The Moral Economy: Why Good Incentives Are No Substitute for Good Citizens - Samuel Bowles (2016)

VII. A Mandate for Aristotle’s Legislator

When the Athenian citizens’ assembly decided in 325 BCE to set up a colony and naval station in the Adriatic, far to the west of Greece, they took on an enormous project requiring thousands of people and 289 ships.¹ And they had little time to spare: the window for safe navigation around the Peloponnesus would close in a matter of weeks. Neither the personnel nor the ships were at the moment under public orders; the settlers, oarsmen, navigators, and soldiers would have to be recruited from their private lives, and the ships outfitted for the mission (some would carry horses, since cavalry were involved). We know how they accomplished this because the assembly’s decree was preserved.

Trierarchs (ship commanders and equippers), appointed from among Athens’s wealthy, were required to bring a fully outfitted ship to the docks at Piraeus by a given date. Those who felt unjustly burdened could appeal their assignment (called a liturgy). They would do this by challenging some other (also presumably wealthy) individual to either take on their liturgy, or else to exchange with the challenger all their real and personal property holdings. If the target of the challenge refused to do either, then a popular jury would determine which man’s estate was the larger and should therefore bear the costs of the liturgy. By allowing citizens to use private information about their own and their neighbors’ wealth to mitigate injustices in the assignment of the liturgy, this ingenious provision limited opposition to the provisioning of the mission.

The decree continues that the assembly would honor the “first [trierarch] to bring his ship [to Piraeus] with a crown of 500 drachmas and the second with a crown of 300 dr[achma]s and the third with a crown of 200 dr[achma]s,” adding that “the herald of the Council [of 500] is to announce the crowns at the contest of the Thargelia [a festival] … in order that the competitive zeal … of the trierarchs towards the demos may be evident.” The daily wage for a skilled worker at the time was about one drachma, so these were substantial rewards, even though they represented a tiny fraction of the total cost of executing a liturgy. Others responsible for the timely dispatch of the mission would also be honored.

Lest there be any doubt about the elevated purpose served by these incentives, the decree spelled out the expected benefits of the Adriatic naval base: “the demos may for all future time have its own commerce and transport in grain” as well as a “guard against the Tyrrhenians” (Etruscan pirates).

And for those unmoved by honors and rewards, there was a warning: “But if anyone to whom each of these things has been commanded does not do them in accordance with this decree, whether he be a magistrate or a private individual, the man that does not do so is to be fined 10,000 dr[achma]s,” with the proceeds going to honor Athena. (The winners of prizes for the timely arrival of the ships at Piraeus most likely would have given them as offerings to Athena, too.)

Collectively, the Athenian polis was an accomplished mechanism designer; and its members would have laughed at the idea that material incentives and moral sentiments were simply additive. They would have found the idea that the incentives they offered might crowd out the Athenians’ civic virtues more risible still.

The “crown” they promised to the first trierarch who outfitted a ship was a prize, not a fee for service; their exhortations and incentives were complements, not substitutes. They were the first Aristotelian Legislators, though there is no evidence that Aristotle himself was involved. (He died three years after the Adriatic mission began.)

Recall the experiment in which imposing fines on parents arriving late to pick up their children at day care centers in Haifa resulted in a doubling of the number of tardy pickups. Now imagine that the Athenians had traveled to Haifa in a time machine and had been asked to help design the day care centers’ policy for dealing with late parents.

The sign that the day care centers posted on their doors read: “Since some parents have been coming late we (with the approval of the Authority for Private Day Care Centers in Israel) have decided to impose a fine on parents who come late to pick up their children. As of next Sunday a fine of NIS 10 [about $3 at the time, in Israeli new shekels] will be charged every time a child is collected after 16.10.”

Had the Athenians been consulted, they certainly would have not approved. Instead, their sign would have announced: “The Council of Parents wishes to thank you for arriving on time to pick up your children, since this reduces the anxiety that the children sometimes feel and allows our staff to leave in a timely manner to be with their own families. We will recognize all parents who have a perfect record unblemished by lateness for the next three months with an award of NIS 500, to be given at our annual parents and staff holiday party, with an option to contribute your award to the school’s Teacher of the Year celebration.”

But that would not have been all: “Those who arrive more than ten minutes late, however, will pay a fine of NIS 1,000, with the payment of the fine publicly transmitted also at the holiday party. In the unlikely event that the occasion for such a fine arises, the payment will also support the Teacher of the Year celebration.” And the message would have ended with: “Of course, sometimes it is impossible, for reasons beyond your control, to arrive on time; and should this occur, you may explain the circumstances before a committee of parents and staff, and if the lateness was unavoidable or if the fine would cause extreme hardship, the lateness will be publicly reported but no fine will be imposed.”

In fairness to the Haifa children’s centers, the cryptic sign informing the parents of the fines was not their idea of good public policy. It was part of an experimental design intended to avoid framing lateness as a moral issue, which would have confounded the effect of the fine. But if the intent was to discover the effect of fines per se on lateness, then the design must have been based on the assumption that the parents’ response to the fines would not depend on how the new policy was explained.

I wonder what would have happened if the doors of the day care centers had carried messages like the one the time-traveling Athenians proposed, explaining the ethical problem of lateness but without announcing a fine. Would parents have been moved to pick up their kids on time? And then, if the message had also warned that a fine would be imposed for lateness, would this have enhanced the salience of the moral message, crowding in social preferences and causing a greater reduction in tardiness than the moral message alone? Would this Athenian version of the experiment have reversed the crowding out that occurred in the absence of moral framing?

It might have.

That is the upshot of what we saw in chapters IV-VI. The problem of crowding out may arise when the information that an incentive conveys is off-putting about the person imposing the incentive, or when it frames the problem as one in which self-interested motives are acceptable or even called for, or when the incentive compromises the autonomy of its target. The problem, we will see, may be the information, not the incentive itself, and there may be ways that the information conveyed by the incentive could be more positive.

We also saw that in the presence of an incentive, generous actions such as helping others may be misunderstood as self-interested, even when they are not, and that this may result in people adopting self-interested preferences to a greater extent than they would in the absence of the incentive. But like the adverse information sometimes conveyed by incentives, this problem might be attenuated by providing ample opportunity for the display of civic-minded motives, as the Athenian assembly did.

The Legislator is beginning to wonder whether incentives per se are really the problem, and whether instead the crowding-out problem may arise from the relationship between the person imposing the incentive and its target, or from the meaning of the incentive. If he is to draw up his mandate as a policy maker and one who “orders the laws,” the Legislator will need to study cases in which incentives and social preferences worked in tandem rather than at cross-purposes. Then he will be ready to update Hume’s maxim about the constitution for knaves.

Getting and Becoming

I asked you to put away for later use the fact (from chapter III) that in a sequential Prisoner’s Dilemma, the second mover most often mimics the first mover, reciprocating cooperation or defection depending on what the first mover did. The fact that second movers reciprocate cooperation means that they place a positive subjective value either on jointly cooperating per se or on the payoffs received by the other player, whose trust and cooperativeness were evident in his first move. This value is sufficient to offset the higher payoff the second mover could receive by defecting on the cooperator, which is why they cooperate.

When second movers defect on defecting first movers, they are taking the action that maximizes payoffs under the circumstances, but evidently are not doing so from exclusively acquisitive motives. The same individuals would have forgone payoffs in order to cooperate with a cooperator, as we have seen. But cooperating with a defector has a different meaning, identifying the second mover as a “loser,” someone easily taken advantage of. Thus, part of the motivation behind the “mimic the first mover” pattern is something that reciprocating says about the second actor herself: “I am the kind of person who rewards those who cooperate and stands up to defectors who would exploit the cooperation of others.”

When people engage in trade, produce goods and services, save and invest, vote and advocate policies, they are attempting not only to get things, but also to be someone, both in their own eyes and in the eyes of others.² Our motives in other words are constitutive as well as acquisitive.

This idea, commonplace among psychologists and sociologists, was missed by most economists until George Akerlof and Rachel Kranton alerted the discipline to the possibility in their Identity Economics (2010).³ Sometimes constitutive and acquisitive motives are closely aligned, as with Adam Smith’s merchant, who (one imagines) would like to act in ways that let him see himself as an honest man, a reputation that would also underwrite profitable exchanges with others. Similarly, the second mover in the sequential Prisoner’s Dilemma game who defects on a defecting first mover is both making a statement about who she is and also maximizing her payoffs.

Distinguishing between acquisitive and constitutive motives and determining which is at work in cases like this can have important implications for the advocacy and design of public policy. Here is an example. Proponents of tax-financed redistribution to the poor often represent their programs as a kind of insurance for voters in the middle of the income distribution, who would not currently be beneficiaries. It is likely that proponents adopt this kind of rhetoric because they think opposition to redistribution programs stems from voters’ self-interest. This supposition is equivalent to viewing a second mover’s defection solely as a payoff-maximizing strategy, and overlooking its constitutive aspect.

But in the United States and elsewhere, much of the opposition to income redistribution is ethical, not self-interested, and is based on beliefs that the poor are undeserving. In a U.S. Gallup survey analyzed by the economist Christina Fong, among poor people who do not expect their economic situation to improve, those who believe that lack of effort causes poverty tend to oppose redistribution.⁴ Analogously, among the securely well off, who expect their incomes to rise in the future, those who believe that poverty is the result of back luck voice strong support for redistribution. Support for redistribution to the poor is also high among white respondents who believe that race is important to one’s chances of getting ahead in America, and to men who feel likewise about gender.

There are two lessons here for the Legislator about political rhetoric and policy advocacy, extending beyond redistribution to matters of climate change, foreign policy, and other highly charged topics. The first is that appeals based entirely on self-interest will fail to tap the social preferences that might lead people to support the policy in question. Canvassers for a signature drive in Albuquerque, New Mexico, to increase the minimum wage were well prepared with arguments that the change would support the local economy. But they found that the quickest way to get the job done was simply to let the resident know the dollar amount of the current minimum wage, a fact commonly met with disbelief, outrage, and a heartfelt signature.⁵

The second lesson is less obvious: an appeal to self-interest is inviting the voter to ask: “What’s in it for me?” The effect may be to reduce the salience of the voter’s ethical and other social concerns. This kind of moral disengagement appears to have occurred in a number of the experiments reviewed in chapter IV. Thus, appeals to self-interest not only may fail to tap citizen’s social preferences, but may even sideline them. This is just another example of the problem of nonseparability of moral sentiments and material interests, with which we began.

There are lessons here too about the use of incentives. Acquisitive and constitutive reasons for actions may sometimes clash. And we know from experiments and from observing ourselves and others that being good sometimes trumps doing well. Responding to an incentive in the manner intended (that is, as a payoff maximizer) may make the responder a victim. But not always. A self-interested response to an incentive may constitute the actor as a good citizen or an intelligent shopper, indicating that constitutive and acquisitive motives were closely aligned. The same reasoning, we will see, suggests how we can make incentives and social preferences synergistic.

How acquisitive ends interact with the constitutive motives that J. S. Mill advised economists to ignore may explain why incentives sometimes work exactly as economists predict on the basis of unmitigated self-interest—and sometimes don’t. Recall that in the Trust game implemented by Fehr and Rockenbach, in which the investor could announce that he would fine the trustee if the trustee’s back-transfer was not sufficient, the announcement reduced the trustee’s level of reciprocity.⁶ Back-transfers from the trustee were lower under the threat of the fine (figure 4.1).

On closer scrutiny, however, the incentive represented by the threat of the fine seems not to have been the problem. When we look at the data from the experiment to see who among the trustees responded negatively to the incentive, it appears that strong crowding out was almost exclusively a reaction not to the incentive per se, but to the apparent greed of the investor. Crowding out occurred when the back-transfer demanded of trustee would have given most of the joint surplus (total payoffs for the two) to the investor. There was no backlash against the fines threatened by investors who asked for back-transfers that allocated both the investor and the trustee substantial shares of the surplus. In these cases, the use of fines by evidently fair-minded investors reduced back-transfers by an insignificant amount compared with back-transfers in the experiment in which fines were not an option.

The key difference was the message sent by the fine. Where the stipulated back-transfer would have captured most of the surplus for the investor, the fine conveyed greed. Where it would have split the surplus more equally, the fine conveyed a commitment to fairness, and perhaps the investor’s desire not to be exploited by the trustee. The use of the fine to enforce a seemingly unfair demand provided an acquisitive motive to comply, but to the trustee, it also may have transformed the meaning of conceding. Complying with the investor’s stipulated back-transfer no longer made the trustee a cooperative and ethical person, as it would have had the investor’s demands been modest. It labelled the trustee as a person easily manipulated, or a victim.⁷

Thus, I suspect that it was the relationship between the investor and the trustee, not the threatened fine per se, that was the source of strong crowding out. That suspicion is reinforced by a diametrically opposite reaction to fines in a Public Goods with Punishment experiment. The imposition of fines by peers who have to pay to levy them, when they had nothing to gain personally from doing so, appear to have crowded in social preferences. We have already seen (in chapter V) evidence that peer punishment works. But this result could have occurred simply because of the direct effect of the incentive on the willingness to contribute, without having any effect on the subjects’ experienced values. We would like to know whether the incentive heightened the salience of the subjects’ social preferences, crowding them in, or whether instead the positive response to the punishment by peers was simply based on self-interest

Recall that in this game once each member’s contributions to a public good are revealed, fellow group members have the opportunity to pay (reduce their own payoff) to inflict a punishment on (reduce the payoffs of) one or more members in their group. In some variants of this experiment, the game is repeated a number of times and group membership is shuffled after each period; in subsequent periods, a punisher it is extremely unlikely to be in the same group with a past target of her punishment. In this so-called stranger treatment, the punisher cannot benefit if the target of his punishment responds by subsequently contributing more. Both the punisher and target know this, and importantly, the target knows that the punisher knows this. The target knows, therefore, that punishment in this case is an altruistic act that benefits others (the members of the target’s group in future periods) at the punisher’s expense; hence, it cannot be seen as a sign of the punisher’s intent to get a larger slice of the pie.

Figure 7.1 Contributions without and with peer punishment in the Public Goods game (stranger treatment) The contribution that maximizes average payoff is 20. In the treatment with no punishment, the contribution that maximizes an individual’s payoff is 0, and this is true irrespective of what others contribute (not contributing is the dominant strategy). (Data from Fehr and Gaechter 2000a.)

Figure 7.1 shows the period-by-period contributions in a stranger treatment of the Public Goods game implemented by Fehr and Gaechter. The first ten periods were the standard game with no punishment option; peer punishment was allowed in the subsequent periods. When punishment was not an option, contributions were initially quite generous but then declined to virtually nothing at the end of ten periods, as is standard in this game. When the punishment option was introduced, contributions started at around the level of the first period of the no-punishment treatment, but then steadily rose. (The results are the same if the ten rounds with the punishment option are played first, followed by ten rounds of the standard Public Goods game.) Recall that this is exactly what happened in the “English-speaking,” “Protestant,” and “Confucian” populations in the cross-cultural study described in chapter V, but not in the “Southern Europe,” “Arabic-speaking,” and “Orthodox/Ex-Communist” populations (figure 5.3).

Why is the fine counterproductive when imposed by an overreaching investor in the Trust game, but so effective in many subject pools when imposed by peers in the Public Goods game? A plausible explanation is that when punished by a peer who had nothing to gain by doing so, players saw the fine as a signal of public-spirited social disapproval by fellow group members. If this were the case, targeted free riders would feel shame, which they would redress by contributing more. If so, the incentive (the prospect of peer-imposed fines) has crowded in social preferences. (The positive effect might be even greater for free riders who escaped punishment, because they get the moral message without being angered by fellow group members’ social disapproval.)

But is this crowding-in explanation true?

The higher contributions in the presence of fines could have occurred because when would-be free riders took into account the likelihood of punishment for low contributions, their self-interest may have dictated their contributing more. If so, this would not be evidence of a fine crowding in a social preference. The increase in contributions instead would suggest that fines sometimes substitute for social preferences.

Curious, my coauthors and I explored this possibility. In our Public Goods with Punishment experiment, we found that the responses of targeted free riders could not be explained by self-interest alone.⁸ The punishments meted out by their peers were not sufficient to offset the gains of free riding and would not have induced an expected payoff maximizer to contribute more. Nonetheless, the experience of being punished strongly affected free riders’ subsequent 'margin-bottom:0cm;margin-bottom:.0001pt;text-align: justify;text-indent:12.0pt;line-height:normal'>Additional direct evidence comes from experiments in which purely verbal messages of disapproval had a substantial positive effect on free riders’ subsequent contributions.⁹ Abigail Barr found this among the rural Tanzanians who played her variant of the Public Goods game. After each round the subjects were given the opportunity to comment on what had transpired in the game. Free riders often received verbal rebukes from other subjects (“Now I know why I never get offered food when I drop by your house!”). The targets contributed more in subsequent rounds. Free riders who escaped rebuke responded even more positively.

Like the fines imposed by seemingly civic-minded peers in the more standard Public Goods with Punishment game, rebukes from neighbors may have raised the salience of the subjects’ constitutive motives. It became more important to affirm one’s generosity, in the eyes of one’s neighbors and oneself alike, than to take home a few more Tanzanian shillings.

Results from a variant of Falk and Kosfeld’s “control aversion” experiment (described in chapter IV) are also consistent with the interpretation that crowding out does not follow from the use of incentives per se but rather from the relationship between the incentive’s target and the designer, and particularly from the presumed intent of the designer. Recall that control per se may not have been what the “employees” found objectionable: when a lower bound on their efforts was imposed by a third party who did not stand to benefit, the negative response of the employees did not materialize.

More relevant to the design of public policy: when agents themselves (rather than a principal) implemented controls on their fellow members, the negative control-averse response also did not occur.¹⁰ Apparently, the imposition of controls by peers, especially if it is the result of a deliberative process legitimated by voting, poses no threat to people’s desire to constitute themselves as autonomous individuals. Indeed, the deliberation and adoption of policies protecting each group member from exploitation by free riders may be an essential part of a constitutive program for a person who values self-direction.

Can this framework provide the Legislator with a guide to policy?

Moral Lessons: Are Incentives to Blame?

At the outset, I cited an early example of the self-interest-based policy paradigm, Jeremy Bentham’s advice to “make it each man’s interest to observe … that conduct which it is his duty to observe.” It is a guide to how proper incentives should harness self-interested individuals’ objectives for public ends. But Bentham also understood the constitutive side of action and the need to design incentives that, unlike the Haifa day care center fines, complement moral sentiments rather than substitute for them: “A punishment may be said to be … a moral lesson, when by reason of the ignominy it stamps upon the offence, it is calculated to inspire the public with sentiments of aversion towards those pernicious habits and dispositions with which the offence appears to be connected; and thereby to inculcate the opposite beneficial habits and dispositions.”¹¹

Before pursuing the idea of punishment as a “moral lesson,” let’s briefly recall its long and checkered history. Public executions and whippings, pillories, ducking stools, branding, and even burning all have been advocated on these grounds. In some U.S. jurisdictions, punishments designed to shame the convicted have been instituted for commercial and sex offenses (embezzlement and prostitution) and minor crimes such as shoplifting. Some people are required to publicly display signs advertising their transgression: “I am a convicted thief” or “I got caught possessing cocaine” or “Convicted: DWI.”¹² A community-access TV channel in Kansas City ran a popular show called John TV, which displayed the names and photographs of men convicted of soliciting prostitutes. In the UK, the National Health Service announced that it would list the names, labeled with a red flag, of general practitioners who repeatedly missed early signs of cancer in their patients; green ratings would be given to doctors with a record of timely recommendations that their patients see a specialist.¹³

In addition to compromising people’s dignity, these punishments seem poorly designed to accomplish the educative role that Bentham advocated. Bentham quite explicitly sought to inspire public aversion not of the transgressors but of the “habits and dispositions” accounting for the transgressions.

Other unconventional forms of punishment are more attuned to this end. At a compulsory “School for Johns” in San Francisco, former prostitutes explain the hardships of that life; a burglar was required by a Memphis judge to permit his victim to enter the burglar’s home unannounced (accompanied by a police officer), choose an item of a value equivalent to the stolen object, and take it. Both punishments made the offense more personal and vivid to the convicted, which, following the reasoning of Loewenstein and Small, should have corrected the moral disengagement that apparently allowed the transgression.¹⁴

Setting aside these eye-catching punishments, the Aristotelian Legislator returns to the seemingly intractable dilemma posed by the fact that incentives are at once a necessary part of good governance yet may undermine its normative foundations. But recall that I told you (in chapter IV) that we would return to the idea put forward by Lepper and his coauthors that these negative effects may not be “a specific function of the use of tangible rewards.” They were referring to the compromised autonomy of the once-avid young artists whose intrinsic pleasure in painting had been diminished by the promise of an award. But the idea has far wider reach.

From an economic point of view, the small tax on plastic grocery bags enacted in Ireland in 2002 resembles the fine for lateness at the Haifa day care centers: it slightly raised the cost of an action that the incentive sought to deter. But its effect could not have been more different: in just two weeks following its introduction, the use of the plastic bags dropped by 94 percent.¹⁵ The tax may have crowded in social preferences: for many Irish men and women, carrying a plastic grocery bag home appeared to have joined wearing a fur coat in the closet of antisocial practices.

The difference between the lateness fine and the plastic bag tax is instructive. In Haifa, the announcement of the fine included no justification for the punishment. There was no “moral lesson.” The absence of an explicit normative justification of the fine invited a default frame: lateness was for sale. Maybe the modest level of the fine conveyed to the parents that their lateness was not really very costly to the school. Moreover, in the eyes of other parents, one might be late for reasons beyond one’s control, not out of deliberate disregard for the inconvenience it caused teachers. The announcement of the fine said that “some parents” were arriving late, which perhaps told parents that they would not be unique were they to arrive late, and therefore that lateness was not a particularly serious transgression of the generally observed social norm of punctuality. Finally, at least among parents, the only observers of one’s tardy arrival at the day care were themselves also late.

The Irish plastic bag tax, by contrast, was preceded by extended public deliberation and a substantial publicity campaign dramatizing the bags’ role in blighting the environment. Unlike an occasional lateness at the day care center, which might be beyond a parent’s control, the use of a plastic bag required a deliberate and highly public act of commission by the shopper. In the Irish case, the monetary incentive was combined with a message of explicit social obligation, and it apparently reminded people of the larger social costs of using and disposing of the bags. In Haifa, the fine seems to have said, “Lateness is okay as long as you pay for it,” while in Ireland the message was something like “Don’t trash the Emerald Isle!”

To sharpen the difference between the two cases, let’s imagine what would have happened if time-traveling members of the Athenian assembly, instead of modern behavioral economists, had designed the day care policy toward tardy parents. Table 7.1 contrasts the actual experiment with my guess about what would have happened had the Athenians made it there in time to help out.

Look at the “Athenian’s moral message” row in the table. While the message alone would have had some positive effect (“better” rather than “status quo”), its impact, I would be willing to bet, would have been enhanced by the fine. And in the “Fine (actual)” column, the message would have changed the fine’s effect from negative to positive. In other words, in the absence of the moral message, the fine crowded out social preferences, and with the moral message, the parents’ ethical concerns were crowded in. In this case, the fine and the moral message were complements rather than substitutes.

Table 7.1. Lateness at the Haifa day care centers: The actual and (imagined) Athenian experiments compared

Note: The “status quo” cell is the situation without fines or messages, that is, before the day care centers sought to address parental lateness. Introducing the fine (as the centers did) delivered the “worst” outcome, namely, more lateness. If the incentives and the moral message were additive in their effects (that is, if the two were separable), then the effects of the fine would be independent of whether it was accompanied by a moral message, and correspondingly, the effect of the message would be independent of whether it was accompanied by a fine.

The entries in the “Athenian’s moral message” row of the table are not entirely fanciful. A similar crowding-in process was apparently at work in a public goods experiment by Roberto Galbiati and Pietro Vertova.¹⁶ Consistent with the crowding in resulting from the presence of a small fine in the Cardenas experiment described in chapter III, Galbiati and Vertova found that the effect of a stated (nonbinding) obligation to contribute a certain amount to a public good was greater when it was combined with a weak monetary incentive than when no incentives were offered. But a stronger monetary incentive did not increase contributions, and also had no effect on behavior in the absence of the stated obligation. The authors’ interpretation is that the explicit incentives enhanced the salience of the stated obligation. This is exactly how Cardenas explained a similar result in his experiment with Colombian villagers (in chapter III.).

A Mandate for Aristotle’s Legislator

These experiments, along with Bentham’s reasoning, suggest that policy makers can sometimes find ways to turn the separability problem on its head, making incentives and morals complements rather than substitutes. Does this provide any practical advice to the policy maker?

I think it does.

Table 7.2 shows the main policy implications of the fact that the material inducements and constraints stressed by the conventional policy paradigm are messages as well as incentives; that incentives may impede the evolution of social preferences in the long run; that these moral sentiments are an essential foundation of good government; and that both crowding out and crowding in may occur, depending in part on the policies adopted.

The Legislator should not expect everyone to applaud his attempts to root out the use of incentives that degrade social preferences. Though hard evidence is difficult to come by, incentives that crowd out social preferences are probably fairly common. It would be surprising if those implementing the incentives were entirely unaware of the nonseparability of social preferences and incentives, or of incentives sometimes reducing the joint surplus available in an economic interaction.

Table 7.2. Advice for the Aristotelian Legislator

Why, then, do we ever observe pie-shrinking incentives in practice? The pie metaphor gives away the answer. The person deploying the incentives is not interested in the pie, but in his own slice. Even if incentives reduce the total expected surplus associated with an economic interaction, such as taking out a loan or employing a worker, the use of incentives may give the principal a sufficiently larger slice to more than compensate for the smaller pie.

This is what occurred in an experiment by Fehr and Gaechter with Swiss students.¹⁷ The experiment was similar to the Fehr and Rockenbach Trust game; it included both a standard (called “trust”) treatment and another treatment in which fines were allowed. The payoffs were such that had subjects responded as if they had entirely self-regarding preferences, the joint surplus (the sum of the payoffs to employer and employee) would have been more than twice as large under the incentive treatment as that under the trust treatment.

But the negative synergy between the incentive and social preferences was so strong that the total surplus was much higher in the trust treatment, that is, without the help of incentives. The counterproductive effect of incentives on the total surplus was true even in those cases in which principals offered exactly the kind of contract that a mechanism designer would recommend when subjects are thought to be entirely self-interested.

But here is why incentives were used anyway. Those employers who used these “optimal” contracts in the incentive treatment received payoffs more than twice as large as the average employer profits in the trust treatment, while the payoffs to employees in the incentive treatment were less than half what they were under conditions of trust. The incentive treatment allowed employers to save enough in wage costs to offset the reductions in work effort and the shrunken surplus. They were better off with a larger share of a smaller pie.

Thus, one of the reasons that constitutive aspirations sometimes lead agents to respond negatively to incentives—namely, that responding to the incentive in the way intended would disproportionately benefit the principal—also explains why these incentives may nonetheless be used by profit-maximizing principals, even when they result in a smaller pie. In this case, simply informing the principals that the result of their actions will be a smaller pie is not effective. If a mutually acceptable division of the pie could be decided on in advance (and enforced afterward), this problem would not arise, because incentives could then be devoted exclusively to increasing the size of the pie rather than being hijacked to enlarge one slice at the expense of the size of the entire pie.

Another challenge for the Legislator arises because populations are made up of individuals with differing mixtures of self-interest and the many forms of social preferences. Can he do better than a “one size fits all” approach? He can, but again he should not expect everyone to applaud.

Suppose, unrealistically, that there are just two types of people, the entirely self-interested and those with a degree of altruism. The Legislator would like to encourage contributions to a public good. A monetary incentive (such as a subsidy for contributing) will encourage contributions from the self-interested, but if incentives and social preferences are substitutes, it may work less well or even backfire among the altruists. The altruists might respond well to hearing about the substantial benefits that the public good will confer on others, but this moral message would be wasted on the self-interested.

An obvious strategy for the Legislator is to separate the two populations, addressing each with the appropriate policy. But this will be difficult because the Legislator does not know each person’s type (it is private information), nor will individuals happily sort themselves into the two obvious categories of citizens. But voluntary-separation strategies can sometimes succeed, at least approximately. The modest salaries at a nonprofit public-service organization may deter those unmoved by the organization’s mission, while the prospect of serving the mission provides enough additional compensation to attract the committed.¹⁸

Coordinating social action in a population with mixed motives grows more challenging when we consider the diversity of motives commonly grouped under the heading “social preferences.” The Legislator, in designing policies for a population with multiple seemingly prosocial motivations, will quickly find that not all good things go together. Here he confronts the opposite of Adam Smith’s invisible hand, which induces self-interested individuals to act in the public interest: instead, in a population with a multiplicity of social preferences—altruism as well as reciprocity, for example—a perverse alchemy can transform good motives into unwanted social outcomes.

Here is how I discovered this. In an experimental Public Goods game, my coauthors and I observed (as expected from figure 7.1) that high levels of cooperation were sustained by peer punishment. We then identified the types of the players: “altruistic,” “reciprocal,” and neither. “Reciprocals” were those who contributed generously when their group members had done the same in previous rounds, but contributed little otherwise; “altruists” contributed generously irrespective of what the others had done; the rest contributed little under any condition.

We found that while the altruists contributed generously to the public good, they were less likely to inflict peer punishment on free riders. Instead, when it came to sustaining the social norm of contributing, it was the altruists who free rode on other group members’ willingness to sacrifice their own payoffs in order to pay to punish low contributors.¹⁹

Stimulated by this result, Sung-Ha Hwang and I studied the case in detail, wanting to know whether the pattern we had found was just a curiosity or instead might be something the Legislator should worry about. Our resulting paper, “Is Altruism Bad for Cooperation?” showed that under quite plausible conditions, it certainly could be.²⁰ (When you see a rhetorical question like this in the title of a paper, you can guess the answer.) When individuals are both altruistic and reciprocal in varying degrees, increasing their degree of altruism can reduce the average level of contributions to the public good. The reason is that as those with reciprocal preferences become more altruistic, their willingness to punish free riders diminishes, and this indirect effect of their greater altruism can offset the direct effect of altruism in increasing contributions.

How can the Legislator use this information?

If, like his Athenian antecedents, he planned to engage in moral suasion, he might consider seeking to inculcate either altruism or reciprocity in the citizenry—but not both, at least not in the same individuals. He might also consider segregation if he could do it without incurring the public’s wrath at his “moral apartheid.” Here is how segregation would help.

He could increase public-goods contributions if he could find a way to segregate the self-interested types and the reciprocators into one subpopulation, and let the altruists have their own group. (I am assuming, probably unwisely, that none of this would affect the long-run distribution of types in the population.) Each group could have its own dedicated policies. But other than the separation itself, no further policies (including incentives) might be required, since the reciprocators’ predisposition to punish self-interested free riding might sustain high levels of contribution in that group, as we have seen. Altruists, for their part, might happily contribute.

However off-putting this illustration of the segregation strategy may be to liberal sensibilities (including mine), do not think that separating populations according to their motivations and then designing incentives and punishments accordingly is entirely whimsical. The legal scholar Lynn Stout has proposed that we think along similar lines about the appropriate compensation for injuries due to faulty products:

Most people have an “internal” incentive, in the form of conscience, to take modest care to avoid harming others. Corporations may lack this incentive… .

… The traditional tort pattern of undercompensating victims … [does] not necessarily pose a problem when we are dealing with humans, the vast majority of whom [have consciences]. For natural persons, partial liability may be enough, when added to internal sanction of conscience, to motivate most to take care of harming others.

But the same undercompensation pattern may produce too little deterrence when applied to corporations… . We may want corporate defendants to pay victims more in damages than human defendants must pay.²¹

Stout’s argument is not that the humans who make corporate decisions are less moral than others (remember the Costa Rican CEOs). It is, instead, that when deciding on an appropriate level of care toward others (in the design of children’s toys, for example), the responsibility of managers to maximize profits on behalf of shareholders should induce them to take into account the expectation of undercompensation of damages, were these to occur. And this would lead them, if they are faithful in their duty to enhance the wealth of the owners of their corporation, to take insufficient account of the costs that a design flaw might inflict harm on a user of the product, perhaps choosing as a result a design cheaper to produce but more likely to do harm. Stout is merely repeating Milton Friedman’s celebrated argument in his essay “The Social Responsibility of Business Is to Increase Its Profits,” namely, that “only people can have responsibilities.”²²

Stout might have added, on the basis of the experimental evidence presented earlier, that the diffused responsibility of the decision-making process and the competitive pressure for a firm’s survival faced by managers would work in the same direction.²³

The final challenge for the Legislator follows from the observation that the outcome of his policy interventions will almost never be a simple average of the behaviors characteristic of each personality type in the population. Instead, the outcome will depend on the composition of the population and the social institutions, including informal rules, that determine how the individual’s actions add up to aggregate outcomes.

This idea, sometimes summarized a little oversimply as “the whole is not simply the sum of its parts,” is not new. It has been around in economics since Adam Smith explained how, by the alchemy of the invisible hand, the self-interest of the brewer, the baker, and the butcher would put someone else’s dinner on the table. And it has been around in political philosophy for much longer: the well-ordered society of Machiavelli was an emergent property of an entire system of government, not a simple aggregation of the qualities of the citizens (as we saw in chapter II).

The challenge that Machiavelli’s “good governance as an emergent property” approach presents to the Legislator is this. Under some rules—for example, the Public Goods game (without punishment)—the actions of the self-interested induce even the civic-minded to act as if they care only about their own gains. The Legislator’s opportunity is that under not very different rules, the reverse occurs.

We have already seen this in the Public Goods with Punishment experiment (figure 7.1). In the absence of the peer-punishment opportunity, even those predisposed to contribute substantial amounts eventually came to act as if they were self-interested. But once the peer punishment of free riders was allowed, the very same population converged on a substantial contribution level, which in the later periods of the game was sustained with very little actual punishment, because egregious free riding had all but disappeared. A combination of the incentive to avoid punishment and the shame of having experienced it apparently led the self-interested to act “as if they were good.”

The Legislator’s aim in situations like this is to design rules—like the Public Goods with Punishment game—allowing the civic-minded, not the self-interested, to determine the outcome. To see the rudiments of what this requires, imagine that there are just two citizens, who will interact once in a symmetrical Prisoner’s Dilemma game (symmetry means that the payoff matrix for the game is identical for each). One of them is known (by both the Legislator and the other player) to wish simply to maximize her payoffs in the game. The other has reciprocal preferences (also known to all): he would prefer to cooperate, but only if the other does.

You already know from what happens in the sequential Prisoner’s Dilemma how the Legislator could tweak the rules of this game so that the efficient and equal-sharing outcome—both cooperate—will occur. In the conventional game (when the two make their choices simultaneously), both will defect (the reciprocator knows that the other will defect because it is the dominant strategy for her, so he does too). But the Legislator could change the rules so that the selfish player goes first. She knows that because the reciprocator will move second and, out of reciprocity, will mimic whatever she does, there are now only two possible outcomes of the game: either both cooperate (and both make higher payoffs) or both defect (and they both do less well). By cooperating, the selfish citizen can bring about the former outcome; she does this, and the other reciprocates.

Have we come full circle? Is this, in the end, what Hume had in mind when he imagined that a good constitution would harness the avarice of knaves? It is not, because essential to the happy outcome in the sequential Prisoner’s Dilemma, and in the Public Goods game too, is the presence of at least some ethical or other-regarding citizens who, under the proper rules, induce the “wicked” to act as if they were not. That is why, as I say in my subtitle, good incentives are no substitute for good citizens.

The possibility that the right incentives, laws, and other rules of the game can sustainably crowd in social preferences, rather than the opposite, suggests that the Legislator might hope to do better than limiting himself to Rousseau’s seemingly prudent injunction with which we began this study (in chapter I) to take “people as they are and laws as they might be.” If the Legislator needs a bumper sticker, he might use the subtitle Stout chose for her book on law and morality: Good laws make good people.

Laws as They Might Be for Citizens as They Might Be

Like Besley’s “creating better people,” Stout’s injunction to “make good people” is a jarring expression, but it is hardly a novel idea. Parents try to do this, as do teachers, religious leaders, and others. It is hard to imagine a viable society in which activities aimed at making good people are not widely practiced. Certainly there is no ethnographic or historical record of a successful society indifferent to virtue.

What is novel in the Legislator’s plans (at least for a liberal society) is the idea that making good people should be a public policy objective. Compulsory schooling continues to be advocated as a way to teach social norms, which it apparently does with some success. The fact that children rather than adults are involved appears sufficient for many to treat compulsory schooling as consistent with the liberal commitment to neutrality in the question of preferences. But the historical experience of state projects of cultural transformation for entire populations is hardly encouraging.

The German Democratic Republic, like many Communist Party-ruled societies, invested considerable resources directed to the creation of more solidaristic and less selfish citizens. But in a recent experiment, adults born in East Germany cheated for monetary gain twice as frequently as did those born in West Germany, and this was particularly true of those who had reached maturity before the fall of the Berlin Wall.²⁴

Nonetheless, the Legislator’s amendment of Rousseau’s injunction is now being taken up in surprising quarters, and in pursuit of objectives quite different from the “new socialist man.”

As the housing bubble burst in 2008 and the financial crisis unfolded, many U.S. homeowners found that their property was worth less than their mortgage obligation to the bank. Some of these “under water owners” did the math and strategically defaulted on their loans, giving the bank the keys and walking away. Unlike the New York Times editorial from two decades earlier (“Ban Greed? No: Harness It,” cited in chapter I), the executive vice president of Freddie Mac, the Federal Home Loan Mortgage Corporation, made a distinctly Aristotelian plea for moral behavior in the economy: “While a personal financial strategy might argue for a strategic default, entire communities and future home buyers can be harmed as a result. And that is why our broader social and policy interests will be best served by discouraging strategic defaults.”²⁵ Rather than trusting that the market, by getting the prices right, would induce people to internalize the external effects of their actions, Freddie Mac urged “borrowers considering a strategic default [to] recognize the damaging impact their actions can have on others.” He was hoping, in short, that morals would do the work of prices.

There was no shortage of moral reasoning on the question. In surveys, large majorities held that strategic default was immoral.²⁶ Most defaults were not strategic at all: they were impelled by job loss or other misfortunes. But Freddie Mac’s plea for morality from underwater debtors could not have been very persuasive for those who accused the financial institutions of adhering to a double standard. After pursuing their own interests single-mindedly for decades, they implored homeowners to act otherwise when their own house of cards tumbled. The main determinant of strategic default was economic: how far underwater the property was. But many people condoning the practice gave moral concerns—such as fairness and predatory banking—as prominent reasons.

Freddie Mac-style moral exhortation alone is also likely to be less effective than policies, inspired by Bentham’s punishments, that convey a more resonant “moral lesson.” Perhaps Bentham had in mind the charivaris of early modern Europe: neighbors, typically women, would surround the home of a philandering husband, a price-gouging baker, or a local dignitary exploiting his status for commercial gain, and then beat pots and pans to express their moral indignation.²⁷ The tradition lives on: the municipal commissioner of the Indian city of Rajahmundry (in Andhra Pradesh) hired ten drummers and directed them to beat nonstop outside the homes of tax evaders.²⁸ The drummers said nothing, but the fact that they were accompanied by tax collectors and other officials conveyed a clear message. The policy worked, apparently by evoking the shame of the tax evaders at their transgression of a social norm.

In Bogota, Antanas Mockus, a two-term mayor, recruited hundreds of mimes in white face and clown outfits and sent them into the chaotic city traffic to poke fun at jaywalkers and to lampoon drivers violating the black-and-white pedestrian crosswalks called “zebras.” Taxi drivers were particularly notorious for their rude driving.²⁹ Citizens were invited to nominate a taxi driver for pedestrian-friendly driving, and the first 140 nominated became the founding members of the Knights of the Zebra. At the induction ceremony, Mayor Mockus—a mathematician and philosopher by training—gave each knight a plastic zebra to hang from his or her rearview mirror (yes, there was at least one Lady of the Zebra.)

During his second term (2001-3), he explained that one of the four key ideas of his Civic Culture Program was to “increase the capacity of some citizens to encourage others towards peaceful compliance with rules.”³⁰ To assist in this, he issued hundreds of thousands of thumbs-down cards, which were enthusiastically flashed to drivers transgressing traffic norms, like soccer referees’ red cards penalizing players for fouls. There were also thumbs-up cards to recognize acts of traffic kindness. At the same time, the city ordered its police to enforce traffic regulations more strictly.

The year the mayor took office, the incidence of traffic-related deaths exceeded the national average by a considerable amount; when he left, it had fallen to well below the national average. It continued falling, in both relative and absolute terms, after his departure, in a few years reaching just one-third of its former level.)

Equally effective was a campaign to save water when a severe shortage resulted from the collapse of a tunnel that carried water to the city. Appeals from Mockus included a television commercial of the mayor and his wife sharing a shower, the water turned off while soaping. The city awarded prizes to those who led in water saving, and called out water hogs, subjecting them to well-publicized (but modest) penalties. Water consumption fell 14 percent in two months.

Though a causal link to the Civic Culture Program of Mayor Mockus would be impossible to establish, the remarkable result of an unpublished experiment by Sandra Polanía-Reyes could indicate the impact of the mayor’s attempt to empower citizens to call out those violating social norms. She implemented a Public Goods with Punishment game identical to that used by Benedikt Herrmann and his coauthors (presented in chapter V; see figures 5.3 and 5.4). Her subjects at the Universidad de los Andes in Bogota had spent a good part of their adolescence with Mockus as their mayor. Unlike the “Arabic speaking,” “Southern Europe,” and “orthodox / ex-Communist” populations who played this game, but like the students in Boston, Copenhagen, and Seoul, the Bogatanos carefully targeted free riders for punishment, who responded positively, resulting in a level of eventual cooperation on a par with that of the most cooperative of the world’s student populations.

Like the tax on plastic grocery bags in Ireland with its surrounding publicity and like drumming back taxes out of evaders in India, the mayor’s highly effective campaign to tame Bogota’s traffic and to save water attached an unmistakable moral message to formal enforcement and material incentives as well as to more informal peer pressure: driving your car aggressively was not cool. Until the water tunnel was repaired, even cleaning a car raised eyebrows (car washes were prominent among the water hogs).

This is what the Athenian assembly did in mounting its Adriatic mission. The objective in all four cases is worthy of the Aristotelian Legislator: to encourage civic action by appealing to both material interests and moral sentiments, framed so that the two work synergistically rather than at cross-purposes. There are limits, of course: not all subsidies can be awarded as prizes, and not all penalties can be flashed like a referee’s red cards.

There are more fundamental limits too.

A world hoping for saints might reflect on how the idea of a constitution for knaves came about in the first place, and why it gained such wide acceptance. An important part of this process (as we saw in chapter II) was the realization that saints are not the only alternative to knaves. There are also zealots, for example, those with unbridled intolerance and hatred of those who fall on the wrong side of the us-and-them divide.

Hume and the economists who followed him were confident (unduly so, we have seen) that a good constitution could harness the avarice of knaves to serve the common good. But harnessing the passions of zealots was an entirely different proposition.

When the economist Charles Schultze wrote that “market-like arrangements reduce the need for compassion, patriotism, brotherly love, and cultural solidarity,” he considered this a feature, not a bug.³¹ One could agree with him, and not only because in an economy of strangers there might not be enough love to make “the world go around,” as Alice had feared in her whispered aside in response to the Duchess. One might also worry that a social order requiring, say, patriotism and cultural solidarity for its proper functioning might also provide justification for divisive and intolerant sentiments, just as the arguments in favor of the invisible hand assisted in the social acceptance of a motivation that had previously been counted among the seven deadly sins.

The idea of a constitution for knaves long predated Smith, but his reasoning about the invisible hand led to something important: a mechanism (or more accurately, a set of them), albeit imperfect, for directing self-interest to public ends through a combination of competitive exchanges of privately held property and public policies aimed at getting the prices right where markets failed to do this. Thanks to economists from Smith to Arrow and Debreu and right up to modern mechanism design, we know what a constitution for knaves looks like.

And now that we understand what it takes to make it work, we find it wanting, not only for its inability to implement an efficient use of resources in a liberal society but also for its likely social and cultural effects. Recent evidence that other-regarding and ethical motives are common in most populations greatly enhances the space of feasible policy interventions, which can include, for example, a wise combination of positive incentives and punishments with moral lessons, such as the mix of motivations appealed to by the decree of the Athenian assembly. But while we thus need an alternative to the constitution for knaves, we do not yet have any similar conception of a constitution for a population made up of knaves, saints, and zealots, in which the constitution will influence not only the quality of governance that will emerge as a result, but also the proportion of these types making up the dramatis personae of our social life.

Considered dynamically in this way, and taking account of the dark side of social preferences, the Legislator’s challenge is far more difficult than heeding Hume’s maxim. Some social preferences may be more difficult than self-interest to channel into socially valued or at least harmless ends. And the positive social preferences of generosity, fair-mindedness, and other civic virtues are a fragile resource for the policy maker, one that may be either empowered by legislation and public policy or irreversibly diminished. This suggests the following extension of Hume’s maxim about knaves: good policies and constitutions are those that support socially valued ends not only by harnessing self-interest but also by evoking, cultivating, and empowering public-spirited motives.

It won’t work as a bumper sticker. But the need for something along these lines is clear in light of Arrow’s point (in chapter I) that social norms facilitate mutually beneficial economic interactions in those cases where contracts cannot cover everything that matters to the parties to the exchange. Examples include the work ethic of the employee, the creativity of the software engineer, or the honesty of the borrower or asset manager. The force of Arrow’s argument is likely to increase as the wealth of nations shifts from steel, grain, and other goods readily subject to contract to producing and sharing intangible knowledge, caring for the young and the elderly, and the other forms of wealth characteristic of what is called the “weightless economy.”

The same conclusion follows from the fact that many of the greatest challenges now facing the world—epidemics, climate change, personal security, and governing the knowledge-based economy—arise from global and other large-scale human interactions that cannot adequately be governed by channeling entirely self-interested citizens to do the right thing by means of incentives and sanctions, whether provided by private contract or government fiat. With economic inequality increasing in the world’s major economies, one may now doubt Dr. Johnson’s reassurance that “there are few ways in which a man can be more innocently employed than in getting money.” The idea of an economy of knaves now appears to be anything but harmless.

I do not know whether an approach to constitutions, incentives, and sanctions adequate to this challenge can be developed. But we have little choice but to try. The Legislator’s mandate is a place to start.