Rationality and Knavery
Daniel M. Hausman
University of Wisconsin-Madison
This paper makes a modest point. Suppose one wants to evaluate alternative policies, institutions or even constitutions on the basis of their consequences. To do so, one needs to evaluate their consequences and one needs to know what their consequences are. Let us suppose that the role of economic theories and game theory in particular is mainly to help us to use information we already possess or that we can acquire at a reasonable cost to judge what the consequences will be. We do not necessarily need true theories or theories that provide perfectly precise predictions. What sort of accuracy and precision we need depends on how the consequences of alternative policies are evaluated. The worth of an economic theory in this context of policy assessment depends on the accuracy of the predictions the theory permits, the costs of gathering the needed information, and the costs of learning and using the theory. My only excuse for uttering these truisms is that well known economists and philosophers have denied them.
My title derives from the following remarks of David Hume:
Political writers have established it as a maxim, that, in contriving any system of government, and fixing the several checks and controls of the constitution, every man ought to be supposed a knave, and to have no other end, in all his actions, than private interest. By this interest we must govern him, and, by means of it, make him, notwithstanding his insatiable avarice and action, cooperate to public good....
It is, therefore, a just political maxim, that every man must be supposed a knave; though, at the same time, it appears somewhat strange, that a maxim should be true in politics which is false in fact. (Hume 1741, pp. 40-42)
When Hume says one should suppose that people are knaves, he means that one should suppose both that they are motivated exclusively by private interest and that they will pursue their private interests competently. He has in mind in particular office holders, who, faced with a conflict between private and public interest cannot be trusted to act in the public interest. But the axiom is stated in a perfectly general way, and one can take it to apply equally to those who govern and those who are governed.(2)) If one seeks to design institutions to secure the public interest, one must incorporate incentives and sanctions that will lead individuals who pursue their private interests to serve the common good. The most obvious way legislators accomplish this is, of course, by the use of rewards and punishments. For example, suppose (as seems likely) it is in the common interest if individuals are secure in the possession of their property. Individuals seeking their own advantage will not automatically cooperate to secure this good. One can lessen the conflict between the common good and the pursuit of individual  interest by making theft more difficult (with the help of security procedures) and less advantageous (with a system of sanctions).
This idea has a long lineage. St. Augustine held that in politics everyone ought to be considered not merely a knave, but a sinner doomed to eternal damnation. Although a few people are saved by God's merciful grace, there is no way to tell who is damned and who is saved, and political institutions need to control the damned. By a fortunate dialectical twist worthy of a deity, the central characteristic of the damned--their egoistic attachment to earthly goods--renders them governable and makes possible peace and imperfect earthly justice (Deane 1963).
This reading of Hume's axiom should be somewhat alarming to liberals, who do not like state and social coercion. Although liberals recognize the need for sanctions to protect individual rights, they would like to circumscribe the functions of government strictly and to permit individuals to enjoy as much freedom as is consistent with the protection of rights. When individual rights, such as the right to property, are not at stake, how can the pursuit of private interest be made to serve the common good without intolerable interferences with individual freedom?
Fortunately, institutions can sometimes lead the pursuit of self-interest to serve public interests in more subtle ways. Competitive markets are the most famous example. As Hume's young friend, Adam Smith maintained, sometimes the invisible hand will do the trick (1776, Book IV, ch. 2). A liberal might put Hume's and Smith's thoughts together as follows: Since we must suppose individuals to be knaves in our design of constitutions, let us rely on markets for the organization of production and distribution. For markets harmonize private and public interests, and at the same time, they minimize coercive state actions. Arguments like this one send a warm glow to the heart of every liberal.
But there are well-known difficulties. The most obvious is that Smith's argument that the pursuit of private interest serves the public is unsound. Nobody with any training in economics would have trouble punching holes in it. The problem is to devise rules for a game that will have the outcomes one wants. When economists have repaired Smith's argument, they have evaded rather than solved this problem. Rather than proving that the outcomes people will achieve meet relevant normative criteria, they have proven that unusually rational and self-interested agents might possibly achieve an efficient outcome in unusual circumstances.
Economists have a great deal to say about the virtues of markets that do not depend on fancy general equilibrium theorems. Markets economize on information. They are highly flexible and encourage individual innovation. Although these considerations do not constitute a demonstration that markets harmonize private and public interest, they give one good reason to believe that markets are a good alternative to coercion. The historical record also provides evidence that markets typically coordinate individual pursuits better than the alternatives.
1 Can economics and rational choice theory be of value even if they are predictively empty?
Some thinkers have held that rational choice models like those in economics are of value regardless of their predictive worth. In his often-cited essay, "Rational Fools," Amartya Sen asks why Edgeworth portrays individuals in his economics as self-interested, despite believing that individuals are not in fact always self-interested. One answer is that Edgeworth believes that in particular domains (war and "contract") individuals are, to a good approximation, self-interested (1977, p. 318). Another answer, Sen maintains, is that Edgeworth -- like many other economists -- is interested in the abstract question of whether egoistic behavior serves the general good (1977, pp. 319-22). General equilibrium theory answers this theoretical question (with respect to idealized circumstances). It does not matter if people are in fact egoists.
From a very different perspective, Alexander Rosenberg also attempts to separate the worth of rational choice theories from the truth of their axioms concerning preferences and choice and from their predictive adequacy. In a series of books and articles over the last fifteen years, culminating in his Economics: Mathematical Politics or Science of Diminishing Returns (1992), Rosenberg has consistently criticized the predictive value of economics. Many of his comments are focused on general equilibrium theory, but his critique applies to all rational choice models and to game theory. I believe that this critique cannot be sustained (see Ross 1995, Rappaport 1995, and Cottrell 1995), but I shall not challenge it here.
I do not accept Rosenberg's indictment of standard economic theory, but if one does, then one needs to explain why so many people respect standard economic theory, work on it, and apply it to new domains, indeed even to new disciplines. One answer Rosenberg gives is that economists (and their fellow-travellers among political scientists, anthropologists, and sociologists) are interested in economic theory for its mathematical structure (1992, ch. 8). Like mathematicians concerned to prove theorems in some system of geometry and uninterested the true structure of space, rational choice theorists, including economists, are concerned to prove theorems employing axioms of rationality, and they are uninterested in whether individuals conform to those axioms. Undoubtedly Rosenberg is right that some economists are really mathematicians.
Rosenberg's second reason why rational choice and economic models are of interest, even if they are empirically empty, is more relevant here. Rosenberg claims that economics forms part of a normative argument for the market.
But at least now we can understand why economists continue to lavish attention on general equilibrium theory. It is not because they think it can be improved in the direction of a descriptively and predictively accurate explanation of economic activity, but because they believe it is already part of the best contractarian(3) argument for the adoption of the market as a social institution and, more important, stands a chance of becoming an even better one, as its assumptions are weakened, changed, and varied. (1992, pp. 219-20)
Rosenberg's assertions here are incredible, if they are read as claims about the conscious beliefs of economists who value orthodox economic theory, including game theory. Presumably the real story has to involve a considerable element of false consciousness, and the account looks much like an ideological critique, though Rosenberg never says whether this reason for lavishing attention on general equilibrium theory is a mistake.
Although the theorems of general equilibrium theory and welfare economics establish conclusions concerning the existence and Pareto optimality of equilibrium (given idealized agents and circumstances), the political argument for the market surely rests as much on informal arguments for the virtues of markets and on claims concerning the freedoms markets permit and protect. Rosenberg does not mention this, in part because he tends to overemphasize the centrality of general equilibrium theory and in part because economics and rational choice theory in general contribute little to our understanding or appreciation of the links between markets and freedom. The normative contribution of economics and rational choice theory apparently lies in their account of the consequences of markets for welfare and growth. Since Rosenberg is interested in the normative role of markets, he accordingly focuses exclusively on welfare arguments for markets.
Rosenberg maintains that rational choice theories can contribute to welfare arguments concerning markets, even if they are largely without empirical content. He cites Brennan's and Buchanan's contrast between the choice of constraints (in a constitutional context) and choices within constraints (in the ordinary course of economic life (Buchanan 1989b, pp. 44-5, Brennan and Buchanan 1983, p. 4; 1985, p. xii). Even though rational choice theory and game theory cannot predict very well how individuals choose within constraints, they can still guide our choice of constraints.
Notice that this argument for the worth of economic theories can be made just as well by someone who questions whether they underwrite laissez-faire policies. General equilibrium theory does not in fact tell us that knaves interacting on unfettered markets in circumstances involving externalities and public goods achieve optimal outcomes. Frank Hahn goes so far as to argue that one of the main virtues of general equilibrium theory is that it enables one to see thorough shoddy arguments for laissez-faire policies (1973, p. 324). Hahn's point (with which I think Rosenberg might agree) does not in any way undercut the normative importance of economics, which Hahn emphasizes as strongly as Rosenberg. Hahn is making a claim about what the normative significance of general equilibrium theory is. He is not denying that it has normative significance. Hahn thinks that economics underwrites the normative conclusion that laissez faire is not always good policy.
 2 An obvious difficulty
On Rosenberg's view, economic theory is supposed to guide the choice of constraints that structure the production and distribution of goods and services. It is supposed to guide this choice by providing an account of the consequences of alternative sets of constraints for welfare and growth. But if rational choice and game theoretic models have little predictive value, how can they successfully tell us what the consequences of markets will be for welfare and growth? If they cannot tell us what the consequences of markets will be for welfare and growth, what role can they play in normative arguments concerning institutional design?
Rosenberg notes the difficulty and asks, "Can a theory that is, as I have argued, predictively weak, carry the normative burden that general equilibrium is being saddled with by this interpretation?" (1992, p. 224) His answer is, in effect, that a theory may be "true in politics which is false in fact." The fact that individuals are not perfectly self-interested, intelligent, or rational and that their information is often biased and mistaken undercuts the predictive worth of game theory and rational choice models.(4) But it does not undercut their normative worth. For what matters when we are considering alternative "constraints" is not what their consequences would in fact be, but what their consequences would be if all men were knaves--that is, if everybody were rational, intelligent, self-interested, and well informed.
(A word about terminology: "Rationality" here involves having complete and transitive preferences and choosing among feasible alternatives what one most prefers. "Intelligence" means being able to figure out anything that the theorist can. "Self-interested" means that the objects of one's preferences ultimately contribute to one's own well-being. To be "well-informed" is to have information that is at least as good as the theorist's or policy maker's. Let us dub someone who is rational, intelligent, self-interested, and well-informed "economically rational." I shall take economic rationality as an explication of Hume's notion of a knave.)
Rosenberg argues that if game theory can tell us what the consequences of institutions and policies would be if everyone were economically rational, then it is of great normative importance, even if it cannot tell us what the actual consequences of institutions would be. Policy-making and the choice of constraints should be guided by knowledge of how ideally rational and self-interested agents would behave. Actual individuals may not choose that way, and a set of institutions designed to work optimally with knaves may not work optimally with actual individuals. With John Stuart Mill as absolute monarch, absolute monarchy might be best. But we cannot be assured that our absolute monarchs will always have sterling characters. As public spirit waxes and wanes, our institutions must be robust. Institutions must be designed to survive the "worst case" in which individuals are driven by nothing but their private interests and are ruthlessly rational in pursuing them. Thus Hume's axiom and thus the oddity "that a maxim should be true in politics which is false in fact."
3 Further arguments for the normative relevance of economics
In developing this argument for the independent normative significance of models of interactions among knaves, Rosenberg cites the work of Geoffrey Brennan and James Buchanan. In their 1985 book and in a joint essay, they offer a lengthy defense of taking economic rationality to be the appropriate model of human behavior for the purposes of "constitutional economics." "Constitutional economics" is the attempt to assess general rules within which economic and political actors pursue their objectives.(5) Like Rosenberg they argue that the appropriateness in the context of constitutional economics of taking individuals to be economically rational -- that is, rational, self-interested, intelligent, and reasonably well-informed -- is independent of the extent to which these claims about individuals are true.(6)
Brennan and Buchanan's first argument (1985, pp. 48-50) is that if agents are modeled as knaves in their economic behavior, then political behavior should be modeled the same way. This argument assumes what Rosenberg wants to show, that economic rationality is the right model in this context for economic behavior, and is of no use in establishing Rosenberg's conclusion. Brennan and Buchanan are right to be dissatisfied with theories that posit different models of behavior in different domains, but wrong to suggest that this methodological commitment takes precedence over evidence. If political theory T1 models people as altruists, while T2 models them as knaves, then those who think that economic rationality is the right model of economic behavior have reason for dissatisfaction with T1. But this reason for dissatisfaction is no argument for accepting T2, if T1 fits the data much better.
Second, Brennan and Buchanan argue that their inquiry concerns how institutions can reconcile or harmonize conflicting interests, and that it thus changes the subject to assume away conflicts of interest. "If we want to discover how institutional rules can turn conflict into cooperation, we cannot simply assume that persons who operate within those rules are naturally cooperative" (1985, p. 53). But to deny that people are economically rational is not, of course, to assume that people's aims and interests are in harmony.
Third, Brennan and Buchanan argue that the truth about people's actual interests and capabilities influences only the significance of the enterprise of constitutional economics, not what the appropriate "analytical method" is (1985, p. 53). It appears that they define constitutional economics as an inquiry into what institutional frameworks are best for a society full of economically rational agents. If constitutional economics is defined this way, then this inquiry must of course model individuals as knaves. But this trivial conclusion is irrelevant to the question of whether one ought to assume that individuals are economically rational when one is addressing problems of constitutional design.
Fourth, Brennan and Buchanan make the assertion (which they bolster by means of an extended example) that models that best fit the data "will be systematically  biased in the direction of inadequate constraints." (1985, p. 55). Their 1983 essay is devoted mainly to this argument. Suppose that actual behavior involves a mixture of economic rationality on the part of some agents and public spiritedness on the part of others. If the welfare loss connected with a failure of public spiritedness is more than a linear function of the divergence from public spiritedness, then one will underestimate the loss if one supposes that every agent pursues the same partly public-spirited policies. Using the best estimate of average behavior leads one systematically to underestimate the welfare loss. One might do better by modeling everyone as a knave.(7)
Even if one concedes that welfare losses are generally convex functions of divergences from public-spiritedness, this argument fails. First, it shows only that one might do better using a model of individuals as economically rational. It does not deny that one might do worse. How is demonstrating that procedure 1 might be better than procedure 2 (though it might be worse) an argument for procedure 1? Second, as Brennan and Buchanan note, it seems that they have offered a critique of one particular misuse of data rather than an argument that an a priori model is better than one informed by data (1983, p. 13). In response Brennan and Buchanan point out that information about the distribution of different modes of behavior may not be available, and they seem to suggest that one is forced to choose between fitting average behavior and supposing everyone is a knave. But if one really has no idea about the distribution of different behavior, why should we assume that everyone is economically rational rather than, for example, assuming some uniform distribution over different motives. (Clearly the reason is empirical!)(8)
Finally Brennan and Buchanan argue that without institutions that constrain the actions of knaves, altruistic behavior may be driven out.(9) Why this argues for modeling behavior as entirely self-interested is never explained.
4 Problems with this argument and Hume's axiom
Rosenberg takes Hume's axiom as a maxim of caution: our institutions ought to hold up even if people were not at all altruistic or public spirited. This argument does not, I think, fare any better than Brennan and Buchanan's. I suggest that game theoretic models of the interactions of knaves have normative significance concerning some domain only if they predict the outcomes people will reach. If a model goes predictively awry, then one should not rely on its advice concerning institutional design.
There are three decisive problems with using Hume's axiom to pull a normative rabbit out of an empty predictive hat. First, institutional choice should not be made on the assumption that every man is a knave. Second, actual economic theories typically fail to model agents as economic rational, ant the attempt to repair this deficiency leads to deep problems. Third, economics is not much  better at predicting what outcomes knaves will reach than it is at predicting what outcomes actual individuals will reach.
4.1 Should we suppose that everyone is a knave?
Suppose it is the case that the optimal institutions (the institutions that best promote the public good, whatever that may be) for real individuals are not the same as the optimal institutions for knaves. Then there are costs to employing institutions and policies designed for knaves.(10) There were costs attached to governing Victorian England by means of imperfectly representative institutions rather than handing over the power to John Stuart Mill. Once one recognizes such costs, one needs to ask whether the benefits in security, resilience or whatever are worth the costs. If one is an expected utility maximizer, one also needs to consider the probabilities of the possible outcomes of different institutions. Even if some institutions and policies would be disastrous if everyone were a knave, they may be superior in all actual circumstances. If the outcomes of institutions designed for knaves are much worse than the outcomes designed for actual individuals and the odds of everybody being a knave were low, then it would be foolish to choose the institutions designed for knaves.
One might argue that if institutions rely upon altruism or public spirit, they will be exploited by those individuals who are in fact knaves, and the exploitation will lead to the same outcomes that a society full of knaves would achieve. For example, individual entrepreneurs might act in a public spirited way by refusing to dump their wastes in adjoining marsh, paying higher wages, refusing to lay off workers in hard times, and so forth. Similarly individual entrepreneurs might be foolish and ill-informed. But public spirited, irrational, and ignorant actions reduce profits, and firms cursed with moral or stupid managers will be unable to compete with firms whose leaders are less squeamish and more intelligent in pursuit of profits. To preserve altruism and public spirit, institutions must place no reliance on them.
This argument relies on questionable empirical assumptions (Hausman 1989). Moral or public-spirited behavior on the part of firms is not always costly, and indeed it can be advantageous. Firms that are known to be "moral" face lower labor costs and enjoy the good will of customers and suppliers (Hausman and McPherson 1996, ch. 3). Furthermore, even if deviations from economic rationality reduce profits, non-profit maximizing behavior will not necessarily be driven out. Markets must be sufficiently competitive and there must be a sufficient amount of profit-maximizing behavior already there. Altruistic behavior can spread, too (Becker 1976).
Suppose one rejects expected utility maximization as the best way to choose institutions and employs a maximin principle instead. If in addition one holds that the worst outcomes would obtain if everyone were a knave, then it would follow that in institutional choice one should suppose that everyone is a knave. This may  be Rosenberg's view. Maximin is, of course, a dubious principle of choice. It could condemn all mankind to stunted lives to protect against a one in a billion chance of a society of knaves.
Even if one accepts maximin, the argument does not go through unless the worst outcomes occur if everyone is a knave. But that claim only needs to be made explicit to be doubted.(11) The worst social disasters--Nazi genocide, Serbian "ethnic cleansing", slaughters of Hutu's and Tutsi's in Rwanda--would never occur in a game played by economically rational individuals. If one ought to design institutions for the worst case, then one ought to prepare for something much worse than universal knavery. In the waning years of the twentieth century, it seems overly optimistic to suppose that everyone is a knave.
Robertson's suggestion (1956) that economists economize on "love" is more moderate and sensible. Instead of supposing falsely that there is no altruism or civic spirit, one should recognize that its quantity is limited, and that institutions that limit conflicts between private and public interest work better.(12) Yet even this more moderate proposal is dubious, because it falsely supposes that the extent to which individuals are economically rational is independent of institutions and policies. But, as Albert Hirschman (1985) points out, altruism and public spirit are not given in fixed quantities like iron ore or petroleum. Their extent depends upon institutions and policies. If there are too many possibilities for free-riding -- if one overestimates the extent to which people will be public-spirited and moral --, altruism and morality may be undermined. If, for example, one has a public transit system with too little enforcement of fares and prevalent cheating, those who pay their fares will feel like "suckers," and the system can undermine the public-spiritedness upon which it depends. But a system with too few possibilities for free-riding can undermine public-spiritedness and moral commitments as well. Workers who have to punch a time clock may be more likely to leave when they have put in their eight hours than workers who are trusted to fulfill their responsibilities. People can become what they are assumed to be, and with too much regulation people may not be able to make trust-inducing overtures to one another (Pettit 1995, p. 225).
4.2 Economic models are not populated by knaves
If one takes a knave to be someone who rationally and effectively pursues his or her material self-interest, individuals as depicted in most economic models are not knaves. First of all, economic theories typically suppose that firms maximize net returns. Just how the internal structure of a firm is supposed to insure that a set of knaves acts so as to maximize the net returns for the firm is deeply mysterious. Second, standard theory says nothing about how knaves are supposed to reach equilibrium and to restore equilibrium after various "shocks." Instead the theory offers an entirely fictitious story. If I've got some extra beaver and you've got some extra deer, I tell an auctioneer how much beaver I'm willing  to part with at the price the auctioneer announces, and you tell the auctioneer how much beaver you're willing to buy and the auctioneer adjusts this price (and all other prices) up or down until quantities demanded and offered equalize.
Third, economic theories generally suppose that agents obey the law. You and I patiently keep the auctioneer filled in on how much we offer and demand. You don't just shoot me and take my extra beaver. Why not? Of course there are the sanctions of the state. You might get caught. But people can get away with murder, sometimes literally and often figuratively. Opportunities abound to take advantage of the inevitable incompleteness of contracts. Over the past generation economists have begun to take all these questions seriously, but as game theorists have made clear, avoiding these short cuts and insisting on modeling how outcomes are reached pose enormous theoretical challenges. The very characterization of rationality is controversial. I would argue that game theory has taught economists modesty. It is not easy to say what knaves are or what games they are playing.
4.3 The predictive problems of economics do not derive mainly from assuming people to be knaves
Suppose, as Rosenberg alleges, that economic theories in general have little predictive value. Then I maintain that they will be of little help in designing institutions, even if everyone were a knave and these models correctly characterized what a knave is. Suppose there were no difficulties with the axioms concerning individual rationality (that people's preferences are complete, transitive and continuous and that choice follows preference) and self-interest (that individuals care only for their own consumption bundles). Would the predictive weaknesses that concern Rosenberg then disappear? Surely not. The predictive problems (which, to repeat, I do not think as serious as Rosenberg urges) do not result entirely or even mainly from its assumptions that people are economically rational. There are still incomplete futures markets, increasing returns to scale, indivisibilities, non-equilibrium trading, and so forth. Furthermore, even the best informed individuals will not have perfect knowledge. One might, of course, maintain that, unlike the assumptions about rationality and self interest, these other assumptions are reasonable approximations that permit one to make accurate predictions concerning the outcomes of the interactions of knaves. But Rosenberg offers no argument for this view. Why should one believe that it is a reasonable approximation to assume that there are complete futures markets, but unreasonable to assume that people's preferences are transitive? The assumptions about rationality and self-interest are apparently better rather than worse approximations than are many of these additional assumptions. If Rosenberg were right about the predictive weakness of economic theories, there would be little reason to rely on them when designing institutions even if everyone were a knave and the theories correctly characterized knaves.
The moral I draw is that economics has normative value in the context of institutional design only if it has predictive value concerning individual behavior within particular institutions. If one believes, as Rosenberg does, that the theories have little predictive value, then one should conclude that they have little normative value as well, and the attention economists and other social theorists lavish on them reflects either mistake or their purely mathematical interest. If one has a less negative appraisal of the empirical content of the theories, one need not have such a negative view of their normative importance or of their overall worth. The bottom line however is empirical content. To what extent do the theories have true implications concerning individual behavior and its social consequences?
Let me emphasize that this paper is no brief for Pollyannaism. I maintain merely that when thinking about institutional design, we should make use of the best knowledge we have of human motivations and capacities and of how they are affected by institutions and policies. What little knowledge we have of human beings does not justify a particularly rosy view of their character. Indeed it may be economists who are the Polyannas.
1. I am grateful for criticisms from Jeffrey Johnson, Julian LeGrand, and audiences at the University of Wisconsin, Duke University, the University of Exeter, and the Conference on Game Theory, Experience, and Rationality sponsored by the Institute Vienna Circle. In Werner Leinfellner and Eckehart Köhler, eds. Game Theory, Experience, Rationality; Foundations of Social Sciences; Economics and Ethics: In Honor of John C. Harsanyi. Dordrecht: Kluwer, 1998, pp. 67-79. Original page numbers in square brackets
2. If Albert Hirschman's reading of the history in The Passions and the Interests (1977) is correct, then it is a misreading of Hume to suppose that he is saying that institutions ought to be designed to withstand a worst-case scenario, in which all citizens pursue their own interests without altruism or public spirit. Hirschman argues that during the 17th and early 18th century political philosophers distinguished between, on the one hand, the passions, such as lust, political ambition, the pursuit of honor or revenge and, on the other hand, the interests, which were identified with material interests. Insofar as institutions enabled individuals to become wrapped up in the pursuit of their interests, they were thought to promote stability. What had been regarded as the vice of avarice came to be seen as a feature of human beings that made their behavior predictable and that reined in their passions. This essay is concerned mainly with this modern misreading of Hume, not with exactly what Hume meant.
3. Rosenberg maintains that general equilibrium plays its normative role as part of a specifically contractarian argument in defense of the market. But the only connection I can see to contractarianism is the irrelevant fact that both microeconomics and contractarian theories portray individuals as rational. The argument that markets are more efficient than the alternatives would be just as relevant to a utilitarian policy-maker as it would be behind a veil of ignorance.
 4. This is an empirical claim, not the fallacious assertion that false premises have only false conclusions.
5. Brennan and Buchanan argue explicitly that one cannot evaluate sets of rules by evaluating their outcomes, on the ground that "there is no external standard or scale through which end states can be "valued" (1985, p. 45). I shall, however, ignore this claim both on account of the difficulties in interpreting it and because Brennan and Buchanan's own arguments conflict with it.
6. "In any evaluation of alternative institutions, therefore, Homo economicus is a uniquely appropriate caricature of human behavior, not because it is empirically valid but because it is analytically germane" (1985, p. 53).
7. Sometimes it seems that Buchanan is only arguing for models that treat individuals as less altruistic, trustworthy, or public spirited rather than for models that treat them as economically rational, "In constitutional choice, therefore, there is an argument for incorporating models of individual behavior that presume more narrowly defined self-interest than any empirical record may warrant" (1987, p. 67).
8. And in his 1989b, after quoting Hume's axiom that every man ought to be supposed a knave, Buchanan writes, "Let me not be misunderstood here. The argument does not defend the use of the narrowly restricted economic model of behavior independently of its descriptive qualities" (p. 48). This is puzzling, because the arguments Buchanan and Brennan make purport to offer a defense of the use of "the narrowly restricted economic model" regardless of how well it fits the data.
9. As their examples show, Brennan and Buchanan are not very precise about what they are assuming about individual behavior. Thus they argue for the claim that individuals ought to be modeled as knaves by pointing out that without social sanctions malevolent behavior is as likely to spread as is altruistic behavior (1985, p. 61). But one who is economically rational is of course neither malevolent nor altruistic. If one is concerned about malevolence, one should not model people as self-interested.
10. Brennan and Buchanan note this point, "In particular, we have not discussed the implications of the elementary fact that more restrictive rules will not only help to prevent the occurrence of disaster but also often preclude actions that may be intended to promote desirable outcomes" (Brennan and Buchanan 1985, p. 54).
11. In their 1985 work, Brennan and Buchanan sometimes equate a society consisting exclusively of economically rational agents with the "worst case" (1985, p. 55), while in their 1983 essay, they write, "Homo economicus by no means represents the worse imaginable character for the social drama" (1983, p. 19). They then argue that the economically rational agent may wind up being only "somewhat worse" than the representative agent, which would be a virtue of their method.
12. Oddly enough, this proposal, which is in many circumstances inconsistent with assuming that everyone is purely self-interested, is cited in support of that assumption by Brennan and Buchanan (1985, p. 63).
Gary Becker, "Altruism, Egoism and Genetic Fitness", in: Journal of Economic Literature 14, 1976, pp. 817-26.
Geoffrey Brennan and James Buchanan, The Reason of Rules. Cambridge: Cambridge University Press, 1985.
_____, "Predictive Power and Choice Among Regimes", in: Economic Journal 93, 1983, pp. 89-105; rpt. and cited in Buchanan (1989a), pp. 3-23.
James Buchanan, "Constitutional Economics", (1987) from the New Palgrave, rpt. and cited from Buchanan (1989a), pp. 57-67.
_____, Explorations into Constitutional Economics. College Station: Texas A&M Press, 1989a.
_____. "Rational Choice Theory", (1989b), in: Buchanan (1989a), pp. 37-50.
Allin Cottrell, "Intentionality and Economics", in: Economics and Philosophy 11, 1995, pp. 159-76.
Herbert Deane, The Political and Social Ideas of St. Augustine. New York: Columbia University Press, 1963.
Frank Hahn, "The Winter of Our Discontent", in: Economica 40, 1973, pp. 322-30.
Daniel Hausman, "Arbitrage Arguments", in: Erkenntnis 30, 1989, pp. 3-22.
Daniel Hausman and Michael S. McPherson, Economic Analysis and Moral Philosophy. Cambridge: Cambridge University Press, 1996.
Albert Hirschman, The Passions and the Interests. Princeton: Princeton University Press, 1977.
_____, "Against Parsimony: Three Easy Ways of Complicating Some Categories of Economic Discourse", in: Economics and Philosophy 1, 1985, pp. 7-21.
David Hume, "Of the Independency of Parliament", (1741) in Essays Moral, Political, and Literary. Rpt. Oxford: Oxford University Press, 1963, pp. 40-47.
Philip Pettit, "The Cunning of Trust", in: Philosophy and Public Affairs 24, 1995, pp. 202-25.
Steven Rappaport, "Is Economics Empirical Knowledge?" in: Economics and Philosophy 11, 1995, pp. 137-58.
Dennis Robertson. "What Does the Economist Economize?" in Economic Commentaries. London: Staples Press, 1956, pp. 147-55.
Alexander Rosenberg, Economics--Mathematical Politics or Science of Diminishing Returns. Chicago: University of Chicago Press, 1992.
_____, "What Is the Cognitive Status of Economic Theory?" in Roger Backhouse, ed. New Directions in Economic Methodology. London: Routledge, 1994, pp. 216-35.
Don Ross, "Real Patterns and the Ontological Foundations of Microeconomics", in: Economics and Philosophy 11, 1995, pp. 113-36.
Amartya Sen, "Rational Fools", Philosophy and Public Affairs 6, 1977, pp. 317-44.