The Formulation of Rational Choice

by Amartya Sen
The Formulation of Rational Choice
Amartya Sen
The American Economic Review
Start Page: 
End Page: 
Select license: 
Select License

Consider the description of a "rational" person in traditional economic theory, or standard game theory. As Roger Myerson (1991 p. 2) lucidly presents the prevailing view of the matter, "a decision-maker is rational if he makes decisions consistently in pursuit of his own objectives." Myerson goes on to explain that "we assume that each player's objective is to maximize the expected value of his own payoff, which is measured in some utility scale." The payoff function is a real-valued representation of the person's preferences over the outcomes. Rationality is seen as intelligently maximiz- ing such a payoff function, using all the available instruments, subject to feasibility.'

This canonical formulation of "rational choice" in standard theory is critically scru- tinized in this note, identifying distinct inad- equacies in different contexts. The need for context-dependent parametric variations in the characterization of rational choice is outlined. The general problems are illustrated in the specific case of monitoring and work motivation.

I. Menu Dependence

In the "utility scale," person i's payoff U,(x) is seen as a function of the outcome x. This is the way that choice behavior is typi- cally axiomatized. This formulation has some difficulty in accommodating "social" behavior such as not picking the last apple when the fruit basket goes around, or not grabbing the uniquely largest slice of cake, since the valuation of outcomes here is

*Department of Economics, Littauer Center, Har- vard University, Cambridge, MA 02138. For helpful discussions, I am grateful to George Akerlof, Kenneth Arrow, Kaushik Basu, Samuel Brittan, Jean Drkze, Robert Nozick, and Emma Rothschild. I am indebted to the National Science Foundation for research sup- port.

'See David Kreps (1988) for a classic exploration of this approach and its implications.

"menu-dependent." Menu-dependence can

arise also from many other types of behav-

ior involving more complex considerations

of ethics and epistemology (see Sen, 1993a).

There is, however, nothing in the idea of

payoffs that prevents a payoff function from

having as arguments not just the actual out-

comes (x) that will emerge, but also the

anticipated set of alternative outcomes (S)

-"the opportunity set" or "the menu,"

which the person reckons she can have

through different choices of her strategy

(given her presumptions about the others'


Menu-dependent behavior, while socially important, would tend to violate some commonly used properties of internal correspondence of choice, or "internal con- sistency of choice" (as it is often called- begging a question about what form consis- tency must take). An example of such an internal-consistency condition is the follow- ing: if x is chosen from a set T and belongs to a subset S of T, then it must be choosable from S. This is violated by menu-dependent behavior (e.g., an apple that is nicely choos- able may cease to be so if the set of alterna- tives is reduced to make it the last apple in the basket). "Binariness" or "rationalizabil- ity" of choice, as they are typically defined (i.e., over sets of outcomes only), would be violated too. With that would go real-valued representation of preference orderings over the outcomes. This does not, however, re- quire abandoning maximization as an oper- ation. Instead, the function to be maximized would have to be defined over the pair (x, S) of outcomes and menus, rather than over outcomes x only.

11. Multiple Interpretations and Preference

Moving from the form to the interpreta- tion of preference, one can ask what the preference ordering stands for. ~~~~f~~ as

the preference ordering is the "support" for predicting a person's behavior, it must be deeply relevant for the person's choice (seen


as maximization based on that preference ordering). I will call it the choice-salience interpretation of preference. But as the term "payoff" suggests, there is also some pre- sumption that the preference relates to the person's own gain from the outcome in question. In economic theory, in particular, the goal to be maximized, reflected in the preference ordering, has been typically seen as the person's own well-being. I will call it the well-being interpretation of preference.

Both interpretations-based respectively on choice-salience and well-being-have some plausibility, but they are not the same. To take them to be the same involves a substantive empirical assumption that a per- son's choices must be governed by the maxi- mization of her own well-being. Many cen- tral results in economic theory depend on the presumed congruence of well-being and choice-salience, such as the Arrow-Debreu "fundamental theorem of welfare economics" (Kenneth Arrow, 1951; Gerard Debreu, 1959), which relates competitive equilibria (seeing preference in terms of choice salience) with Pareto efficiency (see- ing preference in terms of individual wel- fare~).~

A divergence between choice and well- being can easily arise when behavior is in- fluenced by some motivation other than the pursuit of one's own interest or welfare (e.g., through a sense of commitment, or respect for duty). The issue here is not that of altruism as such. Individual well-being can certainly take good note of a person's sympathy (or antipathy) for others (see e.g., Gary Becker, 1976, 1981). Rather, the ques- tion is whether the influence of other-regarding concerns is restricted only to the extent to which they enter the person's own welfare function, either directly (e.g., suf- fering at others' pain) or indirectly (e.g., feeling disappointed or dejected when at- tempts to help others-for whatever motive -are fr~strated).~

o ow ever, by using the choice-salience interpreta- tion only, competitive equilibria can be seen to be "weakly efficient" in freedoms, rather than welfares (Se:, 1993b).

The question here corresponds to the distinction between "sympathy" and "commitment" in Sen (1977).

Can people have reason to act other than the extent to which their own well-being is directly or indirectly enhanced by it? The belief that individuals, as free and rational agents, are capable of going beyond the exclusive pursuit of their respective well- being was strongly asserted by classical au- thors such as Adam Smith (1759, 1776). Smith, in particular, distinguished actions motivated by "prudence" (taking into account indirect as well as direct advantages) from those influenced by "sympathy," "generosity," and "public spirit," to each of which he gave a distinct and differentiated role.4 The variability of motivations was ex- plicitly accepted also by the pioneers of utility theory in economics (such as John Stuart Mill, William Stanley Jevons, Francis

Y. Edgeworth, and Alfred Marshall), though this issue is rarely addressed in modern economic theory.

111. Action Ethics and Behavioral Norms

Another important qualification concerns the relationship between the evaluation of actions and the assessment of consequences. In game-theoretic formulations, a person has a set of clearly defined individual strate- gies, and the exercise of maximization con- cerns picking a maximal strategy from that set (including randomization over these strategies). There may, of course, be com- plicated strategic considerations related to such behavior, which can be particularly important when the game is repeated (see David Kreps, 1990; Myerson, 1991; Drew Fudenberg and Jean Tirole, 1992). But there is, in this framework, no further ethical con- sideration in the choice of actions and strategies (i.e., "further" to what is already incorporated in the person's objective func- tion). Given the feasibility of a strategy, the only issue that influences its choice by a person is its direct and indirect effectiveness in promoting her objective function.

4~eeSmith (1790 [Section IV, Chapter 21); in D. D. Raphael and P. L. Macfie's (1975) edition, pp. 187-93. On Smith's concerns see, among others, Raphael and Macfie's introduction, Sen (19871, Patricia Werhane (19911, and Emma Rothschild (1992).


This formulation of the choice problem contrasts with Adam Smith's analysis of ap- propriate actions in the context of public policy: "In these and in all other cases of this kind, our admiration is not so much founded upon the utility, as upon the unex- pected, and on that account the great, the noble, and the exalted property of such ac- tion~."~There is, in this view, a distinct assessment of actions that is not identified with the evaluation of the useful consequences (what Smith calls the "utility") of that action seen on its own (though Smith does discuss the good consequences of peo- ple in general following such rules of behav- ior).

An important case of such action-judg- ments concerns situations of social interde- pendence such that every member of the group does better by treating as "the unit of choice" their joint strategy. An example is the unrepeated prisoner's dilemma. Choos- ing to be "cooperative" in such a situation (i.e., not confessing) would involve selecting an individual strategy that is strictly domi- nated, and it would entail a violation of standard formulations of individual rational choice.

Yet if some action-ethic, like the one that Smith describes (he relates it, incidentally, to acting according to the view of an "im- partial spectator"), makes the people involved impose restraints on their own instrumental choice, they could all end up having more fulfilment of their unmodiJied preferences6 This line of reasoning relates also to the Kantian requirements of action morality; for example, "Act always on such a maxim as thou canst at the same time will

'smith (1790 [Section IV, Chapter 21); in Raphael and Macfie's (1975) edition, p. 192; italics added.

6~heproblem being discussed here, which arises most unambiguously in the case of unrepeated prisoner's dilemma (though it can coexist with some types of repetitions as well), must be distinguished from the issue of cooperation in repeated prisoner's dilemma. An incentive for cooperation can be directly derived just from the pursuit of one's own objective function if the prisoner's dilemma is repeated indefinitely, and to some extent, even when the repetitions are finite but combined with some specific epistemic conditions (see Fudenberg and Tirole, 1992).

to be a universal law."' Obviously, confess- ing oneself is not such an act in the pris- oner's dilemma, since neither prisoner wants that behavior to be "universal." The Smith- ian-Kantian selfimposed restraints differ from constraints given from out~ide.~

IV. Work Motivation and Behavioral Modes

How does all this relate to the problem of monitoring work performance? One of the general issues to emerge from the preceding discussion is the need to take what one may call a "parametric" view of rational choice. Do the possibilities explored in earlier sec- tions relate to any observed regularities in work, motivation, and monitoring? There are some suggestive empirical data in that direction.

First, there is considerable evidence that the view that is taken of work responsibili- ties varies greatly from one society to another. These cross-section comparisons can be supplemented by evidence of intertem- poral change as a result of movements of social tradition, an old subject discussed, for example, by Karl Marx, Richard Henry Tawney, and Max Weber, among many oth- ers. Social traditions can indeed be relevant in the formation of objective functions as well as in the specifics of action ethics.

Recently, attention has been drawn to the special sense of collective responsibility that

'~mmanuel Kant (1785); in T. K. Abbott's (1907) translation, p. 66. For discussions of this and related issues, see Sen (1973, 1977, 19871, Derek Parfit (1984), and Susan Hurley (1989). On connected matters, see also John Harsanyi (19761, Albert Hirschman (19821, George Akerlof (1984), and Robert Frank (19851, among many other contributions.

'1t is, of course, possible to translate the impact of instrumental constraints-even self-imposed ones-into reconstructed "objective functions" through such devices as Lagrangean multipliers. But then the inter- pretation of the "objectives" would have to be corre- spondingly different and contingent (in the way the values of Lagrangean multipliers must be). The crucial issue here is not whether the "maximizing format" can continue to work in the Smithian-Kantian case (it cer- tainly can), but whether there are reasonable arguments that support self-imposition of action norms, with a deliberate restriction of instrumental possibili- ties.

apparently prevails more abundantly in Japanese industrial relations than in the West, though the explanation given for this agreed phenomenon has varied between distinct theories (see e.g., Michio Morishima, 1982; Ronald Dore, 1987; Masahiko Aoki, 1989). These theories include alterna- tive attempts at explaining why, as The Wall Street Journal put it (stretching a point somewhat), Japan is "the only communist nation that works" (30 January 1989, p. 1).

Second, one of the observed phenomena in behavioral norms is the tendency to imi- tate others, much commented on recently in the context of explanations offered for cor- rupt activities in business and politics in Italy: "I was not alone in doing it" (see Camera dei deputati, Roma, 1993). This type of reasoning, on the one hand, points to the need to bring in the influence of established norms of behavior in explaining what actions will or will not be chosen, and on the other, it indicates that individual choices may often take a menu-dependent form (similar to not picking the last apple). A person may refrain from seizing a unique opportunity of breaking an implicit moral code and yet be willing enough to break that code if there are many such opportuni- ties (either because the departures might then seem more "usual," or because the person expects that some others would pos- sibly grab one or more of the remaining opportunities).

Third, based on a widely observed phe- nomenon in both rich and poor countries, indicating the coexistence of relatively high wages and involuntary unemployment at those wages, there has been an extensive literature on "efficiency-wage models." As Akerlof and Janet Yellen (1986 p. 2) have discussed, the benefits of higher wage pay- ments (above market-clearing wages) can be of different types, including "reduced shirk- ing of work due to a higher cost of job loss, lower turnover, improvement in the average quality of job applicants, and improved morale." These are very diverse benefits9

90n this and related issues, see also Akerlof (1982, 1984).

Whereas connecting the higher cost of job loss to reduced shirking may be fairly ade- quately done through the presumption of self-interested behavior, there would be a more complex connection between "morale" and actual work effort.''

On the basis of a "cross-national" study of monitoring and supervision, David Gordon (1994) has presented an interesting empirical analysis of variations in "the bu- reaucratic burden" among 16 advanced economies in the OECD. He defines that "burden" as the relative proportion of "administrative and managerial workers" among total employees in nonfarm occupa- tions (though he also considers variations of this measure). That "burden," for 1980, ranges between 2.5 percent in Sweden to

11.5 percent in the United States, which is the highest in the list (with the next three positions being taken by the other English- speaking countries, to wit, the United King- dom, Canada and Australia). In the longer, unpublished version of Gordon's paper (available from Gordon upon request), he considers possible explanations of these variations, and while he finds a general con- nection here with the use of higher "ef- ficiency wages" as an alternative to supervi- sion, he also notes that the differences in labor relations appear to reflect structur- al differences in institutions, norms, and culture."

1°1ndeed, as Robert Solow (1990 p. 22) has ob- served, "wage rates and jobs are not exactly like other prices and quantities," and "they are much more deeply involved in the way people see themselves, think about their social status, and evaluate whether they are get- ting a fair shake out of society."

"To comment on a particular issue, the position of Japan in Gordon's (1994) study fits uneasily with the wide literature on the hold of work ethics and shared responsibilities in that country. Japan's "burden" of

4.4 percent is closer to Sweden's (2.5 percent) than to that of the United States (11.5 percent), but it is still a very middling performer (with the tenth lowest "burden" in a list of 16). It appears, however, that given the nature of "life-long" employment relations in Japan, a significant proportion of workers, as they mature in the respective firms, are rewarded by being given "administrative and managerial" positions with- out their being actually involved in monitoring or su- pervising other employees (on this see Dore [1993]).


The wide intercountry variations in be- havior patterns and in the need for monitor- ing certainly do suggest the case for more parametric formulations of rational choice, with varying relationships (i) between choice-salient objectives and personal well- being and (ii) between individual and social instrumentalities.12

V. Concluding Remarks

While the reformulations of rational choice discussed here tend to make more room for "less self-centered" views of ratio- nality, I should avoid giving the impression that more "social" behavior must invariably be more conducive to identifiable require- ments of social justice. Some of the nastiest things in the world happen as a result of "selfless" pursuit of objectives far removed from one's own well-being but also from the well-being and freedoms of others (such as communal rioting or racist violence).13

Also, while social behavior may contribute to group cooperation in such classic cases of atomistic failure as the prisoner's dilemma, that dilemma is very limited in its formulation, in that there is a uniquely Pareto-superior outcome over the atomistic result, so that the question of justice in the division of gains between the parties is sim- ply avoided. When there are alternative di- visional arrangements, some could be taken to be acceptable on grounds that may be far removed from justice (such as historical convention).l4

o ow ever, on the basis of Gordon's (1994) observa- tions alone, it is not yet possible to discriminate be- tween (i) variations within the standard model that would be brought about by different social and economic circumstances and (ii) variations of the basic model itself in the lines outlined earlier in this paper. Gordon's empirical findings strongly suggest the need for follow-up analysis to explore this type of discrimi- nation and also the distinction between the different ways in which "social behavior" may work (e.g., be- tween broadening the objective function and the self- imposition of action constraints).

130n this see Sen (1977) and particularly Samuel Brifian (1993).

The issue of "salience" in coordination or bargaining problems has to be distinguished from that of social optimality or justice; on this see Thomas Schelling (1960).

In those circumstances, the formulation of "selfless" objectives and adherence to "cohesive" action-ethics can, in many cases, help to sustain extensive inequities. For ex- ample, there is much evidence that valuing the cohesion of the family in many gender- unequal societies can actually serve to per- petuate the unequal position of women, by making women themselves give priority to the alleged interest of the family over per- sonal well-being, and also by inducing a view of "jointness" of efforts within the family to sustain unequal provision^.^^ Similar problems arise in many other areas of group cooperation, even in industrial rela- tions.

The need for making more room for other-regarding concerns in the formulation of rational choice would seem to be strong on grounds of better description and greater explanatory and predictive power. Despite their ethical origin, however, the social ef- fects of these concerns need not necessarily be conducive to equity or justice, in any obvious sense. That is a separate issue which calls for a more context-dependent answer.


Akerlof, George A. "Labor Contracts as Par- tial Gift Exchanges." Quarterly Journal of Economics, November 1982, 97(4), pp. 543-69.

.An economic theorist's book of tales.

Cambridge: Cambridge University Press, 1984. Akerlof, George A. and Yellen, Janet L., eds.

Eficiency wage models of the labor mar- ket. Cambridge: Cambridge University Press, 1986.

Aoki, Masahiko. Information, incentive and bargaining in the Japanese economy. Cambridge: Cambridge University Press, 1989.

Arrow, Kenneth J. "An Extension of the Ba- sic Theorems of Classical Welfare Eco- nomics," in J. Neyman, ed., Proceedings of the second Berkeley symposium on mathematical statistics and probability.

'3have tried to discuss this issue in Sen (1990).

Berkeley: University of California Press, 1951, pp. 507-32.

Becker, Gary S. The economic approach to human behavior. Chicago: University of Chicago Press, 1976. .A treatise on the family. Cambridge, MA: Harvard University Press, 1981.

Brittan, Samuel. "Ethics and Economics." Presidential Address to the British Asso- ciation for the Advancement of Science, 1993; in Proceedings of the Annual Con- ference (forthcoming).

Camera dei deputati, Roma. Economia e crimi- nulit;. Proceedings of the May 1993 con- ference of the Anti-Mafia Commission of the Italian Parliament. Rome: Camera dei deputati.

Debreu, Gerard. Theory of value. New York: Wiley, 1959.

Dore, Ronald. Taking Japan seriously: A Confucian perspective on leading economic is- sues. Stanford, CA: Stanford University Press, 1987. . "Involvement and Trust." Mimeo, Massachusetts Institute of Technology, 1993.

Frank, Robert H. Choosing the right pond. New York: Oxford University Press, 1985.

Fudenberg, Drew and Tirole, Jean. Game the- ory. Cambridge, MA: MIT Press, 1992.

Gordon, David M. "Bosses of Different Stripes: A Cross-National Perspective on Monitoring and Supervision." American Economic Reuiew, May 1994 (Papers and Proceedings), 84(2), pp. 375-79.

Harsanyi, John C. Essays in ethics, social be- havior, and scientijic explanation. Dordrecht: Reidel, 1976.

Hirschman, Albert 0. Shifting inuoluements. Princeton, NJ: Princeton University Press, 1982.

Hurley, Susan. Natural reasons. Oxford: Clarendon, 1989.

Kant, Immanuel. Grundlegung zur Metaphysik der Sitten. Riga: J. F. Bartknoch, 1785; English translation [T. K. Abbott, transl.], Fundamental principles of the metaphysics of ethics, 3rd Ed. London: Longmans, 1907.

Kreps, David. Notes on the theory of choice. Boulder, CO: Westview, 1988. . Game theory and economic mod-

elling. Oxford: Clarendon, 1990.

Morishima, Michio. Why has Japan 'succeeded'? Western technology and Japanese ethos. Cambridge: Cambridge University Press, 1982.

Myerson, Roger B. Game theory. Cambridge, MA: Harvard University Press, 1991. Parfit, Derek. Reasons and persons. Oxford: Clarendon, 1984.

Rothschild, Emma. "Adam Smith and Con- servative Economics." Economic History Review, February 1992, 45(1), pp. 74-96.

Schelling, Thomas C. The strategy of conflict. Cambridge, MA: Harvard University Press, 1960.

Sen, Amartya K. "Behaviour and the Concept of Preference." Economica, August 1973, 40(159), pp. 241-59; reprinted in Choice, welfare, and measurement. Cambridge, MA: MIT Press, 1982. ."Rational Fools: A Critique of the Behavioural Foundations of Economic Theory." Philosophy and Public Affairs, Summer 1977, 6(4), pp. 317-44; reprinted in Choice, welfare and measurement. Cambridge, MA: MIT Press, 1982. . On ethics and economics. Oxford: Blackwell, 1987. ."Gender and Cooperative Conflict," in Irene Tinker, ed., Persistent inequalities. New York: Oxford University Press, 1990, pp. 123-49. . "Internal Consistency of Choice." Econometrica, May 1993a, 61(3), pp. 495-522. . "Markets and Freedoms." Oxford Economic Papers, October 1993b, 45(4), pp. 519-41.

Solow, Robert M. The labor market as a social institution. Oxford: Blackwell, 1990.

Smith, Adam. The theory of moral sentiments. London: T. Cadell, 1759; revised Ed., 1790; reprinted in D. D. Raphael and A. L. Macfie, eds. Oxford: Clarendon, 1975. .An inquiry into the nature and causes of the wealth of nations. London: W. Stra- han and T. Cadell, 1776; reprinted, R. H. Campbell and A. S. Skinner, eds. Oxford: Clarendon, 1976.

Werhane, Patricia H. Adam Smith and his legacy for modem capitalism. New York: Oxford University Press, 1991.

  • Recommend Us