Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.
http://www.jstor.org
RICHARDT. ELYLECTURE
By KENNETH J. ARROW*
strategy choices available to individual the formaltheory,at least, no one. They are
actors. determined on (not by) social institutions
Social and historical determinismis not knownas markets,which equate supplyand
as populara viewpointas it used to be, and demand.
an individualisticperspective is a guard I pass over manyother questionsof social
against such theories. Whether in Marxist roles in this description.Tastes may be so-
or other forms, such theories relied heavily ciallycaused;expectationsare influencedby
on disembodied actors such as classes or others; firms are organizations,not individ-
national spirits, rather than on the actual uals. But I want to argue the case most
persons. The newer trends in historical favorableto individualism.
analysis are more concerned with contin- The failure to give an individualisticex-
gency than with determinism,contingencies planation of price formationhas proved to
arising from the flexibilityand freedom of be surprisinglyhard to cure, though there
individualhumandecision-making. are some ingeniousstories. They all depend
In this quick presentation of individual- on the closest realizationof individualismin
ism, I have avoided the term, "rational economic theory, which is to be found in
choice." The individualistviewpoint is in game theory, and indeed Buchanan (1989
principlecompatiblewith boundedrational- p. 82) asserts that, "game theory offers
ity, with violationsof the rationalityaxioms, the appropriate mathematical framework
and with the biases in judgmentcharacteris- that facilitatesan abstractunderstandingof
tic of humanbeings. The additionalstep to economics."
rationalchoice is, of course, of the greatest
practical importance to theory formation, IV. GameTheory
but it is not in principle necessary for the
individualistviewpoint. The currentformulationof methodologi-
cal individualismis game theory.This is not
III. CompetitiveEquilibriumand a particulartheoryof social interaction,but
Methodological Individualism rather a languagefor such a theory. It may
not be surprisingthat its authors are an
I have emphasizedthe desirabilityof an Austrianeconomist and a Hungarianmath-
individualist perspective. I now want to ematician-economistat least closely con-
argue that economic theories require nected culturallyto Austria.
social elements as well even under the In a game, each agent chooses one among
strictest acceptance of standard economic a set of strategiesavailableto him or her. In
assumptions. the usual formulations,the set of available
The prototypicaleconomicmodel, despite strategies is fixed, independent of the
battering, is general competitive equilib- choices of others. The outcome or payoffof
rium. Individualsand firms take prices as the game for each playeris a functionof the
given. Individualschoose consumptionde- strategies of all the players. Hence, all the
mands and offers of labor and other assets, interactionsamongplayersare embodiedin
subject to a condition that receipts cover the payoff functions.The choice of actions
expenditures.Firms choose inputs and out- is totally individualistic.For the game for-
puts subject to the condition that the out- mulation to be meaningful, the outcomes
puts be producible given the inputs. How defined by the payoff functions must be
they make these choices depends on many possible;for example,demandshould never
factors:tastes, attitudes towardrisk, expec- exceed supplyfor any commodity.
tations of the future. But, it is held, these It has proved difficultto define competi-
factors are individual. tive equilibriumas the outcome of a non-
Even if we accept this entire story, there cooperative game. It is not difficult to
is still one element not individual:namely, constructa game whose equilibriumpoint is
the prices faced by the firms and individu- a competitive equilibrium;indeed, Gerard
als. What individualhas chosen prices? In Debreu and I did somethingclose to that in
VOL. 84 NO. 2 RICHARD T. ELYLECTURE 5
our proof of existence of competitive equi- sions of oligopoly theory. The optimizing
librium (Arrow and Debreu, 1954), and an strategy for each firm depends on the
unpublished version uses a game strictu strategies chosen by the other firms. But
sensu. But the game was purely a mathe- how does a firm know what the other firms
matical construct; when nonequilibrium are doing? Indeed, it is of the essence that
strategies were played, the outcomes were they are chosen simultaneously. Each firm
not feasible. Games whose outcomes are has an expectation or conjecture (to use
always feasible and whose equilibria are Ragnar Frisch's term) about the others' ac-
Walrasian have been constructed, but their tions. Why should they be consistent? This
relation to real-life phenomena has varying problem is again actively on the agenda of
degrees of weakness (see e.g., David game theorists. It is a problem of knowl-
Schmeidler, 1980). edge.
The most realistic-sounding noncoopera- Even if we do not question the concept of
tive-game-theory model of competitive equi- an equilibrium point, there is another
librium I know of is due to Douglas Gale difficulty, first raised by Joseph Bertrand
(1986a,b), though in its present form it ap- (1883) in his very belated review of
plies only to a pure exchange economy. Pairs Cournot's book. What after all are the
of individuals meet at random and bargain strategies of the two players? Cournot as-
(bargaining is itself formulated as a sub- sumed that firms choose quantities. Is it not
game). With the endowments obtained as a equally reasonable that firms choose prices?
result of the bargaining, they again meet in If we assume rational consumers, they will
pairs chosen at random, and so forth. The all buy from the sellers with the lowest
bargaining at any time is, of course, affected price. The outcome of this game is entirely
by the (rational) expectations of the results different from Cournot's; under simple as-
of future bargains. These expectations are, sumptions it is in fact the competitive equi-
in effect, prices. Gale's model is a possible librium even if there are only two firms.
formalization of the Austrian viewpoint. What this example shows is that the rules
However, prices never appear as objective of the game are social. The theory of games
phenomena; they are only subjective, that gets its name and much of its force from an
is, expectations held in the agents' minds. analogy with social games. But these have
For a different perspective on the individ- definite rules which are constructed, indeed,
ualist nature of game theory, consider an- by a partly social process. Who sets rules for
other economic model, oligopoly. This prob- real-life games?
lem could be considered the very origin of More generally, individual behavior is al-
noncooperative game-theoretic analysis in ways mediated by social relations. These are
economics (there are, I believe, even earlier as much a part of the description of reality
examples of noncooperative game theory in as is individual behavior.
the analysis of social games); the analysis, in
completely modern form, goes back to V. Externalities
Augustine Cournot (1838). We recall the
formulation. A set of firms are producing Even within standard economic theory, I
the same good. Each chooses a quantity; the have taken the case most favorable to indi-
price is that which clears the market for the vidualistic analysis. Economics has for about
total quantity supplied by all firms. Any one a century (since Alfred Marshall [1920; first
firm then has a profit equal to the revenue edition, 1890]) recognized the importance of
from its chosen output less the cost of pro- what we have come to call externalities.
ducing that output. Each firm maximizes its Roughly speaking, these are social interac-
profit given the outputs of the other firms. tions not mediated through the market. The
This seems straightforward enough, and analysis is currently applied especially to
yet there are some difficulties. One is a environmental issues; air and water pollu-
recurrent question ever since game theory tion, global warming, toxic wastes, but also
was formalized and even earlier in discus- congestion.
6 AEA PAPERS AND PROCEEDINGS MAY1994
This ground is too familiar for discussion, of the goods to corner the wisdom of the
and much study proceeds along individualis- ancients and the accumulated experience of
tic values, though calling for some kind of the race" (p. 185).
collective action. One technical remark is Of course, other forms of knowledge than
useful as a preliminary to the following dis- purely technological are essential to the so-
cussion of knowledge and information as cial system in general and the economy in
social as well as individual characteristics. particular. In many ways, these forms are
For many, though by no means all, external- stressed by Hayek. In his famous 1945 paper
ities, the effect depends on the total of on knowledge in society (Hayek, 1948 Ch. 4),
many individual contributions. For example, he emphasizes the dispersed and tacit nature
individual combustion activities contribute of knowledge and argues that the economy
to a stock of sulfur dioxide in the atmo- solves the problem of allocating resources
sphere, the effect of which on individuals under these conditions. His motive was to
depends only on the total, not on the indi- rebut the possibility of a centrally planned
vidual contributions. Let us call such exter- society, one in which the relevant knowledge
nalities funded. This does not change the is concentrated in one place. But in the
underlying logic of externalities but does course of his argument, he has put obstacles
simplify the analysis. in the way of a better understanding of the
generation of knowledge.
VI. Social Knowledge He does acknowledge (pp. 79-80) that
scientific knowledge differs from the tacit
Among the externalities that Marshall was knowledge held by individuals and not eas-
concerned with, a prime example was infor- ily transmitted to others. In scientific knowl-
mation, especially technical knowledge. The edge, expert opinion may indeed count for
caustic dissenter, Thorstein Veblen (1919 more than the knowledge dispersed
pp. 180-230; originally appearing in 1908), throughout the economy. However, Hayek
in reviewing John Bates Clark's textbook, tends to minimize the role of scientific
identified socially held technical knowledge knowledge and does not really discuss tech-
as a main determinant of economic activity nological knowledge at all, a good deal of
in every economy. In primitive societies, to which is transmittable to others. Let me call
quote, "[t]he 'capital' possessed by such a scientific knowledge and the more transmit-
community-as, e.g., a band of California table parts of technological knowledge to-
'Digger' Indians-was a negligible quantity, gether "reproducible knowledge." (Of
more valuable to a collector of curios than course, a fuller account would have to rec-
to any one else, and the loss of which to the ognize varying degrees of reproducibility.)
'Digger' squaws would mean very little. In many ways, the distinction between re-
What was of 'vital concern' to them, indeed, producible and tacit knowledge is parallel
what the life of the group depended on to that between evolutionary and conscious
absolutely, was the accumulated wisdom of changes in social organization that I re-
the squaws, the technology of their eco- ferred to earlier.
nomic situation." (Earlier, he had said, Suppose we try to pursue the individualist
''women seem in the early stages to have viewpoint about knowledge, that knowledge
been the most consequential factor instead is held by individuals. What is striking is
of the man who works by himself.") Veblen that neither Hayek nor his socialist oppo-
did not deny that the capital goods are nents were concerned with changes in
more significant in modern economies, but, knowledge. The stock of knowledge is given.
"[t]he commonplace knowledge of ways and Hayek takes this knowledge to be dis-
means, the accumulated experience of tributed among economic agents and at-
mankind, is still transmitted in and by the tributes to the socialists the view that it is or
body of the community at large; but, for can be held centrally. (This is in fact a
practical purposes, the advanced 'state of caricature of the position of the market
the industrial arts' has enabled the owners socialists, from Enrico Barone through
VOL. 84 NO. 2 RICHARD T. ELYLECTURE 7
Oskar Lange and Abba Lerner.) More mod- revelation of quality information is much
ern economics is concerned with the acqui- more transparent.
sition of new knowledge, and the test of the With this perspective, it may be easier to
individualist viewpoint has to be an attempt think of information breeding information
to understand how and why new knowledge and to suppress the role of individuals. This
is held by individuals to see if the account is is very much like the role of genetic infor-
adequate. mation. As the late 19th century writer,
New knowledge is acquired in two dif- Samuel Butler, said: "A chicken is an egg's
ferent ways: (1) acquisition from observing way of making an egg."
nature (whether by research or by less for- Indeed, models in which socially available
mal procedures); and (2) learning from other information is a variable are not uncom-
individuals, which in turn can be subdivided mon, especially in the literature on techno-
into (a) intended learning (communication, logical innovation and economic growth.
education), and (b) inferring the knowledge One class includes those models which em-
of others by observing their behavior. phasize the diffusion of technology (for
Let us consider the second mode, learn- well-known examples, see Zvi Griliches
ing from others, first. There are clear empir- [1957], Everett Rogers [1962], and Mansfield
ical problems with maintaining the individu- [1968 Part IV]). The tradition in turn de-
alist orientation. Technical and other rives from earlier work in anthropology and
knowledge exists in social form: books or sociology, in which traits (in particular,
universities (professors in many courses are technologies) in possession of one group
largely interchangeable). These are exam- spread to another at rates proportional to
ples of intended communication, though not the contact between the groups. These in
precisely individualistic as to recipients. In- turn are related to theories of the spread of
ference from the behavior of others is even epidemics.
less individualistic. The existence of new Modern theories of economic growth in-
products is itself a transmission of informa- corporate many of the same elements. Sub-
tion, an externality; it shows that certain tle observation is not needed to see that we
ways of progress are possible, even if the have had great changes in our technological
principles are not made public. When the knowledge. The need for economic analysis
atomic bomb was new, there was great con- is to explain steady or even accelerating
cern about secrecy; but wise physicists ob- rates of growth in advanced economies.
served that the most important information Neoclassical economics without increased
was that the achievement of nuclear fission knowledge should lead to diminishing rates
was possible, and that could hardly be held of growth, even apart from Malthusian con-
secret. In practice, reverse engineering and siderations and exhaustible resources.
the diffusion of basic unpatentable knowl- While dissemination of existing informa-
edge imply a much more rapid transmission tion can certainly account for some gains in
of technical information than merely the productivity, it is clearly necessary for sus-
existence of new products (see Edwin tained growth to have information new to
Mansfield [1985] who estimates that new the entire system, not merely learned from
technology becomes available to other firms others. Where does this new knowledge
within one or two years). come from? The literature has two view-
It is curious, by the way, that so much of points, although they are not really exclu-
the literature of the last 20 years has con- sive. One is that the growth in knowledge is
centrated on using prices as a source of exogenous to the economy (Robert Solow,
information (e.g., Sanford Grossman, 1976; 1956), the other that it is endogenous, a
Roy Radner, 1979; Grossman and Joseph E. result of economic processes (the "new
Stiglitz, 1980). Interesting as this work is, growth economics").
the assumptions needed to imply that prices The first is equivalent to saying that
reveal much of the private information of knowledge produces knowledge. The
others are unrealistically strong, while the second emphasizes that knowledge is pro-
8 AEA PAPERS AND PROCEEDINGS MAY1994