Você está na página 1de 11

Limited Knowledge and Economic

Analysis
By KENNETH J. ARROW*

The content of presidential addresses to It is now more fashionable than it used


this Association provides a fine example of to be for statisticians to be told to take a
a random variable with a high variance. It good look at their data before fitting
might even be a good subject for econo- models. Taking presidential addresses as
metric analysis; the variation might be our data, we find most frequently a review
explained in terms of the economic condi- of the speaker's main research concerns
tions of the moment, previous intellectual but also expressions of methodological or
investments, or even, for boldly interdis- ethical concerns, historical surveys of vary-
ciplinary analysis, the psychological states ing degrees of erudition and humor, and,
or class origins of the speakers or the least frequently, new points of view on sig-
audience. But no doubt captious theorists nificant problems of economics.
like myself will object that the endogenous I am taking a somewhat different tack
variable is not cardinally measurable and today; it will be an expression of discon-
probably not even ordinally measurable; tents and expectations. As I shall try to
tougher-minded econometricians will worry argue, the uncertainties about economics
about collinearity in the predetermined are rooted in our need for a better under-
variables; and practical-minded policy standing of the economics of uncertainty;
analysts will see no discernible effect on our lack of economic knowledge is, in good
the gross national product, the price level, part, our difficulty in modelling the ig-
or the balance of payments through effects norance of the economic agent.
on either fiscal policy or the stock of Critical aspects of this need for reorien-
money. The last group, the policy-oriented, tation of theory have been recognized by
are perhaps the least accurate; at least many scholars in the last quarter-century
according to Keynes, the effect of ideas on and particularly in the last decade. I view
policy is dominant, though the lag may be my remarks today as a summary and per-
as variable as and a good deal longer than spective on a widely shared development
that of the stock of money on money gross of thinking.
national product. The starting point of discussion must
* Presidential address delivered at the eighty-sixth still be the much-abused neoclassical
meeting of the American Economic Association, New theory. No really cohesive alternative
York, New York, December 29, 1973. The views ex-
pressed above are a personal synthesis of a widespread which aspires to the same level of com-
viewpoint which has appeared in many different con- pleteness exists. The neoclassical model is
texts in the published work of and in some cases personal founded on two concepts, which are con-
discussions with G. Akerlof, A. Alchian, G. Calabresi,
R. H, Coase, H. Demsetz, P. A. Diamond, J. Green, siderably different in nature. One is the
E, H. Hahn, A. G. Hart, E. A. von Hayek, J, Hirsh- notion of the individual economic agent,
leifer, L. Hurwicz, C. Kaysen, E, H. Knight, A. Leijon-
hufvud, E. Machlup, J. Marschak, T. Marschak, J, whose behavior is governed by a criterion
Mirrlees, M. Nerlove, R. Radner, M. Rothschild, T. of optimization under constraints which
Schelling, H, A, Simon, A. M, Spence, and G. J. Stigier. are partly peculiar to the agent, such as
Eor some additional aspects of the role of information
problems in economics, see my lecture. Information and production functions, and partly terms of
Economic Behavior, Stockholm 1974. trade with the economic svstem as a whole.
THE AMERICAN ECONOMIC REVIEW MARCH 1974

The other is the market; here, the aggre- logical unemployment to the perpetual
gate of individual decisions is acknowl- surprise of the layman fit in well with the
edged, and the terms of trade adjusted un- neoclassical formulation but have no ready
til the decisions of the individuals are explanation in alternate models.
mutually consistent in the aggregate, i.e., Of course, the implications of neoclassi-
supply equals demand. cal theory have also been conspicuously
The neoclassical theory, especially in its falsified in important ways. Most notably,
competitive form, can be and has been the recurrent periods of unemployment
given a rich formal development. Paren- which have characterized the history of
thetically, one cause for the persistence of capitalism are scarcely compatible with a
neoclassical theory in the face of its long neoclassical model of market equilibrium.
line of critics is precisely that for some A post-Keynesian world in which unem-
reason of mathematical structure, the neo- ployment is avoided or kept at tolerable
classical theory is highly manipulable and levels by recurrent alterations in fiscal or
flexible; when faced with a specific issue, monetary policy is no more explicable by
it can yield meaningful implications rela- neoclassical axioms, though the falsifica-
tively easily. Although I intend to air tion is not as conspicuous.
complaints and desires for change today, Inequality in economic development
I must express my unabashed admiration among countries and among groups and
for the accomplishments of the neoclassical regions within a country provides a second,
viewpoint. In its most formal statement, somewhat complicated difficulty for neo-
we simply use for analysis the equilibrium classical theory. A purely neoclassical an-
conditions of the individual agent and of swer would explain differences in per
the market, without inquiry as to how they capita income by differences in physical
come to hold. Yet even these statements and human assets per capita. This of
turn out to yield revealing insights in the course raises the further question, how
workings of resotirce allocation. Why have this came to be, a question which would
medical costs risen so rapidly relative to require a fully dynamic model to answer;
other prices since 1967? The upward shift but I think the more compelling problem
• in demand due to Medicare and Medicaid is that the differences in income seem
with a price-inelastic supply of physicians much too vast to be explained by factor
and hospitals provides a simple straight- differences. Indeed, in the presence of in-
forward answer; I cannot really imagine ternational trade and especially interna-
how a Marxian or a neo-Ricardian would tional capital movements, wage differences
even approach the question, though I sup- should be very strongly reduced compared
pose they might dismiss it as unimportant. with what would occur in autarchic states
The explanation of environmental prob- where domestic capital is the limiting fac-
lems as due to the nonexistence of markets tor. Hence, we come immediately to the
is similarly an insight of purely neoclassical explanation that there are differences in
origin. The now-demonstrated fact that the production possibility sets of the
flexible exchange rates are a feasible way different countries. This conclusion is a
of conducting international finance is a legitimate and important use of neoclassi-
triumph of theoretical insights over prac- cal analysis; but obviously it raises new
tical men's convictions. More broadly, the questions, to which we will return.
shifts in long-run resource allocation as I pass by the whole tangle of questions
motivated by returns and, in particular, relating to the holding of money and the
the absence of a secular trend in techno- general level of prices. In its pure form.
VOL. 64 NO. 1 ARROW: LIMITED KNOWLEDGE

neoclassical theory is a theory of relative technological conditions as data, possibly


prices. Monetary theories vaguely related varying over time due to exogenous changes
to it in spirit can be grafted on to it, but in scientific knowledge. But here we are
none have succeeded in achieving a asserting that two contemporaries have
genuine synthesis. different access to productive knowledge.
The two failures of the neoclassical ex- Clearly, we are saying something about the
planatory mechanism reflect on its founda- conditions of transmission of knowledge
tions in quite different ways. The existence across national boundaries, and of course
of unemployment is clearly a direct con- the same questions arise among firms or
tradiction to the notion of the smoothly workers within a single economy. The con-
clearing market. One must of course be straints upon the firm's optimization begin
aware that the official measure of unem- to seem more like variables to be explained
ployment is by no means a simple inequal- than like constants exogenously given.
ity between supply and demand; it ag- Let me look now at the two basic con-
gregates a whole range of distinct markets, cepts from the inside, from the point of
it does not separate out voluntary and in- view of our direct perceptions which moti-
voluntary unemployment according to the vate the modelling. The two are far from
tests of economic theory, and it does not parallel. The optimization by individual
take account of unfilled jobs. I do not sub- agents has a sense of concreteness about it,
scribe, I hasten to say, to the sometimes for all the sophisticated mathematical
expressed view that all unemployment is ability with which we theorists endow the
essentially voluntary, an unwillingness to agents. They behave in ways whose logic
search or whatnot; indeed, the oflicial we understand. They seek to achieve goals
measure may underestimate the degree of which are reasonable to postulate, and we
disequilibrium in the labor market, par- can specify constraints which dearly are
ticularly with regard both to underutiliza- real. It can be and has been correctly ob-
tion of advanced skills and discouraged job jected that our models are too simple; we
seekers. With all these qualifications, it is ignore other arguments in the utility func-
clear that statistical unemployment does tion, power, status, social approval, or
correspond to a disequilibrium as that whatnot that also motivate individuals,
term is used in the basic neoclassical and we ignore some constraints, capacity
model; there are two individuals, identical for calculation and social controls. But the
in productive capacity and both willing model is comprehensible, and the motives
to work at a given wage, but one is working and constraints we deal with are real and
at that wage and the other is not. important.
Differential levels of economic develop- The market, on the other hand, is a
ment, on the other hand, point to a difli- much more ethereal construct. Who ex-
culty with the other fundamental concept; actly is it that is achieving the balancing of
the conditions of optimization. If countries supply and demand? Where in fact is the
differ in their production possibility sets, information on bids and offers needed for
then firms, occupying similar economic equilibration actually collected and stored?
positions, are facing different constraints Right from the beginning of neoclassical
on their optimization. This does not con- theory, the difficulty of explaining markets
tradict the fundamental assumption of in terms of individual self-seeking be-
optimizing behavior, but it does raise havior was perceived. Parenthetically, this
severe questions about its interpretation. is one example of the superiority of neo-
The simplest hypothesis is to take the classical analysis to its predecessors, de-
THE AMERICAN ECONOMIC REVIEW MARCH 1974

spite the current fashion for exalting integrate Keynesian viewpoints into stan-
Ricardo over his successors; Ricardo im- dard neoclassical theory, in terms of in-
plicitly equated supply and demand on all dividual motivation, yet this theory, with
his markets without ever realizing the its various modifications, has been a most
problematic nature of this process. Jevons serviceable tool of prediction and control.
felt obliged to enunciate explicitly a Law In fact, it is useful in domains where com-
of Indifference, enforced by arbitrage; but petitive theory fails and vice versa. Neither
this does not really meet the problem when theory is good, however, at predicting dy-
the market is out of equilibrium, for the namic processes, the short-run changes
arbitrage might well not be feasible. which are responses to disequilibria, and it
Menger, at least according to Hayek and is here that the pressure for a more satis-
Streissler, concentrated on individual trades factory model arises.
and ignored the market completely. It is Hold in abeyance for a moment our con-
Walras's auctioneer which has proved to siderations about the market. Let us return
have had the most enduring effect on sub- to the optimization problem of the in-
sequent theoretical development, and the dividual. One aspect on which we put a
stability theory whichflowsfrom that con- good deal of weight, particularly in our less
cept is still the subject of vigorous theo- formal discussions, is that a market system
retical development, though very little is informationally economical. That is, we
empirical application. What is envisioned tend to regard it as a virtue of the system
is a feedback mechanism in which errors that the individual agent need not know
in the price are successively corrected by very much. Specifically, he is supposed to
reference to the disequilibria they gener- know the motivation and production con-
ate. This view specifies and makes feasible ditions which define him, i.e., his utility
the operations of the market. But on one function and production possibiUty set,
hand the stability models are far from together with the prices of the commodi-
adequate representations even of the dy- ties he buys and sells. The economic sys-
namics of the neoclassical models and, tem, taken as a whole, has vastly more in it
what may be connected, the results are by than any one individual knows; it contains
no means necessarily favorable to the the utility functions and production pos-
stability of the adjustment process; and on sibilities of all individual agents. Indeed,
the other hand, the motivations for the the apparent modesty of the information
feedback to operate are obscure. needed is one of the most appealing aspects
Let me be clear on one methodologiGal of the neoclassical model, both in the
point. The fact that our intuitive under- descriptive sense that the individual's de-
standing, our verstehen as the German cision problems appear manageable for him
social methodologists call it, of the market and for the economist studying him, and
as an institution is not entirely satisfactory in the normative sense that the system
does not mean that we should not use the permits its members to spend their time
perfect market as a model, at least pending and effort at producing goods rather than
further development. Certainly, as Popper in unnecessary duplication of information.
and Friedman hold, the acceptability of a But clearly this simplification of the in-
theory is to be judged by its ability to dividual's decision making is made possible
predict and understand phenomena. The only because the markets have supplied
theory of the perfect market is in an in- the information economized on, in the form
teresting way complementary to Key- of prices. In equilibrium, at least, the sys-
nesian theory. We have never been able to tem as a whole gives the impression of
VOL. 64 NO. 1 ARROW: LIMITED KNOWLEDGE

great economy in the handling of informa- to our usual economic calculations an ap-
tion, presumably because transmission of propriate measure of the costs of informa-
prices is in some significant sense much tion gathering and transmission.
cheaper than transmission of the whole set But actually the comparisons between
of production possibilities and utility socialist and capitalist resource allocation
functions. It is this point which emerged systems have tended to overlook some of
in the great debate over the feasibility of the most obvious facts while examining
socialism begun by Ludwig von Mises's finer points closely. As we all know, both
attack and usually thought of as conclud- production and consumption decisions are
ing with the work of Oskar Lange and in fact made with reference to the future
Abba Lerner in the 193O's; though it should as well as to the present. A rational produc-
be added many of the essential points had tion plan includes very importantly deci-
already been made earlier by Vilfredo sions or at least plans about the future; and
Pareto and Enrico Barone. What was ar- similarly with consumption plans. Invest-
gued, in effect, was that a socialist system ment and savings are not only integral
could use the price system and therefore parts of our current decisions but in the
achieve whatever economies in informa- long run shape the possibilities for further
tion it does achieve; and if the equilibrium development. As we know, the formal neo-
conditions are written out they do give the classical model can be extended to deci-
appearance of relative simplicity. But sions over time by dating commodities and
what was left obscure is a more definite regarding the same commodity at different
measure of information and its costs, in dates as different commodities. All pre-
terms of which it would be possible to as- vious conclusions follow; allocative effi-
sert the superiority of the price system ciency, for example, is achieved with the
over a centralized alternative. Though I same appearance of informational efficiency.
feel that current work has brought about a But of course there is a slight problem
considerable clarification, we still have no with this reasoning. The information
definite measure. Indeed, in some respects, about future commodities needed includes
more recent developments have made the their prices. These prices must be those
answers less clear. Several writers, in both found on a suitable market, one in which
Western and socialist countries, have noted future supply and future demand are
that alternative decentralized schemes equated. Unfortunately, no such markets
exist where quantity messages rather than exist. Even the futures markets in certain
price messages are transmitted in the suc- commodities, limited in extent as they are,
cessive stages of approximation and that do not in fact lead to balancing all future
such schemes also have efficient equilibrium decisions. Rather they balance present
points. Indeed, with the development of commitments to the future; but it is un-
mathematical programming and high- derstood by all parties that when the fu-
speed computers, the centralized alterna- ture becomes the present, there will be a
tive no longer appears preposterous. After spot market on which the futures commit-
all, it would appear that one could mimic ments may be undone; and indeed those
the workings of a decentralized system by making no futures commitments at all can
an appropriately chosen centralized al- enter and know now that they will be able
gorithm. While there is more to the story to enter.
than these few remarks, they do make the Even as a graduate student, I was some-
point that if we are going to take informa- what surprised at the emphasis on static
tional economy seriously, we have to add allocative efficiency by market socialists.
THE AMERICAN ECONOMIC REVIEW MARCH 1974

when the nonexistence of markets for fu- do these come from? We may in the first
ture goods under capitalism seemed to me instance regard them as subjective. But
a much more obyious target. the economic agent observes his world and
However that may be, the nonexistence has the opportunity to learn from his ex-
of these markets must be faced. Now in perience, for there is a considerable degree
general equilibrium any part of the system of continuity. By Bayes' Theorem or per-
affects every other part in at least two haps psychological learning theory, the
different ways. Thus, we may ask two probabilities, say of future prices, will
questions about the nonexistence of futures gradually adjust so as to conform to the
goods markets: what are its implications facts. If indeed the economic world ex-
for the rest of the system and what are the hibited the same structure in some sense
reasons for its nonexistence. from period to period, and if everybody
The implication first of all is that the observed everything relevant, then the
information needed by the optimizer is not probabilities ascribed by different individ-
provided by an existing market. It will be uals to the same events might be expected
provided by a market which will exist in gradually to converge to the correct values
the future, but that is a bit too late to help and therefore be the same for all. In fact,
in decisions made today. Hence, the op- of course, the basic economic facts are
timizer must replace the market commit- changing, partly endogenously because of
ment to buy or sell at given terms by capital accumulation in its most general
expectations: expectations of prices and sense, partly exogenously with predictable
expectations of quantities to be bought or and unpredictable changes in technology
sold. But he cannot know the future. and tastes; equally if not more important,
Hence, unless he deludes himself, he must though, is the fact that the dispersion of
know that both sets of expectations may information which is so economical implies
be wrong. In short, the absence of the that different economic agents do not have
market implies that the optimizer faces a access to the same observations. Hence, it
world of uncertainty. is reasonable to infer that they will never
The exact modelling of behavior under come into agreement as to probabilities of
uncertainty is probably not crucial to the future prices.
subsequent discussion; let us use the A further implication is that the past in-
conventional expected-utility hypothesis. fluences the future. Jevons's well-known
When there is uncertainty, risk aversion slogan, "bygones are forever bygones,"
implies that steps will be taken to reduce ceases to be fully accurate. The past is rele-
risks. This partly affects decisions within vant because it contains information
the firm, such as the holding of inventories which changes the image of the future; the
and preference for flexible capital equip- probabilities which govern future actions
ment, and partly leads to new markets are modified by observations on the past.
which will shift risks to those most able It follows that present decisions with im-
and willing to bear them, particularly plications for the future are functions of
through the equity market. The rich de- past values of variables as well as present
velopment of inventory theory and port- values.
folio theory in the last twenty years or so This point of view has been exploited in
reflects growing understanding of these the econometric models which have used
matters. distributed lags in explaining investment
But when we speak of expected utilities, decisions. What still needs to be exploited
we need to have some probabilities. Where more, however, is that the inference to the
VOL. 64 NO. 1 ARROW: LIMITED KNOWLEDGE

future is necessarily uncertain, and the centive to produce such information. We


decisions made still exhibit risk aversion. have then an economic information in-
Expectations for the future are related dustry: data assembly and analysis, busi-
to quantities as well as prices. The im- ness journalism, economic forecasting, with
portance of quantity expectations has been a longer-run perspective business educa-
stressed in macroeconomic models, even in tion. Since information as a commodity
such pre-Keynesian concepts as the ac- does not satisfy all the neoclassical norms,
celeration principle, and most especially in it is not surprising that the government
relation to inventories. It sometimes is plays a large role in this process. Informa-
held that in a neoclassical world only prices tion-acquisition activities and information
matter; in the absence of prices, presum- markets now appear on the economic land-
ably they are replaced by price expecta- scape. Efiiciency in the operation of firms
tions. But that is not strictly true. Under ceases to be purely productive efficiency;
constant returns, at least, quantity in- it involves efficiency in prediction as well.
formation for the individual firm is needed I would conjecture that the incomplete
even when neoclassical assumptions are diffusion of information along the lines just
strictly fulfilled. Neoclassically founded in- sketched has a good deal to do with the
vestment theories usually predict capital- operations of the securities markets and
output ratios or capital-labor ratios; they the decisions on corporate financing. The
still need output forecasts explicitly or im- predominant role of internal financing and
plicitly. This gives considerable, perhaps indeed the whole special importance of the
major weight to past quantity information managerial factor in corporate decision
in predicting the future and therefore in making are clearly connected with differ-
guiding current investment decisions. It is ential access to information about the
perhaps along these lines that Keynesian firm.
theory, with its overwhelming emphasis on You may have forgotten by now, but I
quantity changes as equilibrating vari- earlier promised to consider not only the
ables, can be founded firmly on individual implications of but also the causes for the
optimizing behavior. absence of markets for future goods. One
I have referred to the fact that informa- might wonder why one should explain the
tion is dispersed throughout the economy absence of a phenomenon. Sherlock Holmes
but have not suggested how. In the pure once maintained to the dimwitted local
neoclassical model, each agent knows only police inspector so typical of English de-
his own production possibilities and his tective stories that the significant question
tastes, together with market information in the case at hand was the dog's barking
on the rest of the economy. In the world I at night. "But," said the inspector, "the
have just sketched, however, any variables dog didn't bark." "That," said Holmes,
which improve his ability to predict the "is what is significant." So too is the ab-
future have a very meaningful economic sence of these markets significant for a full
value to him. He will seek to acquire addi- neoclassical theory. A truncated theory of
tional information. Such information is temporary equilibrium in which markets
presumably costly; that is the basis for for future goods are replaced by some form
such great emphasis on the value of in- of expectations, themselves functions of
formational economy. But there is clearly current prices and quantities, has indeed
a great incentive to acquire information of been developed, though its empirical con-
predictive value, and, as neoclassical tent is necessarily meager if the formation
theory would predict, there will be an in- of expectations is left unanalyzed. But the
THE AMERICAN ECONOMIC REVIEW MARCH 1974

true neoclassical spirit is being denied in possible and that it is even socially desir-
such a model. Although we are not usually able to set limits on the penalties for
explicit about it, we really postulate that failure. However, when the exchange of
when a market could be created, it will be. values for values is simultaneous or nearly
I sometimes think that welfare economics so, the contracts may almost be self-
ought to be considered an empirical dis- enforcing. If a good has been sold and not
cipline. Implicitly, if an opportunity for a paid for, it can be recovered; if there is a
Pareto improvement exists, then there will continuing relation of buyer and seller, a
be an effort to achieve it though some so- failure to settle bills can be met by refusal
cial device or another. In our theories and to make further deliveries, in which case
to a considerable extent in practice, the the loss is minimized. With contracts ex-
cheapest way in many cases is the creation tending into the distant future, on the
of a market; and markets do emerge. If a contrary, the possibility of failure to com-
market is impractical for one or another of ply becomes greater, partly because the
the reasons we usually call "market fail- self-enforcement aspects become weaker,
ure," then very likely some other social partly because unexpected changes may
device will at least be tried: government intervene to make even a sincerely in-
intervention; codes of professional ethics; tended compliance difficult or impossible.
or economic organizations with some The outstanding examples of forward
power intermediate between the competi- contracts are credit instruments. The
tive firm and the government. buyer, who is taking the risks of default, is
Thus, the failure of markets for future motivated to protect himself by seeking
goods must be regarded as an analytic more information about the seller. The
problem as well as a presupposition. It lender wants to know the borrower's as-
seems to me there are two basic causal fac- sets, the prospects for changes in them,
tors. One is that contracts are not enforce- possibly even what he is going to do with
able without cost and forward contracts the money. This very individualized in-
are more costly to enforce than contem- formation-seeking relation is quite far
poraneous contracts; the other is that be- from the arm's length impersonal model of
cause of the many uncertainties about the a market. The so-called capital markets
future, neither buyers nor sellers are will- are in many structural aspects very differ-
ing to make commitments which com- ent from our model markets. It is of
pletely define their future actions. Let me course an empirical question how far their
take these two points up in turn. behavior departs from the model. But
The ability to make enforceable con- the recurrent theme of credit rationing and
tracts is a necessary but not sufficient availability doctrines, the essential imper-
condition for a market. However, there is fections of the credit market which un-
no way to insure complete enforceability. derlie monetarist theories of cyclical fluc-
An individual may make a contract which tuations suggest that the incomplete en-
he cannot in fact fulfill. Penalties may in- forceability of credit contracts and the
deed be imposed on failure to live up to protective steps taken by lenders are sig-
one's agreement, but they are not a sub- nificant factors in explaining the working
stitute for compliance from the viewpoint of the market.
of the other party, and there is always a While the enforceability question ex-
degree of cost in enforcing the penalties. plains why those forward contacts that are
The laws of bankruptcy are a social recog- made do not constitute a perfect market,
nition that complete enforceability is im- we need more to understand why even
VOL. 64 NO. 1 ARROW: LIMITED KNOWLEDGE

these are so limited in their coverage of clude that both demand and supply will
future goods transactions. There are for- have a downward bias as compared with
ward contracts in money, some com- the situation in which uncertainty is ab-
modities, real estate, but very little else. sent. A buyer will be unwilling to contract
The explanation lies in uncertainties of for purchase of a good if a superior or
both buyers and sellers about prices and cheaper substitute may be available; and
quantities and about technology and the seller will be unwilling to accept a
tastes. Using uncertainties about prices price sufficiently low to be suitable to the
and quantities as an explanation for mar- buyer, particularly if he thereby precludes
ket failure is a circular argument, though himself from a possible opportunity to
not necessarily a fallacy. That is, if all shift his resources to other closely related
markets for future goods existed and goods. It would seem possible, at least,
cleared all transactions, then there would that there will be no price at which tran-
be no price-quantity uncertainties. But sactions in future goods will take place.
this much is true; if some markets for fu- From a theoretical viewpoint, one might
ture goods do not exist, then the agents say that the market is in a strange sort of
have uncertainties which are relevant to equilibrium; there is some shadowy sort of
their behavior on markets for complemen- price at which supply and demand are
tary or substitute goods. As Hicks showed equated at zero. But this price is not per-
a long time ago, complementarity and sub- forming much of a signalling function.
stitution can occur over time as well as There is one ultra-neoclassical approach
simultaneously. If, as I will argue in a to the market treatment of uncertainties,
minute, uncertainty can tend to destroy in which I take some pride. That is the no-
markets, then we can conclude that the tion of a contingent market. Instead of
absence of some markets for future goods letting uncertainty ruin existing markets,
may cause others to fail. we can take it explicitly into account by
To illustrate, the demand for capital buying and selling commitments to be
goods at any point of time is dependent on carried out only if some uncertain event
the prices and sales of the product at fu- occurs. We could in principle imagine
ture points of time. Therefore the demand agreements to transact which will hold if
for future capital goods will depend on ex- and only a given conceivable technological
pectations about the product at some still innovation does not take place, with a
more removed time. If we assume only second market for transactions valid if the
that we will not have markets for products innovation does take place. Then we can
at some distant point of time, then the restore the possibility of markets.
resulting uncertainty will reflect itself in a Such contingent markets are not en-
failure of the market for capital goods in tirely unknown; insurance contracts are
the nearer future, which will in turn create the purest example, and equity markets
still further uncertainties. and cost-plus contracts provide more
Thus, if some markets for future goods muddied illustrations. But they are rela-
are nonexistent, there will be uncertainties tively rare. Why this should be so follows
on the other markets; in addition, demand again from the general problem of informa-
and supply conditions for the future are tion costs and dispersal. If contracts are
uncertain because of technological and contingent on the occurrence of some
taste shifts. Assume that both buyers and event, then it must be verified whether or
sellers are risk averters. Then without not the event occurred. But this is informa-
going into details it is reasonable to con- tion, and as the example of a technological
10 THE AMERICAN ECONOMIC REVIEW MARCH 1974

innovation suggests, it is information information; in economic terms, this means


likely to be much more easily available to more information per unit of expenditure
one party than to the other. Hence, the of time or money.
range of possible contingent contracts be- The terms of trade with the outside
comes limited to those for whom the events world should not be regarded as freely
are easily verifiable for both parties. The given to the firm. In a world with a large
implications of these limits are known in number of commodities, even knowing the
the insurance literature as adverse selec- prices of relevant commodities involves
tion and moral hazard, and they are of the costly acquisition of certain kinds of
immediate practical significance in such information. This remark has given rise
matters as health insurance. But more to a large literature on search in recent
broadly, they so limit the scope of con- years. One implication which has been
tingent markets in practice that, as argued only slightly explored is that the concept
before in connection with markets for fu- of the market begins to weaken, and
ture goods, they prevent the emergence of Jevons's Law of Indifference becomes
even technically possible markets because more of an equilibrium condition than a
of the large unresolved uncertainties. statement valid about a market even in
I hope enough has been said to indicate disequilibrium. At a moment of time,
the widespread implications of costly, dis- prices of what would usually be thought of
persed information for the process by as the same commodity bought or sold by
which future-oriented economic decisions different firms can differ because buyers or
are made. Let me remark, briefly in view sellers may not, in their ignorance and in
of the length of time I have already taken, the presence of costs of search, find it
that informational costs and values play a worthwhile to shop further. Obviously, the
key role in modifying the structure even of important application of this principle
contemporaneous transactions. The in- may be to the labor market. Clearly, there
dividual optimizing agent is supposed to are important informational differences
know at least his technology or tastes and between the employees currently working
the prices he faces. We have already ar- for a firm and potential substitutes else-
gued a good deal of uncertainty with re- where, although these are interchangeable
spect to the future economic implications in pure neoclassical theory. Indeed, there
of present economic choices. But in addi- are differences both in the information the
tion there is the possibility that techno- firm possesses about its employees as com-
logical information, which would be useful pared with alternatives and the informa-
to him, exists somewhere in the world but tion which employees have about the eco-
outside his firm. There are grounds for nomic opportunities and the specific pro-
engaging in the active pursuit of informa- duction conditions of the firm as compared
tion. We begin to enter the realm of diffu- with outsiders. It appears that considera-
sion of innovations, to which some sociolo- tions of this type must play some role in
gists as well as economists have con- understanding the continued possibility of
tributed. The interesting points here are unemployment and particularly the slug-
the biasses in the information channels, gish response of wages to market dis-
some of which, at least, can be explained equilibria.
in terms of differential costs of acquiring I am far from exhausting the implica-
information. For example, the well-docu- tions of an information-economical view-
mented role of personal influence in ac- point for the economic world. I look for-
cepting innovations can be interpreted as ward to exciting developments in the next
due to a perceived high reliability of such decade.