
See Also
► Environmen
Bibliography
Arrow, K.., R. Solow, P.R. Portney, E.E. Learner, R. Radner, and
H. Schuman. 1993. Report of the NOAA panel on contingent
valuation. Federal Register 58:4601-4614.
Boyle, K..J. 2003. Contingent valuation in practice. In A primer
on nonmarket valuation, ed. P.A. Champ, K..I. Boyle,
andT.C. Brown. Boston: KJuwer Academic Publishers.
Carson, R.T. 2000. Contingent valuation: A user’s guide.
Environmental Science & Technology 34: 1413-1418.
). Carson, R.T., N.E. Flores, K.M. Martin, and J.L. Wright. 1996.
Contingent valuation and revealed preference methodologies:
Comparing the estimates for quasipublic goods. Land
Economics 72: 80-99.
Cummings, R.G., D.S. Brookshire, and W.D. Schulze, eds. 1986.
Valuing environmental goods: An assessment of the
contingent valuation method. Totowa: Rowman and
Allanheld.
Diamond, P.A., and J.A. Hausman. 1994. Contingent valuation -
Is some number better than no number? Journal of
Economic Perspectives 8(4): 45-64.
Freeman, A.M.I. 2003. The Measurement of Environmental
and Resource Values: Theory and Methods. Washington,
DC: Resources for the Future.
Hanemann, W.M. 1994. Valuing the environment through
contingent valuation. Journal of Economic Perspectives
8(4): 19-43.
Holmes, T.P., and W.L. Adamowicz. 2003. Attribute-based
methods. In A primer on nonmarket valuation, ed. P.A.
Champ, K.J. Boyle, and T.C. Brown. Boston: Kluwer
Academic Publishers.
Louviere, J.J., D.A. Hensher, and J.D. Swait 2000. Stated choice
methods. New York: Cambridge University Press.
Manski, C.F. 2004. Measuring expectations. Econometrica 72:
1329-1376.
McFadden, D. 1994. Contingent valuation and social choice.
American Journal of Agricultural Economics 76: 689-708.
Mitchell, R.C., and R.T. Carson. 1989. Using surveys to value
public goods: The contingent valuation method.
Washington, DC: Resources for the Future.
Portney, P.R. 1994. The contingent valuation debate-Why
economists should care. Journal of Economic Perspectives
U4): 3 17.
2184
Smith, R.D. 2003. Construction ofthe contingent valuation market
in health care: A critical assessment. Health Economics 12:
609-628.
Continuity in Economic History
Donald N. McCloskey
Continuity and discontinuity are devices of story-
telling, telling the story of monetary policy over the
past few months or the story of modem economic
growth. They raise certain questions in philosophy
and lesser matters, such as precedence and politics.
It is well to have a case in mind. The most
important is that of the British industrial
revolution.
If it was a ‘revolution’, as it surely was, it
happened sometime. There was a discontinuity, a
before and after. When? Various dates have been
proposed, down to the day and year: 9 March 1776,
when the Wealth of Nations provided an ideology
for the age; the five months in 1769 when Watt
took out a patent on the high pressure steam engine
and Arkwright on the cottonspinning water frame;
or 1 January 1760, when the furnaces at Carron
Ironworks, Stirlingshire, were lit.
Such dating has of course an amateur air. A
definite date looks handsome on a plaque or scroll
but the precision does not fit well with sophisticated
story-telling. The discontinuity is implausibly
sharp, drawing attention to minor details. The Great
Depression did not start on 24 October 1929; the
deregulation of American banking was not
completed with the fall of Regulation Q. Nicholas
Crafts ( ) has pointed out
that the detailed timing of the industrial revolution
should not anyway be the thing to be studied,
because small beginnings do not come labelled with
their probabilities of developing into great
revolutions. He is identifying a pitfall in story-
telling. Joel Mokyr identifies another ( p. 44):
rummaging among the possible acorns
Continuity in Economic History
from which the great oak of the industrial revolution
grew ‘is a bit like studying the history of Jewish
dissenters between 50 BC and 50 AD. What we are
looking at is the inception of something which was
at first insignificant and even bizarre’, though
‘destined to change the life of every man and
woman in the West’.
What is destined or not destined to change our
fives will look rather different to each of us. Each
historian therefore has his or her own dating of the
industrial revolution. Each sees another disconti-
nuity. E.M. Carus-Wilson ( , p. 41) spoke of
‘an industrial revolution of the 13th century’: she
found that the fulling mill was ‘due to scientific
discoveries and changes in technique’ and ‘was
destined to alter the face of medieval England’.
A.C. Bridbury ( , p. xix-xx) found in the late
middle ages ‘a country travelling slowly along the
road ... that [it] travelled so very much more quickly
in Adam Smith’s day’. In the eyes of Marxist
writers the 16th century was the century of
discontinuity, when capitalism set off into the world
to seek its fortune. John U. Nef, no Marxist,
believed he saw an industrial revolution in the 16th
century, centred on coal ( ), though
admittedly slowed in the 17th century. A student of
the 17th century itself, such as D.C. Coleman ( ),
finds glimmerings of economic growth
even in that disordered age. The most widely
accepted period for the industrial revolution is the
late 18th century, especially the 1760s and 1770s
(Mantoux ; Landes ), but recent students of the
matter (Harley ; Crafts ) have found much to admire
in the accomplishments of the early 18th century.
W.W. Rostow ( ) placed the ‘takeoff into self-
sustained
growth’ in the last two decades of the 18th century,
but others have observed that even by 1850 the
majority of British people remained in traditional
sectors of the economy. And later still there was a
second industrial revolution (of chemicals,
electricity, and internal combustion) and a third (of
electronics and biology).
Wider perspectives are possible, encouraging
the observer to see continuity instead. Looking at
the matter from 1907, the American historian Henry
Adams could see a ‘movement from unity into
multiplicity, between 1200 and 1900, ...
Continuity in Economic History
unbroken in sequence, and rapid in acceleration’ (p.
498). The principal modem student of the industrial
revolution, R.M. Hartwell, appealed for continuity
against the jostling throng of dates ( , p. 78): ‘Do
we need an explanation of the
industrial revolution? Could it not be the culmi-
nation of a most unspectacular process, the con-
sequence of a long period of economic growth?’
Such questions of continuity and discontinuity
are asked widely in economics, though sometimes
half consciously. They should not be left to histo-
rians. Economics is mainly contemporary history,
and faces the problem of deciding when a piece of
history has been continuous or not. For instance the
cmcial discontinuity in the growth of big gov-
ernment, as Robert Higgs ( ) points out, might
be placed when the institutions of centralized inter-
vention were conceived (1900-1918) or made
(1930^45) or expanded (1960-70). Even recent
history faces this narrative problem. When, if ever,
did purchasing power parity break down in the
1970s? When did policy on antitrust alter to favour
mergers? When did monetary policy last become
expansionary? Where is the break?
The difficulty in answering the question has
often been misconstrued as philosophical. The
philosophical difficulty was first articulated in the
5th century BC by Parmenides and his student
Zeno: that if everything is perfectly continuous,
change is impossible (Komer ). Everything is so to
speak packed too tightly to move. The economist
will recogn ize the poi nt as analogous to an
extreme form of economic equilibrium, or to the
physicist’s maximum entropy. If human nature
doesn’t ‘really’ change, then history will be a string
of weary announcements that the more things
change the more they stay the same. If the economy
is ‘really’ in equilibrium all the time, then nothing
remains to be done.
Alexander Gerschenkron, the economic histo-
rian who has contributed most to the understanding
of continuity and discontinuity in economics, noted
that such a metaphysics would close the book of
history ( , p. 12). A history of eco
nomics that began with the Parmenidean continuum
would never speak.
For purposes of social science Gerschenkron
rejects the transition from the connectedness of all
2185
change to an absence of change. Tme, if you squint
and fit a curve then no economic change looks
discontinuous in the mathematical sense; but it is
wrong then to deduce that ‘really’ there is no
change at all, or that the industrial revolution is a
mirage. ‘Continuity’ in the strict mathematical
sense must be kept distinct from ‘continuity’ in the
story-telling sense.
Economists have often been muddled about this
philosophical distinction, drawing surprising
ideological implications from it. Alfred Marshall
enshrined on the title page of his Principles the
motto ‘natura non facit saltum’ (nature does not
make a jump; Leibnitz had invented it as ‘la nature
ne fait jamais des sauts’). Marshall himself perhaps
believed that the ability to represent behaviour with
differentiable functions implies that marginalism is
a good description of human behaviour. It is less
sure that he believed that the lack of jumps in nature
(this on the eve of quantum physics) implies people
should not jump either, and should change society
only gradually. Anyway, both implications are non
sequiturs. Though both have been attributed to
neoclassical economics, neither is necessary for it.
Much bitter controversy has assumed that
neoclassical economics depends on smooth curves
and in consequence must advocate smooth social
policies. The peculiar alliance between discrete
mathematics and Marxian economics has this
origin, as does the enthusiasm of some conservative
writers for continuities in economic history.
Gerschenkron cursed both their houses; the social
scientist should study change and continuity
‘unbothered by the lovers and haters of revolutions
who must find themselves playgrounds and
battlegrounds outside the area of serious
scholarship’ (p. 39).
In one sense of ‘continuity’ it is trivial that
economic history is continuous. History has causes
(the fourth of five historically relevant definitions
that Gerschenkron distinguishes). Continuity, then,
can be viewed as being merely an impressively long
causal chain. The exploitation of Scottish iron
deposits in the 18th century was caused by bold
investments, but these depended on a reliable law of
property and commerce, which depended on certain
legal developments in the 16th century, and on the
growth of
2186
political stability in the early 18th century, which in
turn depended on all manner of earlier events.
Establishing continuities, as Gerschenkron remarks,
is the historian’s purpose - or, one might add, the
economist’s, who is doing historian’s work when he
is not doing philosopher’s. The purpose might be to
find a cause of, say, the Great Depression. It would
be to find a chain of events the absence of which
would have made a difference: the international
irresponsibility of the United States, for instance, as
Kindleberger argued; or the domestic
irresponsibility of the Federal Reserve, as Friedman
and Schwartz argued. Finding such chains has its
own philosophical difficulties (see the article in this
Dictionary on “► )”.
The main problems of continuity and disconti-
nuity, however, are not solvable in seminars on
philosophy. They are practical problems in the uses
of measiuement, and must be solved in the
economic or historical workshop. When shall we
say that the industrial revolution happened?
Gerschenkron gives an answer confined to industry,
for in common with most economic historians he
regards agriculture and services as laggards in
economic growth.
In a number of major countries of Europe ... after a
lengthy period of fairly low rates of growth came a
moment of more or less sudden increase in the rates,
which then remained at the accelerated level for a
considerable period. That was the period of the great
spurt in the respective countries’ industrial
development The rates and the margin between
them in the ‘pre-kink’ and the ‘post-kink’ periods appear
to vary depending on the degree of relative
backwardness of the country at the time of the
acceleration, (pp. 33-4)
The level at which such discontinuity is to be explain it with big banks founded in the 1890s.
observed is at choice. As Gerschenkron remarks,
If the seat of the great spurt lies in the area of
manufacturing, it would be inept to try to locate the
discontinuity by scrutinizing data on large
aggregate magnitudes such as national income_________
By the time industry has become bulky enough to affect
the larger aggregate, the exciting period of the great spurt cating a micro foundation for macroeconomics). If
may well be over. (pp. 34—5)
In a footnote to these sentences he remarks that banks appeared, becoming bulky only later, then the
‘Walt Rostow’s failure to appreciate this point has new banks could not have been the initiating force.
detracted greatly from his concept of the take-off,
Continuity in Economic History
which in principle is closely related to the concept
of the great spurt as developed by this writer'.
The point is a good one, and applies to all
questions of continuity in aggregate economics.
Small (and exciting) beginnings will be hidden by
the mass until well after they have become routine.
Joel Mokyr has put it as a matter of arithmetic: if
the traditional sector of an economy is growing at a
slow one per cent per annum, and starts with 90 per
cent of output, the modem sector growing at four
per cent per annum will take three-quarters of a
century to account for as much as half of output (
, p. 5). We may call
it the Weighting Theorem (or the Waiting Theorem,
for the wait is long when the weight is small to
begin with). There are parallel points to be made
elsewhere in economics and in social science
generally. In growth theory, for instance, as was
noticed shortly after its birth, a century of
theoretical time is needed in most models for a shift
to yield growth as much as 90 per cent of its steady
state. More generally, economists have long
recognized the tension between microeconomic
explanations and the macroeconomic things to be
explained. And sociologists have been quarrelling
along similar lines for a century, using even the
same jargon of micro and macro.
In other words, the search for discontinuity in an
aggregate time series raises the question of the level
at which we should do our social thinking, the
aggregation problem. Yet Gerschenkron himself did
not answer the question well, and was hoist by his
own petard. Calculating Italian industrial output he
placed his ‘big spurt’ in 1896-1908, and wished to
Stefano Fenoaltea, once his student, applied the
Weighting Theorem to the case (Fenoaltea ). Surely,
Fenoaltea reasoned, the components of the
industrial index - the steel output and the chemical
output - are the ‘real’ units of economic analysis
(note the similarity of this rhetoric to that advo-
the components started accelerating before the new
Alas, the components did just this. They spoil
Gerschenkron’s bank-led story: the
Continuous and Discrete Time Models
components accelerated not in the 1890s but in the
1880s, not after but before the banks. To paraphrase
Gerschenkron on Rostow, by the time the
progressive components of industry had become
bulky enough to affect the larger aggregate, the
exciting period was well over.
Yet the moral is still Gerschenkron’s: that con-
tinuity and discontinuity are tools ‘forged by the
historian rather than something inherently and
invariantly contained in the historical matter .... [A]t
all times it is the ordering hand of the historian that
creates continuities or discontinuities’ (p. 38).
Gerschenkron nodded, but in nodding made the
point. The multiple datings of the industrial
revolution make it, too. So does any choice of
smoothness or suddenness in economic storytelling.
The point is that history, like economics, is a
story we tell. Continuity and discontinuity are
narrative devices, to be chosen for their storytelling
virtues. Niels Bohr said once that ‘It is wrong to
think that the task of physics is to find out how
nature is. Physics concerns what we can say about
nature’. It is our say. We can choose to emphasize
the continuous: ‘Abraham begat Isaac;
... begat... begat... and Jacob begat Joseph the
husband of Mary, of whom was bom Jesus’. Or the
discontinuous: ‘There was in the days of Herod, the
king of Judea, a certain priest named Zacharias.’ It
is the same story, but its continuity or discontinuity
is our creation, not God’s. That it is out of God’s
hands does not make it arbitrary. Scholars speak of
the industrial revolution as early or late, gradual or
sudden. Other scholars believe or disbelieve their
stories on the usual grounds.
See Also
► Economic History
Bibliography
Adams, H. 1906. The education of Henry Adams. New York:
Modem Library. 1931.
Bridbury, A.C. 1975. Economic growth: England in the later
middle ages. Brighton: Harvester.
Cams-Wilson, E.M. 1941. An industrial revolution of the
thirteenth century. Economic History Review 11(1):
2187
39-60. Reprinted in Essays in economic history, vol. I, ed.
E.M. Cams-Wilson. London: Edward Arnold, 1954.
Coleman, D.C. 1977. The economy of England 1450-1750.
Oxford: Oxford University Press.
Crafts, N.F.R. 1977. Industrial revolution in England and France:
Some thoughts on the question ‘Why was England first?’.
Economic History Review, 2nd series 30(3): 429-441.
Crafts, N.F.R. 1984. Economic growth during the British
industrial revolution. Oxford: Oxford University Press.
Fenoaltea, S. 1987. Italian industrial production, 1861-1913: A
statistical reconstruction. Cambridge: Cambridge
University Press.
Gerschenkron, A. 1962. On the concept of continuity in history.
Proceedings of the American Philosophical Society June.
Reprinted in A. Gerschenkron, Continuity in history and
other essays. Cambridge, MA: Harvard University Press,
1968.
Harley, C.K. 1982. British industrialization before 1841:
Evidence of slower growth during the industrial revolution.
Journal of Economic Histoiy 42(2): 267-290.
Hartwell, R.M. 1965. The causes of the industrial revolution: an
essay in methodology. Economic History Review, 2nd
series 18: 164-182. Reprinted in The causes of the
industrial revolution in England, ed. R.M. Hartwell.
London: Methuen, 1967.
Higgs, R. 1987. Crisis and Leviathan: Critical episodes in the
growth of American government. New York: Oxford
University Press.
Komer, S. 1967. Continuity. In The encyclopedia of philosophy.
New York: Macmillan and Free Press.
Landes, D.S. 1969. The unboundprometheus. Cambridge:
Cambridge University Press.
Mantoux, P. 1928. The industrial revolution in the eighteenth
century. New York: Harper, 1961.
Mokyr, J. (ed.). 1985. The economics of the industrial
revolution. Totowa: Rowman and Allanheld.
Nef, J.U. 1932. The rise of the British coal industry, 2 vols.
London: Routledge.
Rostow, W.W. 1960. The stages of economic growth.
Cambridge: Cambridge University Press.
Continuous and Discrete Time
Models
Christopher A. Sims
Abstract
Most model ling of economic time series works
with discrete time, yet time is in fact continu-
ous. While in many instances simple intuitive
2188
connections exist between results with discrete
time data and the underlying continuous time
dynamics, it is possible for discretization to
create bias or have unintuitive effects. Some
economics literature investigates such distor-
tions. It is also possible to estimate explicitly
continuous-time models, using discrete data.
This approach raises its own difficulties, but has
become more usable as computing power and
the techniques to exploit it have improved.
Keywords
Approximation theory; Continuous and discrete
time models; Distributed lags; Dynamic
stochastic general equilibrium models; Granger
causal priority; Markov chain Monte Carlo
methods; Martingales; No-arbitrage models;
Stochastic differential equations; Vector
autoregressions; Wiener process
JEL Classifications
C3
Discrete time models are generally only an
approximation, and the error induced by this
approximation can under some conditions be
important.
Most economists recognize that the use of dis-
crete time is only as an approximation, but assume
(usually implicitly) that the error of approximation
involved is trivially small relative to the other sorts model, in discrete time.
of simplification and approximation inherent in
economic theorizing. We consider below first the
conditions under which this convenient assumption
may be seriously misleading. We discuss briefly
how to proceed when the assumption fails and the where * stands for convolution, so that
state of continuous time economic theory.
Approximation Theory
Some economic behavioiu does involve discrete
delays, and most calculated adjustments in indi-
vidual patterns of behavioiu seem to occiu fol-
lowing isolated periods of reflection, rather than
continually. These notions are sometimes invoked
Continuous and Discrete Time Models
to justify economic theories built on a discrete time
scale. But to say that there are elements of discrete
delay or time discontinuity in behavioiu does not
imply that discrete time models are appropriate. A
model built in continuous time can include discrete
delays and discontinuities. Only if all delays were
discrete multiples of a single underlying time unit,
and synchronized across agents in the economy,
would modelling with a discrete time unit be
appropriate.
Nonetheless, sometimes discrete models can
avoid extraneous mathematical complexity at little
cost in approximation error. It is easy enough to
argue that time is in fact continuous and to show
that there are in principle cases where use of
discrete time models can lead to error. But it is also
true in practice that more often than not discrete
time models, translated intuitively and informally to
give implications for the real continuous time world,
are not seriously misleading. The analytical task,
still not fully executed in the literature, is to
understand why discrete modelling usually is
adequate and thereby to understand the special
circumstances under which it can be misleading.
The basis for the usual presumption is that,
when the time unit is small relative to the rate at
which variables in a model vary, discrete time
models can ordinarily provide good approximations
to continuous time models. Consider the case,
examined in detail in Geweke ( ), of a
dynamic multivariate distributed leg regression
Y(t)=A*X(t) + U(t), (1 )
OO
A*X{t)= Y, A(s)X(t-s). (2 )
s=—OO
We specify that the disturbances are
uncorrelated with the independent variable vector X,
that is, cov[AT/j, U(s)] = 0, all t, s. The natural
assumption is that, if approximation error from use
of discrete time is to be small, A(s) must be
Continuous and Discrete Time Models
smooth as a function of s, and that in this case ( ) is
a good approximation to a model of the form
y(t) = a* x(t) + u{t) (3) result. Most obviously, the result depends on a in ( )
where
»oo
a* x(t) a(s)x(t
—OO
— s)ds (4)
and y, a and x are functions of a continuous time
parameter and satisfy y(t) = Y(t), x(t) =X(t) and a
(!) A(t) at integer /. In this continuous time model
we specify, paralleling the stochastic identifying
assumption in discrete time, cov[x(0, u (s)] = 0, all
t, s. If the discrete model ( ) corresponds in this way
to a continuous time model, the distributed lag
coefficient matrices A(s) are uniquely determined
by a and the serial correlation properties of x.
We should note here that, though this frame-
work seems to apply only to the case where Xis a
simple discrete sampling of x, not to the time-
averaged case where X(t) is the integral of x(s) from
/ - I to l, in fact both cases are covered. We can
simply redefine the x process to be the continuously
unit-averaged version of the original x process. This
redefinition does have some effect on the nature of
limiting results as the time unit goes to zero (since
the unit-averaging transformation is different at
each time unit) but turns out to be qualitatively of
minor importance.
Roughly speaking, sampling a unit-averaged
process is like sampling a process whose paths have
derivatives of one higher order than the unaveraged
process.
Geweke shows that under rather general
conditions
Y ||A( S )- M ( S T )|| 2 ^0 (5)
s=—OO
as the time unit x goes to zero, where || || is the usual
root-sum-of-squared-elements norm. In this result,
the continuous time process x and lag distribution a
are held fixed while the time interval corresponding
to the unit in the discrete time model shrinks.
2189
This is the precise sense in which the intuition
that discrete approximation does not matter much is
correct. But there are important limitations on the
being an ordinary function. In continuous time,
well-behaved distributed lag relations like ( ) are not
the only possible dynamic relation between two
series. For example, if one replaces ( ) by
y(t) = a(d/dt)x(f) + u(t), (6)
then the limit of A in ( ) is different for different
continuous x processes. In a univariate model with
second-order Markov .r (for example, one with cov
[x(t), x(t — s')] = (1 + d|sj)e~0|‘?l var[x(i)], the
limiting discrete time model, as t goes to zero, is
y(t) = a{ - 0.02X(f +4) +0.06X(r + 3) -0.22X(r+2) +
0.80X(f+l)-0.80X(f-l) + 0.22X(f-2) -0.06X(t-
3)+0.02X{t-4)} + U(t)
(7)
(see Sims ).
This result is not as strange as it may look. The
coefficients on X sum to zero and are antisymmetric
about zero. Nonetheless, ( ) is far from the naive
approximation which simply replaces the derivative
operator with the first difference operator. In fact, if
the estimation equation were constrained to involve
only positive lags of X, the limiting form would be
Y{t) = u.{\.21X(t) - 1.161X(t - 1)
+ 0.43X(t — 2) — 0.\2X(t — 3)
+ 0.03X(t - 4) - 0.01X(t - 5)} + U(t).
(8)
The naive approximation of ( ) by Y(i) = a[X(i )
X(t) — 1] + (7(0 is valid only in the sense that, if
this form is imposed on the discrete model a priori,
the least squares estimate of a will converge to its
tme value. If the resulting estimated model is tested
for fit against ( ) or ( ), it will be rejected.
Although the underlying model involves only
the contemporaneous derivative of x, ( ) and ( ) both
involve fairly long lags in X. If x paths have
2190
higher than firstorder derivatives (for example, if
they are generated by a third-order stochastic dif-
ferential equation) the lag distributions in ( ) and ( )
are replaced by still higherorder limiting forms.
Thus, different continuous time processes for x
which all imply differentiable time paths produce
different limiting discrete A. Here the fact that the
time unit becomes small relative to the rate of
variation in x does not justify the assumption that
approximation of continuous by discrete models is
innocuous. In particular, the notion that discrete
differencing can approximate derivatives is
potentially misleading.
It should not be surprising that the discrete time
models may not do well in approximating a con-
tinuous time model in which derivatives appear.
Nonetheless, empirical and theoretical work which
ignores this point is surprisingly common.
If a is an ordinary function, there is still chance
for error despite Geweke’s result. His result implies
only that the mean square deviation of a from A is
small. This does not require that individual A(t/x)’s
converge to the corresponding aft) values. For
example, in a model where x is univariate and a(/) =
0,t < 0, a(0) = 1, a(s) continuous on [0,oo], the
limiting value for A(0) is 0.5, not 1.0. Thus, \fa(t) e
"" on [0, oo), making a monotone decreasing over
that range, A(t) will not be monotone decreasing. It
will instead rise between i 0 and / I. This is not
unreasonable on reflection: the discrete lag
distribution gives a value at i 0 which averages the
continuous time distribution’s behaviour on either
side of t 0. It should therefore not be surprising that
monotonicity of a does not necessarily imply
monotonicity of A, but the point is ignored in some
economic research.
Another example of possible confusion arises
from the fact that, if the .r process has differentiable
paths, aft) 0 for / < 0 does not imply Ad) 0 for / < 0.
The mean-square approximation result implies that
when the time unit is small the sum of squares of
coefficients on Xft s) for negative s must be small
relative to the sum of squares on X(t — s) for
positive s, but the first few lead coefficients will
generally be non-zero and will not go to zero as the
time interval goes to zero. This would lead to
mistaken conclusions about
Continuous and Discrete Time Models
Granger causal priority in large samples, if signif-
icance tests were applied naively.
Geweke’s exploration of multivariate models
shows that the possibilities for confusing results are
more numerous and subtle in that case. In
particular, there are ways by which poor approx-
imation of otj{s) by Aj(s/T) in some s interval (for
example, around 5 = 0) can lead to contamination of
the estimates of other elements of the A matrix,
even though they correspond to xjs and QjS that in a
univariate model would not raise difficulties.
In estimation of a dynamic prediction model for
a single vector y, such as a vector autoregression
(VAR) or dynamic stochastic general equilibrium
model (DSGE), the question for approximation
theory becomes whether the continuous time
dynamics for y, summarized in a Wold moving
average representation
y(t) = a*u(t) ( 9)
has an intuitively transparented connection to the
corresponding discrete time Wold representation
Y(t)=A*U(t). ( 10 )
In discrete time the U(t) of the Wold repre-
sentation is the one-step-ahead prediction error, and
in continuous time u(t) also represents new
information about y arriving at t. There are two
related sub-questions. Is the A function the same
shape as the a function; and is the U vector related
in a natural way to the u vector? The u vector is a
continuous time white noise, so that U cannot
possibly be a simple discrete sampling of u.
If y is stationary and has an autoregressive
representation, then U(t) A 1 with the expression
interpreted as convolution in continuous time, but
with A 1 putting discrete weight on integers. The
operator connecting U and u is then A '*a. There are
cases where the connection between continuous and
discrete time representations is intuitive. For
example, if a(s) = exp(- Bs) (with the
exponentiation interpreted as a matrix exponential
in a multivariate case), then
 *
w c(s w(t)

>2
■
.HO. 0 b(s)
[HO J

where w and v are white noise processes funda-

mental (in the terminology of Rosanov ), for y and
x. To give b and c a convenient parametric form,
one might suppose them rational, so that ( ) can be
written as a differential equation, that is,
A
AoAo > a(s)a(s) ds, (13) P(D) y(t ) = P(D)a*x{t) + w (t )
Jo
(1 5)
where the inequality is interpreted as meaning that
Q(D)x(t) = v(0, (16)
the left-hand-side matrix minus the right-hand- side
matrix is positive semi-definite. In other words, the where P and Q are finite-order polynomials in the
initial coefficient in the discrete MAR will always derivative operator, Q1 (D)v = b*v, and P 1 (D)
be as big or bigger than the average over (0,1) of W = c*w.
the coefficients in the continuous A discrete time model derived explicitly from a
continuous time model is likely to be nonlinear at
2192
least in parameters and therefore to be more diffi-
cult to handle than a more naive discrete model.
However with modem computing power, such
models are usable. Bergstrom ( ) provides a
discussion of estimating continuous time constant
coefficient linear stochastic differential equation
systems from discrete data, the papers in the book (
) he edited provide related discussions, and
Hansen and Sargent ( ), in some of their own
chapters of that book, discuss estimation of con-
tinuous time rational expectations models from
discrete data.
Estimating stochastic differential equation
models from discrete data has recently become
easier with the development of Bayesian Markov
chain Monte Carlo (MCMC) methods. Though
implementation details vary across models, the
basic idea is to approximate the diffusion equation
dyt = a(yt)dt + b(yt)dWh (17) where Wt
is a Wiener process, by
yl = e-a^)y,-5 + b{yl^l. (1 8)
Such an approximation can be quite inaccurate
unless 6 is very small. But one can in fact choose
<5 very small, much smaller than the time interval
at which data are observed. The values ofyt at times
between observations are of course unknown, but if
they are simply treated as unknown ‘parameters it
may be straightforward to sample from the joint
posterior distribution of the y’s at non-observation
times and the unknown parameters of the model.
The Gibbs sampling version of MCMC samples
alternately from conditional posterior distributions
of blocks of parameters. Here, sampling from the
distribution of y at non-observation dates
conditioning on the values of model parameters is
likely to be easy. If the model has a tractable form,
it will also be easy to sample from the posterior
distribution of the parameters conditional on all the
y values, both observed and unobserved.
Application of these general ideas to a variety of
financial models is discussed in Johannes and
Poison ( ).
Another approach that has become feasible with
increased computing power is to develop
Continuous and Discrete Time Models
numerical approximations to the distribution of yt +
s conditional on data through time t. Ait- Sahalia (
) surveys methods based on this
approach.
Modelling in continuous time does not avoid the
complexities of connecting discrete time data to
continuous time reality - it only allows us to
confront them directly. One reason this is so seldom
done despite its technical feasibility is that it forces
us to confront the weakness of economic theory in
continuous time. A model like ( )-( )
makes an assertion about how many times y and x
are differentiable, and a mistake in that assertion can
result in error as bad as the mistake of ignoring the
time aggregation problem. Economic theory does
not have much to say about the degree of
differentiability of most aggregate macroeconomic
time series. When the theory underlying the model
has no believable restrictions to place on fine-
grained dynamics, it may be better to begin the
modelling effort in discrete time. As is often true
when models are in some respect underidentified, it
is likely to be easier to begin from a normalized
reduced form (in the case the discrete time model)
in exploring the range of possible interpretations
generated by different potential identifying
assumptions.
Recent developments in financial economics
have produced one area where there are continuous
time economic theories with a solid foundation.
Stochastic differential equations (SDEs) provide a
convenient and practically useful framework for
modelling asset prices. These SDE models imply
non-differentiable time paths for prices, and it is
known (Harrison et al. ) that differentiable time
paths for asset prices would imply arbitrage
opportunities, if there were no transactions costs or
bounds on the frequency of transactions.
However, there are in fact transactions costs and
bounds on transactions frequencies, and no-
arbitrage models for asset prices break down at very
fine, minute-by-minute, time scales. Successful
behavioural modelling of these fine time scales
requires a good theory of micro-market structure,
which is still work in progress.
It is worthwile noting that a process can have
non-differentiable paths without producing white
Continuous-Time Stochastic Models 2193

noise residuals at any integer order of differentia- Sims, C.A. 1971. Approximate specifications in distributed
lag models. In Proceedings of the 38th Session, Bulletin of the
tion: for example, a model satisfying ( ) with a(s) =
International Statistical Institute 44, Book 1.
s'^e Such a process has continuous paths with Sims, C.A., and S. Maheswaran. 1993. Empirical implications
unbounded variation and is not a semimartingale. of arbitrage-free asset markets. In Models, methods and
That is, it is not the sum of a martingale and a applications of econometrics, ed. P.C.B. Phillips. Oxford:

process with bounded variation, and therefore Blackwell.

cannot be generated from an integer- order SDE.

Similarly, if a(s) s°'5e ', the process has non-
differentiable paths but is nonetheless not a
semimartingale. The existence of such non-
semimartingale processes and their possible Continuous-Time Stochastic Models
applications to financial modelling is discussed in
Sims and Maheswaran ( ). Robert C. Merton

See Also
► Time Se
Bibliography
A'it-Sahalia, Y. 2007. Estimating continuous-time models
using discretely sampled data. In Advances in economics and
econometrics, theory and applications. Ninth World Congress , ed. R.
Blundell, T. Persson, and W.K. Newey, vol. 3.
Cambridge: Cambridge University Press.
Bergstrom, A.R., ed. 1976. Statistical inference in continuous time economic
models. Amsterdam: North- Holland.
Bergstrom, A.R. 1983. Gaussian estimation of structural
parameters in higher order continuous time dynamic
models. Econometrica 51: 117-152.
Geweke, J. 1978. Temporal aggregation in the multiple
regression model. Econometrica 46: 643-662.
Hansen, L.P., and T.J. Sargent, eds. 1991. Rational expectations
econometrics. Boulder and Oxford: Westview Press.
Harrison, J.M., R. Pitbladdo, and S.M. Schaefer. 1984.
Continuous price processes in frictionless markets have
infinite variation. Journal of Business 57: 353-365.
Johannes, M., and N. Poison. 2006. MCMC methods for
continuous-time financial econometrics. In Handbook of
financial econometrics, ed. Y. A'it-Sahalia and L.P. Hansen.
Amsterdam: North-Holland.
Marcet, A. 1991. Temporal aggregation of economic time
series. In Rational expectations econometrics, ed. L.P. Hansen and
T.J. Sargent. Boulder and Oxford: Westview Press.
Rozanov, Yu. A. 1967. Stationary random processes, trans A.
Feinstein. San Francisco/Cambridge/London/
Amsterdam: Holden-Day.
Models in which agents can revise their decisions
continuously in time have proved fruitful in the
analysis of economic problems involving
intertemporal choice under uncertainty (cf.
Malliaris and Brock ). These models frequently
produce significantly sharper results than can be
derived from their discrete-time counterparts. In the
majority of such cases, the dynamics of the
underlying system are described by diffusion
processes, whose continuous sample paths can be
represented by Ito integrals. However, in selected
applications, this assumption can be relaxed to
include both non-Markov path- dependent
processes and Poisson-directed jump processes.
An early application of this mode of analysis
was the lifetime consumption-portfolio selection
problem (Merton , ). Under the assump
tions of continuous trading and asset returns gen-
erated by diffusion processes, the derived structure
of optimal portfolio demands produce portfolio-
separation or mutual fund theorems like those
derived in the static Markowitz-Tobin mean-
variance model, but without the objectionable
assumption of either quadratic preferences or
Gaussian-distributed asset prices. Indeed, in the
special, but prototypical, case of lognormally-
distributed asset prices, the intertemporal optimal
rales are identical to those of the mean-variance
model. The continuous-time analysis thus provides
a reconciliation of this classic model, with models
of general expected utility maximization
2194
in an environment where asset ownership has
limited liability.
Using these same assumptions of continuous
trading and lognormality of security prices, Black
and Scholes ( ) derived a formula for pricing
options that provided the foundation for subsequent
development, of a unified theory of corporation
liability evaluation and general contingent- claim
pricing. Cox et al. ( ) use the
continuoustrading methodology with diffusion
processes to derive a general theory for the term
structure of interest rates.
Building on the continuous-time model of indi-
vidual choice, Merton ( ), Breeden ( ),
and Cox et al. ( ) develop intertemporal
models of equilibrium asset prices. Huang ( )
provides a stronger foundation for these
models by showing that if information in an econ-
omy with continuous-trading opportunities evolves
according to diffusion processes, then equilibrium
security prices will also evolve according to
diffusion processes.
In the intertemporal version of the Arrow-
Debreu general equilibrium model with complete
markets, the markets need only to be open ‘once’
because agents will have no need for further trade.
The continuous-trading model is in this respect at
the opposite extreme. Economies in which the
dynamics of the system are described by diffusion
processes will have a continuum of possible states
over any finite interval of time. Thus, in the strict
sense, to have complete markets in the continuous-
time diffusion model requires an uncountable
number of pure Arrow-Debreu securities. The
continuoustrading model with diffusions,
nevertheless, appears to have many of the important
properties of the Arrow-Debreu model, but without
nearly so many securities.
As is well known, in the absence of complete
Arrow-Debreu markets, a competitive equilibrium
does not in general produce Pareto optimal
allocations. However, Radner ( ) has shown
that an Arrow-Debreu equilibrium allocation can be
achieved without a full set of pure time-state
contingent securities if agents can use the available
securities to implement dynamic trading
Continuous-Time Stochastic Models
strategies which replicate the payoff structure of the
missing pure securities. There is much analysis to
suggest that continuous-trading opportunities
together with diffusion representations sentations
for the evolution of the economy provide a
particularly fertile environment for fulfilling the
Radner conditions.
Under reasonably general assumptions about
agents’ preferences and endowments Breeden ()
among others has shown that the
intertemporal equilibrium allocations generated in
economies with continuous trading in a finite
number of securities can be Pareto efficient. In the
analysis of the individual portfolio selection prob-
lem underlying these equilibrium models, the
derived portfolio-separation theorems show that the
set of individually-optimal portfolios can be
generated by combinations of relatively few com-
posite securities or mutual funds.
The extensive literature on options and
contingent-claims pricing provides further evidence
that continuous-trading opportunities make possible
a large reduction in the number of securities
markets without loss of efficiency. Although
typically partial equilibrium in nature, these
analyses show that continuous-trading dynamic
portfolio strategies using as few as two securities
can replicate a wide range of state and time-
dependent payoff structures.
In perhaps the most general analysis to date,
Duffie and Huang ( ) study the role continu
ous trading plays in successfully implementing
Arrow-Debreu equilibria with infinite dimensional
commodity spaces, using only a finite number of
securities. In particular, they derive necessary and
sufficient conditions for continuous-trading
portfolio strategies with a finite number of
securities to effectively complete markets in a
Radner economy. By working with martingale
representation theorems, Duffie and Huang show
that the class of dynamics for which these results
obtain extends beyond vector diffusion processes to
include some non-Markov path dependent
processes. They also show that having
heterogeneous probability assessments among
agents provides no important difficulties with the
results, provided all agents’ subjective
Continuous-Time Stochastic Models
probability measures are uniformly absolutely
continuous. Although there remain further technical
issues to be resolved, it is evident that the
continuous-trading models provide a strong foun-
dation for the belief that a good substitute for
having many markets and securities is to have fewer
markets which are open for trade more frequently.
A sketch of the derivation of the portfolio
separation theorem along the lines of Merton ( ,
, ) and Breeden ( ) is a
follows:
At each time t, each consumer-investor acts so
as to
U[c{i:),S{x),x\dx + B[W(T),S(T), T] j
Max E,
(l)
where E, is the conditional expectation operator,
conditional on information available at time t. S (f)
= [Si (f), . . . , Sm(t)] is a finite-TM vector set of
state variables which together with the consumer’s
current wealth W(t) is sufficient to describe the state
of the economy at time t. c(r) denotes the
instantaneous consumption flow selected at time t.
U is a strictly concave, statedependent utility
function for consumption and B represents utility
from bequests at date T.
The evolution of the state variables S is
described by a Markov system of fto stochastic
differential equations
d.S’,(7j = Gi(S, i)dt + ll,(S, t)dq„ i
= 1,2, . . . , m (2)
where G, (S, t) is the instantaneous expected change
in St(t) per unit time at time t, Hj is the
instantaneous variance of the change in where it is
understood that these statistics are conditional on
S(t) = S. The dq, are Wiener processes with the
instantaneous correlation coefficient per unit time where
between dq, and dqj given by the function rj,j(S, t),
i,j = 1 , . . . , m. At each point in time, the consumer
chooses a consumption flow and allocates his
wealth among n risky securities and a riskless
security whose instantaneous rate of
2195
return per unit time is the interest rate r(t). The rate
of return dynamics on risky security j can be written
as
dp p
i/ j = Xj(S, t)dt + (Tj(S, t)dzj,
7= 1, 2, ...,n
where a; is the instantaneous conditional expected
rate of return per unit time; af is the conditional
variance per unit time; and dzy is a Wiener process.
Denote by pJk(S, t) the instantaneous correlation
coefficient per unit time between dzj and dzk, and
denote by p,/S, t) the instantaneous correlation
coefficient between dq, and dzj, i = 1 , 2 , . . . , m
a n d j = 1 , 2 , . . . , n.
The accumulation equation for the consumer’s
wealth can be written as
d W = [rW + y — c\dt
+ Y/wjW[dPj/Pj-r dt] (4 )
7=1
where y = y(S,!) is the consumer’s wage income;
Wj is the fraction of his wealth allocated to risky
security j at time t, and is the
fraction allocated to the riskless asset.
The optimal consumption and portfolio mles,
c*(W, S, t) and w*(W, S, t), are derived by the
technique of stochastic dynamic programming.
Among the first-order conditions to be satisfied by
these optimal rules are the n conditions for the
optimal portfolio holdings at time t, which can be
expressed as j = 1, 2,..n
0 = jo, - r - (^hw'aiGjPfjW + /R :|
dU dc*
x
dcdWW
(5)
K = - dU/dc./ [d2U/dc2 ■ dc*/dW}
and
2196 Continuous-Time Stochastic Models

At = - dc*/dSt + {d2U/dcdSi)/{d1U/dc1) / efficient set). Fund no. (2 + /) provides the maxi-

mum feasible correlation between its return and the
{dc*/dW), *'= 1,2,
stochastic component of the instantaneous change in
state variable S,-(f), i = 1 , . . . , m. As discussed in
By inspection, the manifest characteristic of the detail in the cited Breeden and Merton papers, these
system of Eq. is that it is linear in the optimal latter portfolios serve the function of providing the
demands for risky assets. Therefore, if none of the best feasible hedges against utility losses caused by
risky assets is redundant, then standard matrix unanticipated changes in the state variables of the
inversion can be used to solve Eq. explicitly for economy.
these demands. That is, In the important case where the set of available
securities is such that the return on fund no. (2 + i)
= KYvkj{zK ~ r) + is perfectly correlated with the change in state
6)
l l v
’ variable St(t) for each i, i = 1, . . . , m, Breeden (
j = 1,2 n ) shows that the resulting intertemporal
equilibrium allocations are Paretoefficient. This is
where o kj is the k -/'th element of the inverse of the also the condition under which it is possible to
variance- covariance matrix of returns replicate the payoff structure for the complete set of
pine Arrow-Debreu securities using continuous-
trading dynamic portfolio strategies with a finite
and number of securities.
WPij Pit-
The dynamic strategies for replicating the pay-
offs to pure Arrow-Debreu securities can be derived
By inspection, K, Au . . . , Am are the only ele- in a similar fashion to the derivation of contingent-
ments in Eq. that depend on the individual investor’s claim prices in Merton ( ). Sup
preferences or endowment. As an immediate pose that among the available traded securities,
consequence, it follows that there exist (m + 2) portfolios can be constructed whose returns are
portfolios (‘mutual funds’) constructed from linear instantaneously perfectly correlated with changes in
combinations of the available securities such that, each of the state variables, [Sj(/), . . . , Sm(t)\ Without
independent of preferences, wealth distribution, or loss of generality, assume that these portfolios are
planning horizon, all investors will be indifferent the first m risky securities (i.e. dz, = dqt, i = 1 , 2 . . . ,
between choosing their portfolios from combinations m).
of just these (m + 2) funds or combinations of all n Let F(S, t) satisfy the linear partial differential
risky securities and the riskless security. This equation
portfolio- separation theorem is, of course, vacuous
if m > n + 1. If, however, m « n, then it implies a -i m m 2

nontrivial reduction in the number of securities °=OTHH‘HJ%11 d F

required to generate the set of optimal portfolios.
Although not unique, a set of funds which meets
the criterion of the theorem is: fund no. 1 holds the
riskless asset; fund no. 2 holds fraction y;, vkl(ak — r)
in security jj = 1,..., n and the balance in the riskless
asset; for i = 1, 2,..., n, fund no. (2 + i) holds fraction
Q.j in security j, j = 1 , . . . , n and the balance in the
riskless asset. Funds nos. 1 and 2, together generate
the set of portfolios with maximum expected return
for a given variance of the return (i.e. the mean-
variance
dSidSj
imi
fdp flL'
+J2lGJ - Hj(aJ - r)/ai\ + -Qt~rF
(7)
subject to the boundary conditions: 0 < F(S,t) < oo
for all S and t < r; F(S, T) = <5 [5i — Si ( r ) j . . .
<5[5,„ — S,„(T)] , where <5[] is the Dirac delta
function and Sk are given parameters, k = 1 , . . . , m.
Under mid regularity conditions on the functions
and r, a solution to Eq. exists and is unique.
Continuous-Time Stochastic Models
Consider the continuous-trading portfolio
strategy which allocates fraction x; (?) = (dF/dSj)
Hjl\ajV{i)\ to security j, j = 1 , . . . , m a n d
1 to the riskless security at time t,
where V(t) denotes the value of die portfolio. It
follows from Eq. and the prescribed allocation that
the dynamics of the portfolio value can be written as
J2xj(aJ ~ r ) + r At + '^jXjOjAzj
dV = V
f-irp
Y.Q^ H j( a J- r )/^J + rV
because dZj = dq/J = 1, 2 , . . . , m.
As a solution to Eq. , F is twice-continuously
differentiable. Thus, Ito’s Lemma can be used to
describe the stochastic process for F as
1 „dF
2
clF dS,dS, +
where F is evaluated at S = S(t) at each time t.
Because F satisfies Eq. , Eq. can be rewritten as
dF = )/«J + rF
From Eqs. and , dF = dV = r(F - V)dt,
which is an ordinary differential equation with
solution F[S(t), t\ - V(t) = [F(,S'(0), 0) - E(0)]
exp f r(u )d u If, therefore, the initial invest-
Jo
ment in the portfolio is chosen so that F(0) =
F[S(0),0] then V(t) = F[S,(t),t] for 0 < t < r.
Thus, a dynamic portfolio strategy using (m + 1)
available securities has been constructed that has a
payoff at t x of ri [S/ - S'l (t)] ... dp,,, — Sm(r)]. By
inspection of this payoff
2197
structure, it is evident that this security is the natural
generalization of Arrow-Debreu pure state
securities to an environment where there is a
continuum of states defined by St and x. By
changing the time and state parameters x and St, one
can generate all of the uncountable number of pure
securities. Moreover, F, the solution to Eq. used to
implement each strategy, w i l l also be the
equilibrium price for the corresponding pure Arrow-
Debreu security.
Continuous trading, like any other continuous-
revision process, is of course an abstraction from
m
dF physical reality. If, however, the length of time
At +
Y.FFHjAclj
YdSj between revisions is very short, then the
8
( )
continuous-trading optimal solutions will be a
reasonable approximation to their discrete-time
counterparts (see Samuelson and Merton , ).
From the work of Magill and
Constantinides ( ), this conclusion appears
to be robust even in the presence of transactions
costs, which cause trading to be discrete almost
certainly.
dF'
V '!E Whether the length of time between revisions is
short enough for the continuous solution to provide
a good approximation must be decided on a case-
by-case basis by making a relative comparison with
(9) other time scales in the problem. The continuous-
trading assumption appears to be especially
appropriate for the analysis of security markets
where the aggregate trading volume is large, the
minimum unit-size for a transaction is relatively
small, and the length of calendar time between
At successive transactions is quite short.
The continuous analysis may also provide a
valid approximation in problems where the calendar
(10)
length of time between revisions is not short. For
example, Bourguignon ( ), Bismut
( ), and Merton ( ) use this mode of
analysis to extend the Solow model of economic
growth to an uncertain environment and to analyse
the stochastic Ramsey problem. It is the practice in
such models to neglect ‘short-run’ business cycle
fluctuations and to assume full employment.
Moreover, the exogenous factors usually assumed
to affect the time path of the economy in these
models are either demographic or technological
changes. Since major changes in either factor
typically take rather long periods of time, the
2198
length of time between revisions in the capital
stock, although hardly instantaneous, may well be
quite short, relative to the time scale of the
exogenous processes.
See Also
► Continuous-Tiny Stochastic Processes
► 'inance
► Options
Bibliography
Bismut, J.M. 1975. Growth and optimal intertemporal
allocation of risks. J o u r n a l of Economic
T l i e o n ’ 10: 239-257.
Black, F., andM. Scholes. 1973. The pricing of options and
corporate liabilities. J o u r n a l of Political
E c o n o m y 81: 637-654.
Bourguignon, F. 1974. A particular class of continuoustime
stochastic growth models. J o u r n a l o f E c o n o m i c
T h e o r y 9: 141-158.
Breeden, D.T. 1979. An intertemporal asset pricing model
with stochastic consumption and investment opportunities.
J o u r n a l o f F i n a n c i a l E c o n o m i c s 7: 265-296.
Cox, J.C., J.E. Ingersoll Jr., and S.A. Ross. 1985a. Atheory of
the term structure of interest rates. E c o n o m e t r i c a 53:
385-408.
Cox, J.C., J.E. Ingersoll Jr., and S.A. Ross. 1985b. An
intertemporal general equilibrium model of asset prices.
E c o n o m e t r i c a 53: 363-384.
Duffie, D., and C. Huang. 1985. Implementing Arrow-Debreu
equilibria by continuous trading of a few long-lived securities.
E c o n o m e t r i c a 35: 1337-1356. Huang, C. 1985.
Information structure and equilibrium asset prices.
J o u r n a l o f E c o n o m i c T h e o t y 35: 33-71. Magill,
M.J.P., and G.M. Constantinides. 1976. Portfolio selection
with transactions costs. J o u r n a l o f E c o n o m i c
T h e o i y 13: 245-263.
Malliaris, A.G., and W.A. Brock. 1982. S t o c h a s t i c
methods in economics and finance.
Amsterdam: North-Holland. Merton, R.C. 1969. Lifetime
portfolio selection under uncertainty: The continuous-time
case. R e v i e w o f E c o n o m i c s a n d S t a t i s t i c s 51:
247-257.
Merton, R.C. 1971. Optimum consumption and portfolio
rules in a continuous-time model. J o u r n a l o f
E c o n o m i c T h e o r y 3: 373-413.
Merton, R.C. 1973. An intertemporal capital asset pricing
model. E c o n o m e t r i c a 41: 867-887.
Merton, R.C. 1975a. An asymptotic theory of growth under
uncertainty. R e v i e w o f E c o n o m i c S t u d i e s 42:
375-393.
Merton, R.C. 1975b. Theory of finance from the perspective
of continuous time. J o u r n a l o f F i n a n c i a l a n d
Q u a n t i t a t i v e A n a l y s i s 10: 659-674.
Continuous-Time Stochastic Processes
Merton, R.C. 1977. On the pricing of contingent claims and
the Modigliani-Miller theorem. J o u r n a l o f F i n a n -
c i a l E c o n o m i c s 5: 241-249.
Merton, R.C. 1982a. On the microeconomic theory of
investment under uncertainty. In H a n d b o o k o f
m a t h e m a t i c a l e c o n o m i c s , vol. 2, ed. K.J. Arrow
and M.D. Intriligator. Amsterdam: North-Holland.
Merton, R.C. 1982b. On the mathematics and economics
assumptions of continuous-time models. In F i n a n c i a l
economics: Essays in honor of Paul
C o o t n e r , ed. W.F. Sharpe and C.M. Cootner.
Englewood Cliffs: Prentice-Hall.
Radner, R. 1972. Existence of plants, prices, and price
expectations in a sequence of markets. E c o n o m e t r i c a
40: 289-303.
Samuelson, P.A. 1970. The fundamental approximation
theorem of portfolio analysis in terms of means, vari-
ances, and higher moments. R e v i e w o f E c o n o m i c
S t u d i e s 32: 537-542.
Continuous-Time Stochastic Processes
Chi-Fu Huang
Applications of continuous-time stochastic pro-
cesses to economic modelling are largely focused
on the areas of capital theory and financial markets.
In these applications as in mathematics generally,
the most widely studied continuous time process is
a Brownian motion - so named for its early appli-
cation as a model of the seemingly random move-
ments of particles which were first observed by the
English botanist Robert Brown in the 19th century.
Einstein ( ), in the context of statistical
mechanics, is generally given credit for the first
mathematical formulation of a Brownian motion
process. However, an earlier development of an
equivalent continuous-time process is provided by
Louis Bachelier ( ) in his theory of stock
option pricing. Framed as an abstract mathematical
process, a Brownian motion {B(t); t e R+} is
described by the following properties: (1) for 0 : (s
< t < oo,B(t) — B(s) is a normally distributed
random variable with mean zero and variance t - s;
(2) for 0 < to < t\ < • • • < t\ < oo,
Continuous-Time Stochastic Processes
{ B ( t 0 ); B ( t k )- B ( t k ^ ), k = l , .. . , / }
is a set of independent random variables.
From this construction, Doob, Feller, Ito, Wie-
ner, among others went on to develop the general
theory of continuous-time stochastic processes.
During the half century of this development of
the theory, its application in economics was con-
fined primarily to the formulation and testing of
hypotheses concerning time series properties of
economic variables. It was not until the 1950s and
early 1960s that the theory of continuous-time
stochastic processes found its way into economic
theory. Motivated by the rediscovery of Bachelier’s
work on options by L.J. Savage, Sam- uelson ( )
presents a theory of rational warrant
pricing. Unlike Bachelier’s assumption of a
Brownian motion for a stock price process, Sam-
uelson posits that the logarithm of a stock price
follows a Brownian motion, and thereby, ensiues
that model stock prices exhibit non-negativity as
required by limited liability. This process, called a
geometric Brownian motion by Samuelson, remains
to this day the prototypical process used by
economists to describe stock price behaviour.
Working with Samuelson, McKean ( ) uses
the theory of optimal stopping to provide a rigorous
derivation of the warrant price in Samuelson’s
theory.
Although it is the standard mode of analysis for
warrant and option pricing theory today, the cele-
brated work on the stochastic integration by K. Ito (
, ) was not introduced into economic
analysis until the late 1960s. Merton ( , )
was the first to use Ito’s stochastic calculus in
economics. He analysed an agent’s optimal con-
sumption and portfolio policies in a continuous time
economy where asset prices are Ito processes.
Ito’s contribution to the theory of stochastic
processes lies in a definition of an integral with
desired properties when the integrator is a Brownian
motion. A pathwise definition in the Stieltjes sense
may fail since a Brownian motion has sample paths
that are nowhere differentiable with probability one.
For a classical treatment of the Ito integral, see also
Ito and McKean ( ) the General Theory of Processes. Very abstract, and
and McKean ( ). A good reference for modem
2199
treatments can be found in Chung and Williams (
).
Ito’s definition of a stochastic integral, in con-
trast to that of Stratonovich ( ), is much better
suited for analysing intertemporal economic deci-
sion making. The non-anticipating integrand in
Ito’s definition captures the economic constraint
that agents cannot anticipate future speculative
price movements.
The most useful result of Ito’s stochastic cal-
culus is the so-called Ito’s lemma: any twice
continuously-differentiable function of an Ito pro-
cess is itself an Ito process. This implies that the
agent’s wealth process is an Ito process and there-
fore the Bellman equation in the stochastic dynamic
programming problem becomes a second-order
partial differential equation. The latter allows one to
analyse the portfolio problem by looking at just the
first two moments of price processes and to achieve
sharp characterizations. Merton ( ) applied this
technique further to
study equilibrium relations among risky asset prices
and arrived at the Intertemporal Capital Asset
Pricing Model.
The introduction of Ito’s stochastic calculus
opened a whole new world for economists. With it,
most of the static utility maximization models are
readily extended to a dynamic setting with
uncertainty. The continuous time set-up allows one
to work with differential equations rather than with
difference equations. For applications to capital
theory and economic growth, see Bismut ( ),
Brock and Magill ( ), and
Merton ( ); to asset pricing models, see Cox
et al. ( , ).
Ito’s work was later extended by Kunita and
Watanabe ( ) to the case where integrators are
square-integrable martingales. They also proved a
martingale representation theorem for a Brownian
motion: any square-integrable martingale adapted to
a Brownian motion filtration (see below) is
representable as an Ito integral.
The most general notions of a stochastic process
and a stochastic integral to date are in the terrain of
the so-called French School Probability Theory, or
surely developed for intrinsic intellectual reasons, it
nevertheless seems to have been
2200
invented for the study of financial markets; for
references, see Dellacherie and Meyer ( , take the time span of the economy to be [0,1], We
1982), Jacod ( ), and Meyer ( , ).
Although making no explicit use of French
probability theory, the seminal paper of Black and
Scholes ( ) in the pricing of stock options
nevertheless opens up the possibility of its appli-
cation in financial economics. This work was sub-
sequently generalized and formalized by Merton (
, ). The idea is that the payoff of a
stock option can be replicated by continuous trad-
ing in its underlying stock and a riskless asset. The
replicating strategy is self-financing in that after the
beginning of this strategy there are neither
additional funds invested into it nor funds with-
drawn out of it. Thus, to rule out arbitrage oppor-
tunities, the stock option must sell for the exact
value of the replicating portfolio at any point in
time.
Black and Scholes’s theory provided a strong
incentive for financial economists to study contin-
uous time stochastic processes. The key observation
in this literature was made by Cox and Ross ( ).
They noted that since the expected rate of
return of the stock does not enter the Black and
Scholes pricing formula for a stock option, the price
of an option must be determined as if investors were
risk neutral and had probability beliefs such that the
stock earns an expected rate of return equal to the
riskless rate. Harrison and Kreps ( ) formalized
this observation in showing
that any arbitrage-free price system can be
converted into a martingale through a change of an
equivalent probability after a suitable normal-
ization. Note that this martingale connection of an
arbitrage-free price system was vaguely
foreshadowed in Samuelson and Merton ( ).
Harrison and Kreps ( ) and Harrison and
Pliska ( ) make clear that the answer to
whether a contingent claim can be replicated by
dynamic trading is intimately related to the mar-
tingale representation theorem. A sketch of their
arguments will be given.
Taken as primitive is a complete separable
probability space (Q, W- P) and a filtration F {7Ft; i
C [0, 1]}. A filtration is an increasing family of sub-
sigma-algebras of W representing
Continuous-Time Stochastic Processes
information revelation over time. For simplicity, we
assume that agents are endowed with the same
information structure F. Readers can think of a
filtration to be like an event tree in a discrete time
finite state setting. We also assume that agents at
time zero knows that the frue state of the nature is
an element co e Q, which they will learn at time
one.
Agents can only consume at time one. For
simplicity again, we take the commodity space to be
the space of square-integrable random variables
defined on (Q, fty P), denoted by L2(P).
There are N + 1 long-lived securities traded
indexed by n 0 , 1 , . . . , N. A long-lived security is a
security available for trading all the time in [0,1]
and is represented by a price process. {>S„(0} It
pays a dividend only at time one and is equal to
S„( 1) almost surely. Price processes are semi-
martingales (adapted to F) and S„( 1) C_ L2(P). In
modelling a dynamic asset trading economy, before
anything interesting can be said, one has to
formulate a budget constraint. That naturally
involves stochastic integrals. Jacod ( ) has
shown that for stochastic integrals to have desired
properties, it is necessary that integrators be semi-
martingales. Thus, semimartingale price processes
can be assumed without loss of generality.
In a Walrasian economy, only relative prices are
determined. Thus we can assume that the price
system has been normalized such that So(t) = 1 V/
C [0, 1], We will call the 0th security the riskless
security and the rest risky seciuities.
A trading strategy is an (N - 1 (-dimensional
predictable process 0 = {0„ (t); n = 0, 1, - N}, where
we interpret 0„ (/) to be the number of shares of
security n held from t - to t before trading at time t.
A process is predictable if its values at time t
depend only upon the information available strictly
before time t. Given the interpretation of 0,
predictability is a natural information constraint.
A trading strategy is said to be simple if it is
bounded and changes its value at most at a finite
number of time points in [0,1].
A trading strategy 0; is said to be self-financing
if the stochastic integral
Continuous-Time Stochastic Processes
is well-defined and if
T r
f
d(t) S ( t ) = 0(O) S(O) + 6 ( s ) r d S ( s ) Vt e [0, 1] a.s.,
JO
(l)
r
where denotes transpose. That is, the value of the
portfolio 9 at time / is equal to its initial value plus
accumulated capital gains or losses from time zero
to time t. There are neither new investments into nor
withdrawals of funds out of the portfolio. This is
just a natural budget constraint.
Harrison and Kreps ( ) and Kreps ( ) expectation under g by the Cauchy-Schwarz
show that if all the simple of self-financing trading
strategies are allowed and if arbitrage opportunities
are absent, then there exists a probability measure g
equivalent to P such that the Radon- Nikodym
derivative E, = dQ/dP is an element of L2(P) and
that S is a martingale under g, or a g-martingale. Fix
g and note that since P and g are equivalent, all the
a.s. statements to follow apply to both.
Now we can specify the space of admissible
strategies 0 [5]. A self-financing trading strategy 9
is admissible if
d (s ) T dS (s ) t e [0,1]
Jo
is a g-martingale and 9 (1 ) 7 .ST 1) C L2(P). [See
Jacod ( ) for sufficient conditions for this to be
true.] Then one can show that given 0 [5] indeed
there are no arbitrage opportunities.
A contingent claim is an element of L2{P). A
contingent claim .r is said to be marketed if it can be
dynamically manufactured by an admissible trading
strategy. Formally, .r is marketed if there exists 9 e 0
[*S] such that
1 representable is an Ito process had been an open
x = d(0)T S(0) + d ( t ) T d S( t ) a . s.
JO
The value of .r at time t is 9 (t)T S(t). By the
definition of admissibility, we have
2201
[ 6(s)TdS = N rt
Jo 51 en(s) dS„ (i). (2)
Now here is the key observation. Let x be a
contingent claim. We know from relation ( ) that if
it is marketed, its value over time is equal to its
initial value at time zero plus a stochastic integral
with respect to N g-martingalcs. Conversely, a
contingent claim x marketed if the conditional
expectation E [x | y(], which is a g-martingale, can
be represented by a stochastic integral with respect
to the N g-martingales {Sn(t); n = 1, 2, ..., N}. [Here
we should remark that any x e L2(P) has a finite
inequality.] This observation turns on the
machinery of the martingale representation theorem
in the study of market completeness.
The security markets are said to be dynamically
complete if all contingent claims are marketed.
From the above discussion, it follows that markets
are dynamically complete if all g-martingales are
representable as stochastic integrals with respect to
the N risky g-martingale prices. In such event, the
N g-martingales are said to have the martingale
representation property. Readers might be curious
by now that the riskless asset seems to disappear
from the story. Indeed, whether a contingent claim
is generated by a (not necessarily self-financing)
trading strategy does not depend upon the riskless
asset after time zero. The riskless asset, however, is
a vehicle through which the budget is balanced over
time.
The contribution made by Harrison, Kreps, and
Pliska is methodological. They make available a
powerful machinery for the study of financial/cap-
ital markets: the theory of martingales. Now we
shall present some consequences of their work.
Since Merton’s ( , , ) analyses
of optimal intertemporal consumption-portfolio
policies and their implications on equilibrium asset
prices, the conditions under which a price system is
question for more than a decade. A short answer
found in Huang ( ) is as follows:
Take the set-up of the economy as above and
2202
assume henceforth that there are no arbitrage
opportunities. Moreover, assume that the informa-
tion structure F is a Brownian motion filtration. We
know S is a 0-martingalc, so we can write
S„(f) = £*[S„(l)|r] a.s.
_E[S,{1)Zfr]
E[^] a.s.,
where the second equality follows from the Bayes’
rule, and where we recall that c = d0/dP, which is
strictly positive by the fact that Q and P are equiv-
alent. The numerator and the denominator of the
above relation are both /’-square integrable martin-
gales. By the martingale representation theorem of
Kunita and Watanabe ( ), we know that any
P-square integrable martingale is representable as
an Ito integral. Then S„ is an ltd process by ltd
lemma. Hence any arbitrage-free price system is an
ltd process when the information structure is a
Brownian filtration.
We can also study the sample path properties of
a price system, which relates to examining empir-
ically the so-called efficient market hypothesis.
Much of the empirical work in financial economics
and accounting concerns the response of cap-
ital/financial asset prices to information. The null
hypothesis in this work is typically that the capi-
tal/financial markets are efficient in the sense that
prices rapidly adjust to new information. But is it
true that prices only make large adjustments at
surprises and what exactly is a surprise, mathe-
matically? Here we turn to the classification of
stopping times in the general theory of processes. In
this context, a surprise is a non-predictable stopping
time. We also know that a martingale must be
continuous at predictable stopping times (provided
that a minor technical condition is satisfied). Thus,
S can make discrete changes only at nonpredictable
stopping times or at surprises. This and other related
issues can be found in Huang ( , ). A ization, a price system plus the accumulated
reference for the classification of
stopping times is Dellacherie and Meyer ( ).
So we discover that a price system must be ltd
process when the information is a Brownian motion
filtration and when there are no arbitrage
opportunities. There still remain further questions:
Continuous-Time Stochastic Processes
does there exist an equilibrium where equilibrium
price processes are ltd processes? More impor-
tantly, does there exist an equilibrium where
although there are only a finite number of long-
lived securities traded, the markets are dynamically
complete and thus the equilibrium allocation is
Pareto optimal? Note that in the Arrow-Debreu
equilibrium theory, markets for all contingent
claims are available at time zero. Agents trade to a
Pareto optimal allocation. There is no need and no
incentive for the markets to reopen after time zero.
Of course, this does not conform with actual market
structures. We do not have a complete set of
contingent markets. What we do have are
constantly-open financial markets where a finite
number of long-lived assets are traded. Thus it is
important to know whether there exists an equi-
librium in such a world and to know the efficiency
of the resulting allocation.
It follows from the earlier discussion on the
martingale representation property of risky price
processes that what is needed for an affirmative
answer to the above questions: is that there be a
riskless security with unit price throughout and a
finite number of risky long-lived securities that
have the martingale representation property. What
complicates the story, however, is that the demand
and supply of the long-lived securities must be
equal in equilibrium. Thus those securities must be
picked carefully. Moreover, it is not true that a
finite number of martingales having the martingale
representation property can always be found. Duffie
and Huang ( , ) and Dufifie
( ), in exchange as well as production econo
mies, demonstrated a procedure to select long-lived
securities having the desired properties and condi-
tions under which the number is finite.
The martingale connection of an arbitrage-free
system has been generalized to economies where
securities can pay dividends and agents can con-
sume at any time in [0,1], After a suitable normal-
dividends form a martingale under an equivalent
probability measure. This is done in Huang ( ).
Similar theory is also valid in economies
where agents have differential information. Inter-
ested readers are referred to Duffie and Huang (
) for details.
Continuous-Time Stochastic Processes
Although the focus of research has been on
capital theory and financial markets, applications of
the theory of continuous time stochastic processes
to economic problems outside these areas can be
found. For example, Dufifie ( ) securities. Econometrica 53: 1337-1356.
applies classical potential theory as in the context of
Markov processes to valuation of securities, and Li
( ) examines the stochastic theory of
the firm in continuous time. He uses point processes
to model stochastic demands for commodities and
endogenizes a firm’s demand for inventories,
among other things. For applications of the theory
of optimal stopping to game theory, see Hugues (
) for zero-sum stopping games,
and Huang and Li ( ) for nonzero-sum stop
ping games.
See Also
► Capital Asset Pricing Model
► Continuous-Time Stochastic Models
► Options
Bibliography
Bachelier, L. 1900. Theorie de la speculation. Paris: Gauthier-
Villars.
Bismut, J. 1975. Growth and optimal intertemporal allocation
of risks. Journal of Economic Theory 10: 239-257. Black, F.,
andM. Scholes. 1973. The pricing of options and corporate
liabilities. Journal of'Political Economy 81: 637-654.
Brock, W., and M. Magill. 1979. Dynamics under uncertainty.
Econometrica 47: 843-868.
Chung, K., and R. Williams. 1983. An introduction to
stochastic integration. Boston: Birkhauser.
Cox, .1., and S. Ross. 1976. The valuation of options for
alternative stochastic processes. Journal of
Financial Economics 3: 145-166.
Cox, .1., .1. Ingersoll, and S. Ross. 1985a. An intertemporal
general equilibrium model of asset prices.
Econometrica 53: 363-384.
Cox, I, I Ingersoll, and S. Ross. 1985b. A theory of the term
structure of interest rates. Econometrica 53: 385-408.
Dellacherie, C., and P. Meyer. 1978. Probabilities and
potential A: General theoiy of process. New York: North-Holland.
Duffie, D. 1986a. Stochastic equilibria: Existence, spanning
number, and the ‘no expected gains for trade’ hypothesis.
Econometrica.
2203
Duffie, D. 1986b. Price operators: Extensions, potentials, and
the Markov valuation of securities. Research Paper No.
813. Graduate School of Business, Stanford University.
Duffie, D., and C. Huang. 1985. Implementing Arrow-Debreu
equilibria by continuous trading of few long-lived
Duffie, D., and C. Huang. 1986a. Multiperiod securities
markets with differential information: Martingales and
resolution times. Journal of Mathematical
Economics 15: 283.
Duffie, D., and C. Huang. 1986b. Stochastic production-
exchange equilibria. Stanford: Graduate School of
Business, Stanford University
Duffie, C., and P. Meyer. 1982. Probabilities and
potential B: Theory of martingales. New York:
North-Holland.
Einstein, A. 1905. On the movement of small particles
suspended in a stationary liquid demanded by the
molecular kinetic theory of heat. Annals of Physics
17: 549.
Harrison, M., and D. Kreps. 1979. Martingales and arbitrage
in multiperiod securities markets. Journal of Eco-
nomic Theory 20: 381—408.
Harrison, M., and S. Pliska. 1981. Martingales and stochastic
integrals in the theory of continuous trading. Stochas-
tic Processes and their Applications 11: 215-260.
Huang, C. 1985a. Information structure and equilibrium asset
prices. Journal of Economic Theory 35: 33-71.
Huang, C. 1985b. Information structure and viable price
systems. Journal of Mathematical Economics 14:
215-240.
Huang, C., and L. Li. 1986. Continuous time stopping games,
Working Paper No. 1796-86, Sloan School of
Management, MIT.
Hugues, C. 1974. Markov games. Technical Report No.33,
Department of Operations Research, Stanford University.
Ito, K. 1944. Stochastic integrals. Proceedings of the
Imperial Academy 22: 519-524. Tokyo.
Ito, K. 1951. On stochastic differential equations ,
Memoirs of the American Mathematical Society. Rhode
Island: The American Mathematical Society.
Ito, K., and H. McKean. 1965. Diffusion processes and
their sample paths. New York: Springer.
Jacod, J. 1979. Calcul stochastique etproblemes de
martingales, Lecture Notes in Mathematics, vol. 714.
New York: Springer.
Kreps, D. 1981. Arbitrage and equilibrium in economies with
infinitely many commodities. Journal of Mathe-
matical Economics 8: 15-35.
Kunita, H., and S. Watanabe. 1967. On square-integrable
martingales. Nagoya Mathematics Journal 30:
209-245.
Li, L. 1984. A stochastic theory of the firm. Unpublished PhD
thesis, Northwestern University
McKean, H. 1965. Appendix: A free boundary problem for
the heat equation arising from a problem in mathemat-
ical economics. Industrial Management Review 6:
32-39.
2204
McKean, H. 1969. Stochastic integrals. New York: Academic.
Merton, R. 1969. Lifetime portfolio selection under uncer-
tainty: The continuous case. Review of Economic and Statistics 51:
247-257.
Merton, R. 1971. Optimum consumption and portfolio rules
in a continuous time model. Journal of Economic Theory 3: 373-
413.
Merton, R. 1973a. An intertemporal capital asset pricing
model. Econometrica 41: 867-888.
Merton, R. 1973b. Theory of rational option pricing. Bell Journal
of Economics and Management Science 4: 141-183.
Merton, R. 1975. An asymptotic theory of growth under
uncertainty. Review of Economic Studies 42: 375-393.
Merton, R. 1977. On the pricing of contingent claims and the
Modigliani-Miller theorem. Journal of Financial Economics 5:
241-249.
Meyer, P. 1966. Probability and potentials. Waltham: Blaisdell
Publishing Company.
Meyer, P. 1976. Un corns sur les integrales stochastiques. In
Seminaires de Probability X . Lecture Notes in Mathematics 511.
New York: Springer.
Samuelson, P. 1965. Rational theory of warrant pricing.
Industrial Management Review 6: 13-32.
Samuelson, P., and R. Merton. 1969. A complete model of
warrant pricing that maximizes utility. Industrial Management
Review 10: 17-46.
Stratonovich, R. 1966. A new representation for stochastic
integrals and equations. SIAM Journal of Control 4: 362-371.
Contract Theory
David Martimort
Abstract
This article offers a brief overview of contract. It
focuses on the theory of complete contracts and
the three associated paradigms of adverse selec-
tion, moral hazard and non-verifiability. By
showing difficulties in allocating resources
between asymmetrically informed partners,
contract theory has deeply changed our view
ofthe functioning of organizations and markets.
Keywords
Adverse selection; Asymmetrical information;
Bayesian-Nash equilibrium; Collusion; Contract
theory; Cost observability; Free-rider
Contract Theory
problem; Incentive compatibility; Incentive
constraints; Incomplete contracts; Informative-
ness principle; Insurance; Laffont, J.-J.; Limited
liability; Monotonicity; Moral hazard; Multi-
agent organizations; Non-verifiability; Optimal
contract; Pontryagyn principle of optimality;
Principal and agent; Revelation principle; Risk
aversion; Risk neutrality; Sharecropping;
Spence-Mirrlees condition; Tournaments
JEL Classifications
DO
As with so many major concepts in economics,
contract theory was introduced by Adam Smith
who, in his monumental Wealth of Nations ( ,
book ID, ch. 2), considered the relationship between
peasants and farmers through this lens. For instance,
he pointed out the perverse incentives provided by
sharecropping contracts, widespread in 18th-century
Europe. However, it is fair to say that the issues of
incentives and contract theory were largely ignored
by economists until the end of the 20th century. By
then, the focus of economic theory was on the
working of markets and price formation. Firms were
viewed only as production technologies, and the
issue of the separation between ownership and
control was most often put aside. This black-box
approach was, of course, quite unsatisfactory. At the
turn ofthe 1970s, with the methodological
revolution of game theory, more emphasis was
placed on strategic interactions between a small
number of players in a world where informational
problems matter. From this new perspective, the
allocation of resources is no longer ruled by the
price system but by contracts between
asymmetrically informed partners. Contract theory
has deeply changed our view of the functioning of
organizations and markets.
This article aims to provide a brief overview of
contract theory, stressing a few major insights and
illustrating them with useful applications. Due to
space constraints, it does not do justice to several
aspects of contract theory, and will mostly reflect
my own tastes in the field. In particular, I focus on
the so-called theory of complete contracts, leaving
Contract Theory
aside the burgeoning theory of incomplete contracts
which is covered elsewhere in this dictionary.
Successive sections deal respectively, with adverse
selection, moral hazard and non-verifiability: the
three different paradigms which have been used in
the field of complete contract theory. Since the
distinction between complete and incomplete
contracts is easier to draw once these notions have
already been explained, I will postpone such
discussion to the end of the article.
Adverse Selection
Consider the following buyer-seller relationship as
the archetypical example of contractual relationship
between a principal (the buyer) and his agent (the
seller) who produces some good or service on his
behalf. The mere delegation of this task to the agent
gives the agent access to private information about
the technology. This adverse selection environment
is captured by assuming that a technological
parameter 9 is known only by the agent. It is drawn
from a distribution in an exogenous type space ©
which is common knowledge. Neither the principal
nor a court of law observes this parameter.
Contracts cannot specify outputs and prices as a
function of the realized state of nature.
The buyer enjoys a net benefit S(9, q) -1 when
buying q units of output at a price t. The seller
enjoys a profit t —C{9, q) from producing that
good. We will assume that these functions are
concave in q. Notice that the state of nature 9 might
affect both the agent’s and the principal’s utility
functions. This can, for instance, be the case if this
parameter also determines the quality of the good to
be traded.
Under complete information, efficiency requires
that the buyer and the seller trade the first-best
quantity q (0) such that the buyer’s marginal benefit
from consumption equals the seller’s marginal cost
of production:
^ (9 ,q f 9 ) ) = ^( 9 ,q *( 9) ) . ( 1) the seller’s participation constraints
2205
Many mechanisms or institutions lead to this
outcome. Both the price mechanism and a take-it-
or-leave-it offer by one party to the other would
achieve the same allocation, although with different
distributions of the surplus between the traders. If
the principal retains all bargaining power (for
instance, because there is a competitive fringe of
potential sellers), he could offer a forcing contract
stipulating an output q (0) and a transfer /(()) which
just covers the seller’s cost. This forcing contract
maximizes the buyer’s net gains from trade and
leaves the seller just indifferent between
participating or not.
In what follows, we mostly focus on the case
where the uninformed principal has full bargaining
power in contracting. In this framework, the con-
tract between the buyer and the seller does not only
have the allocative and distributive roles it has
under complete information. It also has the role of
communicating information from the informed party
to the uninformed party. This communication role
suggests that the informed party should be given a
choice among different options and that this choice
should reveal information about the adverse
selection parameter.
A first step in the analysis consists of describing
the set of allocations which are feasible under
asymmetric information. The basic tool for doing so
is the revelation principle (see Gibbard ; Green and
Laffont ; Dasgupta et al. ; Myerson , among others),
which states that there is no loss of generality in
restricting the analysis to revelation mechanisms
that are direct, that is, of the form Jv & with 9 a
message (‘report’) sent by the informed seller to the
uninformed buyer, and truthful, that is, such that the
agent finds it optimal to report his true type.
Therefore, incentive feasible contracts satisfy
the following incentive constraints
t { 9) - C ( 9, q( 9 )) > t ( e )
-c[9,q(9^v(9,9^je& 2 . (2)
To be acceptable, a contract must also satisfy
2206 Contract Theory

t (9 ) - C (9 ,q ( 9) ) > 0 0 ( 3) The result of this optimization is straightforward.

Inducing information revelation by the most efficient
which ensure that, irrespective of his type, the agent types requires giving up an information rent to those
by contracting gets at least his reservation payoff types. The basic intuition of most adverse-selection
(exogenously normalized to zero). models is that reducing this rent requires production
Once the set of incentive feasible allocations is to be distorted. For instance, when efficient types
described, the analysis may proceed further. want to mimic less efficient ones, the latter’s
Keeping in mind that the uninformed buyer designs allocation should be made less attractive. This is
his offer under asymmetric information, we might obtained by distorting their production downward
characterize an optimal contract. Such a contract and modifying transfers accordingly.
maximizes the uninformed buyer’s expected net To see more formally the nature of the output
surplus subject to the feasibility constraints ( ) and distortion, consider the case where types are dis-
( ). tributed over a compact set [9,0] according to the
Much of the theoretical literature developed cumulative distribution function F( ) (with a positive
over the 1980s and early 1990s has investigated the density/(•)). The second-best optimal output qSB(0)
structure of the set of incentive feasible allocations under adverse selection is the solution to:
and its consequences for optimal contracting. A key
property is the so-called Spence-Mirrlees condition
(see Spence
; Mirrlees ) for early contributions which put
|(e,^ 8)) = ^(8Y»(«))
(
forward that condition). This condition is satisfied
when the slope of the agent’s indifference curves
can be ranked with respect to his type. In our
example, this condition holds when jj^r/ > 0, that is,
Condition ( ) states that, for any type 9, the buyer’s
when higher types also have higher marginal costs
marginal benefit must equal the seller’s marginal
and should thus produce less. Therefore, the
virtual cost (see Laffont and Martimort , chs 2 and 3,
monotonicity condition
for details). The virtual cost of a given type takes
into account not only its cost of production but also
q( d) > q{ & ) fo r 9 < 9 ' ( 4) the cost of deterring other types (here more efficient
types) from mimicking that type. The allocation is
is a direct consequence of the incentive constraints. no longer efficient, as under complete information,
The Spence-Mirrlees condition can be viewed as a but interim efficient in the sense of Holmstrom and
regularity assumption making the incentive problem Myerson ( ).
well-behaved. It ensures that only incentive Condition ( ) is crucial, and is found in various
constraints between ‘nearby’ types matter in the forms in any adverse-selection model. It states that,
optimization. Intuitively, this means that the seller under asymmetric information, there is a
with a given marginal cost may be tempted to fundamental trade-off between implementing
overstate slightly its costs, receiving the higher allocations close to efficiency and giving informa-
transfer targeted to less efficient types but produc- tion rents to the most efficient types to induce
ing at a lower marginal cost. By so doing, this more information revelation. This trade-off calls for
efficient type receives an information rent Once distortions away from efficiency.
these local constraints are taken into account and Provided that the output schedule defined by ( )
when the Spence-Mirrlees condition holds, the satisfies the monotonicity condition ( ), this is the
incentives to mimic more distant types are no longer exact solution of our problem. To guarantee
relevant. With this reduction of the set of relevant monotonicity, on top of assumptions on the
incentive constraints, the principal’s optimization concavity of S(-) and % (•), convexity of C( ) and
problem is significantly simplified. 9}C
dC(\ d 2 C >0, IfiTtO >° and^(0 <°>
( 08 W’ dBdq l
08 dq
V
Contract Theory 2207

one needs also to impose a property on the type organizations. Such complex organizations emerge
distribution, the so-called monotonicity of the haz- because of the need to share common resources,
ard rate (see Bagnoli and Bergstrom ). Otherwise, produce public goods, internalize production
the optimal contract may entail some area of externalities or enjoy information economies of
pooling such that all types belonging to a set with scale. Although any such reason calls for a specific
positive measure produce the same amount and are analysis, a few common themes of the literature can
paid the same price. The optimal solution may then be highlighted by remaining at a rather general
level.
be obtained using ‘ironing techniques’ (see for
Regarding the implementation concept, different
instance Guesnerie and Laffont )•
notions of incentive compatibility may be used
depending on the context. First, agents may know
Direct Extensions
each other’s types and play a Nash equilibrium of
Adverse-selection methodology has been success-
the direct revelation mechanism offered by the
fully extended in various directions allowing for
principal (see Maskin , and the discussion of the
multidimensional types (Armstrong and Rochet ),
non-verifiability paradigm below). Second, agents
and/or multiple outputs (Laffont and Tirole , ch. 3),
may only know their own type, form beliefs on each
and type-dependent reservation utilities (Lewis and
others’ types and play a Bayesian-Nash equilibrium
Sappington ; Jullien ). There, the analysis is
(see D’Aspremont and Gerard-Varet ). Third, one
substantially more complex as types can no longer
may also
be ranked as easily as in the model sketched above.
insist on dominant strategy implementation because
The Spence-Mirrlees condition might fail to hold
it does not depend on the specification of beliefs
and global incentive constraints may bind, leading
(see Gibbard ; Groves ; Green and Laffont ). To
to pooling allocations being optimal. Another
each implementation concept corresponds a notion
interesting extension is the case of hidden
of incentive feasibility. Once the set of incentive
knowledge, in which contracting takes place before
feasible contracts is defined, one can proceed to
the agent becomes informed. The logic of such
optimization. It is a trivial observation that, the
models is very close to that we discuss below in the
more restrictive the implementation concept, the
section on moral hazard. In a nutshell, the trade-off
lower is the principal’s payoff at the optimum.
between allocative efficiency and rent extraction is
In some cases, such as the provision of public
now replaced by the trade-off between insuring the
goods within a society of privately informed agents
agent against shocks on costs and inducing him to
or in bargaining models between a buyer and a
reveal his cost once it is known. Output distortions
seller with equal bargaining power, the goal is no
still arise (see Laffont and Martimort ch. 2, for
longer to design a multilateral contract which would
details). Others have endogenized the asymmetric
extract the rents of all agents but, instead, to
information structure and examined the incentives
maximize some ex ante efficiency criterion under
to leam about the unknown parameter (see, for
incentive constraints. Groves ( ) showed that
instance, Cremer et al. ). Finally, there exists a
dominant strategy mechanisms suffice to implement
literature that considers the case where the principal
the first-best decision in a public good context. One
is the informed party (Maskin and Tirole , ).
caveat is that the budget generally fails to be
New difficulties
balanced. D’Aspremont and Gerard-Varet ( )
arise from the fact that the mere offer of the
proposed a Bayesian incentive-compatible
contract may signal information.
mechanism which implements the first-best and still
satisfies budget balance. As argued by Laffont and
Multi-agent Organizations
Maskin ( ), such a mechanism may con
The most important extensions of the adverse
flict with the agents’ participation constraint. In a
selection paradigm certainly concern multi-agent
bargaining environment, Myerson and Satterthwaite
( ) showed in a similar vein
2208
that there exists no Bayesian bargaining mechanism
that is efficient, budget-balance and individually
rational.
The optimal multilateral contract can be very
sensitive to the information structure. In environ-
ments where risk-neutral agents have correlated
types but know only their own type, the principal
can condition one agent’s compensation on
another’s report. By doing so, the principal can
fully extract the rent from both agents in a
Bayesian-Nash equilibrium. One may view this
result as a strong rationale for relative performance
evaluation, yardstick competition, benchmarking
and internalization of similar activities within the
same organization. This puzzling insight of Cremer
and McLean ( ) no longer
holds when one introduces risk-aversion, ex post
participation constraints or limited liability con-
straints. These assumptions reintroduce information
rents in the multi-agent organization, and the
standard trade-off between efficiency and rent
extraction reappears.
When the agents’ types are independently dis-
tributed, yardstick competition is ineffective and the
agents derive information rents. However, the
externality that one agent’s task may exert on
another can shape the distribution of these rents. In
competitive environments, such as procurement
auctions among sellers, it is no longer the distribu-
tion of the agents’ marginal costs but the distribu-
tion of their virtual marginal costs (see Myerson )
which determines who should produce and how
much. Because virtual costs may be ranked
differently from true costs, inefficiencies arise
under asymmetric information. Moreover, compe-
tition may help reduce rents by putting each agent
under the threat of being excluded from production
if he overstates his cost too much. There is then a
positive externality among competing agents.
Instead, more cooperative environments, such as
public good problems or procurement of com-
plementary inputs by several suppliers, involve
negative externalities between agents. Given that
each agent has a limited impact on the organiza-
tion’s overall production, the incentives to overstate
costs and thereby receive greater transfers are
exacerbated. ‘Free riding’ arises in such orga-
nizations (see Mailath and Postlewaite ).
Contract Theory
When competition between agents or between
agents and the supervisors supposed to monitor
them would benefit the principal, one must consider
the possibility of collusion aimed at securing more
rent. Reducing the scope for collusion requires
using mechanisms that are less sensitive to
information and reducing supervisory discretion.
Incentive contracts look more like inflexible
bureaucratic rules (see Tirole ; Laffont and
Martimort ). The optimal response to collusion may
also entail more delegation to lower levels of the
hierarchy, as in Laffont and Martimort ( ) and
Faure-Grimaud et al. ( ).
Dynamics
Different extensions of the static framework cor-
respond to different abilities of the contractual
partners to commit themselves inter-temporally
and/or different ways for the cost parameters to vary
over time. Under full commitment, the lessons of
the static rent-efficiency trade-off can be easily
extended, although the precise features of the
optimal contract depend on how types evolve over
time (see, for instance, Baron and Besanko , for the
case of persistent types). The case of limited
commitment is more interesting. Longterm
contracts may either be renegotiated (Dewatripont ;
Hart and Tirole ; Laffont and Tirole ) or even are
not feasible, in which cases the parties resort to spot
contracts (Laffont and Tirole ). The rent-efficiency
trade-off must be adapted to take into account how
information is revealed progressively over time.
However, the basic idea still holds. As past per-
formances reveal information about the agent’s
type, the optimal contract trades off ex post effi-
ciency gains in contracting against the agent’s
desire to hide information in the earlier periods of
the relationship so as to secure more rent in the later
periods.
Applications
Since the mid- 1980s, models of optimal contracting
under adverse selection have spanned the economic
literature. Let us quote only a few major applica-
tions. Mirrlees ( ) analysed optimal taxation
schemes when the agent’s productivity is privately
observed. He introduced the Spence-Mirrlees
Contract Theory
condition and derived die implementability condi-
tions. He also used optimal control techniques
(Pontryagyn Principle) to compute the optimal tax-
ation scheme. (The taxation problem differs from
our buyer-seller example because participation in
the mechanism is mandatory and the state’s budget
constraint must be added to the characterization of
feasible allocations. )
Mussa and Rosen ( ) studied the problem
of a monopolist selling one unit of a good to a
continuum of consumers vertically differentiated
with respect to their willingness to pay for the
quality of this good. This was the first model using
adverse selection techniques in a framework
without income effect. Maskin and Riley (
were interested in characterizing the optimal non-
linear price used by a monopolist in a second-
degree price discrimination context.
Baron and Myerson ( ) applied the meth
odology to the regulation of natural monopolies
privately informed about their marginal costs of
production. Laffont and Tirole ( ) extended
this analysis to allow for cost observability but also
introduced moral hazard elements (the possibility
for the regulated firm to reduce its costs by
undertaking some non-observable effort). They
derived cost-reimbursement rules and pricing pol-
icies. They showed that menus of linear contracts
might implement the optimal contract.
Green and Kahn ( ) and Hart (
studied labour market contracts and discussed
distortions towards overemployment or underem-
ployment that may arise depending on the con-
tractual environment considered.
Finally, Townsend ( ) and Gale and
Hellwig ( ) analysed optimal financial con
tracts in a framework where the borrower’s income
is observable only ex post and at a cost. Optimal
contracts may look like debt in such environments.
Moral Hazard
To return to our buyer-seller example, we now
assume that there is only one unit of a good to be
Risk Neutrality
traded whose quality q is random and which yields
A first case of interest is when the agent is risk-
a surplus S(q) to the buyer. The distribution of
neutral (u(t) = t). The simple ‘sell-out’ contract,
quality
t(q) = S(q) — C where C is a constant, implements
2209
is affected by an effort <? undertaken by the agent
at a cost ij/(e) (where i// > 0 and ij/" > 0). The
cumulative distribution is F(q\e) (with
density/(q\ej) on a support Q \q, q\ independent of
the agent’s effort. To simplify, the agent’s
preferences are separable in money and effort: U =
u(t) — ij/(e) where w(-) is increasing and concave
(u' > 0, u" < 0). The agent’s outside option is not to
produce, which gives him a payoff normalized to
zero.
The agent’s effort is observable neither by the
principal nor by a court of law. This is a moral
hazard setting. Contracts stipulate the agent’s
payment as a function of the realized quality
) assumed to be observable and verifiable
(contractible) by a court of law. Therefore, con-
tracts are of the form {t(q)}-, n.
If the effort were observable, its value could
also be specified by contract. Therefore, the seller
can at the same time be forced to exert the first-
best level of effort and be frilly insured against
uncertainty on realized quality with a flat payment
independent of his performance:
u(t*) = i// (e*).
This is no longer the case when the agent’s effort is
non-verifiable. The first step of the analysis is to
describe the set of feasible incentive contracts
) implementing a given level of effort e.
In amoral hazard setting, incentive constraints
write as:
“W?))/(<7\ e)dq - He)
1
> f u(t (q))f (q\ ^)dq- i l / (e') V(e,e' ). (6)
Jq
The agent’s participation constraint is:
-H ) > °-
e
(7)
2
2210
the first-best level of effort e . Provided that 5
(q)f(qI e*) ~ ^is scheme also extracts all
the surplus from the agent who is just indifferent
between producing or not.
Intuitively, with such a ‘sell-out’ contract, the
agent’s private incentives to exert effort are aligned
with the social incentives. This efficient outcome is
obtained by, first, having the agent pay a bond
worth C for the right to serve the principal, and
second, having the principal pay an amount S(q)
contingent on the quality realized.
Such a ‘sell-out’ contract requires that the agent
bears the full consequences of a bad performance. It
might not be feasible when the agent has limited
liability and cannot be punished for bad
performances. (For details, see Laffont and
Martimort , ch. 4). The conjunction of moral hazard
and limited liability allows the agent to derive a
limited liability rent. Intuitively, only rewards, not
punishments, can be used to provide incentives, and
this restriction on instruments is costly for the
principal. This rent creates a trade-off between
efficiency and rent extraction, as in the adverse
selection framework. Effort is distorted below the
first-best level.
Risk Aversion
Let us turn to the more complex case of risk
aversion. A first concern of the literature has been
to ‘simplify’ the set of incentive constraints ( ) by
replacing it with a first-order condition:
u{t{q))fM\e)dq = i//'(e). (8)
Denoting by X (resp. /i) the positive multiplier of
the incentive (resp. participation) constraint ( )
(resp. ( )), the optimal second-best schedule tsil(q)
satisfies
1 Milgrom (
„ | JM\e) (9.
u'i^iq)) f{q\e) '
This condition yields two important insights. First,
the contract must simultaneously provide the risk-
averse agent with insurance, which requires a fixed
payment, and with incentives to
Contract Theory
exert effort, which requires that payments be linked
to performance. There is now a trade-off between
insurance and incentives.
Second, the monotonicify of the agent’s com-
pensation with respect to the qualify level (a priori a
quite intuitive property) is obtained only when the
monotone likelihood ratio property holds,
namely, when fe{Ct\e)
f(q\e)
> 0 . This property
_
means that higher levels of performance are
more informative about the agent’s effort.
Finally, the optimal contract must use all sig-
nals which are informative about the agent’s effort
but no uninformative signals. Using them would
only let the agent bear more risk without any
beneficial impact on incentives. This is the
so-called informativeness principle of
Holmstrom ( ).
Extensions
In a model with a finite number of qualify and effort
levels, Grossman and Hart ( ) offered a
careful study of the set of incentive constraints and
its consequences for the shape of optimal contracts.
There is no general result on the ranking between the
first-best and the second-best effort levels in such
environments. The discrete version of the first-order
approach requires that only nearby constraints matter
in the agent’s problem. This concavity of the agent’s
problem is ensured when F(q\e) is itself convex in q.
In models with a continuum of effort levels and
outcomes, this first- order approach was suggested in
Mirrlees ( ),
more rigorously justified in Rogerson ( ) and
Jewitt ( ) and applied in Holmstrom ( )
and Shavell ( ).
The moral hazard methodology has been used to
justify the optimality of linear incentive schemes in
well-structured environments (Holmstrom and
Milgrom ); an often found feature of real world
contracts. Equipped with this tool, Holmstrom and
, ) investi
gated how multiple tasks and jobs should be
arranged in an organization.
To avoid the complexity of models with a
continuum of effort levels, modellers have found it
useful to focus on simplified environments with two
levels of effort. This approach was
Contract Theory 2211

instrumental in the work on corporate finance of of his performances. This may eliminate any agency
Holmstrom and Tirole ( ). problem, at least when parties do not discount too
much the future (see Laffont and Martimort , ch. 8,
Multi-agent Organizations for an example). More generally, the insurance-
When applied to multi-agent organizations, the incentives trade-off may be relaxed when the risk-
‘informativeness principle’ suggests that an agent’s averse agent’s rewards and punishments can be
compensation should be linked to another’s smoothed over the whole relationship, as shown in
performance if it is informative about his own effort Spear and Srivastava ( ). A direct consequence
(see Mookherjee ). Relative performance evaluation of inter-temporal
and benchmarking can help eliminate common smoothing is that the optimal dynamic contract
shocks affecting all agents’ performances. Of exhibits memory, good (resp. bad) performance
particular importance in this respect are today will also affect positively (resp. negatively)
tournaments which use only the ranking of the future compensations. This insight has been used to
agents’ performances to determinate their formalize a theory of the wage dynamics inside the
compensations. Tournaments provide agents with firm (Harris and Holmstrom ).
insurance against common shocks, which has a Fama ( ) argued that reputation in the
positive incentive effect. More generally, the labour market exerts enough discipline on managers
properties of tournaments and how they compare to alleviate moral hazard even in the absence of
with (a priori suboptimal) linear schemes have been explicit contracts. Holmstrom ( ) built a
investigated in Nalebuff and Stiglitz () and Green model of career concerns where the manager’s
and Stokey ( ). interest in influencing the labour market’s beliefs
In more cooperative environments where dif- concerning his or her quality provides incentives to
ferent agents contribute to a joint project, the exert effort. Career concerns are nevertheless in
fundamental difficulty is how to share the proceeds general not enough to induce first-best effort levels,
of production among agents of the team and still and some inefficiencies remain.
provide some incentives. Since each agent enjoys
only a fraction of those proceeds but bears the full
cost of his effort, he reduces his effort supply. This Non-verifiability
leads to a free-rider problem within teams, which is
analysed in Holmstrom ( ). Let us return to the buyer-seller model above.
If we remain in cooperative environments but Although we now assume that it is observable by
allow now for a principal acting as a budget both the principal and the agent, the state of nature
breaker, this principal may find it worthwhile to 6 may still not be verifiable by a court of law, in
reduce the agency cost of implementing a given which case it cannot be part of the contract. This
effort profile by having agents behave cooperatively shared knowledge stands in sharp contrast with the
(Itoh ). Even when agents do not cooperate, mutual asymmetric information structures examined in
observability of effort levels can also help to previous sections.
eliminate agency cost, as in Ma ( ). The first difficulty consists of building a mech-
This last argument relies on the logic of non- anism based only on verifiable variables (namely,
verifiability models, developed below. the quantities traded and corresponding payments)
which implements the first-best quantity q*(0) and
Dynamics
transfers f*(0). This problem was addressed by
The basic issue investigated by dynamic models of
Maskin ( ). He demonstrated
moral hazard is the extent to which repeated
that the first-best quantities and transfers can easily
relationships alleviate the moral hazard problem.
be implemented with a direct revelation mechanism
The intuition is that the principal should filter out
| t(6a,6b^,q^6a,6b^ g) e 0 2 where
the agent’s effort by looking at the whole history
both the buyer and the seller report
2212
simultaneously the state of nature they commonly
know. Truth-telling is obviously a Nash equilibrium
of this mechanism provided that both traders are
severely punished when making different reports,
since such cases would be inconsistent with the
underlying information structure.
A more subtle issue is how to design a mech-
anism such that this truthful Nash equilibrium is
unique. Maskin ( ) proposed a condition for
players’ preferences such that this is the case.
Moore and Repullo ( ) significantly extended
the domain of preferences by hardening the imple-
mentation concept, replacing Nash behaviour by
subgame-perfection in a sequential moves mech-
anism (see Laffont and Martimort , ch. 6, for an
example, and Moore , for an exhaustive survey of
the literature).
The basic thrust of the non-verifiability para-
digm is that a court of law can get around non-
verifiability by building such revelation
mechanisms, at least as long as the non-verifiable
state is payoff-relevant. If one sticks to that inter-
pretation, non-verifiability does not present a
significant limit on contracting.
A second issue of the literature is the impact of
non-verifiability on the incentives of traders to
perform specific and non-verifiable investments.
Given our previous claim that non-verifiability is
generally not a constraint, the model resembles the
standard moral hazard model. Providing incentives
for investments meets the same difficulties as in the
previous section.
Extensions
In practice, revelation mechanisms have been
criticized as overly complex, as relying on threats
which may either be non-credible or violate limited
liability constraints. The so-called incomplete
contracts literature has thus focused on cases where Acknowledgment I thank D. Gromb and J. Pouyet for
such revelation mechanisms are not feasible. In
such environments, either no contract at all or only
a very rough one can be written ex ante. For Bibliography
instance, parties can agree ex ante on a simple
fixed-price/fixed quantity contract which serves as a Armstrong, M., and J. Rochet. 1999. Multidimensional
threat point for the bargaining which takes place ex
post when the state of nature is realized (see Ed I in
and Reichelstein , among others).
Contract Theory
Alternatively, this threat point may be deter-
mined by the allocation of ownership rights where
such a right gives the owner the opportunity to use
assets as he prefers in case bargaining fails (see
Grossman and Hart ; Hart and Moore ). The issue is
then to derive from those exogenous constraints
distortions of investments and optimal
organizations which may mitigate those distortions.
The incomplete contracts paradigm is similar to
the complete contracts one (adverse selection,
moral hazard and non-verifiability) in the sense that
it also imposes limits on what a court may verify. It
differs from it because it also imposes exogenous
restrictions on the set of mechanisms available to
the parties. The justification for these restrictions is
found either in the bounded rationality of players or
the difficulties in describing or foreseeing
contingencies, all theoretical issues which remain
high on the agenda of economic theorists and are
still unsettled. The relevant literature on incomplete
contracts is too large to be summarized in this short
article. The interested reader may refer to Tirole (
) for an
overview or to the entry for this term in this
Dictionary.
See Also
► p 2..
► Ag»,::r;j '-'ms
► Incomplete Contracts
► Mechanism Design
► i/X Design. J Wtr'l e v * . n t s )
► i/'orc.CFaza: d
helpful comments on an earlier version
screening: A user’s guide. European Economic Review
43: 959-979.
Bagnoli, M., and T. Bergstrom. 2005. Log-concave probability
and its applications. Economic Theory 26: 445-469.
Contract Theory
Baron, D., and D. Besanko. 1984. Regulation and information
in a continuing relationship. Information Economics and
Policy 1: 447-470.
Baron, D., andR. Myerson. 1982. Regulating a monopolist
with unknown costs. Econometrica 50: 911-930.
Cremer, J., andR. McLean. 1988. Full extraction of surplus in
Bayesian and dominant strategy auctions. Econometrica
56: 1247-1257.
Cremer, J., F. Khalil, and J. Rochet. 1998. Strategic infor-
mation gathering before a contract is offered. Journal of
Economic Theory 81: 163-200.
D’Aspremont, C., and L. Gerard-Varet. 1979. Incentives and
incomplete information. Journal of Public Economics 11:
25-45.
Dasgupta, R, R Hammond, and E. Maskin. 1979. The
implementation of social choice rules. Review of Economic
Studies 46: 185-216.
Dewatripont, M. 1989. Renegotiation and revelation infor-
mation over time: The case of optimal labour contracts.
Quarterly Journal of Economics 104: 489-520.
Edlin, A., and S. Reichelstein. 1996. Hold-ups, standard
breach remedies, and optimal investments. American
Economic Review 86: 478-501.
Fama, E. 1980. Agency problem and the theory of the firm.
Journal of Political Economy 88: 288-307.
Faure-Grimaud, A., J.-J. Laffont, and D. Martimort. 2003.
Collusion, delegation and supervision with soft
information. Review of Economic Studies 70: 253-280.
Gale, D., and M. Hellwig. 1985. Incentive-compatible debt
contracts: The one-period problem. Review of Economic
Studies 52: 647-663.
Gibbard, A. 1973. Manipulations of voting schemes: A
generalized result. Econometrica 41: 587-601.
Green, J., and C. Kahn. 1983. Wage-employment contracts.
Quarterly Journal of Economics 98: 173-188.
Green, J., and J.-J. Laffont. 1977. Characterization of
satisfactory mechanisms for the revelation of preferences
for public goods. Econometrica 45: 427-438.
Green, J., and N. Stokey. 1983. A comparison of tournaments
and contracts. Journal of Political Economy 91: 349-364.
Grossman, S., and O. Hart 1983. An analysis of the principal-
agent problem. Econometrica 51: 7-45.
Grossman, S., and O. Hart. 1986. The costs and benefits of
ownership: A theory of lateral and vertical integration.
Journal of Political Economy 94: 691 -719.
Groves, T. 1973. Incentives in teams. Econometrica 41: 617-
631.
Guesnerie, R., and J.-J. Laffont 1984. A complete solution to a
class of principal-agent problems with an application to
the control of a self-managed firm. Journal of Public
Economics 25: 329-369.
Harris, M., and B. Holmstrom. 1982. A theory of wage
dynamics. Review of Economic Studies 49: 315-333.
Hart, O. 1983. Optimal labour contracts under asymmetric
information: An introduction. Review of Economic Studies
50: 3-35.
2213
Hart, O., and J. Moore. 1988. Property rights and the nature
of the firm. Journal of Political Economy 98: 1119-1158.
Hart, O., and J. Tirole. 1988. Contract renegotiation and
Coasian dynamics. Review of Economic Studies 55: 509-
540.
Holmstrom, B. 1979. Moral hazard and observability. Bell
Journal of Economics 10: 74-91.
Holmstrom, B. 1982. Moral hazard in teams. Bell Journal of
Economics 13: 324-340.
Holmstrom, B. 1999. Managerial incentive problems: A
dynamic perspective. Review of Economic Studies 66: 169-
182.
Holmstrom, B., and R Milgrom. 1987. Aggregation and
linearity in the provision of intertemporal incentives.
Econometrica 55: 303-328.
Holmstrom, B., andR Milgrom. 1991. Multi-task principal
agent analysis. Journal of Law, Economics, and Orga-
nization 7: 24-52.
Holmstrom, B., and R Milgrom. 1994. The firm as an
incentive system. American Economic Review 84: 972-991.
Holmstrom, B., and R. Myerson. 1983. Efficient and durable
decision rales with incomplete information. Econometrica
51: 1799-1819.
Holmstrom, B., and J. Tirole. 1997. Financial intermediation,
loanable funds, and the real sector. Quarterly Journal of
Economics 112: 663-691.
Itoh, H. 1993. Coalition incentives and risk-sharing. Journal
of Economic Theory 60: 416-427.
Jewitt, I. 1988. Justifying the first-order approach to
principal-agent problems. Econometrica 56: 1177-1190.
Jullien, B. 2000. Participation constraints in adverse-
selection models. Journal of Economic Theory 93: 1-47.
Laffont, J.-J., and D. Martimort 1998. Collusion and del-
egation. Rand Journal of Economics 29: 280-305.
Laffont, J.-J., and D. Martimort 2000. Mechanism design
with collusion and correlation. Econometrica 68: 309-342.
Laffont, J.-J., andD. Martimort 2002. The theory of incen-
tives: The principal-agent model. Princeton: Princeton
University Press.
Laffont, J.-J., and E. Maskin. 1979. A differentiable approach
to expected utility maximizing mechanisms. In
Aggregation and revelation of preferences, ed. J.-J.
Laffont. Amsterdam: North-Holland.
Laffont, J.-J., and J. Tirole. 1986. Using cost observation to
regulate firms. Journal of Political Economy 94: 614-641.
Laffont, J.-J., and J. Tirole. 1988. The dynamics of incentive
contracts. Econometrica 56: 1153-1175.
Laffont, J.-J., and J. Tirole. 1990. Adverse selection and
renegotiation in procurement. Review of Economic
Studies 57: 597-626.
Laffont, J.-J., and J. Tirole. 1993. A theoty of incentives in
procurement and regulation. Cambridge, MA: MTT
Press.
2214 Contracting in Firms

Lewis, T., and D. Sappington. 1 989. Countervailing incen- Tirole, J. 1986. I lierarchies and bureaucracies: On the role of
tives in agency problems. Journal of Economic Theory 49: collusion in organizations. Journal of Law, Economics and
294 313. Organization 2 : 1 8 1 2 1 4 .
Ma, C. 1988. Unique implementation of incentive contracts Tirole, J. 1999. Incomplete contracts: Where do we stand?
witli many agents. Review of Economic Studies 55: 555 Econometrica 67: 741 782.
572. Townsend, R. 1979. Optimal contracts and competitive
Mailath, G., and A. Postlcwaitc. 1690. Asymmetric infor- markets w i t h costly state verification. Journal of Eco-
mation bargaining problems witlr many agents. Review of nomic Theory 2 1 : 4 1 7 425.
Economic Studies 57: 35 1 638.
Maskin, E. 1999. Nasli equilibrium and welfare optimality.
Review of Economic Studies 66: 23-38.
Maskin, E.,and J. Riley. 1684. Monopoly with incomplete
information. Rand Journal of Economics 15: 171 196.
Contracting in Firms
Maskin, E., and J. Tirole. 1990. The principal-agent rela-
tionship with an informed principal. I: Private values.
Econometrica 58: 37l> 410. Canice Prendergast
Maskin, E., and J. Tirolc. 1992. The principal-agent rela-
tionship with an informed principal. II: Common values.
Econometrica 60: 1 42.
Mirrlees, .1. 1071. An exploration in the theory of optimum
income taxation. Review of Economic Studies 38: 175 208. Abstract
Mirrlees, .1. 1999. The theory of moral hazard with This article provides an overview of recent
unobservable behaviour. Part I. Review of Economic
advances in theoretical and empirical work on
Studies 66: 3 22.
Mookheijee, D. 1684. Optimal incentive schemes with many incentive contracting in firms. The specific focus
agents. Review of Economic Studies 51: 4.33 446. is on a variety of reasons why the prediction of
Moore, .1. 1992. Implementation in environments with the early literature on contracting -suggesting a
complete information. In Advances in economic theory, ed.
strong relationship between performance and
J.-J. Laffont. Cambridge: Cambridge University Press.
Moore, .1., and R. Repullo. 1988. Subgame-perfect imple- pay - has not been borne out.
mentation. Econometrica 56: 1191-1120.
Mussa, M., and S. Rosen. 1978. Monopoly and product
Keywords
quality. Journal of Economic Theory 18: 301-317.
Myerson, R. 1979. Incentive compatibility and the bargaining Agency theory; Contracting in firms; External-
problem. Econometrica 47: 61-73. ities; Free rider problem; Incentive contracts;
Myerson, R. 1981. Optimal auction design. Mathematics of Input monitoring; Measurement error; Multi-
Operations Research 6: 58-63.
tasking; Performance-related pay; Principal and
Myerson, R., and M. Satterthwaite. 1983. Efficient mech-
anisms for bilateral trading. Journal of Economic Theory agent; Private information; Risk
28: 61 7.3.
Nalebuff, B., and .1. Slight/. 1983. Prizes and incentives:
Towards a general theory of compensation. Bell Journal
JEL Classifications
of Economics 14:21 4.3.
J410
Rogerson, W. 1985. The first-order approach to principal-
agent problems. Econometrica 5.3: 1357 1368.
ShavcM, S. ll>71). On moral hazard and insurance. Quarterly
Journal of Economics 93: 541 562. In many realms of economic life, the actions of
Smith, A. 1776. The wealth of nations, 1991. New York: individuals affect the welfare of others. Nowhere is
Prometheus Books. this more relevant than in films, where employees
Spear, S., and S. Srivastava. 11>87. On repeated moral hazard act on behalf of owners or shareholders to provide
witli discounting. Review of Economic Studies 54:599-617.
Spence, M. 1673. Job market signaling. Quarterly Journal of
services for customers and clients. This separation
Economics 87: 355 .374. of the interests of employees from those whose
Spence, M. 1974. Market signalling: Informational transfer in actions they benefit has generated a large literature
hiring and related processes. Cambridge, MA: Harvard on incentive contracting, where the overarching
University Press.
objective is the alignment of such interests. The
early literature on agency theory, described in the
first edition
Contracting in Firms
of this volume by Lazear ( ), conceptually
mimics that on externalities - the other area of
economics that deals with welfare consequences of
actions on others - by showing a variety of ways in
which the compensation of agents can be
constructed to internalize the effects on one’s
actions on others. There are two ways of doing this.
First, one could simply tell employees what to do
and to penalize them if they fail to do so. In the
literature, this is referred to as input monitoring.
While this can sometimes help, it is often hard to
monitor either what workers do, or the intensity
with which they do so - a salesman on the road
would be a good example. Similarly, while over-
seers can sometimes identify what it is that agents
are doing, they may not know what they should be
doing - a board of directors monitoring a CEO
would be apposite here. Accordingly, the second
solution to misaligned incentives is to design
compensation plans such that the agent’s pay
depends on her contribution - ‘output’ - so that the
concerns of other parties are internalized.
A simple model can illustrate this point, and is
useful to describe other complications that can arise.
The agent is assumed to take some action (‘effort’)
e > 0, which is unobserved by the principal. She is
averse to exerting effort. Consider a simple
parameterization of the agent’s utility function,
where the agent cares about wages w and effort;
assume that the agent has exponential utility V =
cxp[ r(w — C(e))], where w is the worker’s wage, r
> 0 is the constant rate of absolute risk aversion, the
worker’s cost of supplying effort is C(e) = and her
reservation utility is U*. To focus attention on the
role of output contacting, assume that the principal
cannot observe effort e (so monitoring of inputs is
not possible), but instead only observes a signal on
effort y = e + £, where e ~ N(0, a2), with a2
representing measurement error. Assume also that
the principal chooses to reward the agent in a linear
fashion on output - a piece rate: w = fj0 + ft vv.
(There is a large literature on the optimal shape of
compensation contracts - see Prendergast ;
Gibbons , for an overview). Then there is a simple
solution to attaining efficient effort: choose the
contact to internalize
2215
the benefit to others by setting [>v = 1. In words,
efficiency arises when the agent is residual claimant
on the benefit of others.
This solution, providing a simple prescription
for how compensation contacts should be designed,
is both simple and intuitive. And empirically false.
There are, of course, some occupations where one
can find evidence of such ‘high- powered
incentives’, where agents are essentially residual
claimants on output. Indeed, the literature on
agency theory is replete with references to such
occupations - taxi cab drivers, franchisees, share-
cropper farmers and the self-employed. Yet these
are exceptions; instead, ‘low-powered incentives’ in
firms are more the norm (see Prendergast for
details). Consequently, one of the quandaries of the
literature has become why so few workers seem to
have contracts where their pay is strongly linked to
their performance, and much of the subsequent
literature to that outlined in the first edition of the
New Palgrave has identified relevant constraints on
incentive contracting.
The earliest candidate to explain why high-
powered incentives are rare is that high-powered
contracts impose risk on workers (Holmstrom ).
Consider the contract that induces efficient effort
above: /?„ = 1. The objective of the firm is to
maximize profits subject to the worker’s willingness
to take the position. This implies that the fixed
component, /i0, is changed to guarantee that agents
earn their reservation utility, so the principal’s
objective becomes a surplus maximization exercise.
When the worker is risk neutral, the fixed
component is reduced sufficiently such that the total
compensation cost is U * + %. In words, the only
cost that the employer incurs in addition to U* is the
effort cost. This is not true when the worker is risk
averse. In the context of the preferences V above,
compensation costs increase when incentive
contracts are used for two reasons - the cost of
increased effort as above, but also a risk cost
imposed on workers. Both costs are increasing in
[y. With exponential preferences and linear
contracts, this trade-off results in the optimal
contract being /?* = l+lrcai ■ This approach to
studying incentive contracting has become knows as
the ‘trade-off of risk and
2216
incentives’, where firms trade off the benefits of
great effort with higher compensation costs induced
by a risk premium, such that the chosen level of
effort falls below the level that internalizes benefits
to others. Only in the case where there is either no
measurement error (a2 = 0) or risk neutrality (r = 0)
does efficient effort arise.
At its most general, this costliness of exposing a
worker to large degrees of risk (or its analogue,
liquidity constraints) surely explains some part of
the absence of high-powered incentives. In much
the same way as financial assets with higher
undiversifiable risk require higher expected returns,
so also are risky jobs likely to demand higher
compensation. Despite this, the empirical literature
on how compensation contracts trade off such risk
issues against higher effort has shown little
evidence in its favour. There are two principal
empirical implications of the theory. First, riskier
environments should have lower incentives /;,*
declines with a2. There have been many studies of
the relationship between risk and the strength of
incentives in a variety of occupations. If anything,
this literature suggests that the relationship between
risk and the provision of incentives is positive rather
than the negative relationship posited by this theory.
See Prendergast ( ) for details and an explanation
as to why
this may be. Second, the trade-off of risk and
incentives implies that compensation should not
depend on measures that workers cannot control.
Again, this has found little support in the data. For
example, Bertrand and Mullainathan ( ), have
examined executive contracts in the United States,
and found little evidence that contracts reward
executives any less for measures that they cannot
control (say, where an oil company’s profits change
simply because the price of crude changed) than for
those that they can (such as a merger). More
evidence on this failure to filter out uncontrollable
factors concerns the infrequency of relative
performance evaluation. Consider two sales-force
workers (or executives) who carry out a similar job.
If demand for the products that they sell varies for
common reasons beyond their control, an efficient
way of limiting risk exposure is to (at least
partially) reward the workers on how
Contracting in Firms
well they do relative to each other. Yet empirically
there is relatively little evidence of such
benchmarking (for example, see Janakiraman et
al. ).
A second limitation on incentive contracting
arises when measures do not reflect the objectives
of the principal. Workers often carry out a host of
activities in their jobs, yet measures of performance
may not reflect all these aspects. A good example of
this would be measuring the performance of a
teacher. While measures may be available on some
component of what they do - such as test scores for
a teacher - many important aspects may remain
unmeasured. When contracts are designed on the
subset of things that can be measured, there is a
danger that they ignore the unmeasured aspects. For
instance, there is evidence of teachers ‘teaching for
the test’ or cheating to achieve higher test scores
(Jacob and Levitt ). This phenomenon has become
known as multitasking (Holmstrom and Milgrom ),
which becomes potentially important when there is
no single measure that reflects the contribution of
an agent. Accordingly, it is not surprising that a
consistent empirical finding is that jobs which are
described by firms as complex tend not to offer
significant incentive pay (see Prendergast , for
details).
Another limitation on the ability of firms to
provide incentives to workers comes from team
production. Measures of performance for most
workers reflect not only what they do but also the
contributions of others. In itself, this does not
change the calculus above in any conceptual sense,
other than that the measurement error now includes
the actions of others. As an example, assume that
two agents (1 and 2) work on a team and that output
measures the true contributions of both plus an error
term y = ei + e2 + e. Efficient effort arises as before
by setting [iv = 1 for each worker. However, there is
now a potential problem of budget breaking, where
marginal payments exceed marginal output. In this
example, when total output rises by one dollar,
compensation costs increase by two dollars. In
many firms - for instance, partnerships - such
budget breaking is not possible. If instead the
principal
Contracting in Firms
can pay out no more than one dollar for every dollar
extra on output, this naturally places an upper
bound of /(,. ^ on average for the agents. Hence,
budget balancing places a natural limitation on firm
incentives. This also leads to a free rider problem in
teams, where maximum incentive compensation in
an N member team mechanically declines as N
increases (This is known as the ‘^problem’). There
is also considerable empirical evidence (such as
Gaynor and Pauly ) on such free riding - mostly
from legal and medical partnerships - illustrating
how various measures of performance disimprove
as the size of the team being rewarded increases.
Many measures of output are not denominated
in dollar terms, but instead come in the form of
evaluations by others. For instance, it would be
difficult to measure the contribution of a social
worker or a customer service representative without
using feedback from supervisors or clients. Another
limitation on contracts arises when such subjective
measures can be corrupted by evaluators with
vested interests. Two particular sources of such
vested interests have been considered in the
literature. First, information on performance often
originates with clients as they are the only ones with
first-hand experience of the agent’s efforts. For
instance, compensation for many customer service
representatives depends on client evaluations. When
clients have relatively similar preferences to the
principal - such as that the agent should be
courteous and efficient - contracts based on
evaluations can mimic the objective contracts
above. Yet in other instances, the vested interest of
clients can render incentives difficult to implement.
A good example of this arises in occupations such
as police or immigration control, whose objective is
not necessarily to make their clients happy. Making
pay depend on evaluations in these instances can be
harmful as it gives agents incentives to keep clients
happy when they should not, such as a police
officer not arresting a suspect. In these cases,
incentive contracting on evaluations typically needs
to be curtailed to avoid such incentives (see
Prendergast , for details).
The second example of vested interests with
subjective evaluations is where the principal has
2217
an incentive not to implement the (ex ante) efficient
contract by reneging on a promised payment to save
costs. Thus, even though an agent exerts effort and
performs well, the supervisor claims otherwise to
keep costs down. This can arise either by outright
lying or perhaps by manipulating whatever
measures are available. A relevant example here is
the movie industry, where actors are sometimes
paid on the ‘net profits’ of a film. As a result, there
have been numerous court cases regarding firms
using creative transfer pricing arrangements to
reduce profits for very successful movies. See
Cheatham et al. ( ) for more
details on this. Such incentives to renege are likely
even worse when there are no objective measures of
performance. Because the desire to renege is greater
when discretionary incentive payments are higher, it
follows that the only credible contracts often
involve few incentives (Clive Bull considers a role
for repeated interaction between the principal and
agent as a means of reducing incentives to renege.
While repeating the relationship can result in
sufficient incentives for complete honesty by the
principal, it remains the case that, if the
relationship’s value is not sufficiently great,
incentives must be muted to reduce incentives for
cheating).
It is incorrect to assume that the ability to
manipulate measures of performance always mute
incentives - sometimes it can result in incentive pay
being inefficiently high. Consider again two
occupations where agents are typically residual
claimants - taxi drivers and sharecroppers. At first
blush, it would seem odd that they have such
extreme incentives. Aren’t these as likely candi-
dates for trading off risk against incentives as any?
However, one characteristic of each of these occu-
pations is that they have opportunities for hiding
output, either by taking fares without using the
metre (in the case of cab drivers) or selling crops
privately (in the case of farmers). In both cases, the
only outcome that makes this incentive irrelevant is
to render them residual claimants, even if risk
considerations would suggest otherwise.
Another issue which can constrain efficient
incentives, yet which has received almost no
attention in the empirical literature, is where
2218
agents hold private information. Take a specific
instance - real estate agents. In Chicago, real estate
contracts take a simple form - agents make three per
cent of the sales price of the house. Assume that my
home is worth $500. This linear contract not only
offers only three per cent on the relevant margin for
improving the selling price of the house, but
predominantly rewards the agent for selling the
house for say $450. Yet anyone could sell the house
for $450 and it seems highly inefficient to reward in
this way. So why not renegotiate to something
better? An example of such an improvement
(subject to risk issues) would be to offer nothing on
the first $450, but to pay a piece rate of 30 per cent
on anything over $450. In this way, the agent has
more incentives on margin, yet breaks even relative
to the original contract if the house sells for its
original price.
One reason why such renegotiation does not
arise is that the agent may privately know the true
value of the home, while the owner believes it to be
worth $500 on average. Consider a homeowner who
offers the new contract above to the agent. It is clear
that the agent rejects the new contract if it is truly
worth less than expected, and accepts it if worth
more. But this implies that the agent earns
information rents on average. As a result, on aver-
age the homeowner loses money from the renego-
tiation unless effort increases enough. This option
available to the agent limits the ability of contracts
to attain efficiency. Instead, in the usual monopoly
fashion, the homeowner would offer a contract to
trade off the efficiency gains of increased effort
with infra-marginal losses of the type described
above, resulting in lower-powered incentives (There
is a large mechanism design literature on this topic
that has largely been ignored by the empirical
literature on incentives; see Laffont and Tirole , for
example. This is surely partly because of the
empirical conundrum as to why mechanisms are so
rare in reality).
Much of the recent literature has been focused
on how incentive contracts can cause adverse
behavioural responses. Another possible mechanism
for such responses is where intrinsic motivation can
be crowded out by the use of incentive contracts.
The premise of this literature has been
Contracting in Firms
that in many occupations agents enjoy carrying out
the activity or care about the outcomes of their
actions. As a result, they will exert effort beyond
that which they can get away with even in the
absence of incentive contracts. This, in itself, is not
enough to limit incentive contracting. However,
there is some psychological evidence that agents
enjoy their jobs less when incentive contracting is
used. In effect, they feel that they are only doing it
‘for the money’ and hence lose interest. A
commonly cited example is the willingness of
people to donate blood, where the warm feeling
from donating declines when payments are made. In
some instances, this can imply that incentive
contracting can reduce effort if these crowding out
effects are strong enough. As a result, it can be
optimal to provide no incentives even when effort is
one-dimensional. This area of research, whose
empirical testing has largely been restricted to the
laboratory, is still in its early stages and is likely to
see much refining over the coming years. See Frey
and Jegen ( ) for a
survey.
Another likely fruitful area of future research
concerns non-monetary ways of motivating
workers. This literature largely began as an exercise
in how workers could be motivated to internalize
the benefits of others, yet has almost exclusively
become an exercise in how to motivate through
monetary contracting. Yet it is clear that there are a
myriad of means of motivating workers - sense of
achievement, ‘doing good’, status, and so on - that
firms tap into. How such mechanisms operate, and
the way in which they interact with monetary
contracts, remain an unstudied topic of research,
though see Besley and Ghatak ( ), for some
theoretical work on
this issue.
It is worthwhile to note a caveat before con-
cluding. The discussion above concerns the absence
of observed incentive contracts. Yet workers often
have unobserved carrots and sticks that can
motivate them. For instance, many workers exert
effort in the hope of attaining a promotion (Lazear
and Rosen ), or a better job offer (Holmstrom ).
Many ofthese mechanisms for inducing desired
behaviour are dynamic, where good performance
today results
Contradiction
in a greater likelihood of promotion, or better job
offers in future. Such incentives are clearly impor-
tant for workers. However, it remains the case that
explicit incentive payments remain limited even in
those cases where the above types of career concern
are negligible (For example, it is well known that
promotion prospects become very limited for
workers who remain in a job grade for a long
period. Yet explicit incentives are no more common
for those workers than for any other). The
interaction of unobserved (typically career)
incentives with the more exphcit set of piece rates
and bonuses that have been considered above is
surely of first-order importance to firms, though it
remains surprisingly unexplored in the literature
(see Baker et al. , for an exception).
To conclude, perhaps the central foundation of
modem economics is the idea that appropriate
prices guide behaviour in efficient ways. Despite
this, one of the defining characteristics of the
employment relationship in many firms is the
absence of the kind of exphcit prices whereby
wages depend in a clear way on observed outcomes.
The early incarnations of agency theory were
concerned with designing prices in a way that could
serve to fully internalize the effects of agents’
actions on the welfare of their employers. Yet this
initial optimism has now been tempered with a
somewhat more nuanced view that shows trade-offs
that will ultimately help in defining more precisely
the nature of labour market relationships.
Bibliography
Baker, G., R. Gibbons, and K.J. Murphy. 1994. Subjective
performance measures in optimal incentive contracts.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 109: 1125-1156.
Bertrand, M., and S. Mullainathan. 2001. Are CEOs
rewarded for luck? The ones without principals are.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 116: 901-932.
Besley, T., and M. Ghatak. 2005. Competition and incentives
with motivated agents. A m e r i c a n E c o n o m i c
R e v i e w 95: 616-636.
Bull, C. 1987. The existence of self-enforcing wage contracts.
Q u a r t e r l y J o u r n a l o f E c o n o m i c s 102: 147-
159.
Cheatham, C., D. Davis, and L. Cheatham. 1996. Hollywood
profits: Gone with the wind? C P A J o u r n a l 12: 32-34.
2219
Frey, B., and R. Jegen. 2001. Motivation crowding theory: A
survey of empirical evidence. J o u r n a l of
E c o n o m i c S u r v e y s 15: 589-611.
Gaynor, M., and M. Pauly. 1990. Compensation and pro-
ductive efficiency in partnerships. Evidence from medical
group practice. J o u r n a l o f P o l i t i c a l E c o n o m y
98: 544-574.
Gibbons, R. 1996. Incentives and careers in organizations. In
Advances in economics and
econometrics: Theory and applications,
ed. D. Kreps and K. Wallis. Cambridge: Cambridge
University Press.
Holmstrom, B. 1979. Moral hazard and observability. B e l l
J o u r n a l o f E c o n o m i c s 10: 74—91.
Holmstrom, B. 1999. Managerial incentive problems: A
dynamic perspective. R e v i e w of Economic
S t u d i e s 66: 169-182.
Holmstrom, B., andP. Milgrom. 1992. Multi-task principal
agent analyses: Linear contracts, asset ownership and job
design. J o u r n a l o f L a w , E c o n o m i c s , a n d
O r g a n i z a t i o n 7: 24-52.
Jacob, B.A., and S.D. Levitt. 2003. Rotten apples: An
investigation of the prevalence and predictors of teacher
cheating. Q u a r t e r l y J o u r n a l o f E c o n o m i c s
118: 843-877.
Janakiraman, S.N., R.A. Lambert, and D.F. Larker. 1992. An
empirical investigation of the relative performance
evaluation hypothesis. J o u r n a l o f A c c o u n t i n g
R e s e a r c h 30: 53-69.
Laffont, J.-J., and J. Tirole. 1986. Using cost observation to
regulate firms. J o u r n a l o f P o l i t i c a l E c o n o m y
94: 614-641.
Lazear, E.P. 1987. Incentive contracts. In T h e N e w P a l -
g r a v e : A D i c t i o n a r y o f E c o n o m i c s , ed. J.
Eatwell, M. Milgate, and P. Newman, vol. 2. London:
Macmillan.
Lazear, E., and S. Rosen. 1981. Rank order tournaments as
optimal labor contracts. J o u r n a l o f P o l i t i c a l
E c o n o m y 89: 841-864.
Prendergast, C. 1999. The provision of incentives in firms.
Journal of Economic Literature 37: 7-63.
Prendergast, C. 2002. The tenuous trade-off between risk and
incentives? J o u r n a l o f P o l i t i c a l E c o n o m y
110: 1071-1102.
Prendergast, C. 2003. The limits ofbureaucratic efficiency.
Contradiction
Michael Dummett
The fundamental form of a contradiction is a pair of
propositions, ‘A’ and ‘Not A’, one the negation of
the other. If such an explicit contradiction is
2220
part of a body of propositions asserted by some
individual or group at a given time, it follows that
not all those propositions can be true: by thus
impairing the reliability of the proponent, the
occurrence of the contradiction throws doubt upon
the truth of all the other propositions asserted.
Far more frequent than an overt or explicit
contradiction is a hidden contradiction. A hidden
contradiction is contained in a body of propositions
when, by logically valid deductive reasoning, an
explicit contradiction can be derived from them.
Two quite different cases arise, according to the
stafris of the original propositions from which the
contradiction was derived. The first is that in which,
as before, those propositions were all asserted by an
individual or body of people. In this case, the hidden
contradiction is as fatal to the joint correctness of
those assertions as is the explicit one, although, of
course, considerable work may have had to be
expended in bringing it to light.
The second ease is that in which at least some of
the original propositions were not asserted, but
merely advanced as suppositions to be considered.
The formal conclusion now remains exactly the
same: given that the contradiction was validly
derived, not all the original propositions can be true.
The effect, of course, is very different. Suppose that
the original propositions - the premisses from which
the contradictory pair ‘A’ and ‘Not A' have been
derived - were four in number: ‘Bf, ‘B 2\ ‘CI’, and
‘C2’; and suppose that ‘BC and ‘B 2’ were asserted
outright, but that ‘Q ’ and ‘C 2’ were merely
advanced for consideration. The contradiction
shows that not all four can be true: without having
to withdraw anything that he asserted, the
proponent, still maintaining ‘Bi’ and ‘B2’, is now in
a position to assert 'If Ci, then not C2’ (or its
equivalent, ‘If C2, then not Cf).
For this reason, derivation of a contradiction can
be employed, not merely as a means of refuting the
assertions of another, but as a method of
demonstrating negative propositions: this is the
celebrated mode of argument reductio ad
Contradiction
absurdum. In conjunction with premisses, ‘BC and
' B 2 ‘ , say asserted outright, a proposition ‘C’ is
presented as a hypothesis: not as a conjecture or
supposition to be seriously entertained, but with an
eye to proving its negation ‘Not C’. From the
premisses ' B ^ and ‘ B 2 ’ and the hypothesis ' C \
two contradictory propositions 'A' and 'Not A' are
then derived: on the basis of the two premisses, the
hypothesis now being dropped, its negation ‘Not C’
can now be definitely asserted. Two special cases
fall under this general description. One is that under
which ‘A’ is actually one of the two premisses, say
‘B] If from premisses ‘ B 2 ’ and the hypothesis ‘C\
the conclusion 'Not B \ can be derived, one may
assert 'Not C ' on the strength of the two premisses.
The other special case is that in which 'A' is the
hypothesis ‘C\ itself. If from premisses ‘ B i ’ and
' B 2 ‘ and the hypothesis ‘C\ the negation 'Not C’ of
the hypothesis can be derived, ‘Not C ' may be
asserted outright on the strength of the two
premisses.
As is well known, reductio ad absurdum argu-
ments are frequent in mathematics. One of the
simplest, as well as historically most important, is
the proof that 2 has no rational square root. We
may take as the premisses: (Bi) any rational number
may be represented as a fraction whose numerator
and denominator have no common factor; and (B2)
the square of an odd number is odd. As the
hypothesis we may take: ( C ) 2 has a rational
square root. From (BO and ( C ) it follows that, for
some integers m and n, m and n have no common
factor and (m/n)2 2; hence
m2 = 2rf. Hence, by the definition of ‘even’, m2 is
even, and from (B2) it can be inferred that m is not
odd, and hence is even. Thus m = 2k for some k,
and we have: m2 = 4Ir - In2, and so 2k2 = n2.
Applying (B2) once more, wc infer that n is also
even. Since m and n arc both even, they have 2 as a
common factor, which contradicts the earlier
stipulation that they have no common factor. By
arriving at this contradiction, wc have achieved a
demonstration that 2 has no rational square root,
that is, a proof of the negation of our hypothesis
(C).
Contradiction
The derivation of a hidden contradiction from
premisses all definitely asserted may also be illus-
trated from mathematics. The most celebrated
examples are the set-theoretic paradoxes which
provoked the first ‘crisis in the foundations’ of
mathematics in the early years of this century. Some
of these, such as Burali-Forti’s paradox and
Cantor’s paradox, demand some technical
background; but Russell’s paradox has the advan-
tage of being statable in very simple terms. The
propositions leading to this contradiction appear at
first sight entirely harmless. They are: (1) to every
property there corresponds a class, the class of
things having that property; (2) the class of things
having a given property F contains as members all
and only those things that have the property F.
Russell’s contradiction arises from considering the
property G of being a class that does not contain
itself as a member. By assumption (1), there exists a
class, which we may call W, of things having the
property G. We now ask whether W contains itself
as a member. It may then be argued by reductio ad
absurdum that it does not: we have first to suppose,
as a hypothesis, that it does. By assumption (2), it
therefore has property G. But properly G is the
properly of being a class that does not contain itself:
it follows, contrary to the hypothesis, that W does
not contain itself. We may thus discard the hypothe-
sis, and assert outright that W does not contain
itself. Applying assumption (2) once more, we may
then conclude that W lacks property G. It therefore
is either not a class or does contain itself as a
member; since it is by definition a class, it contains
itself as a member. This now contradicts our earlier
conclusion that it does not contain itself as a
member: the two intuitively reasonable assumptions
(1) and (2) have led to a contradiction.
Russell’s contradiction is so simple to state and
so easy to derive that those who are unaware of the
importance in mathematics of the notion of a class
or set are liable to think it no more than a trivial
verbal puzzle. It is very far from trivial. It over-
threw the life’s work of the great logician Gottlob
Frege, namely to provide unquestionably firm
2221
foundations for the theories of natural numbers and
of real and complex numbers, just when he believed
that he had accomplished it; and it cast the
foundations of mathematics into confusion for some
years. By showing that assumptions (1) and (2)
cannot both be consistently made, it challenged
mathematicians and logicians to find weaker
assumptions that would be consistent: for which
properties F do assumptions (1) and (2 ) hold good?
This question proved exceedingly difficult to
answer: hence the ‘crisis in the foundations’.
Why is the discovery of a hidden contradiction
so devastating? One answer is that, if a pair of
contradictory propositions are both consequences of
a given set of assumptions, then every proposition is
a consequence of them. This holds good in virtue of
the logical law, known to the medievals as ex fa Iso
quodlibet, that, from a pair ofpremisses ‘A’ and ‘Not
A’, any proposition ‘B’ may be inferred. This law
appears at first glance both useless and implausible;
a natural first reaction is therefore to suggest
repudiating it. That, however, is not easily done,
since it is a consequence of two other, seemingly
inescapable, laws concerning the logical constant
‘or’. The first is the law now usually known as ‘or’-
introduction or disjunction-introduction, that, for
any propositions ‘A’ and ‘B’, ‘A or B’ follows from
‘A’. The second is that known in the traditional
logic as modus tollendo ponens, that ‘B’ follows
from the premisses ‘A or B’ and ‘Not A’. It is quite
obvious that, from the premisses ‘A’ and ‘Not A’ of
the ex falso quodlibet, its conclusion ‘B’ can be
reached by first applying ‘or’-introduction and then
modus tollendo ponens: but it is very hard to see
how we could possibly reject either of the two latter
laws, for they appear absolutely constitutive of the
meaning of the connective ‘or’. The best possible
way to establish the truth of a disjunction - a
statement of the form ‘A or B’ - is to establish the
truth of one or other of the disjuncts ‘A’ and ‘B ’: if
the truth of ‘A’ does not guarantee the truth of ‘A or
B ’, what does ‘or’ mean? Conversely, the truth of a
disjunction appears to demand that at least one of
the disjuncts be true: so, if ‘A or B’ is