Cost and Choice

A century has elapsed since the subjective-value revolution in economic theory, but the subjective theory of value has not been fully reconciled with the classically derived objective theory. As the notes on the development of the concept of opportunity cost indicate, economists have not drawn carefully the distinction between a predictive or scientific theory and a logical theory of economic interaction. As subsequent chapters will demonstrate, this methodological confusion is the source of pervasive error in applied economics. The treatment and discussion of cost, especially in its relation to choice, provides a usefully specific context within which the more general methodological issues can be examined.

The Predictive Science of Economics

From its classical origins, economics has laid claim to classification as a predictive science. This means that conceptually refutable hypotheses are embodied and that the refutations of these hypotheses can command ultimate recognition by competent professional scientists. To qualify under this restriction, the science must have objective, empirical content. Something must be measurable—at least conceptually—which will allow either the corroboration or the refutation of the central propositions. The basic elements of economic theory are, of course, the actions of human beings. The science consists in the efforts to predict the effects on human behavior induced by specific changes in environment. Operationality dictates that the behavioral responses be objectively measurable.

Consider the elementary proposition that relative prices rise when relative costs increase, subject to the standard ceteris paribus qualifications. This proposition is derived from the postulate that individuals behave "economically," that they seek to minimize "costs" and to maximize "benefits" or "revenues." But this postulate remains empirically empty until specific descriptive content is given to "costs" and to "benefits" or to "revenues." The behavioral postulate is that of economic man. If this is dropped, the predictions are drained of power.

It is important that the limitations as well as the strengths of this predictive theory be noted. There is no implied presumption that men should behave economically. Properly interpreted, the theory's claim is limited to making predictions on the "as if" assumption that men do so behave in some average or representative sense. The motivational assumption is vital in that this allows the scientist to use the objectively observable magnitudes of money cost and money revenue streams as representations of the subjectively evaluated alternatives of choice in individuals' behavior patterns. As experience has shown, considerable success has been achieved in this genuinely scientific theory of economic behavior. Men do behave economically in sufficient degree to allow many predictions to be corroborated. But the oversight of the basic limitations of the predictive theory has led to major error in normative application.

The orthodox neoclassical model of market process is one in which the acting units behave economically. To the extent that the model approximates reality, the objectively observed cost and revenue streams accurately represent the dimensionally different and subjectively evaluated alternatives among which choices are actually made by individuals. To this same extent, and only to this extent, specific relationships among costs, as objectively observed, and prices, as objectively observed, can be predicted to describe the equilibrium toward which the whole process converges. Note especially that these relationships, these conditions of equilibrium, are themselves derivative predictions based on the motivational postulates of the model. For example, equalities between prices and marginal costs, as objectively observed quantities in fully competitive equilibrium, are inferred predictions which depend on the behavioral assumptions upon which the whole theory is constructed. These equalities have no normative significance, and they have no direct relationship to allocational efficiency. The methodological muddle in modern economics is perhaps most clearly revealed by the unwarranted crossing of the bridge between the inferential predictions of the genuinely scientific theory and the normative conclusions about efficiency that are so often drawn.

This may be illustrated in a variation, similar to that used by Knight in his papers previously cited, upon the simplest of models, Adam Smith's deer and beaver model. The objective conditions of the model remain as before. One day's labor is required to kill a deer and two days' labor to kill a beaver. Objectively measurable costs are in a one-for-two ratio. The prediction is made that exchange values will settle in a two-for-one ratio, a ratio which will be described by the equalities between marginal costs and prices. Let us suppose, however, that the relative price ratio exhibits no tendency to move toward the equilibrium level that is predicted; prices do not tend toward equality with marginal costs. Only the most naive of welfare economists should conclude from this that the allocation of resources is inefficient. Instead he probably would introduce, as Knight did, the possibility that hunters, generally, may have some nonpecuniary or noneconomic arguments in their utility functions. Marginal costs, as these affect choice behavior, may then not be the same as the simply observed ratios of labor time. The welfare economist, presuming only that the market is competitively organized, then concludes that the price-marginal cost equalities are satisfied in the equilibrium that he observes, despite the variations from objectively based predictions.

In resorting to noneconomic arguments in the utility function to rectify his falsified predictions, however, the economist has shifted the whole analysis from a predictive to a nonpredictive and purely logical theory. Objectively observable cost-revenue streams cannot serve as surrogates for the subjectively evaluated alternatives in which noneconomic elements are influential. The inferred predictions of relationships that characterize equilibrium positions are falsified. No potential gains-from-trade are indicated if these predictions are not fulfilled. No welfare improvements could, therefore, be expected from rearrangements designed to ensure that the predicted relationships will be produced.

The implications of all this for modern welfare economics could scarcely be underestimated. My argument suggests that almost all of this subdiscipline has been based on simple methodological confusion. It has converted predictive propositions into allocative norms which have then been used to make policy proposals. Some of the more specific instances of this confusion will be discussed in subsequent chapters.

In one sense it might be said that the neoclassical economist has succumbed to the temptation to make his whole theory more general than its methodology warrants. This temptation has been increased by the parallel, and equally confused, logical theory of economic choice, which itself is completely general but which lacks predictive content. This purely logical theory, sharply distinct from the classical in its predictive implications, finds its origins in the subjective-value theorists, but its more explicit sources are Wicksteed, the later Austrians, and the economists associated with the London School of Economics. In full flower, this is the "subjectivist" economics espoused by Hayek and Mises to which I earlier made reference. Some reconciliation between the genuinely scientific theory of economic behavior and the pure logic of choice is required. The achievement of this reconciliation is one of the major purposes of this exploratory study in which the notion of opportunity cost becomes the analytical coupling device.

Cost in the Predictive Theory

If we remain strictly within the predictive science of economics, cost can be considered to be properly defined in most of the modern textbooks, and there is little need to elaborate on these standard definitions. This is the cost of the familiar textbook diagrams, the objectively identifiable magnitude that is minimized. It is the market value of the alternate product that might be produced by rational reallocation of resource inputs to uses other than that observed. This market value is reflected in the market prices for resource units; hence, cost is measured directly by prospective money outlays.

For whom is this cost relevant? This becomes a critically important question. Cost as just defined is faced in the strict sense only by the automaton, the pure economic man, who inhabits the scientist's model. It is the behavior-inhibiting element that is plugged into the purely mechanistic market model. The conversion of objective data reflecting prospective money outlays into the subjective evaluations made by real-world decision-makers is of no concern to the predictive theorist. In the strict sense, this theory is not a theory of choice at all. Individuals do not choose; they behave predictably in response to objectively measurable changes in their environment.

Cost in a Theory of Choice

The distinction between the concept of cost in the predictive context, as sketched out above, and the concept of cost in a more general theory of choice, as articulated—though not fully—in Chapter 2, can best be emphasized at this point by elaborating this second concept. The essential element in this concept is the direct relationship between cost and the act of choice, a relationship that does not exist in the neoclassical predictive theory. In the London-Austrian conception, by contrast, cost becomes the negative side of any decision, the obstacle that must be got over before one alternative is selected. Cost is that which the decision-taker sacrifices or gives up when he makes a choice. It consists in his own evaluation of the enjoyment or utility that he anticipates having to forego as a result of selection among alternative courses of action.

The following specific implications emerge from this choice-bound conception of cost:

In a theory of choice, cost must be reckoned in a utility dimension. In the orthodox predictive theory, however, cost is reckoned in a commodity dimension. This distinction can be applied to each of the six attributes listed above. In a theory of choice, cost represents the anticipated utility loss upon sacrifice of a rejected alternative. Because utility functions are necessarily personal, cost is tied directly to the chooser and cannot exist independently of him. In the predictive theory of economic behavior, the cost of producing one "good" is the amount of another that could be produced instead. Cost, as such, exists independently of the choice process, and there is no direct linkage between choosing and bearing cost.

Cost, then, is purely subjective in any theory of choice, whereas cost is objective in any theory that involves genuine prediction. If cost is to influence choice, it must be based on anticipations; it cannot be based on realized experience—at least directly. On the other hand, once cost is divorced from choice, it is a physical concept; it becomes irrelevant whether cost is measured before, at the moment of, or after actual commitment. In the Smithian model, the cost of a beaver is two deer, and this holds so long as the postulated physical relationships hold; there is no point in distinguishing ex ante and ex post measurements. Because of the technological or physical nature of cost, derived from the transformation function in commodity space, the alternatives facing the actor can be "costed" by an external observer. There is no need for the observer to psychoanalyze the hunter in Smith's model. And the problem of dating does not arise in the objective definition implicit in the predictive theory. On the other hand, cost must be precisely dated in any theory of genuine choice because it is tied to the moment of choice as such. Before choice, cost exists as a subjective experience. After choice, cost vanishes in this sense. What happens to the chooser after he has chosen remains to be considered.

Choice-Influencing and Choice-Influenced Cost

The six attributes of cost listed above are relevant to any particular choice. If cost is to be influential in affecting that choice, it must be defined in terms of these attributes. Nonetheless, we must also recognize that choice has consequences: things happen as a result of decisions. Having committed himself to one course of action rather than another and having presumably made some rational estimation of the costs that this would embody, the individual "suffers" the consequences. He may not regret his prior decision, but, at the same time, he may undergo "pain" or "sacrifice" when he is required to reduce his utility levels. Whether or not choices were rightly or wrongly made has little direct relevance to the existence of this choice-influenced "cost."

It is this "cost" consequent to choice which helps to create a part of the confusion between cost in the predictive theory and cost in the logic of choice. That which happens after choice is made is what economists seem to be talking about when they draw their cost curves on the blackboards and what accountants seem to be concerning themselves with. It is necessary that this choice-influenced "cost" be more thoroughly examined.

If we bend linguistic principle to accommodate orthodox usage here, it seems best to allow the word "cost" to be used in these two quite separate senses within any theory of choice, while continuing to employ this same term in its single usage in the predictive theory of economic behavior. Hence, we have both "choice-influencing cost" and "choice-influenced cost" within the theory of choice and defined in utility space, and we have "objective cost," defined strictly in the commodity space of the predictive theory. Let us now neglect the objective cost of the predictive science and concentrate on choice-influencing and choice-influenced cost. Every choice involves both of these. First, there is the genuine obstacle to choice, the opportunity cost that was central to the thoughts of those economists whose contributions were summarized in Chapter 2. Second, there are the utility losses that are always consequent to choice having been made, whether these be suffered by the chooser or by third parties and whether there may or may not be objectively measurable surrogates for these losses, e.g., payouts. These losses are the result of decision and never its cause. In the one case, cost inhibits choice; in the other, cost results from choice. These concepts of cost can be discussed more fully in connection with several of the familiar, but ultimately misleading distinctions.

Opportunity Cost and Real Cost

Strictly speaking, only choice-influencing cost represents an evaluation of sacrificed "opportunities." It might therefore be reasonable to limit the term opportunity cost to this conception and to invent other descriptive appellations to refer both to choice-influenced cost in a logic of choice and to the objective cost of the predictive theory. In a more general sense, however, any one of the three conceptions may be meaningfully treated in opportunity-cost terms. In the orthodox price-theory conception where cost is measured objectively by money outlays, it is helpful, for explanatory purposes, to equate these outlays to the values that members of society place on the alternate end products that might have been produced by the same outlays differently directed. In a certain ambiguous sense, therefore, cost here does reflect "opportunities lost." But it is noteworthy that the "opportunities lost" in this context more accurately reflect the value of potential alternatives as judged by others rather than by the chooser himself.

The notion of "opportunities lost" can also be applied to the results of choice, or to choice-influenced cost. Here the concept is tied to the choice and the opportunities represent those things that "might have been," as these are viewed after decision has been made. Given this hindsight, alternatives can be viewed differently than they were viewed before a commitment was made. Within a before-choice, or choice-influencing context, the opportunities lost are those that "might be," as considered and evaluated at the moment of choice itself and as reflected in the presently anticipated value of utility losses expected to be incurred. Within the post-choice or choice-influenced context, by comparison, the opportunities lost are those that might have been enjoyed, as these are reflected in experienced utility losses or sacrifices. There can be an important psychological difference in the utility losses involved in choice-influencing and in choice-influenced cost. At the moment of choice itself, cost is the chooser's evaluation of the anticipated enjoyments that he must give up once commitment is made; it is also that which he can avoid by choosing another alternative. Cost in this setting must be and remain a purely mental event. The chooser's utility is reduced only in the sense that it is functionally dependent on expected utility in post-decision time periods. After the choice is made, cost may still reflect the evaluation of enjoyments that were sacrificed and cost remains a mental event, but there is more to it than this. Among the experiences that might have been avoided may be those requiring an explicit submission to pain, to suffering, to deprivation, in some physically relevant meaning of the terms. Having made a charge-account commitment, the buyer must pay his bills when they come due. Despite his possibly rational anticipation of this cost at the moment of choice, he still suffers pain when these bills must be met. This purely physical exposure to negative choice-determined effects enters into his subjective evaluation of the alternative that might have been. In a certain sense, therefore, the nature of cost is different in the choice-influencing and the choice-influenced contexts, although both remain in utility space.

So long as we treat cost in either a cost-influencing or a choice-influenced sense, that is to say, so long as we remain with the theory of choice itself, we are closer to the classical notion of real cost than is the neoclassical conception. Either as an obstacle to choice or as an undesirable consequence of choice, cost represents utility loss. In relatively sharp contrast, when cost is divorced from the choice process, as in the neoclassical predictive setting, there is nothing "real" about it. No pain, suffering, or utility loss is involved. This seems to have been the basis for Knight's conceptual distinction between opportunity cost and real cost which led him to say that "... all references to 'sacrifices' (should be) simply omitted."^*48

The Subjectivity of Sunk Costs

I have referred to cost in any logic of choice as "subjective" and to cost in any predictive science as "objective." In a preliminary discussion in another volume,^*49 I employed the subjective-objective terminology ambiguously, because I failed at that time to distinguish the separate dimensions of cost within these related but quite different settings. Cost in the predictive models of economics must be objective. If cost is introduced into a logic of choice, however, it is obviously subjective. This has been repeatedly emphasized by some of the LSE scholars whose works were mentioned earlier, and notably by G. F. Thirlby.

The consequences of choice, the results of decision, enter the individual's experiences as subjectively valued events, even though, as noted, there may also be physical repercussions to the decision. If a commitment is made and things happen, such happenings affect the individual's utility—quite independently of the fact that they cannot be avoided. The individual suffers utility loss as a consequence of a prior decision even if, on balance, the decision was itself fully rational. This suffering is a subjective event whether it be regrets at what might have been or pain at what is. Strictly speaking, only this subjective choice-determined cost squares fully with the economist's concept of "sunk cost" or with Jevons' "bygones." Since choice has been made, this cost is irrelevant excepting insofar as the experience may modify anticipations about choice alternatives in the future.

In this choice-influenced sense, cost is related to choice ex post, but it is not personally tied to the chooser or decision-maker. This is an important distinction that has contributed its own share in the general cost-theory confusion. As we noted earlier in connection with the first-listed attribute of choice-influencing cost, the opportunity cost must be borne by the decision-taker himself if choice is to be affected at all. Indeed, in this context, cost can be borne only by the chooser; the whole notion becomes meaningless otherwise. By contrast, the consequences of choice—the utility losses suffered as a result of a decision—need not be endured only by the chooser. Because these consequences are always realized after choice, the chooser himself may be considered a different person once the consequences of choice are realized. Even when this is neglected, however, there remains no formal connection between the person taking a decision and the person or persons who suffer utility losses as a result. Those who "bear the burden"—even though bearing this burden is a subjective experience—need not be those who undergo the "agony of choosing."

Cost and Equilibrium

Given presumably objective data drawn from nonutility space, neoclassical economics makes predictions about properties of the equilibrium relationships that will tend to be established behaviorally by participants in the market-interaction process. To what extent does the equilibrium emphasis allow for some reconciliation between the two cost conceptions, between the objective cost of the predictive science and the purely subjective cost in the logic of choice? Do objectively measurable outlays reflect foregone opportunities only under conditions of full equilibrium?

If the whole economy is not operating at full competitive equilibrium, profits-losses may occur and, hence, observed outlays cannot be taken to reflect foregone opportunities of the actual decision-takers in any general setting. In full equilibrium, on the other hand, observed outlays directly represent the maximum contribution of resources in different uses. Therefore, to the extent that decision-takers behave economically, the observed outlays reflect genuine "opportunity costs," even if somewhat indirectly. The apparent reconciliation here verges on the tautological, however, since the whole purpose of the economic theory in which cost is relevant is to demonstrate how choices made in nonequilibrium settings will generate shifts toward equilibrium. And choices in disequilibrium must be informed by opportunity costs that cannot, even indirectly, be represented by measured outlays. In disequilibrium, the opportunity costs involved in taking the "wrong" decision must include the profits foregone in the rejection of the alternative course of action.

Marginalism provides only a partial rescue here. If an individual behaves economically, and if no profit opportunities exist elsewhere in the whole system, and if all decisions can and must be made marginally, the marginal-cost derivation of orthodox theory can be taken to represent the genuine "opportunity cost" of an output decision. This means that all choices are made at equilibrium in the short-term planning context where output decisions within the firm are conceptually divorced from the rest of the economy. It is essential, however, for each of the qualifying conditions to be satisfied if measured marginal cost is to be employed as an objective representation of the subjective element that actually enters the individual's choice calculus.

If, on the other hand, the individual incorporates nonpecuniary or noneconomic considerations in his decision, if there are profits to be secured elsewhere than in the activity in question, if discrete rather than marginal adjustments are possible, then objectively measured marginal outlay is not a veritable expression of genuine opportunity cost, because these "ifs" may represent inhibitions upon choice behavior which are not susceptible to objective measurement.

There is necessarily a close correlation between the relevance of objectively measured costs for a theory of choice in either long-term or short-term equilibrium and the presence or absence of uncertainty. In the face of uncertainty, the evaluation of alternatives by the actual decision-taker may differ from the evaluations of any external observer, even if the qualifying conditions are met. The inherent subjectivity of cost in any theory of choice reasserts itself here.

The equilibrium concepts introduced in this section up to this point are those of the predictive neoclassical theory. This implies that descriptions of equilibrium take the form of objectively defined relationships among variables in nonutility dimensions. Prices must bear specific relationships to costs. If we are content to remain within a more general, but ultimately nonpredictive and purely logical theory of economic choice, the concept of equilibrium may be modified. The equilibrium of the "subjectivist economics" espoused by Hayek is described behaviorally. It is attained when the plans of participants in the economic interaction process are mutually satisfied. Although prices continue in this equilibrium to bear some relationship to costs, such costs carry no objective meaning and cannot, therefore, be employed as criteria for determining prices in some welfare or efficiency sense.