Draft 1.1: 6 June 1995 Comments Welcomed Not for Quotation For presentation at the PEGS Conference, Washington D.C. February 10 and 11, 1994. Values, Policies & Citizen Competence: An Experimental Perspective by Norman Frohlich Department of Business Administration University of Manitoba Winnipeg, Manitoba R3T 2N2 and Joe A. Oppenheimer The University of Maryland Collective Choice Center & Department of Government and Politics College Park, Md. 20742 ABSTRACT Roles for experiments in identifying aspects of citizen competence are examined. The link between individual preferences, choices and welfare is explored. Experiments are shown to provide evidence that there can be a decoupling of preference, choice and welfare due to framing effects. Moreover, experiments reveal the existence and impact of other-regarding preferences and behaviors which improve social outcomes. These preferences are also shown to be subject to framing effects. Experimental mechanisms for identifying ethically acceptable policies are identified. Finally, experimental evidence is adduced to argue that incentive compatible mechanisms for achieving better social outcomes may have pernicious effects on citizens' ethical motivations. Some institutional design implications are drawn for democratic and just societies. Table of Contents ABSTRACT i EXPERIMENTS AND CORRECTABLE KNOWLEDGE CLAIMS 3 FRAMING AND PREFERENCES: AN INITIAL PROBLEM 4 Questions of Probabilities: Some Slippage between Preferences, Choices and Welfare (4) THE POSSIBILITY OF DEMOCRACY WITH IGNORANCE: SOME AGGREGATION ISSUES 6 Multi Dimensionality and More Complicated Environments for Democracy (7); Experiments Regarding Collective Action (8) QUESTIONS OF ETHICS AND OTHERS' ENTITLEMENTS 10 Balancing Everyone's Interests: The Ethical Problem (11); The Economists' Solution: Incentive Compatible Devices and Some Problems of Implementation (12) DISCUSSION 14 BIBLIOGRAPHY 15 VALUES, POLICIES AND CITIZEN COMPETENCE: AN EXPERIMENTAL PERSPECTIVE Norman Frohlich and Joe A. Oppenheimer Citizen competence has been a contentious issue in the scholarly literature since at least the time of Plato and his student Aristotle. The teacher took a dim view of the "democratic man" characterizing him in the Republic as being self-indulgent and whimsical. The student, by contrast, considered the citizen to be the ultimate arbiter of public policy. In one of the earliest and most famous defenses of consumer (citizen) sovereignty, he argued that the guest at the banquet and the resident were better judges of the meal and the house than were the cook and architect. In the modern world the Aristotelian view has, with the spread of democracy, tended to predominate. After all, in a democracy the citizens are presumed to be competent to judge what is best for themselves. Indeed, much of the new theory of non-market economics (also known as public, collective, social, or rational choice) presumes that individuals are, in their choices, merely indicating their preferences and thereby, their welfares. If we accept the argument that citizens do indeed know their preferences, that their preferences are truly reflective of their welfare and that the choices made in a democracy (of any sort) aggregate the preferences appropriately, we are done. Experiments aren't needed. Indeed, this conference isn't needed. Of course there are a few problems with accepting these premises even in the matter of private choices by competent individuals. Leave out the incompetents: the children, the insane, etc. Many people choose to smoke, abuse alcohol, or snort cocaine. Consider, the following action, by a presumably competent individual - Jane. She smokes! We can ascertain her preferences from her choices. She prefers the action and its consequences to abstinence. Yet many of us tend to say, "We know that smoking isn't good for Jane." How could we be so presumptuous? The answer lies in our rejection of some (or all) of the premises about the bases and consequences of Jane's actions. As a start, we may wish to claim that 1) Jane may be ignorant of her true preferences. She may not be able to imagine accurately how she'd feel after a few weeks of having stopped smoking. She might not know how well she would feel. Were she but to know what awaited her, she would indeed prefer not to smoke. Ignorance of the welfare that accompanies a different choice misleads her. Or we might argue that 2) Jane doesn't know that there are relatively painless ways to quit smoking and that she is only continuing because she incorrectly fears the temporary discomfort of stopping. She does not have good information about a good program that will allow her to stop. Finally, we might assert that 3) Jane is not taking into account the externalities of her behavior on the health of her husband, her young daughter and her unborn child. (Did we forget to mention that she was pregnant?) She lives in a social context and many aspects of that context need to be taken into account in any presumably individual decision. Each of these caveats - which call into question the Jane's competence regarding an individual choice - might well be applied to the question of citizen competence in general. Democracy, as a form of government is premised on the notion that citizen participation in decision making leads to better results than does lack of participation. There are no claims of perfection - recall Churchill's famous dictum: Democracy is the worst form of government except for all others. Yet in most instances citizens' competence is hindered by problems such as those that beset Jane. Usually citizens' incompetence is discussed in terms of knowledge or information problems - much as we described Jane's difficulties. Phrased generally: 1) Citizens lack knowledge about the relationship between their preferences for possible states of the world and the welfare they would experience under those states (e.g., the case of Jane's smoking). 2) Citizens lack knowledge about the empirical consequences for themselves of specific political structures and public policies. Different voting (aggregation) institutions and public policies may affect the outcomes achieved by any actions and hence the ultimate welfares gained by any changes. The real consequences of any policy decision are difficult to gauge. 3) Citizens lack knowledge regarding how to evaluate the consequences of different states of the world for others: they have difficulty in aggregating the welfare of all concerned. These failings can drive a wedge between a citizen's choices and subsequent welfare. They call into question the citizen's competence. Most of this daylight can be viewed as "informational" (i.e. caused by bad or imperfect information) by the neo-economists who talk of preference driven non-market behavior. What exactly is the status of information in the question of citizen competence? Some would argue that it is quite strong. For example, consider a typical policy question: citizen safety. In America today, many see their environment as crime infested, and violent. It is easy to get citizens to indicate their desire for less crime. The problem is not getting the simple expression of the preference: "Less violence, please." Two further difficulties appear. First, how much is "less violence" worth? This is not a trivial question since "less violence," like all other public goods, requires a more precise evaluation so that a sense of balance can be developed in policy development. In other words, what is the welfare gain of a specific reduction in the crime rate. Second, there is the issue of instrumentality: what does it take to get "less violence?" This is also not an easy question to answer. But perhaps these are not issues of citizen competence? Rather, it could be seen as an issue of governmental competence. Citizens do not need to know how to fashion public policy. Citizens are only required to judge which of the alternatives from which they choose most assuredly meet their needs. At extraordinary times they might be a bit more involved in the direct fashioning of policy, through mass movements, a greater sensitivity toward some objective or another. But even then (e.g. the civil rights, or the environmental movement) the nuts and bolts of the policy are not theirs to put together. But even if it is the politicians who must be able to "structure the alternatives," the citizen's choice problem still hinges on questions of knowledge and information. While we accept these traditional threats to citizen competence, we argue in this paper, based on a variety of experimental findings, that not all, questions associated with citizen competence can be reduced simply to questions of information processing and assessment. Why might this be so? There are a number of reasons. First, we believe that the issue of individual citizen competence and expression of preference must be put in the perspective of the aggregation problem. This is certainly an old tale, told and retold both by game theorists and the collective choice crowd (Arrow, Sen, etc.) and one which we review later (see page 8). Suffice it to say for the moment, that the way in which a particular democracy takes individual citizen preferences into account in determining a policy can have profound effects on, and be profoundly affected by, citizen competence. Much of the issue of getting a better society may be a question of the adoption of good institutional rules (see page 6, below). Second, we argue (see page 4) that there is solid evidence, mainly from experiments, but also now in the theories of neuro-biology and even choice theory itself, that preferences and choice do not give us a foundationally solid basis for the development of a link between information, individual choice, and individual welfare. With substantial information, individuals can be shown to be relatively incompetent in developing stable, and sensible, choices when the questions involve probabilities, likelihoods and even morals. This difficulty has been at times been referred to in the literature as problems of framing. As will become clear in our discussion of this, we believe that this framing problem is perhaps more general than has been acknowledged. But we also believe that some experiments which identify these additional problems of citizen competence, when examined closely, offer clues as to how the difficulties might be overcome. EXPERIMENTS AND CORRECTABLE KNOWLEDGE CLAIMS Experiments, as we all know from our high school days, involve the careful construction of environments. One constructs these environments to control for extraneous factors and to isolate the effects of single variables on the phenomenon of interest. The method is useful for the testing, debunking, and/or improving specific hypotheses. Putting it slightly differently, experiments are a useful tool in developing correctable, (and corrected) knowledge claims. And because experiments occur in carefully controlled, environments, they are primarily useful for the testing of closely specified hypotheses, the sort which are found in deductive arguments: where care in the crafting of the hypothesis is of the essence. How then does this relate to the problem of citizen competence? The connection is not immediately obvious. The link is the evidence that experiments provide us regarding the bases of individual choice. Experimental studies yield understanding about what individuals preferences are and how they are linked to choices and welfare. In this way they can help identify the nature of some of the informational problems discussed above. But they can demonstrate how such factors as voters' ignorance affect the quality of democratic outcomes. They can demonstrate how different policies affect outcomes. They can identify how framing effects impact upon choices. And they can even yield insight into how many individuals' welfares might be aggregated in a normatively justifiable way. FRAMING AND PREFERENCES: AN INITIAL PROBLEM Our concern with preferences stems from the belief that for democracy to function well there must be a strong relationship between an individual's choice and the individual's welfare. Any factors which weaken this link undermine a citizen's competence. We therefore look at how individuals' choices and preferences reflect (or fail to reflect) the best interests of the individual (see above, p. 1). Specifically, we will consider sources of doubt which have crept into the welfarist linking of preference and welfare. The effects of framing constitute a major challenge to the traditional view. Framing can affect decisions in relatively simple situations: when a single individual evaluates alternatives in which there is no tension between the decision maker's welfare and that of others. It has been shown that one's preferences can be changed, in a totally arbitrary fashion, from the point of view of welfare. In more complex situations, when there is a conflict between the individual's welfare and the welfare of others, individuals' preferences can also be shifted by framing. That problem will be discussed after a preliminary consideration of the aggregation issue. There we will address how experiments both allow for the identification of underlying preferences in complex situations and reveal how framing effects can profoundly affect group outcomes and welfare. Questions of Probabilities: Some Slippage between Preferences, Choices and Welfare Most alternatives do not involve outcomes which are "sure bets" of ones' choices. Rather, most choices lead to outcomes on a probabilistic basis, like the purchase of a lottery ticket. This is so even though we don't usually think of things this way. For example, choosing to cross a busy street to see a friend could be seen as a lottery in which one prize is a social visit with the friend, and one of the many others is hospitalization for injuries incurred by being hit by a car. A series of experiments (see footnote 4), have established that individuals have a number of peculiarities in making choices concerning risky gains or losses. Consistency across these domains was shown to be weak. These problems were often lumped together and called problems of "preference reversal." Although Kahneman and Tversky stressed the role of gains and losses in the generation of these inconsistencies, others showed that the issues were more complex and involved more generalized difficulties in consistent handling of probabilistic alternatives in choices. The first step in the theoretical generalization of the problem of preference reversal was made in a careful series of experiments by Goldstein and Goldstein and Einhorn. These were later written up by Goldstein and Einhorn. Most basically, they discovered that often when one is confronted with a gamble, one evaluates it on two quite different scales. First, one examines the stakes, looking at the positive gains (or winnings, W) and the losses (or L), as well as their associated probabilities. Individuals then evaluate those in terms of some personal evaluation [call these utilities, such as U(W), U(L)]. Now the individual's personal evaluation of the risky alternative or gamble [U(G)] is ordered as U(L) < U(G) < U(W) in a manner reflecting the probabilities of receiving L and W. When the individual has to identify a price for the gamble, they often use the same probabilistic discounting, but on a different scale! The scale now is the actual losses and gains (rather than the subjective evaluation of these values). This can lead to substantial inconsistencies of choice, or preference reversals. Of course, there are implications of these reversals. Since it has long been noted that individuals gamble and buy insurance, we also note that it is not easy to categorize such a person as "risk neutral," or "risk averse." The standard argument (see Quattrone) has been that individuals handle risky losses differently than risky gains. Or, that the status quo matters. Since the status quo is not definitively and unambiguously specified, the description of one's current position can influence what one sees as gains or losses. Such framing effects clearly affect one's choices. Experiments abound in which the story's frame determines the choice of the individual. For example consider a policy choice which must be made in a situation where 600 persons are likely to die (see Tversky and Kahneman p. 453). Policy alternative set #1 has policy à with 200 persons being saved for certain, or á, a chancy policy with a possibility that all will be saved with a probability of 1/3 and 2/3 that none will be. (The results here are that 72% of the persons asked chose à). Policy alternative set #2 has policy à with 400 persons dying for certain or á, a chancy policy with a possibility that all die (with a probability of 2/3) and 1/3 that none will die. (The results here are that 70% of the persons asked chose á). In these two sets of policies, it should be clear to the reader that à is always the same (200 live, 400 die), and similarly á is invariant. Yet the choices are exactly reversed. Accepting this as common knowledge among the social psychologists would imply that if we wished to place a policy proposal in front of the voters, the masters of communication should be able to frame it to their advantage. The outcome may have more to do with the frame than with the substance of the proposal. Some would have us believe that the results are driven by naivety and inexperience. But tests run in our own advanced classes of game theory on PhD students studying utility theory (as well as by Grether et. al.) lead one to a greater appreciation of the generalizability of the result. At the very least, the influences of the frame of a discussion on the choices of risky alternatives makes it difficult to accept a tight fit between welfarism, choices, and preference. This requirement of citizen competence is clearly at risk. A strong link between citizen choice and social welfare is clearly threatened. A competent citizenry must be able to cope with this problem. THE POSSIBILITY OF DEMOCRACY WITH IGNORANCE: SOME AGGREGATION ISSUES Despite the caveat offered by this preliminary discussion of the framing effect, there are models of democracy which yield a surprising result: that good outcomes can be achieved with relatively little citizen knowledge. One conception of modern representative democracy is captured in Downs (1957): competition causes the government to represent the majority by picking, as the winner, the competitor sitting on the median voter's ideal point. This model is the implicit basis of the oft-heard political pundit's comment that party X won the election because they captured the center. How this rough and ready version of the democratic life relates to reality is an open, and interesting, question. One way of getting a preliminary insight is to test whether competition does indeed take this form and achieve the predicted result in the laboratory. The important experiments which test Downsian political competition on a continuum were conducted by Collier, et. al. (1987 and 1989). In those experiments a group of subjects was divided up: Two (leaders) were put in one room, and the rest (voters) in another. The voters all had single-peaked preferences over a range of numbers. They didn't know much about what generated their payoffs. Each of them only knew that he or she got a payoff as a function of the leader's declaration of a number in the range. And the leader only knew that his or her choice of numbers would generate a payoff to each of the voters. Voters had to decide to keep, or throw out, leaders. In other words, no one knows the underlying payoff functions. After the leader chose a number, the voters got their payoffs. They subsequently got to vote (by majority rule) on whether to change the incumbent. Each of the two leaders get paid only for the sessions in which they were incumbents. Leaders knew the history of the leaders' positions, the aggregate vote outcomes, and the dimensions of the strategy space. The result: unbelievable convergence to the median voter; great confirmatory outcomes for Downs. If the world is a one-dimensional, single peaked affair, then a minimal democratic structure will yield an adequate outcome for the system, even with rotten information and almost no competence on the part of voters. Multi Dimensionality and More Complicated Environments for Democracy Alas, much of the world would seem to be more complicated than the simple world of those experiments. What happens as we increase the complexity of the world inhabited by the voters? The first, and in many ways greatest, hurdle is the move from single to multiple dimensional issue spaces. The reasoning for the complications are simple: in general (i.e. except under wildly improbable conditions) there is no equilibrium outcome for a preference driven choice in issue spaces of more than 1 dimension. Thus, when the world is multi-dimensional, the theoretical outcomes of democratic competition based on competing platforms are not definitive winning policy positions. Indeed, from a theoretical point of view, the predictions are confusing. The outcomes one can expect often depend on minor ancillary institutions or perturbations of behavior. Consider, for example, majority rule and simple voting for a point on a policy plane. There are some surprising experimental results. In an early experiment, Fiorina and Plott found that when there was no core, the outcome of the votes still clustered around the area where the core would emerge were one to exist. Further experiments also showed that there were pockets of unpredicted stability in the majority rule world of multi-dimensional issue spaces. Some of the stability was shown to be generated by coalitions even when there was no core. This has been conjectured to stem from the bargaining patterns of coalitions with their members. So informal, or extra-constitutional institutions such as political parties help to generate the sorts of order or predictable relations between choice and likely outcomes which create the substance for citizen's competence. One other line of reasoning followed the logic of some of the existing structures in democracies. Begun by Romer and Rosenthal, and followed up by Shepsle, and Shepsle and Weingast, the role of structures in inducing equilibria was examined. These structures included legislatures, committees, and more. Again, experiments were conducted. These experiments were run, not in a representative democratic context, but rather in a context of a committee reporting to a legislature. They showed that although the world did not follow the theory precisely, it was because the "citizens" of the world were more forgiving than the theory would have predicted. Thus, for example, in a test of a structure-induced equilibrium model based on an argument by Niskanen, the experimenters Eavey and Miller (1984a) discovered that the outcomes reflected a less than cut-throat form of bargaining among the individuals. Other experiments showed that when the theoretical prediction derived from an assumption of extreme self-interest was an extremely one-sided set of payoffs (viewed as unfair to some) the theoretical prediction was proven inaccurate in the laboratory. The results obtained were not because individuals were smarter or less ignorant than theoreticians had assumed, but they were morally superior. They were less self-interested, more interested in the impact of their choices on others' welfare than is traditionally assumed by economists. The assumption of individual self-interest which seems to be a reasonably close approximation of reality in markets seems to be shaky when applied to non-market behavior. Behavior incorporating other regarding elements means that the non-market institutions of society have a more forgiving environment in which to work than had been conjectured. This increases the likelihood of sensible outcomes in the absence of either tyranny or high costs of citizen involvement in politics. Experiments Regarding Collective Action There is yet another form of citizen decision which is central to the democratic process. It is the decision of how much to contribute to a group effort to obtain a group benefit. To begin with, we must note the fundamental difference between individual competence in choices regarding private and shared, or public, goods. With private goods, especially in cases of a market, one makes one's own bed. Having to lie in it provides an incentive to hone one's skills. But when a collective effort is made, one's own role in the outcome can be small (Olson). If one doesn't make one's own bed, or at least not very much of it, the importance of bed making skills to one's own night sleep is pretty small and can easily be overrated. With public goods the relationship between the competence of an individual citizen and the quality of outcomes would seem to be diluted. It is precisely with regard to collective action that theory predicts that self-interest plays a role in limiting the quality of what society can achieve. Individual self interest, and instrumental rationality, lie behind the free rider problem that bedevils collective action. How do the findings of the experimental studies treat this conjecture? Not too well, as it turns out. Specifically, the modern Olsonian argument regarding free-riding in collective action situations is usually characterized as one shot (i.e. non repeated) n-person prisoners' dilemma (NPD) problem (Hardin, 1971 and 1982). The current theory of the NPD game is that the only rational outcome is for all to defect and for a massive sub-optimality to result. Early students of the game noted the role of the motivational frame (e.g. trust, and suspicion) on the outcomes for the group. Later experiments, primarily by political scientists and economists, were conducted in more "sterile" environments. With no cues for cooperation, or other motivational structure to get in the way of the Nash (selfish) outcome, they expected an easy corroboration of what has come to be known as Olson's "strong free rider hypothesis." They didn't succeed. The strong free rider hypothesis was falsified. Significant levels of cooperative behavior were found in one-shot standard 2- person PD's. (See Flood as well as Rappoport and Chammah). For NPD's with many rounds, cooperation is usually found at the beginning but it tends to deteriorate quite quickly in the first 6-7 rounds. In contexts which most resemble markets, the non-cooperative outcome is chosen in overwhelming numbers. But in general, and especially in non-market contexts, there remains a significant residual of cooperation under virtually all laboratory conditions testing the repeated (or even more so the one - shot) NPD. The residual was actually quite substantial (i.e. more than 20%) in NPD's which did not involve market like transactions. Among experimentalists the strong free-rider hypotheses gave way to the "weak free rider" hypotheses. This hypothesis, inductively defined to conform to the experimental results (see Marwell and Ames, 1979 & 1980; Isaac et. al. 1984 & 1985), was that the modal behavior would be one of free riding. In the laboratory, we repeat: the strong free rider hypothesis was falsified. This result had its echoes in field research by many social scientists. What is to be learned from this branch of experimental work? A small proportion of the individuals in a society can be expected to be cooperators in most circumstances. They can make quite a difference to the overall welfare level of the group. We conjecture that these are also the individuals who are not focused on pure self interest (see footnote 10). In a series of experiments involving choices of payoffs for oneself and an unknown other, a substantial proportion of subjects were shown not to be strictly self-interested (Frohlich and Oppenheimer, 1984). Subjects' preferences were a function of factored into their decision the payoffs of the anonymous other for whom they were making a decision. Moreover, three varieties of interacting utilities were found in the subjects. Some were altruists, some were malevolent, and others oriented towards equity. By maintaining expectations of civility it could well be that altruistic individuals might be encouraged and make a substantial difference in the quality of life in the society. Further, differing institutional contexts of the society are likely to generate differing quantities of cooperative behavior. QUESTIONS OF ETHICS AND OTHERS' ENTITLEMENTS These differences in the way in which another's welfare enters into an individual's decision may be, as we hinted above, subject to framing effects. The framing problem may be more general than had original been understood by Kahneman, Tversky and the other pioneers in the field. A variety of experimenters have performed so-called dictator and ultimatum experiments. These experiments have shown that the degree of other regarding behavior varies drastically by context, in a manner which also shows that choice is triggered not only by the "outcomes" but by the "social quality of the process." The dictator experiments have the following structure: Two individuals are paired, anonymously. One of them is chosen (completely without contact with the other individual and with only a minimum of contact with the experimenters) to receive an envelope with money in it (usually $20). This individual can then leave any proportion of the money for the other (anonymous) subject. The ultimatum experiments are similar, but the second individual has the right to refuse the share left for her. Any such refusal results in no one getting anything. What are the characteristic results? The more care there is in constructing a condition of anonymity, the more self-interest is manifest (i.e. the less is left in the envelope for the second individual). As anonymity and non-observability break down, equality and other-regardedness become more manifest. The conclusions which are usually drawn? Other regardedness may not be a function of a taste for fairness but rather may be a matter of social reputation (Hoffman, et. al. p. 26), and hence a social exchange. The experimenters are quick to point out that this also supports Axelrod's notions that reciprocity makes cooperation pay (p. 27). This view (somewhat inconsistent with that of Frohlich and Oppenheimer, 1984, above) has most recently been challenged directly by Grossman and Eckel who argue that the reason for the results must be sought elsewhere. They point out that the anonymity and sterilization of the process drive out any incentive to leave any money for the second party. In one experimental treatment they changed the environment by indicating that the second party is a well known charity (e.g. the Red Cross) representing needy others. They also varied the productivity of giving (e.g. a $1 gift might produce $1 or $2 for the recipient under different conditions). These changes radically altered the results. Many gave where in the original experiments very few gave. In other words, there seems to be a deep seated other regarded preference held by many which can be largely choked off only by relatively severe changes in the normal social, informational, and motivational environment. The framing of the social choice situation affects the degree to which other regarding preferences are brought into play. Balancing Everyone's Interests: The Ethical Problem How should the welfare of others be taken into account in an individual's choices of a social outcome? What should be the appropriate balance between an individual's own interests and the interests of others? The absence of any generally accepted consensus regarding that ethical question poses a threat to citizens' competence. If one does not know what is right then one certainly will have difficulty factoring ethical concerns into a policy decision. In such cases, as well as in many other circumstances involving ethical matters, philosophers have long advocated the adoption of a perspective of impartiality to identify the just or right choice. One of the surprising findings in the experimental literature is how simple it is to develop a relatively effective means for motivating impartial reasoning and choice based upon it (see Frohlich and Oppenheimer, 1992 and forthcoming). These attempts to engender impartial choice reveal a pattern of concern which appears to remain constant across cultures. In experiments in Communist Poland, Canada, Australia, Japan, and the United States, individuals were asked to choose an arrangement of distributing income when they did not know to which income class they would be assigned. Not knowing their individual status ex-ante meant they could not reliably estimate their own fate under any particular income distribution scheme. In the experiments subjects overwhelmingly (about 3 out of every 4 groups) chose to set a floor income (and taxes to support the floor) to insure that there was some acceptable level of earnings. They further agreed not to cap after tax income (taxes were needed to support the floor). The experiments showed that individuals seem, almost universally, to hold three general classes of concerns for the needs of others, for rewarding effort, and for efficiency. The consensus is revealed in carefully constructed experimental environments. The difficulty posed by such results is clear: given the importance of the context, when the subject is not in a constructed environment, the relevance of the results is indirect. But the fact that there appears to be an almost universal, deep seated, response to questions of fairness in distribution, under approximations of the conditions defined as relevant by major schools of philosophic inquiry is a promising, if somewhat surprising development. As Jencks notes (1990) "One of the classic puzzles - perhaps the classic puzzle - of social theory is how society induces us to behave in ways that serve not our own private interest, but the common interest of society as a whole. The device of placing individuals in decision contexts in which they have an incentive to consider policy issues from an impartial point of view may be a promising way of gaining insight into the ethical imperatives associated with various policy options. The Economists' Solution: Incentive Compatible Devices and Some Problems of Implementation Economists took quite a different tack in attempting to achieve optimal group outcomes. Working within their framework of self-interested behavior a number of micro-economists set themselves the task of designing mechanisms - often referred to as incentive compatible devices - to overcome these tendencies and to achieve optimal outcomes. (see Clarke, 1971, 1977; Groves, 1973, Groves and Ledyard, 1977, and the special volume of Public Choice). The fundamental idea behind their effort was to find some sort of institutional structure (such as a tax scheme) that would align individual interests and group interests so that each individual's incentives would correspond to what would be required to achieve the best result for the group. We have argued (Frohlich and Oppenheimer, 1992) that many of the conditions identified in Rawls' argument can be approximated in the laboratory and that experimental methods can be used to identify what constitute fair outcomes. Moreover, we maintain that impartial reasoning can be invoked experimentally to align self-interest and ethical imperatives so that it acts as an incentive compatible device to achieve optimal and fair outcomes. It is possible to construct an experimental setting which can implement impartial reasoning in an n-person prisoner's dilemma. By aligning individual and group incentives, this arrangement was expected to lead to more cooperative behavior and to invoke ethical motivation in individuals. In recent work, (Frohlich and Oppenheimer, forthcoming a and b) we tested these conjectures in a 5-person Prisoner's Dilemma experiment. Subjects chose a strategy knowing that, after their choices, they would be randomly assigned to one of the five positions and would be given the payoff associated with the choice made by the person who had previously occupied that position. This "impartial transformation" of the Prisoner's Dilemma has a dominant strategy of complete cooperation and may be viewed as an incentive compatible device for aligning individual and group interests. In these experiments we demonstrated that this device was successful in moving groups towards optimal provision of benefits. Groups playing a 5-PD from an impartial point of view in a first phase of an experiment outperformed groups playing a regular 5-PD. We had also anticipated that the invocation of impartiality would promote ethically motivated behavior. To our surprise the use of the impartiality device had unanticipated consequences for both the role of the ethical motivators and for subsequent behavior. Subjects playing the 5-PD from an impartial point of view, although they were more successful in achieving cooperative outcomes, evidenced no relationship between their reported ethical concerns and their behavior. By contrast, individuals in a control group playing a regular 5-PD showed a strong and significant relationship between their ethical concerns and behavior. In a second phase of the experiments - when both experimental and control groups played regular 5-PD's - especially when subjects had been allowed to discuss their strategies in phase 1 - higher levels of cooperation persisted in the group that played the regular 5-PD. In other words, the effect of greater cooperation due to impartial reasoning was not only transient - in that it disappeared after phase 1 - it seemed to undermine subsequent cooperation and leave the group worse off than those in the control group who had played a regular 5-PD. One possible interpretation of these findings is that the incentive compatible device of impartial reasoning - by virtue of the very fact that it aligns individual and group interests - may blind participants to the ethical dilemmas inherent in the situation. As Professor Steve Turnbull commented at a presentation of the results - "It prevents subjects from flexing their ethical muscles." By removing the opportunity to wrestle with the dilemma, the device may be cueing individuals simply to follow their individual interests, and may cause self-interested behavior to be reinforced and carried over into subsequent decisions. This effect can be interpreted as another framing effect. By characterizing a problem in a particular way, individuals' both proximate and even subsequent choices are affected. Indeed, which elements of their preference structures are actually evoked appear to be affected by the framing. DISCUSSION What then are the conclusions we can bring away from this "trek" through the experimental forest? It would appear that the simple welfaristic link between the choices of the individual and the welfare generated by social policy is suspect. And the problem appears to be deeper than the simple informational or knowledge limitations of the individual citizen. It has to do with the general lack of determinate relationship between preferences and choices stemming from a variety of generalized "framing" effects. The existence of a sub-current of other-regardedness in preference and choice must make one ask, "What is the effect of this preference on the quality of social outcomes?" and "What are the contextual and environmental bases for maintaining these preferences?" This problem brings to the fore one possible effect of framing on citizen competence. If one is to achieve better outcomes one must frame the choice to encourage the emergence of beneficent preferences. Those appear to be the sorts of choices which are consistent with the society people truly want. Of course if preferences and choices are a function of the environment, and we choose our environment there is a big regress problem. Can we unravel this so to get to the primordial egg which tells us what we ought to want? Probably not. The issue must be resolved with different vocabulary, a different focus. And this may require a meta-individual perspective: one which goes beyond the methodological individualism inherent in the rational and other psychological choice models discussed here. The use of experimental "veils of ignorance" may cast some light on such issues. But that experimental methodology has not been developed extensively and is subject to a number of caveats. Another difficulty stems from the fact that the link between individual and group requires some sort of aggregation mechanism. These mechanisms do more than sum welfares mechanically. Indeed, as shown by Arrow, there is no simple aggregation mechanism that can guarantee an outcome which is acceptable according to minimal democratic standards. Any such mechanism either can be expected to yield no deterministic result, or generates stability by disregarding some of the major implications of the preferences held by the individuals. Experiments show that there may be substantially more other regardedness in individuals' preferences than most standard models assume. The allocations of weights to other citizens' interests in a social choice appears to be sensitive to the framing of the decision context. When the framing takes the form of an incentive compatible device, the problem of decoupling preferences and welfare may be magnified. It is the essence of an incentive compatible device designed to achieve optimal outcomes that it transform a situation from one in which individual incentives lead to sub-optimal results to one in which the individual incentives are aligned with group interests. Our results suggest that any such device might blind those who are subject to it to the ethical dilemma they face and might undermine ethical reasoning. The general argument is straightforward. If selfish and ethical interests coincide, there is no cognitive tension and no need to go beyond considering what is best for the individual. Individuals subject to such a device in an ethically problematic situation forego the opportunity to confront the ethical dilemma which they face. Their subsequent behavior in similar situations in which the device is absent are therefore likely to yield worse outcomes than would have been obtainable had they actually initially wrestled with the ethical dilemma. Given that economists advocate the use of incentive compatible devices, one can imagine some potential implications of this argument. Consider, for example, the possible effects of introducing more and more incentive compatible devices (such as the auctioning of pollution rights) to get industry to generate optimal outcomes. Such a system clearly removes opportunities for the management of industry to confront the ethical dilemmas they face in making decisions on effluents. Their ethical behavior in other spheres, in which clever devices have not yet been devised, may therefore suffer (from a societal point of view). Citizens, accustomed to such devices might have their ethical competence undermined in a similar fashion. If there has been a lesson to learn from all these experiments it is the importance of environmental or framing effects on the manifestation of values, and hence the maintenance of consistent choice. Framing effects may need to take their place alongside informational and knowledge problems as a threat to citizen competence. BIBLIOGRAPHY Arrow, Kenneth (1963). Social Choice and Individual Values, 2nd ed. New Haven: Yale. Aumann, R. J. and Maschler (1964). "The Bargaining Set for Cooperative Games." in M. Dresher, L.S. Shapley, and R. Tucker, ADVANCES IN GAME THEORY. Princeton, N.J.: Princeton Univ. Press, pp. 443-447. Axelrod, Robert. (1984). The Evolution of Cooperation. Basic Books. New York. Baumol, William J. and Wallace E. Oates (1979), ECONOMICS ENVIRONMENTAL POLICY AND THE QUALITY OF LIFE. Prentice Hall, Englewood Cliffs, NJ. Binmore, Ken. (1994). Game Theory & the Social contract 1: Playing Fair. MIT Clarke, Ed. (1971). "Multipart Pricing of Public Goods, PUBLIC CHOICE, Fall, 2: 17-33. Clarke, Ed. H., "Some Aspects of the Demand Revealing Process," PUBLIC CHOICE, 29, - supplement Spring 1977, 37-51. Chong, Dennis (1991), Collective Action and the Civil Rights Movement. Chicago: Chicago University Press. Collier, Kenneth E., Richard D. McKelvey, Peter C. Ordeshook, and Kenneth C. Williams. (1987). "Retrospective Voting: an Experimental Study." Public Choice. vol. 53: 101-130. Collier, Kenneth, Peter C. Ordeshook, and Kenneth C. Williams. (1989). The Rationally Uninformed Electorate: Some Experimental Evidence." Public Choice vol. 60: 3-29. Deutsch, Morton. (1958). "Trust and Suspicion," Journal of Conflict Resolution, v. 2, no. 4, pp. 265-279. Downs, Anthony (1957). An Economic Theory Of Democracy. New York: Harper and Row. Eavey, C. & G. Miller, (1984a). "Bureaucratic Agenda Control: Imposition or Bargaining?" American Political Science Review, 78, September, 1984: 719-733. Eavey, Cheryl L. and Gary J. Miller. (1984). Fairness in Majority Rule Games with Core. AMERICAN JOURNAL OF POLITICAL SCIENCE, 28: 570 - 586. Enelow, J. and M. Hinich. (1984). The Spatial Theory of Voting. Cambridge Univ. Press: Cambridge, UK. Fiorina, M. P. and C. R. Plott (1978). "Committee Decisions under Majority Rule: An Experimental Study." American Political Science Review, 72 (June, 2): 575 - 598. Flood, Merrill M. (1952). "Some Experimental Games," Rand Corporation Research Monograph, RM 789-1, June 20. Frohlich, N. and Irvin Boschman (1984),"Partisan Preferences and Attitudes Towards Income Redistribution: Cross-National and Cross-Sexual Results", Canadian Journal of Political Science, March, pp. 53-69, Frohlich, N. and J. Oppenheimer, w Pat Bond and Irvin Boschman. (1984). "Beyond Economic Man." J of Conflict Resolution v. 28, no. 1, March, 3-24. Frohlich, Norman and Joe A. Oppenheimer (1992). CHOOSING JUSTICE: An Experimental Approach to Ethical Theory. Berkeley: California University Press. Frohlich, Norman and Joe A. Oppenheimer (forthcoming a)."Experiencing Impartiality to Invoke Fairness in the n-PD." Public Choice. Frohlich, Norman and Joe A. Oppenheimer (forthcoming b). The Incompatibility of Incentive Compatible Devices and Ethical Behavior: Some Experimental Results and Insights. Public Choice Studies. Goldstein, William M. and Hillel J. Einhorn (1987), "Expression Theory and the Preference Reversal Phenomena." Psychological Review, Vol. 94, No.2 pp. 236 - 254. Grether, David M. and Charles R. Plott, (1979). "Economic Theory of Choice and the Preference Reversal Phenomenon," American Economic Review, 69 (September): 623 - 638. Grossman, Philip and Catherine C. Eckel (1994), "Anonymity and Altruism in Dictator Games" mimeo. Presented at the Econometrics Society, Loew's Le Commodore Hotel, Quebec City, Quebec, June 28, 1994. Groves, T. (1973). "Incentives in Teams," ECONOMETRICA, July: pp. 617 - 631. Groves, T. and J. Ledyard (1977), "Optimal Allocation of Public Goods: A Solution to the Free Rider Problem," ECONOMETRICA, (May) v. 45: 783-809. Hardin, Russell (1971). "Collective Action as an Agreeable N-Prisoners' Dilemma" Behavioral Science 16 (no. 5): 472 - 479. Hardin, Russell (1982). Collective Action. Johns Hopkins Univ Press (for Resources for the Future): Baltimore, Md. Hoffman, Elizabeth, Kevin McCabe, Keith Shachat, and Vernon Smith (1992). Preferences, Property Rights and Anonymity in Bargaining Games. mimeo. Presented at Annual Meetings of the Public Choice Society, New Orleans, La. March, 1992. (for AER) Isaac, R. Mark, James M. Walker, Susan H. Thomas, (1984). "Divergent Evidence on Free Riding: An Experimental Examination of Possible Explanations." PUBLIC CHOICE, 43, 113-149. Isaac, R. Mark, Kenneth F. McCue and Charles R. Plott, (1985), "Public Goods Provision in an Experimental Environment," Journal of Public Economics, 26, pp. 51 - 74. Jencks, Christopher (1990). "Varieties of Altruism," in Jane J. Mansbridge, Beyond Self Interest. pp. 53-67. Chicago: Chicago U. Press. Kahneman D. and A. Tversky (1982), "The Psychology of Preference," Scientific American, 246 (January): 160-173. Kahneman, Daniel and Amos Tversky (1979), "Prospect Theory: an Analysis of Decision Making Under Risk." Econometrica 47 (March): 263-291. Konow, James (1994). A Positive Theory of Economic Fairness, Working Paper 94081 Loyola Marymount University: Los Angeles, CA. Loomis, James L. (1959). "Communication, The Development of Trust, and Cooperative Behavior." Human Relations, 12: 305-15 Machina, Mark J. (1989). "Dynamic Consistency and Non-Expected Utility Models of Choice Under Uncertainty." Journal of Economic Literature Vol. XXVII (Dec. 1989): 1622 - 1668. Marwell, Gerald and Ruth E. Ames, (1979), "Experiments on the Provision of Public Goods (I): Resources, Interest, Group Size, and the Free Rider Problem." American Journal of Sociology, v. 84. No. 6: 1335- 1360. Marwell, Gerald and Ruth E. Ames, (1980), "Experiments on the Provision of Public Goods (II): Provision Points, Stakes, Experience, and the Free Rider Problem." American Journal of Sociology, v. 85. No. 4: 926 - 936. McKelvey, R. D. Peter C. Ordeshook, and Mark D. Winer (1978). "The Competitive Solution for n-Person Games without Transferable Utility." American Political Science Review. 72: 599-615. Miller, Gary J. and Joe A. Oppenheimer (1982),"Universalism in Experimental Committees." AMERICAN POLITICAL SCIENCE REVIEW, 76 (2, June): 561 - 574. Mueller, Dennis C. (1989). Public Choice II. Cambridge Univ Press, Cambridge, UK. Muller, Edward N. and Karl-Dieter Opp (1986). "Rational Choice and Rebellious Collective" Action, APSR v. 80, (2: June) pp. 471-487. Niskanen, William A. Jr. (1971). Bureaucracy and Representative Government. Aldine: Chicago. Olson, Mancur, (1965). The Logic of Collective Action. Harvard University Press: Cambridge. Ordeshook, Peter C. (1986). Game Theory and Political Theory. Cambridge University Press. Plott, Charles (1967). "A Notion of Equilibrium and its Possibility Under Majority Rule." American Economic Review, 57 (Sept.): 787 - 806. Quattrone, George A. and Amos Tversky, (1988). "Contrasting Rational and Psychological Analyses of Political Choice." AMERICAN POLITICAL SCIENCE REVIEW. (82, NO. 3 SEPT.) 719-736. Public Choice, Vol. XXIX, no. 2. Special supplement to Spring 1977, on alternative demand revealing procedures. Rapoport, Anatol and Albert M. Chammah (1970), Prisoner's Dilemma, Ann Arbor: University of Michigan Press. Rawls, John (1971). A Theory Of Justice, Cambridge: Harvard University Press. Romer, Thomas and Howard Rosenthal (1978). "Political Resource Allocation, Controlled Agendas and the Status Quo," PUBLIC CHOICE 33 (4): 27 - 43. Schofield, Norman (forthcoming) "Party Competition in a Spatial Model of Coalition Formation." In W. Barnett, M.J. Hinich, and N Schofield (eds.), Advances in Scoial Choice and Cooperative Games. Cambridge: Cambridge Univ. Press. Schofield, Norman (1993) "Political Competition in Multiparty Coalition Governments," European J of Political Research, v. 23, pp. 1-33. Shepsle K. & Barry Weingast (1981), "Structure Induced Equilibrium and Legislative Choice," PUBLIC CHOICE, 37, no. 3, 503-520. Shepsle K. (1979), "Institutional Arrangements and Equilibrium in Multidimensional Voting Models," AJPS, (February, No.1) 23: 27 - 59. Sen, A.K. (1970). Collective Choice and Social Welfare. New York: North Holland. Smith, Vernon L. "Experimental Economics: Induced Value Theory." American Economics Association. Papers and Proceedings. May 1976, v. 66 No. 2. pp. 274-279. Tversky, A. and D. Kahneman (1981). "The Framing of Decisions and the Psychology of Choice," SCIENCE 221 (Jan. 30): 453-458. Yew-Kwang Ng (1990). "Welfarism and Utilitarianism: A Rehabilitation." Utilitas, v. 2, No.2 (November): 171-193.