The theory of perfect competition is unscientific because, by assuming a world of perfect knowledge in which firms cannot interact to change their economic environments, such a theory imposes pompous preconditions on our subject matter: competitors are so constrained in the behavior in which they can engage … we are precluded from understanding economic reality or developing a testable theory.
Burton H. Klein (1977, p. 71)
Consider a sports game for which spectators and participants agree that the rules are fair. The judges of individual sports events profess, however, to know beforehand what the outcome of individual contests should be. Failure of the sports events to produce the predicted outcome results in condemnation of them by these judges. Judicial nullification sometimes involves substituting the preferred or predicted winner for the actual winner of the contest. Sometimes, however, the judges require that the sports event be replayed over and over again, until the “correct” outcome is produced. Plays are called back, races rerun and contests repeated, not because of untoward conduct or fouls committed by participants, but solely because the “wrong” outcome results. If, contrary to judicial expectations, a particular participant or team persists in winning, that person or team is handicapped or perhaps even forbidden from playing in the future. The judges assert that they are enforcing rules, but the evidence that rules have been broken is inferred from the fact that other than the predicted winner came out first in the game. Some creative judges produce theories about what unfair activity must be going on to produce the unwanted results. However, these theories generally involve hypothetical actions not observed by anyone. Observable behavior that is blamed for bad outcomes cannot be distinguished by participants or spectators from approved behavior.
Clearly, no such sports contest exists. The enforcement of such rules would make a mockery of playing the game or running the race. Almost everyone recognizes that the winner of a sports event cannot be defined apart from who finishes first or scores the most points. “Winning” means nothing but playing fairly and coming in first. “Playing fairly” is defined before the fact in terms of observable behavior during the contest, and not, of course, in terms of after-the-fact outcomes. Teams or individuals scoring highly are not penalized for doing so, and certainly are not implicated in cheating merely because they frequently win by large margins. There is no sports foul of “excess points.”
Indeed, sports contests are played precisely because we cannot and do not know before the event who is the better player or team. Prior expectations of winners and losers are frequently falsified. To know the outcome with certainty would be to render the playing of the game unintelligible. To rig the results after the fact would be to perpetrate an injustice on players and spectators alike.
We claim that this procedure, which would be rejected out of hand for sports competition – or any other kind of competitive “race” in life, including that for most scientific analysis – is precisely that adopted by the vast majority of economists in thinking about competition. What would be an obvious paradox is suppressed by redefining “competition” to mean nearly the opposite of the behavior and phenomena it obstensibly denotes.
In this chapter, we argue that economic competition has more in common with other competitive activities in life than it has with economists’ standard conceptualization of economic competition. Competition is most fruitfully viewed as a process rather than a state of affairs. Viewing competition in static terms causes numerous analytical problems, some of which we highlight in this chapter. Among other things, static competition theory ignores the fact that only if there is a rivalrous, competitive process will the desirable normative properties of competition tend to be produced. For instance, statutory monopolies are objectionable not only because they produce a given product at higher prices, but also because they fail to produce the range and quality of products preferred by consumers.
More generally, we argue that the theory of perfect competition denies the very reality it purports to study, is a poor predictive theory and is untenable in normative analysis. As well as offering a critique of current approaches, we present our own approach. We try to be concrete and offer examples, for we believe that only by offering numerous examples can we suggest just how unsatisfactory the dominant theory of competition is. Moreover, in many cases we view these problems as exemplary of the future research agenda of a process theory of competition.
It is well known that in economic theory “competition” means the opposite of its meaning either in ordinary language or in commonsense economic discussions of competition. This ironic use of “competition” is seen as a virtue rather than a vice. Theorists argue that perfect competition teaches us about important relationships among economic variables in competitive equilibrium and about certain normative properties of equilibrium. Before examining these professed benefits, we shall look at what is lost by using the theory.
Perfect competition is a theory of states, not of processes; it tells us nothing about the adjustment from one competitive equilibrium state to another. In fact, if we consider what we know from general economic theory, we must be pessimistic about the system’s ability to move from one hypothetical equilibrium state to another. Equilibrium positions are not path independent. False trading, for example, produces wealth effects, which in turn generate a new implied equilibrium, different from the original one.
Not only does theory tell us nothing about adjustment processes, but, when taken seriously, it implies that there ought to be no process of adjustment at all (see Chapter 3 above, pp. 54–5). Once a new equilibrium is known, agents move to it immediately.
Competition in fact is a continuous process and not a set of conditions. As Hayek (1948a, p. 94) observed of competition, its “essential characteristics are assumed away by the assumptions underlying static analysis.” The received theory of competition is comparative static, focusing on beginning and end points. Economic agents are interested in neither the beginning nor end points, but in coping with never-ending adjustments. The theory of perfect competition analyzes the state of affairs or equilibrium conditions that would exist if all competitive activity ceased. It is not an approximation but the negation of that activity.
All theories abstract from part of reality. Theorists must determine in each case the appropriate degree of abstraction. What is essential and permanent to the phenomena ought to be part of the analysis. It would, for example, be a pretty poor economic theory that abstracted from scarcity. All genuinely economic, as opposed to purely computational, problems arise, however, because of the passage of time and concomitant changes in knowledge and the data. Economics must analyze the process of adaption to change as surely as it analyzes scarcity. Under some conditions, this adaptive process is competitive. We need, therefore, a theory of the competitive process. Neoclassical economics contains no such theory.
The orthodox theory of competition postulates a situation in which a large number of buyers and sellers of a homogeneous good transact in an environment of free entry, parametric pricing and perfect knowledge. We have already added our own criticisms of perfect knowledge to those of numerous other authors. Here we would emphasize that the problem of incomplete knowledge and the necessity of adaptation exist only to the extent that the data change unexpectedly. “Economic problems arise always and only in consequence of change. So long as things continue as before, or at least as they were expected to, there arise no new problems requiring a decision, no need to form a new plan” (Hayek, 1945, p. 523).
The existence of scarcity necessitates agents making a set of choices – a plan. In a static world of scarcity, however, each individual would need to form only one plan per lifetime. Plan revision occurs not because goods are scarce but because of changes in the environment, or in the individual agent himself.1 If the static model were a close approximation of reality, we would have exhausted our subject long ago. There is just so much that can be said about the Pure Logic of Choice in an unchanging world. Yet in the theory of competition, which ought to be preeminently a theory of change, the static model of pure competition reigns supreme.
For example, contrast a process view of product heterogeneity to the orthodox one. Neoclassical economic theory treats product differentiation as an equilibrium phenomenon, the outgrowth of consistent plans between consumers and producers. Yet product differentiation may also be the outcome of a process in which entrepreneurs try to mesh their plans with those of consumers. Because of changing conditions (including but not limited to changes in consumer tastes), producers are not sure what buyers want. By a process of trial and error, producers change nonprice variables in an attempt to discover how best to serve consumer wants. This producer-generated trial and error process is likely also to generate consumer experimentation with heterogeneous offerings. The observable result, which is not part of anyone’s explicit intention, is “product differentiation.” The attempt to discover by trial and error the actual set of consumer tastes is bound to produce greater diversity in product offerings. Uncertainty and a high degree of competition, not market power and imperfect competition, produce this result. Heterogeneous expectations and tastes characterize the process.
The treatment of product heterogeneity is an instance of a more general issue. Concepts and problems whose existence derives from change and the passage of time are often analyzed in static terms. The positive analysis is deficient, and will almost inevitably mislead the analyst in policy application. We say more about this issue in what follows.
In our analysis of competition, plan coordination is the norm. As we have seen, the plan coordination concept was the outcome of Hayek’s early attempt to reformulate equilibrium analysis for a multi-person economy in time. The Pure Logic of Choice is abstract and deductive. The degree to which an economy actually tends toward complete plan coordination (“equilibrium”) determines the applicability of equilibrium models to real-world economies. For Hayek (1937a, pp. 43–4), the analysis of equilibrating tendencies constituted the “empirical element” in economics.
The focus on forces tending to equilibrium explains the concern with entrepreneurship in modern Austrian work, since the entrepreneur is the active coordinating agent in market economies. Prices are signals or indicators, but not unfailing guides to economizers. Institutions provide a background for decision-making, and these set practical limits to the divergence of expectations. Nonetheless, it is entrepreneurs who are alert to opportunities, and, indeed, whose creativity is the very source of many of these opportunities.2 In static equilibrium, there can be no profit opportunities. But only by acting on their hunches and forecasts, so as to grasp profits, could entrepreneurs bring about a situation in which equilibrium is approached.
Hayek originally defined as an equilibrium a situation in which there is both ex ante plan consistency, and no information disruptive of plans that agents are bound to learn in the course of executing their plans. Exogenous disturbances might occur before these plans are executed, and upset the equilibrium. As long as agents did not themselves bring about these disturbances by the very execution of their plans, their plans were coordinated and consistent. Endogenous market forces would then tend to bring the system to the original equilibrium position.
In his work on entrepreneurship, Kirzner (1973) has consistently adhered to Hayek’s early view. Yet by focusing on entrepreneurship, we can understand better the reasons that surely entered into Hayek’s revised approach to competition, coordination and equilibrium. The fundamental problem is that the “tendency to equilibrium” view does not take time seriously. The latter is, of course, as serious an internal criticism as one could levy against a subjectivist analysis.
Competition is a dynamic process, a process in time. Since, as we have seen, knowledge must change with the passage of time, individual agents will alter their plans as time passes. This changing of plans disrupts the plans of other agents. Conventional learning models are, however, “clockwork” models, whose most significant characteristics have been succinctly described by Littlechild:
The agents are equipped with forecasting functions and decision functions to enable them to cope with uncertainty. Indeed, the agents are these functions. But though their specific forecasts and decisions may change over time in response to change in economic conditions, the functions themselves remain the same. The agents never learn to predict any better as a result of their experience. Nothing will ever occur for which they are not prepared, nor can they even initiate anything which is not preordained. They are clockwork Bayesians, wound up with prior distributions and sent on their way, to attain eventually, if circumstances permit, that everlasting peace in which they never need to move their posteriors.
(Littlechild, 1977, pp. 7–8)
The theory problem is modeling competition as a continual process. One could formally model the economy as a clockwork mechanism, and then hypothesize a never-ending stream of exogenous shocks wrought by entrepreneurs outside the system. In such a model, the passage of time would not be accompanied by learning. To prevent the system from “winding down,” one would need to bring in the entrepreneur as a deus ex machina. Schumpeter chose this latter course in his Theory of Economic Development (1934). The circular flow is then the natural state of the economy, into which it settles unless disturbed by disruptive entrepreneurs. The Schumpeterian gambit maintains the Newtonian conception of time and places the entrepreneur outside the system. Yet surely neither entrepreneurship nor learning ought to be completely exogenous (i.e., unexplained) in a process analysis, even if one accepts that there are exogenous components to both.
In “Competition as a Discovery Procedure,” Hayek (1978) explicitly integrated true learning (i.e., changes in the learning functions themselves) into a process theory. He brought the entrepreneur in from the shadows and made him part of the economic system. In so doing, Hayek offered a genuinely novel view of the function of competitive institutions. He treated a competitive market as a spontaneously evolved set of institutions and customs facilitating information acquisition. To simplify only slightly, nearly everything assumed to be data in orthodox constructions is the object of a trial and error process in Hayek’s approach. As Hayek put it,
Competition is valuable only because, and so far as, its results are unpredictable and on the whole different from those which anyone has, or could have, deliberately arrived at. Further, … the generally beneficial effects of competition must include disappointing or defeating some particular expectations or intentions.
(Hayek, 1978, p. 180)
Hayek’s view embodies endogenous learning and entrepreneurship. Moreover, it captures that essential element of competition that is absent from alternative economic conceptualizations: the element of surprise or the unexpected. It thereby meets the criterion suggested in our sports contest example; at least some of the time, the outcome of a competitive struggle must be other than what we expect.
There are five general characteristics of knowledge with which a Hayekian view of competition is concerned. Knowledge is (1) private, (2) empirical, (3) often tacit, (4) not all gained through price signals and (5) often the source of surprise. Each of these characteristics is important, and together they distinguish a process from a static conception of competition.
A large and significant part of advanced economic theory explicitly or implicitly rejects the privacy of knowledge. The strong version of the “efficient markets” hypothesis explicitly denies that any knowledge remains private, and not just part of the data of the system. With relatively few exceptions,3 rational expectations theorists fashion models as though information available to anyone is available to all. While it certainly would be an exaggeration to suggest that privacy of knowledge is generally accepted, there is nonetheless some recognition of this issue in other areas of theory.
In the literature on localized information,4 some knowledge is private and not datum. Privacy models can be used to analyze numerous issues, including the ability of individuals to exploit the knowledge on the market. We consider this literature to be salutary, and we have no disagreement with it as far as it goes. Our problem with the direction it has taken comes out in the following discussion of the other characteristics of knowledge.
The knowledge sought by economic agents is empirical in the following sense. They are primarily seeking “knowledge of the particular circumstances of time and place” (Hayek, 1945, p. 52). When acquired, this knowledge does not consist of abstract scientific propositions, which form the basis of logical deductions to certain conclusions. Empirical knowledge consists of information of temporary and fleeting significance, which may be factual (i.e., profitable) only so long as others do not also know it. Almost any profit opportunity fits this characterization.
The value of incomplete information partly depends on the processor of the information. Each actor must exercise judgment about information and its place in his overall plan. What an agent ought to do cannot be determined by an outside observer possessed of different information, judgments, tastes and plans. There is, consequently, no uniquely rational or “correct” course of action, as may be the case when dealing with a scientific-deductive problem. This latter point requires emphasis, given economists’ inclination to pronounce on “the rational” course of action in numerous situations.
In a situation of given means and given ends, there may be only one course of action that will maximize the relevant objective function. This would be true by virtue of our construction of the problem. We may choose to label this course of action “rational.” We can deduce the “rational” or “correct” action only because we have converted an empirical or trial and error learning problem into a logical or deductive problem. This conversion is acceptable if and only if all required information is available simultaneously to the decision-maker. Only in this case could one logically deduce an optimal course of action. It is misleading to conceive of decision-making, which involves trial and error discovery of incomplete information, as though it were a scientific problem, about which agents “theorized.” Hayek stated the point clearly and concisely:
Implication is a logical relationship which can be meaningfully asserted only of propositions simultaneously present to one and the same mind … Only to a mind to which all these facts were simultaneously known would the answer necessarily follow from the facts given to it. The practical problem, however, arises precisely because the facts are never given to a single mind, and because, in consequence, it is necessary that in the solution of the problem knowledge should be used that is dispersed among many people.
(Hayek, 1945, p. 530; emphasis added)
The nondeductive or nonscientific quality of much economic knowledge is obviously related to its privateness. Both are also related to the tacit quality of much economically important knowledge. In all areas of human endeavor, individuals employ knowledge that either they are not aware they possess or they cannot characterize precisely enough to communicate to others. In philosopher Gilbert Ryle’s terminology, they “know how” to do something but do not “know that” so and so is true of what they do (Ryle, 1949).
Scientific progress occurs as the stock of interpersonal and communicable knowledge increases. In this regard, however, scientific knowledge and progress are poor models of economic knowledge and progress (Sowell, 1980, pp. 8–11). Thus, tacit knowledge may take the form of a skill or may be embodied in a custom or unarticulated rule of behavior. Cycling and swimming are two examples. Relatively few accomplished at either activity understand the principles involved, or are even adept at teaching others how to ride or swim. Cyclists and swimmers know how but not that. They may very well be fit subjects for imitation. Apprenticeship, not a reference work or university short course, is the learning model. Michael Polanyi illustrated the theoretical point with reference to swimming:
The decisive factor by which the swimmer keeps himself afloat is the manner by which he regulates his respiration; he keeps his buoyancy at an increased level by refraining from emptying his lungs when breathing out and by inflating them more than usual when breathing in; yet this is not generally known to swimmers.
(Polanyi, 1962, p. 49)
Two important implications follow from the above considerations. First, discovering tacit knowledge involves time-consuming processes that may never be successful. Other market participants may not be able to discover the source or reasons for an entrepreneur’s success. His profits may accordingly persist for a substantial length of time. Second, some of what individuals do and why they do it cannot be successfully communicated or explained to third parties. Third-party observers ought not to expect agents to be able to rationalize their conduct. We develop some of the further implications of this for public policy on competition and monopoly.
Since much information is tacit and cannot be communicated, even in equilibrium, not everyone will know everything. Economic systems do not move toward a situation in which information is fully disseminated, at least not explicitly. Some knowledge will remain private.
When it comes to the role of prices in allocating resources, Hayek’s message seems to have been learned almost too well. The proposition that “the price system [is] a mechanism for communicating information” (Hayek, 1945, p. 526) plays an especially important role in the localized information literature. But the context in which Hayek presented the idea has been forgotten. The proposition has been transformed into the entirely different one that “nothing but” price signals communicate information on markets. Its corollary is that, unless price signals accurately reflect equilibrium scarcity values, we cannot or ought not rely on them.
A world in which prices were always at their general equilibrium level would be a world in which prices were not needed. To understand this point, consider the function of the Walrasian auctioneer. He centralizes information and ensures that individuals do not act on disequilibrium data. Equilibrium prices are generated via the tatonnement process. No real time passes and no false trading occurs. Though nominally in time, the Walrasian world is really static. The most interesting feature, however, is the centralization of information. Formal economic theory provides no argument for decentralization. Based on general equilibrium theory, perfect competition theory is more applicable to centralized than to decentralized economies. Static models of centralized information render real competition superfluous and literally wasteful. In this world, we can forget prices and just have the central processor of information issue production orders instead. The prices of general equilibrium theory do not provide information of the type we have been discussing. These prices are simply statements of equilibrium rates of trade-off.
What of the idea that prices guide behavior? First, prices are useful guides or signals because, and insofar as, they reveal discrepancies, previous maladjustments and errors. It surely misses the point to ask if they are now “correct.” Prices reveal what people want relatively more urgently now, and in the future, not what they would want in a hypothetical and unattainable equilibrium. No known system accomplishes the latter, and it is pointless if not misleading to make this a normative reference point.
Second, prices and markets function as part of a social system, not in isolation. A social system generates many kinds of signals and rules besides prices. Unless all these other guides are superfluous, it is erroneous to suggest that prices alone are sufficient guides. A theory of passive response to prices and only to prices is not a theory of human action, but a physics of automatons.
Nonprice constraints are as much part of a decentralized economy as are the prices they help to generate. These constraints are reference frameworks and orientation points, in terms of which actors form expectations. Prices are formed on markets composed of contracts, rules and customs, which are part of the constraints and basis for observed behavior.5
The tendency in the industrial organization and applied price theory literature is to view nonprice constraints as either extraneous or suspicious intrusions on “competition.” We are arguing to the contrary, namely that these constraints are often necessary accompaniments to markets. For example, it is strictly impossible to imagine a “price system” devoid of contracts and property rights. Yet much of the focus of the applied literature involves casting suspicion on all nonprice constraints on behavior (e.g., resale price maintenance). There is a presumption against them, and agents must justify their existence. We think this attitude follows from the presumption that prices “ought to” allocate resources, because they do so in perfectly competitive models. Not prices but people allocate resources, and flesh and blood human actors depend on all these nonprice variables in their decision-making.
Third, to varying degrees agents are endowed with entrepreneurial ability. Entrepreneurs do not merely respond to, but also create, change. They outguess market prices when these prices do not seem consistent (Rothbard, 1970, II, pp. 464–9). Whether we call this entrepreneurship a capacity to find out “particular circumstances” (Hayek, 1945, p. 52) or “alertness” (Kirzner, 1973, pp. 65–9), it is a sine qua non of a market economy. Yet this “driving force” of market economies is absent from models of perfect competition. Schumpeterians and Austrians have tried to fill this gap with theories of entrepreneurship.
Competitive market processes must produce results surprising (at least in part) to market participants and observers. Any interactive social process in real time produces unintended and hence unforeseen results. This will be true if for no other reason than the conflicting goals of diverse members of society. Whenever there is goal conflict, there must be a social mechanism to reconcile these conflicts. The mechanism may involve peaceful or nonpeaceful reconciliation. It may entail either private or collective action. If private, it may involve market or nonmarket approaches. Regardless of the mechanism, some agents must experience disappointment and revise their plans. This plan revision must involve their making choices other than the ones they had originally wanted to make.
The outcomes of social processes are unintended for three interrelated reasons. First, there is the need to reconcile conflicting plans. Second, there are always unintended by-products to individual actions in a society. Acting individuals inevitably produce results that were, in Adam Smith’s phraseology, “no part of their intention.” This realization is surely the basis for any social theory and is, in any case, the core principle of modern economics (O’Driscoll, 1977). Any economically interesting analysis of attempts to engineer ex ante economic and social outcomes necessarily involves specifying unintended consequences of human action. For instance, the results of imposing below-market rental prices on housing units is an example of the principle involved. The intended outcome (plentiful and cheap housing) is the given for the problem; the analysis entails deducing the unintended outcome (expensive and depleted housing stock). What is true of economic theory is likewise true of sound sociological and political analysis. Each discipline studies the unintended consequences of human action.6
Third, actions are not merely “additive.” As the number of buyers and sellers of a good increases, this may result in not merely more production and trade, but also a more highly developed market for the good. For instance, if an increasing number of traders seek liquidity and try to economize on holding inventories of real goods, they may succeed not only in achieving their goals but in facilitating the emergence of a medium of exchange (Menger, 1892). As Marx emphasized, quantitative changes evolve into qualitative differences.
The unintended consequences of human action must often be as surprising to the theorists as they are to market participants. Similarity of problems renders pattern prediction feasible (as in the rent control case). Yet the seeming precision of such predictions masks basic theoretical ambiguity. Myriad outcomes are possible. Will the controls be enforced or not enforced? Evaded (how?) or not evaded? Will the government eventually subsidize enough construction to meet the excess demand (Sweden) or not (the USA and UK)? The attempt to predict precisely the unintended by-products of human action would involve, inter alia, a predictive theory of human institutions. Nothing short of discovering the “laws of history” would make this endeavor possible. Past efforts in this area have been conspicuous failures. Our only prediction of the future is that this record will continue unchanged.
The final element of surprise concerns expectations formation. We have more to say on “objectivist” theories, such as rational expectations, in the chapter on money (Chapter 9). Here we simply observe that a subjectivist approach emphasizes the diversity of expectations. This diversity is a function both of the diversity of human beings and of the effects of change and the passage of time. Pre-reconciliation of plans and forecasting of the future would necessitate each individual’s predicting the mental states and choices of large numbers of unknown people, whose decisions affect his environment and his choices. Theories that entail or assume this ability violate the basic requirements of a subjectivist and methodologically individualistic analysis. They fail to take differences among individuals seriously. The theories proceed “as if” there were only one decision-maker. These theories produce unintelligible results when applied to decentralized economies. Yet, to reiterate, the modern theory of competition assumes centralized information (in the auctioneer), and (usually) uniformity of expectations. It would thus be only a slight exaggeration to say that neoclassical economics has produced an elegant static theory of centralized resource allocation, but no theory of competition in a decentralized economy.
A theory of competition as a process will necessarily be at odds with static theories of competition. These latter equate competition with the attainment of certain conditions, the existence of certain market structures, and the presence of stereotypical behavior (e.g., price-taking) by transactors. If markets fail to replicate these conditions or to produce certain results (e.g., P = MC), then this is taken as imperfection of competition. This conception underlies the proclivity, illustrated in our parable, to condemn actual market outcomes.
If competition serves a social purpose, it must produce something that we could not have in its absence. To the degree that competition does what we could have done equally cheaply in its absence, it is wasteful. Competition in fact leads to the discovery of opportunities that would otherwise go unnoticed. It thus generates a spontaneous discovery process, the exact course of which is unpredictable. This process includes, inter alia, both the discovery of hitherto unsatisfied wants and the products to satisfy those wants, and the invention of lower-cost methods of satisfying preferences. It also encompasses the creation of new economic forms, customs and structures.
None of this suggests, of course, that competition does not also fulfill its traditional functions of creating incentives to keep one’s prices and costs in line with other producers. Even in this case, however, the discovery process plays a crucial role. Costs are not a well-defined given; they change as productive techniques vary. Even in highly competitive industries that best fit the model of pure competition, such as agriculture, different production techniques exist literally side by side. The differences are sometimes dramatic, as in tending of vineyards and production of wine. They exist also, however, in activities as prosaic as the growing of feed corn. If we move to industries of heterogeneous goods, then the relevant price is no longer a given, and ascertaining its profit-maximizing level is part of the competitive discovery process. It is not, then, what competition does to fulfill our expectations that recommends it: it is what competition does that we would not have expected it to do that recommends it.
Normative analysis of competition must differ in process and static theories of competition. In its positive analysis, process theory analyzes the competitive discovery procedure inherent in the coordination of plans. Accordingly, process economists focus on the capacity of social systems to discover and innovate. Contrast the emphasis on change and discovery with the approach of standard welfare economics. Choices are pre-reconciled under known conditions, which generate predictable outcomes. As Kaldor (1934, p. 147) pointed out, “the formation of prices must precede the process of exchange and not be the result of it.” In effect, the outcomes of trade must be known by agents before the trades are consummated. Further, the analyst-observer knows the outcome in the guise of well-known equalities at the margin.
Emphasis on the unintended consequences of human actions leads to another decisive difference between process and neoclassical normative analyses. Standard optimality criteria are foreign to a process approach. Optimality criteria are static rather than dynamic, and judge performance or outcomes rather than processes or operation. They presume that social processes ought to produce outcomes that can be specified in advance. This argument is taken up in more detail in the next chapter. Here we will just observe that there is no straightforward way of applying standard optimality concepts to institutions or processes. To know whether an institution or process is optimal, we would need to know the very information whose discovery is the object of that process or the operation of that institution. Once again, if we could independently ascertain the information, the institution or process in question would be superfluous.
Even casual inspection of the literature reveals a systematic conflation between general equilibrium theory and classical arguments in favor of free competition or laissez-faire. Consider the following argument of Kornai:
The modern equilibrium theory is nothing else than a mathematically exact formulation of Smith’s “invisible hand” which harmonizes the interest of egoistic individuals in an optimal manner. At the time of Smith … this description of the functioning of a capitalist economy was not unrealistic (though not exact either). More than a hundred years were required for Smith’s intuition to be expressed in a faultlessly exact form; by the time it was achieved, it became utterly anachronistic.
(Kornai, 1971, p. 349)
Or compare Hahn:
When the claim is made – and the claim is as old as Adam Smith – that a myriad of self-seeking agents left to themselves will lead to a coherent and efficient disposition of economic resources, Arrow and Debreu show what the world would have to look like if the claim is to be true. In doing this they provide the most potent avenue of falsification of the claims.
(Hahn, 1973, p. 324)7
In fact, Smith’s defense of competition involves a theory of unplanned or spontaneous order. This tradition, in which Smith forms a middle link between the scholastics and natural law theorists on the one hand and modern theorists like Mises and Hayek on the other, is an alternative to modern general equilibrium theory and not an early or crude anticipation of it. A theory of evolved orders is not a theory of optimality or efficiency, precisely because it is a process and not an end-state theory. Social efficiency or global optimality concepts must be foreign to process theories. Individuals economize and, in this sense, attempt to allocate their own resources efficiently. There is no social choice process, however, in which society chooses so as to maximize “social utility.” Outside of a static context and absent the stringent assumptions of general competitive theory, it cannot be demonstrated that the pursuit of individual optimality results in a well-defined social optimum (unless all that is meant by the latter is the process of individual optimization). As a corollary, however, one cannot apply static welfare criteria to demonstrate that there is suboptimality.
It is simply anachronistic to attribute modern welfare concepts to eighteenth- or even to most nineteenth-century writers. No such concepts existed then. Many modern interpreters apparently believe that Smith et al. “must” have been groping toward modern welfare analysis, but lacked the necessary training in calculus to articulate their views. Nonetheless, careful reading of The Wealth of Nations provides little support for the thesis that Smith was a crude neoclassical welfare theorist.
First, Smith wrote of advancing (not maximizing) the material well-being (not utility) of the common man (not society as a whole). He thus frequently advocated uncompensated property rights transfers, as when he recommended removing monopolistic trading privileges for the benefit of consumers and to the detriment of the monopolists.
Second, Smith relied heavily (though not exclusively) on arguments in terms of rights and liberty. For example, he argued that
[t]he property which every man has is his own labor, as it is the original foundation of all other property, so it is the most sacred and inviolable. The patrimony of a poor man lies in the strength and dexterity of his hand; and to hinder him from employing this strength and dexterity in what manner he thinks proper without injury to his neighbour, is a plain violation of this most sacred property. It is a manifest encroachment upon the first liberty both of the workman, and those who might be disposed to employ him.
(Smith, 1937, pp. 121–2)
Smith’s argument is typical of the classical arguments for free and competitive markets. Competition would permit individuals to achieve those goals most important to them. For Smith, this would lead to the maximal attainable gain for most, given side constraints of which individual liberty was the most important. He did not suggest that the result of this would conform to any particular, preconceived outcome of nonindividualist welfare criteria.
Arguments in terms of liberty formed important parts of the classical liberal case for competition. Liberty was valued both in its own right and for instrumental reasons. Paralleling the common law, liberal political economy justified outcomes, at least in part, because they resulted from a system of voluntary trade and political freedom.8 The modus vivendi of a process of free exchange and production created a presumption in its favor. Only if this process were operative would we know what individuals valued most. In other words, the classical liberal argument for free exchange and competition – an argument inextricably intertwined with the classical political-economic case for competition – is a process argument. It is also a fundamentally different argument than modern ones in terms of optimality criteria and social utility maximization.
Our major point is not doctrine-historical but substantive. Any dynamic analysis of competition must have criteria alternative to those of static welfare theory. It would be strictly inconsistent to fall back on static welfare theory in assessing markets. Whatever can be claimed on their behalf must depend on the alternative criteria. Elements of a normative analysis are presented in the next section.
Theorists have traditionally utilized exact or deterministic equilibrium concepts. Indeterminism and unpredictability show up in fitting or applying the concept and analysis to concrete problems. By emphasizing the unpredictability and indeterminateness of social process, we are not raising a new problem so much as proposing a new solution. As Coddington (1975, p. 156) has said, “When it comes to being put to some use, the static method abandons its own formalization anyway. The choice then becomes less dramatic: between abandoning formalization openly or abandoning it in a surreptitious way.” We have argued above that our alternative approach would permit theorists to incorporate time and change in a meaningful way in models.
The passage of time makes it inevitable that some expectations will not be met, some plans not fulfilled. Some disappointment of expectations and a degree of frustration in implementing plans are inevitable in any social system. No set of policies or institutions can insulate us from the effects of time’s passing. The relevant question is how different institutions and policies affect individuals’ adaptations to unexpected outcomes.
Once we abandon the static framework, our theoretical and policy focus changes. For instance, owners of existing entitlements always prefer policies that insulate them as far as possible from the undesirable effects of a changing environment. These preferences may sometimes by justified by a static analysis (e.g., by a favorable “equity-efficiency” trade-off). These claims automatically carry diminished weight in a nonstatic framework. Policies protecting existing entitlements inhibit adaptation of agents to past and future changes. To the extent that this happens, the probability that a market participant chosen at random will be able to fulfill his plans diminishes. Even those protected owners of current entitlements will find adaptation in the future more, not less, difficult because of their seeking current protection. Thus, “sick” industries become sicker not better as they are protected more thoroughly.9
As we saw in the previous chapter, exact coordination of individuals’ activities is not only practically impossible but also conceptually self-contradictory. Acting takes place in real time, and, as time passes, agents learn and alter their behavior. Complete coordination of activities thus cannot be a state toward which social systems are moving. It is possible, however, to postulate a tendency toward pattern coordination (a true dynamic equilibrium). This depends, inter alia, on the degree to which typical features persist with the passage of time. If, on the contrary, there were no typical features, then we could not speak of a tendency to equilibrium in any sense of the term.
Pattern coordination consists of coordination among the typical but not the unique aspects of individual behavior. In this context, two variants of a normative criterion suggest themselves. The first variant ranks different pattern equilibria, while the second deals with properties of the transition process from one equilibrium to another.
With respect to the first case, it is important to realize that any given instance of concrete behavior can be described in myriad ways. Many patterns of typical features can in principle be identified in a set of actions. The same vector of actions can thus comprise numerous pattern equilibria. Some are, however, “better” from agents’ perspectives than others. To refer to the illustration of the previous chapter, Professor B may identify several patterns of his co-author’s (Professor A’s) behavior. The realization that A comes into the office on Mondays and Wednesdays may be a more useful insight than the fact that he carries his briefcase whenever he comes into the office. Some pattern equilibria enable agents to coordinate their activities more effectively than others. To repeat, no form of pattern coordination will permit exact coordination of activities because there will always be unique features of events and actions. Our criterion of evaluation must, therefore, relate to the degree of coordination consistent with the endogenous change within the system and thus with the existence of real time.
With respect to the second case, the criterion is essentially the same. Suppose that the system is exogenously shocked, so that the typical features of agents’ behavior change. In the movement from one pattern equilibrium to another, some attempts at coordination will be frustrated. The actions of A may be predicated on the no-longer-typical features of B’s behavior (and vice versa). Nonetheless, we can still assess the performance of an economic system on the basis of its adjustment to change. Here the criterion is the relative amount of coordination consistent with the system’s exogenous change.
More concretely, we can attempt to base our judgments of various policies on the likelihood that a given change will result in a randomly chosen individual’s fulfilling his plans (Hayek, 1978, pp. 183–4). This criterion has two interrelated features. First, some social systems or policies adapt to changes with greater or lesser plan frustration. Second, other systems completely or partially block adaptation to change, thus also resulting in plan frustration.
We illustrate our point by referring again to the example of rent control. First, we argue for plan coordination as the preferable equilibrium concept. Second, we explain why a market-process approach requires adopting the concept of pattern rather than that of exact coordination.
The economically crucial effect of setting rents below market levels is surely not that they create a market excess demand for housing. The crucial and more general effect of these controls involves their effect on plan coordination and market signaling of relative scarcity values. If a public housing authority were created to supply the requisite housing units, then any excess demand would be temporary. Yet frustration and discoordination of plans would persist. We will show this first by considering the standard case of no public housing, and then by considering the case in which excess demand is supplied governmentally.
After effective controls are imposed, housing services will be temporarily in excess demand. Lessors and lessees cannot make their plans mesh. Over time, however, the housing stock will deteriorate until housing services supplied satisfy observed demand at the controlled rental prices. Even with market excess demand eliminated, plans continue to be frustrated, however, for renters cannot bid higher prices for the higher-quality units that they prefer.10
Now we consider the case in which a public housing authority supplies the unsatisfied demand at controlled prices. Market excess demand is eliminated in the long run, and plans are apparently fulfilled. In reality, however, plan frustration (but not market excess demand) appears in other sectors and under other guises. Taxpayers must shoulder part of the housing cost of entitlement-holding tenants. Net taxpayers (i.e., those paying more in taxes than the value of subsidized housing received) must now curtail their planned consumption of other goods. Moreover, satisfaction of renters’ preferences will be more apparent than real. Renters will be satisfied with their existing housing stock only because they are unable to implement other plans, such as moving to more desirable areas. Entrepreneurs wishing to respond to price signals in nonhousing markets (e.g., in manufacturing) and workers wishing to take advantage of higher wages will be frustrated in their attempts to move to more desirable cities or regions by the inflexibility of the housing market. Eventually, these planned moves will be canceled and the best will be made of the current location. There is frustration and lack of coordination, but the market excess demand is eliminated. The state housing authority will perceive that it has “solved” the problem. Instead, it has added to the discoordination of plans. This can be seen by considering why a region would be attractive to firms and workers.
Consumers’ preferences would be better satisfied by immigration of firms and individuals into the hypothetically more desirable region. Without this immigration, wants that could be satisfied will go unfulfilled. Plans cannot be executed. The mechanism that would normally facilitate adaptations – rising rental prices in growing regions – is rendered inoperative by the controls. There is probably no government with sufficient command over resources to supply excess housing demand of a mobile population (the source of inflexibility in the housing stock alluded to above). Even if public housing authorities were able to draw on the necessary funds, no effective system of signaling and incentives exists to inform managers of the relative importance of different housing demands. The very reason for the housing authority, the rent controls, has eliminated market-generated signals (i.e., flexible price changes).11
Thus far we have been arguing the case for adopting plan coordination as the equilibrium concept. We now will explain the reasons for adopting a process approach and pattern coordination concept. The rent control case is vastly more complex than even we have made it thus far. Textbook analysis is misleading in a number of respects. For example, the New York City housing market has not yet fully adapted to a system of rent controls first put in place during World War II. This situation certainly does not stem from the housing stock’s failure to deteriorate sufficiently! Rather, controls generate a process of response, where each stage generates changes in the environment that cause further responses. In turn, market participants’ responses produce further political responses. Public choice theorists would (correctly) assure us that the political changes are endogenous to the system. This endogeneity can be best captured in a model of rent control incorporating a dynamic conception of time.
Each control breeds evasions of the control. Each evasion produces further changes in the market environment and additional adaptations. In the political arena, those evasions produce demands for new controls, some of which are supplied. There is literally no end to this process so long as there are calculating entrepreneurs, economic and political, who can profit from existing opportunities. There simply is no static equilibrium to which anything will settle down. The rent control case is a prime example of how a process may never cease or stabilize. If we were to predict the future course of events in the New York City housing market, we could predict only possible patterns. The only sure prediction is that the details will be upset endogenously.
For any applied problem, theorists can and do cut off the analysis at a point at which they have reduced the unexplained phenomena to second-order effects. It is misleading, however, to suggest that they have identified an equilibrium, in the sense of a complete state of rest. They have just delimited an analytically convenient place to end one chapter of a story. What is a second-order effect often becomes a first-order problem subsequently. For instance, no analysis of the economics of converting rental units to cooperative apartments (a legal form that is economically similar to condominium ownership) would be complete if it did not relate the process to the longstanding system of rent controls in New York City. Yet such conversions belonged to the category of second-order effects until comparatively recently.
Our process view is fully consonant with cutting off analysis in accord with the problem at hand (writing “chapters” of analysis). By not employing an exact or deterministic equilibrium concept, however, we enhance the chance that related and dependent events will be seen as such, rather than as the product of unrelated exogenous shocks (such as the case of cooperative apartment conversion and rent controls). This is one major advantage, as we see it, of abandoning the search for determinateness at the conceptual rather than at the application stage.
We agree that a fruitful equilibrium concept is necessary for developing a systematic analysis. As we saw in the last chapter, equilibrium analysis is a type of causal reasoning. Causal reasoning about process in time differs from static analysis, however, and the equilibrium concept must change accordingly. Maximal possible plan coordination is the most straightforward adaptation of the plan coordination concept to dynamic problems. Like any other normative economic criterion, this one leaves some issues unresolved. For instance, one could increase the likelihood that some plans would be fulfilled by decreasing the likelihood that others will be fulfilled. Likewise, one would presumably wish to minimize the chance that some plans, such as those of a would-be murderer, would be implemented. This is to recognize that the basic questions of right and wrong, of the justice of entitlements, and of the role of the private and governmental sectors must be resolved before economic reasoning can be used in policy analysis. If this fact is made more obvious by adopting the criterion of maximal plan coordination, so much the better.
Our claim here is that a well thought out economic analysis can contribute to public policy discussions. As in the rent control example, economic analysis may serve largely to clarify fact patterns for policy-makers. Indeed, the rent control case is a paradigmatic example of Arrow’s observation (1968, p. 376) that
the notion that through the workings of an entire system effects may be very different from, and even opposed to, intentions is surely the most important intellectual contribution that economic thought has made to the general understanding of social processes.
More than any other modern school, Austrian subjectivists have consistently applied this insight to social processes. One’s attention is drawn to the unintended effects of human actions by systematically analyzing these as part of a process in time. Economic models in which agents foresee all events (even in a probabilistic sense) render unintelligible any concept of unintended consequences of human action. Insofar as models postulate perfect foresight, they obscure “the most important” contribution of the economic way of thinking.
We have argued that neoclassical economic theory misconceives the problems when it abstracts from the passage of real time, a tendency that has been particularly harmful in the theory of competition. In the next section, we develop our point further. We focus on the technique of assuming that agents engage in continuous utility maximization. We suggest that they cannot do this in a world of uncertainty and changing knowledge. Not only is the requisite knowledge for continuous utility maximization absent, but the techniques and institutions people actually use to cope with uncertainty are seriously misunderstood if continuous utility maximization is assumed. We argue that agents follow rules and use institutions as a substitute for continuously choosing at the margin. These rules and institutions are not themselves entirely the product of rational choice. We then briefly relate our argument to some recent work on the theory of the firm.
Neoclassical attempts to explain rules in the widest sense, including customs, law and other institutionally embodied complexes of rules, are marred by consistent misdiagnosis. A single error characterizes much recent work on such varied topics as the theory of the firm, the efficiency of legal rules and rules versus authority in monetary policy. In no other area does a subjectivist process approach yield such different conclusions from orthodox models and shed so much light on an important topic.
Any model that explains social rules solely in terms of maximizing behavior fundamentally misconstrues the phenomena. Men follow a rule when they respond in the same way to perceptions of a recurrent pattern. They thereby exhibit regularity of behavior in typical situations. Rule-following behavior is the product, however, not of knowledge or omniscience but of ignorance. This is certainly true for the large class of social rules, examined here, which are the product of a process of evolution rather than maximization.
Superficially, it might appear that continuous utility maximization would also result in a pattern of decisions similar to rule-following behavior. Nevertheless, the opposite is the case. What makes an event “typical” is that it shares certain abstract properties with other events of this type. Nonetheless, events of a type or class differ in details, often significantly. If, in fact, individuals were to know enough to discriminate among events of a class, then they would not follow rules or adopt consistent decision patterns in dealing with these events. Rule-following agents decide on the basis of the abstract properties of members of a class. When an agent knows enough about the details or particulars of a case, and of the differential effects of alternative decisions, then he decides not according to rules but on a “case-by-case” basis (or on “the merits of the case”). Case-by-case decisions involve treating similar cases differently, while rule-guided decisions involve treating similar cases identically.
The rule of law is perhaps the best illustration of our point. Stated simply, the rule requires judicial decision-makers to treat cases of the same class the same way (Leoni, 1961, pp. 59–76). In adhering to a concept of the rule of law, judges are under no illusion that cases are all the same in every detail. Nor would most legal commentators deny that, in some sense, there may be a net social gain from occasionally relaxing a rule. The practical problem is that, while we know that there are probably cases in which (for instance) it would be better not to punish the guilty, we can never know concretely which cases conform to this hypothetical situation. What jurists and legal commentators have discovered is that in the preponderance of cases justice is served by consistent application of relevant rules. And similarly for rules generally.
Rules evolve and are adopted because they work, whether these be rules of justice or of procedure within a firm. In the case of evolved rules of conduct or behavior, people obeying the rule often do not know why they work. In many cases, they may not even be aware that they are following a rule. Rule-following behavior precedes knowledge of this behavior. Understanding this behavior and being able to articulate or rationalize the rule is an even later development.
To be able to articulate why rules work presupposes more knowledge about the processes governed by rules than rule followers often can or do have. Justice is much more difficult to articulate than it is to practice. Were individuals to know enough to rationalize a rule, they would generally know enough to abandon it. If, for example, we had a scientific explanation (theory) of what a successful monetary rule accomplishes, we could in principle dispense with the rule. Indeed, monetarists emphasize this point in defending their money-stock rule. It is ignorance of the lag structure that necessitates adoption of a monetary rule.
The aforementioned character of rules derives from their being evolved. Without doing violence to the phenomena, we cannot conceive of evolved rules and institutions as the product of a maximization process. We reject maximization models here for two interrelated reasons. First, agents arrive at patterns of action through a trial and error process. Successful procedures are adopted and unsuccessful ones are rejected. More precisely, those agents who do not adopt successful rules cannot adapt to the relevant environment. In a market situation, this would mean losses and eventual forced withdrawal from the unsuccessful activity. Individuals stumble upon rules unconsciously, and, to repeat, are often unaware that they have done so. Not only do rules involve tacit knowledge, but their adoption is also often tacit. Maximization models cannot incorporate this insight. Hayek stated the issue concisely:
Man acted before he thought and did not understand before he acted. What we call understanding is in the last resort simply his capacity to respond to his environment with a pattern of actions that help him to persist. Such is the modicum of truth in behaviorism and pragmatism, doctrines which, however, have so crudely oversimplified the determining relationships as to become more obstacles than helps to their appreciation.
(Hayek, 1973, p. 18)12
Our second reason for rejecting maximizing models as explanations of evolved rules relates to the optimality of these rules. The outcomes of evolutionary processes will not generally be optimal in any nontrivial sense of the term. There is no “rule set” from which we draw. Evolved rules are not the product of choices over known alternative rules. Rule-governed behavior is the unintended outcome of trial and error procedures. As in biological evolution, starting along one evolutionary path closes off options. In a trial and error procedure, knowledge of alternatives is often gained only as they are tried. The “choosing” of a rule is consequently done without knowledge of alternative rules. Indeed, adopting one rule diminishes the likelihood that the adopter will learn about the alternatives. To adopt a rule is to regularize behavior, and to abandon further trial and error activities along certain lines. “Workable,” not “optimal,” is the appropriate modifier for evolved rules.13
If our analysis is correct, then the question of the “efficiency” of economic, political and legal rules must be reconsidered.14 Such rules represent the framework within which rational or maximizing behavior takes place. The adoption of such frameworks is itself not always the product of a maximizing procedure. It is, then, not so much an error as a misnomer to talk of the “efficiency” of political, legal and market rules and of rule-based institutions. Given these rules, we may analyze the efficiency of choices. But the rules themselves are adopted, at least in part, by an a-rational process. To suppose that, for instance, the legal framework is the product of rational choice only pushes back the question of the framework within which that choice was made. One would eventually be led into an infinite regress, a problem avoided by adopting an evolutionary approach to rules and institutions.
“Economic imperialism” is widely interpreted as the application of continuous utility maximization models to all human problems. If carried through, this approach will yield not knowledge but the pretence of knowledge. By claiming more for the theory of maximizing behavior than it can deliver, economists will only put their theory into disrepute in those situations for which it is well suited. In pointing this out, we are only suggesting that economists take their own protestations seriously. Economics analyzes marginal adjustments. We surely can explain the effects of changes in constraints on institutions at the margin; this, however, is not sufficient to generate a complete social theory of institutions. Likewise, we may feel confident of the effects of economic factors on voting; this does not, however, imply that economics yields a deterministic theory of voting.
In this chapter, we have outlined a subjectivist theory of competition. We highlighted a few topics on which subjectivism casts new light. We also attempted to indicate how concretely a subjectivist analysis might differ from standard formulations. At this point, we would like to focus on a subject of significant interest to economists in recent years: the theory of the firm. Surprisingly, there is no subjectivist or Austrian theory of the firm. This is true even though the subjectivist approach is particularly appropriate for analyzing firms as evolved social institutions. The absence of a subjectivist theory of the firm is even more remarkable, since Coase’s (1937) article has a subjectivist flavor.
The theory of the firm encompasses the question of both why firms exist and how they behave given that they exist. With respect to the first question, the situation as first described by Coase (1937) still represents the classic statement of the problem:
An economist thinks of the economic system as being coordinated by the price mechanism and society becomes not an organization but an organism. The economic system “works itself.” This does not mean that there is no planning by individuals. These exercise foresight and choose between alternatives. This is necessarily so if there is to be order in the system. But this theory assumes that the direction of resources is dependent directly on the price mechanism. Indeed, it is often considered to be an objection to economic planning that it merely tries to do what is already done by the price mechanism.
(Coase, 1937, p. 387)
The existence of firms represents a paradox for formal economic theory because it represents “nonprice planning.” This paradox derives from the exclusive reliance of the theory on prices to allocate resources, a reliance we criticized above. In Coase’s analysis, individuals compare the costs of using the price system to the costs of nonprice planning. Firms represent nonprice market institutions within which decisions are made and resources allocated. On the market, there is an “optimum” amount of this nonprice planning (Coase, 1937, p. 389 n.). The limit to vertical integration is set by the calculational chaos that may infect nonprice planning. If, for instance, the market for an input were to disappear entirely through vertical integration, then all firms using the input would be caught in an economic calculation problem. There would be no market to yield transfer prices within the firm. The firms could no longer calculate profits and losses in this line of activity (Rothbard, 1970, II, pp. 554–60).
We have long found Coase’s approach to the existence of firms congenial. It incorporates the essential conclusion of the economic calculation debate. That is, calculation of profits and losses is impossible without competitive markets for inputs. Gains from hierarchical organizations can be captured only so long as they do not completely eliminate factor markets. Coase’s approach is an excellent static conceptualization of the problem.15
The Coasean solution does not, however, address the question of firm behavior in response to change. Neoclassical production deals with this issue but does so in “Newtonian time.” Subjectivists cannot be comfortable with the theory, but they have not offered a substitute approach. We believe, however, that Richard Nelson and Sidney Winter (1982) have given us the elements of a dynamic theory of the firm. In effect, they apply a Hayekian theory of rules and evolved market institutions to firm behavior. Their explicit debt is to Schumpeter, but to exactly those parts of the Schumpeterian system that have most in common with the work of Menger and Hayek.
Nelson and Winter view firms as generated by an evolutionary process. Existing firms can be best explained by reference to prior adaptations of the environment. This adaptation is revealed as rule-following behavior, or “routines” in their terminology. Routines characterizing firm behavior correspond to genes in biology. A firm’s routine largely but not completely determines its behavior, as well as its ability to cope with environmental change. Thus, firms adapt to the environment in different ways. The way in which they adapt may determine their survival characteristics for future environmental changes. If a firm’s routine is inappropriate to a changing environment, and the firm does not hit on a new routine quickly enough, it will be “selected out,” i.e., may suffer losses and disappear.16
We read Nelson and Winter as avoiding the pitfall of taking the evolutionary analogy too literally. Entrepreneurs can consciously alter their firm’s routine, as can human and some nonhuman species. There is also a Lamarckian element in their story; acquired characteristics are inheritable.
An important advantage of Nelson and Winter’s approach is the ease with which the entrepreneur is integrated into the analysis. Indeed, the entrepreneur is absolutely indispensable to their story. He is a force for change but is not the product of ad hoc theorizing. He disturbs firms’ routines by changing the environment. In turn, conscious entrepreneurial adaptation to a changing environment is sometimes the only way a firm, locked into a now-inappropriate routine, may survive. Entrepreneurial innovation may result eventually in a new routine, an adaptation to the new environment. Nelson and Winter focus their analysis on innovation, invention and the upsetting of firm routines.
Clearly, much more work needs to be done on a subjectivist or Austrian theory of firm behavior. Any theory along these lines must take seriously the Menger-Hayek distinction between designed and undesigned social institutions. Certainly, firm owners attempt to maximize their profits through the vehicle of their firms. The pattern of firm survival and the character of surviving firms is not simply the result, however, of conscious planning and rational profit-maximization. The industrial landscape reflects both the results of conscious planning and the unintended consequences of entrepreneurial interaction in the marketplace. (Firms do not, for instance, normally plan their own demise.) We find an evolutionary approach like Nelson and Winter’s to be attractive because it allows for both planning and profit-maximization, and unplanned consequences. Entrepreneurs can, to some extent, influence their environment and affect the fortunes of their firms. But firms are ultimately either “selected” or not. Evolutionary forces in the environment reflect the unplanned consequences of others’ actions (both purchasers of a firm’s products and competitive firms).
Though consciously an essay in Schumpeterian economics, Nelson and Winter’s work may have the unintended consequences of providing a building block for a subjectivist theory of the firm. In any case, their work exemplifies the sharp contrast between neoclassical theory and one incorporating the effects of the passage of time. Their contribution is a first step toward an alternative theory of the firm. More generally, it offers an approach to modeling institutional change.
In modeling learning, change, error and expectations, theorists increasingly employ the concept of a stochastic equilibrium. In such models, there is an underlying stochastic process generating the data. Expectations of rational agents will tend to the mathematically expected value. “Learning” involves updating of priors until the subjective probability distribution conforms to this objective probability distribution. Apparent forecast error is optimal. Since information is costly to acquire, it does not pay agents to acquire more information about the underlying stochastic process. There is variance in the outcomes, but experienced errors in forecasts will not lead agents to revise their optimal forecasts or to alter their behavior.
Some of our misgivings with this approach were stated above in reference to Hahn’s “learning functions.” In the dynamic or pattern equilibrium that we have articulated, there is genuine learning. The competitive market process is a never-ending learning process. In an explicitly evolutionary theory this learning is captured in the process of adaptation and innovation. There is no learning in a stationary stochastic equilibrium (Hahn, 1973, pp. 18–20; Littlechild, 1977, pp. 6–8).
Similarly, “imperfections” in the market process do not necessarily arise from participants’ choosing an “optimal level” of error. The errors on which we are focusing arise from the very attempt to arrive at individual adjustment to the environment. They are unintended and unplanned consequences of individuals’ actions and interactions. These market errors or imperfections may take the form of “too much” product heterogeneity, as firms try to discover and satisfy consumer tastes. The errors may also show up as poorly invested capital (Chapter 8 and Chapter 9). In our theory, agents will be dissatisfied with outcomes, will revise expectations and will alter their behavior in the face of error. Error is part of the very market process itself, part of the stimulus to further adjustments. In general equilibrium models, none of this should occur, since all observed error is “optimal.” At the descriptive level, we think our framework provides a better understanding of a wide range of phenomena, from the behavior of firms on competitive markets to agents’ desperate attempts to cope with “ragged (uneven) inflation.” At the normative level, our approach and that of neoclassical economics frequently generates diametrically opposed policy prescriptions. We examine some of these in the next chapter.
As noted above, neoclassical models to which we are referring postulate an underlying objective stochastic process. This assumption is very helpful if one is studying Brownian motion, the decay of radioactive particles, and, to a lesser extent, the weather. Social science surely must also take account of how physical processes affect human decisions. But there is no given, underlying process to which individuals must conform. In an open system, the process itself is largely the product of individuals’ tastes, aspirations and expectations. They help make this “objective” process whatever it is. It is not something entirely apart from human beings, but is partly of their creation. In forming expectations, they are not guessing about the collision of heavenly bodies. They are forming expectations about other individuals forming expectations about them, and so forth, as in the classic Keynesian beauty contest example. Every shift in their expectations changes others’ expectations, thereby changing the data. There may be no unique, objective value to which expectations “ought to” conform. Sometimes there may be such an “objective” value, as in the cycle of weather and crop failures; other times there may not be, as in certain aspects of inflationary expectations.
As in a number of other areas, we are not claiming that there are no applications of stochastic equilibrium in economics. Our objection is to the implied assumption that it is the one, uniquely correct way of understanding informational issues, even if the approach does violence to the phenomena itself. For instance, in their effort to explain learning, neoclassical economists are one step short of “proving” that there is no error in the world. Before falling with them into this abyss, we may want to step back and consider alternatives.
1 “Man wants liberty to become the man he wants to become. He does so precisely because he does not know what man he will want to become in time. Let us remove once and for all the instrumental defense of liberty, the only one that can possibly be derived directly from orthodox economic analysis. Man does not want liberty in order to maximize his utility, or that of the society of which he is a part. He wants liberty to become the man he wants to become” (Buchanan, 1979, p. 112).
2 Kirzner (1973, 1979a) emphasizes entrepreneurial discovery of existing opportunities. Shackle (1979) emphasizes entrepreneurial creativity; entrepreneurial actions themselves make the profit opportunities.
3 Lucas (1975) is one conspicuous exception.
4 Hurwicz (1973) provides a useful bibliography up to the early 1970s. Recent work includes Grossman (1976, 1977) and Grossman and Stiglitz (1976), which also contain partial bibliographies. See also O’Driscoll (1981).
5 The importance of contractual and nonprice constraints in competitive markets arose in Chicago Board of Trade v. United States, 246 U.S. 231 (1918). Apparently the court did not understand that price restraints could, at least in principle, foster competition. We would have to examine this case more carefully to make a definitive judgment. But it does seem to represent an example of our point here. See the interesting discussion in Bork (1978, pp. 41–7).
6 Menger (1963, p. 146) perceived the basic question of all social sciences as, “how can it be that the institutions which serve the common welfare and are extremely significant for its development can come into being without a common will directed towards establishing them?” Marx’s conception of social science was, of course, similar to Menger’s.
7 We submit that Hahn is logically in error in this argument. Arrow and Debreu have given us sufficient conditions for the attainment of competitive equilibrium. Hahn’s argument would be strictly correct only if the Arrow-Debreu conditions were necessary (see the discussion in Chapter 5).
8 See Epstein (1982, p. 55) on the common law system’s concern with private property, individual liberty and corrective justice: “The initial ethical premise of the system is that possession is the root of title, or that taking possession of an unowned thing is prima facie evidence of individual ownership, as lawyers say, in a fee. The rule is stated as a presumption that can be rebutted only by a person whose earlier possession gives rise to a higher title.” Justice depends, then, on the process producing ownership, and not on presumed benefits of the pattern of ownership of property.
9 Two examples come to mind. Milk price supports (and agricultural price supports generally) have not solved or alleviated farm problems, but have perpetuated and even exacerbated them. Current problems of US steel producers have a history going back to nineteenth-century protective tariffs. Each round of protection creates more dependency on existing and future protection.
10 Here and in what follows, one could recast our analysis in terms of “notional” excess demands. We have no desire to quibble over words. Our substantive point would remain: even when there is no effective excess demand, plans may be frustrated. This point is recognized in macroeconomics; see Clower (1965).
11 A more complete property rights analysis could demonstrate that managers of the housing authority would not be as likely to respond to market signals as would entrepreneurs who could appropriate the profits of correct decisions. Our case is not entirely hypothetical. We understand that the extensive and longstanding system of rent controls has inhibited regional growth and mobility in Great Britain.
12 Friedman has argued that “natural selection” explanations and continuous utility maximization explanations are complements rather than substitutes. Thus, successful firms are selected by the market. But as theorists we can suppose that it is “as if” firms are conscious profit-maximizers (Friedman, 1953). Langlois effectively refutes Friedman’s position. He argues that it is dubious methodologically, and, building on Sidney Winter’s work, points out that profit-maximizers who are best adapted to an equilibrium state may be selected out during the equilibration process. Survivors might then not be true “optimizers” (Langlois, 1982a, pp. 19–22).
13 Our anthropomorphic language presents a barrier in discussing evolutionary processes. In speaking of “adopting” a rule, we do not wish to imply conscious choice by the adopter; quite the contrary.
14 Such a reconsideration was started in a special volume of the Journal of Legal Studies, to which both of us contributed (March 1980).
15 None of what we have said here is intended as an implicit criticism of Alchian and Demsetz’s theory of the firm (1972), which we view as dealing with one aspect of the firm. This aspect is not of particular concern to us here.
16 A distinction between the maximizing and evolutionary approaches can be illustrated in this case. If there is a “correct” adaptation to the new environment, rationality and profit-maximization are invoked to ensure that the firm will survive. The evolutionary approach supposes no such thing. For instance, a firm may not survive environmental change even were it true that an infinitely long search would ensure its survival and even prosperity. The evolutionary approach thus implicitly incorporates the passage of real time, which in this case incorporates cash and other relevant constraints. Trial and error discovery takes time, and firms may run out of resources before they discover a profitable response to change. Optimization models do not incorporate real time. Thus constraints such as liquidity constraints, which operate in real time, are ignored. See note 12 (above).