The Moral Economy: Why Good Incentives Are No Substitute for Good Citizens - Samuel Bowles (2016)
V. A Liberal Civic Culture
If incentives sometimes crowd out ethical reasoning, the desire to help others, and intrinsic motivations, and if leading thinkers celebrate markets as a morality-free zone, it seems just a short step to Karl Marx’s broadside condemnation of capitalist culture: “Finally there came a time when everything that men had considered as inalienable became an object of exchange, of traffic and could be alienated. This is the time when the very things which till then had been communicated, but never exchanged, given but never sold, acquired but never bought: virtue, love, conviction, knowledge, conscience—when everything passed into commerce. It is the time of general corruption, of universal venality.”1
But a century and a half later, “universal venality” fails to describe the cultures of northern Europe, where capitalism was born, or the North American and other offshoots of these populations. Or at least so it must seem to the parking enforcement division of the New York Police Department.
Diplomatic immunity from prosecution for traffic violations in New York City provides a natural experiment for testing Marx’s prediction.2 In deciding where to park and whether to pay any fines incurred, the diplomats of 146 nations differ markedly in how often they willingly break the law and inconvenience others by parking illegally and by not paying the resulting tickets.
In the five years before November 2002, the average number of violations per diplomat was 19, with the most flagrant violators being those from Egypt (140 violations per diplomat), Bulgaria (117), Albania (84), and Pakistan (69). Over the same period, the 31 diplomats from the United Kingdom, where capitalism was born, committed exactly zero violations, as did those from Sweden, Norway, Canada, and the Netherlands, the second birthplace of capitalism. Those from other early capitalist nations posted modest numbers of transgressions during this period: 1 per diplomat for Germany, 2.7 for Belgium. Some latecomer capitalist nations were likewise paragons of parking probity: Japan’s 47 diplomats posted not a single violation, and Korean diplomats averaged just 0.4 violations.
Do not read too much into this. It is far from an ideal experiment. And as we will see, Adam Smith warned that unlike merchants, diplomats were not to be trusted. So perhaps the people with D license plates are not representative of their cultures.
But the puzzle is real in light of the experimental evidence suggesting that explicit economic rewards and penalties sometimes drive out social preferences. We will shortly turn to data more convincing than the parking transgressions of the world’s diplomats, evidence based on cross-cultural comparisons of behavior. These experiments show that the oldest capitalist societies have sustained vibrant civic cultures characterized by widespread conformity to cooperative and generous social norms.
Were this a lecture, a person would at this point likely jump up and ask me something like “what planet are you on?” and follow that with a litany of counterevidence, everything from the unethical behavior of hedge fund managers to the fact that Americans bowl alone. I do not wish to exaggerate the cultural virtues of the oldest capitalist societies, but only to point to some apparent differences between them and many other societies in which the reach of market institutions has been more recent and more restricted.
The puzzle would be easily solved if the behavioral effects observed in experiments were either ephemeral or strictly limited in domain, if, for example, people’s moral disengagement when offered incentives for a particular task at work did not spill over to other domains, such as family life or citizenship. But we will see that for better or worse, the economy is a great teacher, and its lessons are neither fleeting nor confined within its boundaries.
I will also offer reasons why living in a highly incentivized economy might have long-term adverse effects on the process of cultural evolution beyond those stemming from the negative situational cues sometimes associated with incentives as messages. The upshot is that the extensive use of incentives may adversely affect the evolution of civic preferences in the long run. In the terminology of the previous chapter, “endogenous preferences” are analogous to the sometimes-adverse effect of incentives on “situation-dependent” preferences.
These further thoughts about the puzzle thus only deepen the mystery. But perhaps the admirable civic cultures of many of the longest-standing capitalist economies owe more to the liberal social order in which these economies are embedded than to the extensive role of markets and incentives per se. This is the resolution of the puzzle that I propose.
By a “liberal society,” I mean one characterized by extensive reliance on markets to allocate economic goods and services, formal equality of political rights, the rule of law, public tolerance, and few barriers to occupational and geographic mobility based on race, religion, or other accidents of birth. Some examples of liberal societies in the experimental studies I will introduce are Switzerland, Denmark, Australia, the United States, and the United Kingdom. Examples of societies that I do not term liberal (lacking at least one of the above attributes) are Saudi Arabia, Russia, Ukraine, and Oman as well as small-scale societies of hunter-gatherers, herders, and low-technology farmers.
The Economy Produces People
How people interact in markets and other economic institutions—who meets whom, to do what, with what rewards—durably shapes social norms and preferences, and these are then generalized to noneconomic domains of life. This has long been recognized. Marx was not alone in holding this view.
The royalist Edmund Burke lamented that the French Revolution had ushered in “the age of sophisters and economists”: “Nothing is left which engages the affection on the part of the commonwealth … so as to create in us love, veneration, admiration or attachment. All the decent drapery of life is to be rudely torn off.”3
Others have taken a more benign view of the cultural consequences of markets. The baron de Montesquieu wrote that “where there is commerce the ways of men are gentle.”4 But the idea that the economy produces people as well as goods and services is not in question.
The long-term effects on preferences to which Marx, Burke, Montesquieu, and others refer are quite different from the incentive effects we saw in the previous chapter, in which fines and subsidies affect preferences because they alter the situation in which a person finds herself, increasing the salience of some preferences and diminishing others. Making sense of Marx, Burke, or Montesquieu is easier if we recognize that incentives may also alter the process by which people come to acquire new tastes, habits, ethical commitments, and other motivations.
The way in which we come to have our particular preferences is much like the way in which we come to have our particular accents. The process takes place early in life, is for the most part unwitting, and depends critically on our social interactions with others. While midlife changes in preferences (like acquiring a new accent) may occur, the learning process is strongly attenuated after adolescence.
When it comes to the effects of incentives, the key difference between endogenous and situation-dependent (or framing-sensitive) preferences is that in the former case, incentives may affect a long-term learning process whose results persist over decades, even entire lifetimes. By contrast, when preferences are situation-dependent, a new situation—such as the withdrawal of an incentive—changes which preference in a person’s repertoire will motivate behavior. Incentives affect preferences in both situation-dependent and endogenous preference cases, but the effects differ. In the former case, the incentive is a reversible signal about the principal or the situation; in the latter, the incentive alters the long-term, not easily reversed preference-learning process.
The developmental processes involved in learning new preferences typically include the effects of interactions over long periods with large numbers of others, such as the processes that occur in schooling, religious instruction, and other forms of socialization not readily captured in experiments. Conformism to widely practiced behaviors (and the adoption of the preferences that motivate them) is typically involved. We cannot, therefore, hope for the kind of experimental evidence on the evolution of preferences that is possible when we study the effects of incentives as messages about situations. But historical, social survey, and ethnographic data, while not directly related to the use of incentives, are quite consistent with the view that economies structured by differing incentives are likely to produce people with differing preferences.5 Here is some of the evidence.
Over four decades, the social psychologist Melvin Kohn and his collaborators have studied the relationship between, on the one hand, a person’s position in the authority structure of the workplace—giving as opposed to taking orders, designing incentives rather than being their target—and, on the other, one’s valuation of self-direction and independence in one’s children, as well as one’s own intellectual flexibility and self-directedness.6 They found that “the experience of occupational self-direction has a profound effect on people’s values, orientation, and cognitive functioning.”7 The studies take account of the problem of reverse causation—the possibility that personality affects the occupational situation one ends up in rather than vice versa—and they provide convincing evidence that there is a causal arrow from job to preferences.
Kohn’s collaborative study of Japan, the United States, and Poland yielded consistent findings across cultures: people who exercise self-direction on the job also value self-direction more in other realms of their life (including child rearing and leisure activities) and are less likely to exhibit fatalism, distrust, and self-depreciation.8 Kohn and his coauthors reason that “social structure affects individual psychological functioning mainly by affecting the conditions of people’s own lives.” They note in conclusion: “The simple explanation that accounts for virtually all that is known about the effects of job on personality … is that the processes are direct: learning from the job and extending those lessons to off-the-job realities.”9
Additional evidence comes from a study by the anthropologists Herbert Barry, Margaret Child and Irvin Bacon. They categorized seventy-nine mostly nonliterate societies according to the prevalent form of livelihood (animal husbandry, agriculture, hunting, or fishing). They also measured whether food could easily be stored—a common practice in agricultural and herding economies but not among foragers—and other forms of wealth accumulation, the latter being a major correlate of dimensions of social structure such as stratification.10
Barry and his coauthors also collected evidence on forms of child rearing, including obedience training, self-reliance, independence, and responsibility. They found large differences in child-rearing practices. These varied significantly with economic structure, controlling for other measures of social structure such as extent of polygyny, levels of women’s participation in the predominant subsistence activity, and size of population units.
Where food storage was common, parents valued obedience rather than independence in children to a far greater extent than where storage was absent. The authors concluded, “Knowledge of the economy alone would enable one to predict with considerable accuracy whether a society’s socialization pressures were primarily toward compliance or assertion.”11 The causal arrow is unlikely to run from child rearing to the type of economy, since the latter is dictated primarily by what combination of hunting and gathering, herding, or cultivation best provides a livelihood in the geographic area concerned.
These society-level studies cannot isolate the effects of incentives per se. The most that cross-cultural ethnographic studies can provide is evidence that preferences vary with economic structure. It is not difficult, however, to explain why differences in economic structure—and the extensive use of incentives in particular—might affect the evolution of preferences.12
Incentives and the Evolution of Preferences
Incentives and other aspects of economic organization affect the evolution of preferences because they influence both the types of people one encounters and the set of behaviors that are feasible and rewarding, given the kinds of tasks that people undertake.13 The extensive use of incentives—say, a subsidy given to those who contribute to a public good—may impede the learning of prosocial preferences because of two uncontroversial aspects of the process of cultural evolution. First, people tend to adopt ways of behaving (including the preferences that motivate them) that they perceive to be common, independently of expected material payoffs of these behaviors. Second, the presence of incentives may lead people to interpret some generous and other-regarding acts as instead being expressions of self-interest induced by the subsidy.
The conformist element in cultural transmission (adoption of the behaviors that most others are doing) is in part the result of the powerful effect of mere exposure on social learning, which has been documented by the psychologist Robert Zajonc and subsequent authors.14 For example, U.S. students were exposed, with low or high frequency, to a list of twelve nonsense English “words” (“kadirga,” “zabulon”), as well as to an equal number of invented Chinese characters, and asked to rate them on a good-bad scale. With a single exception (one of the “Chinese” characters), the more that subjects were exposed to a word or character, the more likely they were to think that it referred to something good. The exposure effect is one of the many reasons that cultural transmission may have a conformist element, favoring the numerous over the rare, independently of the economic success associated with the behavior involved.
The second element in this explanation is that the presence and extent of incentives to contribute to a public project (or to engage in similar activities that benefit others) make the behavior (contribution) a less convincing signal of an individual’s generosity, resulting in observers interpreting some generous acts as being merely self-interested. To see why this is so, we return to the psychologist Lepper and his coauthors: “When an individual observes another person engaging in some activity, he infers that the other is intrinsically motivated … to the extent that he does not perceive salient, unambiguous, and sufficient extrinsic contingencies to which to attribute the other’s behavior.”15The presence of incentives, however, may lead observers to think that a seemingly generous action was done not for the intrinsic pleasure of helping others but as an instrumental response to the incentive.
There are two reasons why the presence of an incentive may lead people to mistake a generous act—helping another at a cost to oneself—for a self-interested one. The first is that the incentive provides a competing explanation of the generous act: “he did it for the money.” The second is that incentives often induce individuals to shift from an ethical to a payoff-maximizing frame (even, as we have seen, relocating neural activity to different regions of the brain). Knowing this, the presence of an incentive for an individual to help another may suggest to an observer that the action was self-interested.16
Taken together, these two facts—that incentives may reduce the perceived level of generosity in a population and that people tend to adopt common behaviors (and the preferences motivating them)—have an important implication. By reducing the perceived frequency of individuals who act out of generous preferences, the extensive use of incentives may lead, via the conformist effect on how we learn new behaviors, to the disadvantaging of generous traits relative to self-interested ones in the selection processes by which a culture persists and evolves.
The two causal mechanisms accounting for these possible adverse endogenous preference effects of incentives—conformism and “he did it for the money”—are empirically plausible. But I do not see any practical way of testing this explanation with historical data, since doing so would require finding something that almost certainly does not exist: a sample of otherwise similar societies with measurably different incentive structures, combined with data over a period of generations on social norms. But surprisingly, in light of their inability to capture long-term learning effects, experiments can isolate what appear to be short-term learning effects of incentives per se.
The Persistent Effects of Incentives
We already know that the adverse effects of incentives can sometimes persist even after the incentive is removed. Examples include the lasting tardiness of the Haifa day care parents and the reluctance of the child artists to take up painting long after having previously been promised a reward for their artwork. This is consistent with the idea that the effects of incentives on preferences go beyond situational cues and constitute part of a learning environment in which preferences are durably modified. This appears to be the case in other experiments, too.
Donning the hat of Aristotle’s Legislator, Joseph Falkinger and his coauthors designed an incentive system to induce experimental subjects to contribute to a public good.17 Recall that the Public Goods game is a multiperson version of a Prisoner’s Dilemma; the dominant strategy for someone motivated solely by material self-interest is to contribute less than the amount that would yield the greatest payoffs for all were all to do the same.
Figure 5.1. Effective incentives apparently crowd out social preferences in subsequent periods (Data from Falkinger et al. 2000.)
The Falkinger incentive system worked: as can be seen in figure 5.1, it induced subjects to contribute almost exactly the amount that would be predicted for someone who cared only about maximizing her own material gains in the experiment. In panel A, the horizontal line at 30 is the predicted contribution level of an own-payoff-maximizing individual under the incentive system. The dots, indicating the actual mean contributions in each of the periods, show that the subjects acted almost exactly as the self-interest assumption predicts. This is not unexpected. We saw this in chapter III: when Irlenbusch and Ruchala offered a high incentive for a public-goods contribution, subjects contributed almost exactly what the economics textbooks predict for Homo economicus (figure 3.4). Cardenas et al. saw a similar result in the later periods of the treatment in which they imposed fines for overextraction from the “forest” (figure 3.2).
At this point (and without looking at the rest of the figure), it would be tempting to conclude that the subjects were indeed material-payoff maximizers. But this conclusion would be mistaken. Look at what happened in the Falkinger experiment during periods 21-40 (in panel A): in the absence of the incentive, subjects initially contributed significantly more than 10, the amount that a self-interested payoff maximizer would contribute, even though by period 40 the subjects were contributing exactly what an own-payoff maximizer would give. This shows what by now should not come as a surprise, given the experiments discussed in chapter III: the motives that induced the subjects to contribute generously in the absence of incentives during the second 20 periods (at least initially) were entirely absent when, in the first 20 rounds, the incentives were offered.
Even more interesting, from the standpoint of the durable influence of incentives on preferences, is how the previous experience of an incentive system affects subjects’ behavior after the incentives have been withdrawn. If the effect of an incentive were simply to provide a situational cue, then behavior in the absence of an incentive should not depend on one’s prior experience with an incentive system. In Falkinger’s experiment, however, this was not the case. In the absence of the incentive, the subjects who had previously experienced the incentive system (panel A, periods 21-40) contributed 26 percent less than those who had never experienced the incentives (panel B, periods 1-20). It could be that the framing effect of the incentives persisted long enough to affect subsequent play, or it could be that the incentive actually altered the individual’s preferences in some durable way. Because of the experiments’ short duration, we cannot distinguish between these possibilities.
A durable negative effect of the experience of incentives occurred, too, in an experimental Gift Exchange game implemented by Simon Gaechter and his coauthors.18 In the standard version of the game (before the introduction of fines and bonuses), the subject in the role of a principal like an “employer” chooses a “wage” to offer to his agent, the “employee.” The employee then accepts the wage or not. “Workers” who accept the wage then select a level of “production” that is costly for the worker to provide (providing more reduces the worker’s payoffs) and beneficial for the employer The game lasts just a single period, but it is implemented in a succession of periods, employers and employees being randomly rematched from the pool of subjects after each period.
In this setup, selfish workers would obviously accept any positive wage and then produce nothing, for there would be no way for the employer to retaliate or discipline a nonperforming worker. Knowing this, an employer who cared only about his own payoffs and who attributed the same preferences to the worker would offer a wage of zero; and the worker would reciprocate in kind. Both would then receive a payoff of zero, forgoing the positive profits and wages that would have occurred had a generous wage been offered and had the worker reciprocated with a substantial amount of effort. In light of the other experiments, you will not be surprised to find out that this is not at all what happened.
In addition to the standard (no-incentives) game, there were two “incentive” treatments. In both, the employer offered a contract specifying not just the wage but also a required level of output, with the stipulation that the wage would be paid only if the worker met the employer’s output target. In the “fine” treatment under the incentive setup, failure to meet the target would be penalized by a wage reduction, while in the “bonus” treatment, meeting the target would be rewarded by a wage increase. The standard setup, without targets, bonuses, or fines, is called the “trust” treatment because a principal would offer a positive wage only if he trusted the agent to reciprocate by providing sufficient production.
The authors suspected—based on earlier experiments—that in the standard game (that is, without incentives of any kind) employers would indeed trust and workers would reciprocate. And initially they did, as the dashed line in the phase 1 panel of figure 5.2 shows. But the workers’ production declined over time. By contrast, incentives (either fines or bonuses, shown by the thin and thick lines) sustained substantial levels of production over the entire first phase.
Figure 5.2. Effects of the prior experience of incentives on “effort” During phase 2, no incentives were offered to any of the three groups, and those who had experienced an incentive (either a fine or a bonus) in phase 1 provided significantly lower effort. (Data from Gaechter, Kessler, and Konigstein 2011.)
But the authors wanted to know whether the experience of having first played the game under the fine or bonus treatments would affect how the subjects behaved in the absence of incentives. So in phase 2, all three groups played under the trust treatment, that is, with no incentives in place. The results are in the phase 2 panel of the figure.
In the absence of incentives, the subjects who had experienced either the bonus or fine treatment for the first ten periods produced far less than they had in the incentive treatments. Those who had not experienced incentives in the first phase (the dashed line, again) produced much more than those who had previously been exposed to incentives. For any wage offered by the principal, workers who had experienced incentives in the first period offered substantially (and statistically significant) lower levels of production when, in the second period, the incentives were removed. The experience with incentives appears to have diminished the subject’s motivation to reciprocate the employer’s trust or generosity in offering the wage. Remember: the employer who believed the employee to be entirely selfish would maximize his payoffs by offering a wage of zero.
I say “appears to have “because in this experiment as in most, we have no observations on the subjects’ motivations. Remarkably, the best clues about motivational learning under the influence of incentives come from what may be the first experiment conducted on how incentives may crowd out ethical reasoning, designed by two political scientists.
Norman Frohlich and Joe Oppenheimer implemented a five-person Public Goods game under two conditions. In addition to the standard game (for example, the Falkinger experiment just described, but without incentives), they introduced an ingenious Rawlsian “veil of ignorance” treatment that aligned individual self-interest with the interests of all members of the group, thus abolishing the social dilemma normally constituted by a public goods game. In the Rawlsian treatment, each of the five players chose how much to contribute to the public good, but then received the payoffs of a randomly chosen player. Whatever he contributed, each player had an equal chance of receiving his own payoff or that of one of the other four players. In the usual game, as in the Prisoner’s Dilemma, an individual maximizes her payoffs by contributing nothing. But in the Rawlsian treatment of the game, the best one can do is to contribute the amount that will maximize the average payoff of all five players, since doing so will maximize each player’s own expected payoff. Players had no difficulty figuring out that this required contributing the maximum.
Ten groups played each of these treatments (conventional Public Goods Game and the Rawlsian “veil of ignorance” game), half of them with a brief period of discussion among the players before their (anonymous and simultaneous) play. In the other groups, no communication was allowed. All the groups then entered a second phase in which no communication was allowed and only the standard Public Goods game was played.
Not surprisingly, in phase 1 and in the absence of communication, subjects in the “veil of ignorance” game contributed much more than those in the standard game. Also unsurprisingly, when communication was allowed, contributions in both the “veil of ignorance” game and the standard game were significantly higher.
The authors, however, were more interested in phase 2 of the experiment, when everyone played the standard game. They wanted to know whether having played the “veil of ignorance” game would induce subjects to act more generously toward others. Those who had played behind the veil of ignorance, the researchers reasoned, would recognize the fair and socially optimal outcome (namely, full contribution) and would be more motivated than other players by fairness considerations when they came to play the normal game.
But this is not at all what happened.19 In the groups that had not communicated in phase 1, there was no difference in contributions in phase 2 between those who had played the “veil of ignorance” game and those who had played the standard game. But among those who had communicated during phase 1, in phase 2 those who had played the standard game contributed twice as much as those who had played the “veil of ignorance” game.
The experience of having played a game in which self-interest was a good guide to what was best for all group members (the “veil of ignorance” treatment) apparently made people less able to articulate or respond to arguments of fairness when, in phase 2, they faced a real social dilemma and were allowed to communicate with one another. This may explain why the effect of communication in phase 1 was much greater in the standard game than among those playing behind the veil of ignorance. In fact, those who played the standard game with communication contributed slightly (but statistically significantly) more than those playing behind the veil of ignorance with communication. It may be that those whose play featured a real ethical dilemma (the standard game) had more to communicate than those who had been spared the dilemma by the veil of ignorance.
When surveyed, subjects who had played behind the veil recognized that they were playing fairly. But in phase 2, when they played the standard game, their self-reported concerns about fairness were uncorrelated with (and apparently had no effect on) how much they contributed. Notice that the effect of the veil of ignorance was not to eliminate fair-minded sentiments, but rather to sideline them, in the sense that those who were more fair-minded did not contribute more than those who were indifferent to fairness. By contrast, those who had played the standard game in phase 1 self-reported that they had played less fairly, but in phase 2, fairness concerns were a strong predictor of contributing larger amounts.
The Rawlsian treatment, like the Falkinger incentive system for the Public Goods game, is what economists call an incentive-compatible mechanism, meaning that under the set of incentives and constraints represented by the mechanism, self-interested individuals will implement a socially desirable outcome. In any such a mechanism, as in the Rawlsian veil of ignorance treatment, prices do the work of morals: the material incentives facing each player lead them to implement the social optimum. Here is how the authors explain the results: “The [Rawlsian] mechanism or any other incentive compatible device renders the need to invoke ethical concerns … moot … Both ethically motivated and selfishly motivated players can agree on the best strategy … As a result, when … players [who had experienced the Rawlsian treatment] subsequently have to make ethical decisions they are more likely to downplay the ethical components than are those regular players who have had practice confronting ethical issues.”20
This “use it or lose it” interpretation of the eclipse of moral reasoning is entirely at odds with the view, common among economists, that incentive-compatible mechanisms such as ideal competitive markets or the veil of ignorance treatment in the experiment are to be admired precisely because they economize “the scarce resources of altruistic motivation,” which might otherwise be used up.21
Taking account of the fact that ethical motivation is not just a resource to be economized but also one that can be diminished through nonuse, returns us to Marx’s dismal prognostication about the cultural consequences of capitalism. Competition among self-regarding individuals for goods in competitive markets with complete contracts is widely advocated as an incentive-compatible mechanism with attributes similar to the Rawlsian veil of ignorance. This is the reasoning behind both Buchanan’s indifference toward the condition of his fruit seller, and Gauthier’s claim that morality has no role in evaluating competitive market outcomes. If Frohlich and Oppenheimer are right that citizens in such settings do not have to “flex their ethical muscles,”22 then one wonders why Marx’s prediction that the flowering of capitalism would be a “time of general corruption and universal venality” turned out to be mistaken.
The experimental evidence suggests that ethical crowding-out effects can be substantial and that the lessons of our economic experiences are sometimes long-lasting and tend to be generalized to other domains of life. This knowledge does not much help us resolve the puzzle with which we began.
My research took a surprising turn, one that only further deepened the puzzle, when, with a group of anthropologists and economists, I began to investigate whether societies where markets play a large role suffer a deficiency of social preferences, as would be expected from the above reasoning.
Markets and Fair-Mindedness
Three large cross-cultural behavioral experimental studies in populations with a broad range of economic and political systems have given us behavioral measures concerning individuals’ cooperativeness, fair-mindedness, and other social preferences. In addition to the Third-Party Punishment game, Dictator game, Trust game, and Ultimatum game already described, the Public Goods with Punishment game (described below) provides behavioral measures of generosity, willingness to sacrifice personal benefits to uphold social norms, and readiness to contribute to a public good. These studies show that these behaviors flourish in market-based societies, though to varying degrees.
The most surprising evidence comes from an experimental Ultimatum game played by subject pools in fifteen isolated small-scale societies (not the same fifteen in the study described in chapter IV).23 Recall that in this game one player proposes a division of a sum provided by the experimenter, and the responder can either accept her proposed share or reject it, neither player getting anything in the latter case. Entirely self-regarding proposers who believe that respondents are also self-regarding will anticipate that no positive offer will be rejected, so they will offer the smallest possible amount. This prediction from the assumption of self-interest has rarely been observed in hundreds of experiments in dozens of countries. Our study was no exception. Most proposers offered substantial amounts to the responder; low offers were frequently rejected.
In our study of hunter-gatherers, herders, and low-technology farmers (horticulturalists), the groups with greater average exposure to markets made more generous offers as proposers and were more willing to reject low offers as responders—meaning they were willing to receive nothing rather than accept a highly unequal division of the pie. The two least market-exposed groups—Tanzanian Hadza hunter-gatherers and Amazonian Quichua horticulturalists—offered a quarter and a third of the pie respectively, in contrast with the highly market-integrated Indonesian Lamalera whale hunters, who offered on average a bit more than half the pie to respondents. Considering all the groups, a standard deviation difference in our measure of market exposure was associated with about half a standard deviation increase in the mean Ultimatum game offer.
These were eyebrow-raising findings among anthropologists and other social scientists outside of economics, many of whom think that markets make people selfish. The Wall Street Journal, unsurprisingly, saw things differently. It headlined its January 24, 2002, front-page story about our results, “The Civilizing Effect of the Market.” I spent the next couple of days answering agitated telephone calls from friends.
A second phase of this project (in which I was not involved) studied primarily rural peoples in Africa, Oceania, and South America.24 (This was the project that produced the evidence about the crowding out of religion in the Third-Party Punishment game in Accra). Using improved data and techniques, the positive correlation between Ultimatum game offers and market exposure was replicated (at about the same magnitude), and the authors found a similar positive market correlation for offers in the Dictator and Third-Party Punishment games.
These results are not inconsistent with the experimental evidence presented in the previous chapter. The Accra workers for whom monetary incentives apparently reduced the salience of religion, resulting in less generous behavior, were among the most market-exposed in this study (they purchased all their food, for example, rather than obtaining it from hunting, gathering, or barter) and were also among the most generous, offering well above the average of the fifteen subject pools in the Dictator and Ultimatum games.
Unlike the first phase of the project, the second included one market-based liberal society, the United States, represented by a rural population in Missouri. We can gauge the Missourians’ fair-mindedness in the Ultimatum game by the minimum offer (fraction of the pie) they would accept if the proposer offered it, as reported to the experimenter at the outset of the game. This “minimum acceptable offer,” or MAO, can also be understood as the greatest amount the subject is willing to forgo in order not to accept an unfair offer. The MAO thus captures at once the subject’s “willingness to pay” for fairness and the least advantageous division of the pie the subject considers fair enough not to trigger a rejection.
The MAO of the highly market-exposed Missourians was the third highest among the fifteen subject pools. Controlling for subjects’ age, sex, schooling, and average income, the Missourians’ minimum acceptable offer was 2.6 times the average of the other groups. In the Dictator game, virtually all the Missourians offered half the pie, making them, at least by this measure, the most egalitarian of all the populations. Rural Missourians tend to vote Republican; but from this experimental evidence, they appear to be more concerned about economic inequality than the Hadza hunter-gatherers, whose practices of food sharing and lack of political hierarchy were the inspiration for James Woodburn’s classic paper “Egalitarian Societies.”25 Recall that the Hadza subjects offered a quarter of the pie on average in the Dictator game, and their MAO in the Ultimatum game was less than half of the Missourians’.
Cultural Differences in Cooperation and Punishment
It would help address our puzzle if we had an explanation why rural Republican and highly market-exposed Missourians are more oriented toward fairness in the Ultimatum game than the group that perhaps more than any other has exemplified the egalitarian foraging way of life. Evidence that may help comes from experiments with an unusually diverse set of (also, coincidentally, fifteen) subject pools, including some from liberal societies (the United States, the United Kingdom, Switzerland, Germany, Denmark, Australia). Other pools came from Turkey, Russia, Saudi Arabia, China, Oman, and South Korea. In addition, this study provides an idea that may help resolve the puzzle with which we began. The cultural differences among the subject pools may be somewhat attenuated, however, because unlike the herders, hunter-gatherers, and farmers in the other cross-cultural studies, the subjects here were university students, who may be more culturally similar around the world than those not exposed to the same set of life experiences.26 The experiment implemented (by the same experimenter) in these sites is a Public Goods with Punishment game.
This is a modification of the Public Goods game, the n-player Prisoners’ Dilemma introduced in the previous chapter. The players are each awarded an endowment by the experimenter and given the opportunity anonymously to contribute some, all, or none of it to a common pot (the public good). After all the contributions are made, the amount in the pot is doubled (or in some experiments, tripled) and then distributed equally to the players, irrespective of their contributions. In most versions of this game, the group size and the multiplication factor (that is, doubling or tripling the amount in the pot) are set so that each individual will maximize his or her payoff by contributing nothing, irrespective of what the others do. Yet the total payoff for the entire group is greatest if everyone contributes his or her entire endowment.
For example, if there are five members of the group and the multiplication factor is two (the amount in the pot is doubled by the experimenter before being redistributed to the subjects), then by contributing 1 to the public pot, the pot to be distributed increases by 2, of which I will get my 1/5 share. So by giving up 1, I would increase my payoff from the eventual distribution out of the pot by 2/5, which clearly does not justify forgoing the 1 if one is an own-payoff maximizer. Yet if everyone contributes 1, each will receive 2.
The punishment modification of this game is that after all players have made their allocations, each is provided with information about the other players’ contributions (the identities are not given, just an ID number known only to the experimenter). Each player is then given the opportunity to sacrifice some of his or her own payoff in order to reduce the payoff to any other member in the group. This is the “punishment option,” but it is of course not described in those terms, in order to avoid framing the action as morally motivated.
This procedure is followed in each of the periods of the game (often ten). The presence of an option to punish a player who violates a norm of generosity toward others makes the Public Goods with Punishment game similar to the Third-Party Punishment game played in Accra, which we saw in the last chapter. In that game, however, the punisher was a third party, not one whose payoffs were directly affected by the behavior of the possible target of punishment.
This game provides information on three behaviors motivated by social preferences: willingness to contribute to a public good (public generosity) and to penalize those who do not (upholding social norms), both at a cost to oneself, and the degree of positive response to punishment by others (shame at one’s violation of a social norm). Where all three of these dispositions are present, contributions to the public good are substantial.
The results of this experiment when played around the world are summarized in figure 5.3. The grouping of the subject pools by the authors of the study (though somewhat arbitrary) dramatizes the considerable cultural differences evident in the experimental play. In each of the six panels, the leftmost set of (mostly downward trending) lines of dots shows the average contributions in the first ten periods of the game, when there was no punishment option. The line of dots on the right of each panel records average contributions when the punishment option was introduced.
Figure 5.3. Cultural differences in mean contributions with and without a punishment option in a Public Goods game In each of the panels, the left observations are the treatment without punishment, and those on the right are from the same game with the punishment option. (Data from Gaechter, Herrmann, and Thoni. 2010; the cultural categories are from the original source.)
As expected, cultural differences in game play among the subject pools were significant. But in all of them (as is common in other experiments with the Public Goods game), subjects contributed substantial amounts in the first period.27 In the absence of the punishment option, however, cooperation soon unraveled. When available, the punishment option was widely used, especially in the early periods, and the contributions stayed consistently high in all fifteen subject pools as a result. In the treatment with punishment, the subject pools with the highest average contributions were (in order) Boston, Copenhagen, St. Gallen (Switzerland), Zurich, and Nottingham; the lowest average contributions were in Athens, Riyadh, Muscat (Oman), Dnipropetrovs’k (Ukraine), and Samara (Russia).
Average contribution levels in the subject pools correlated positively with measures (for the populations in which the experiment as conducted) of rule of law (the correlation coefficient between these two measures was r = 0.53), democracy (r = 0.54), individualism (r = 0.58), and social equality (r = 0.65). Positive correlations were also found, as expected, with survey measures of trust (r = 0.38).28
Voluntary contribution to a public good is surely a plausible measure of the civic virtues that Marx thought would die out in a market economy. The same is true of rejections of stingy offers in an Ultimatum game, because they show that people will sacrifice their own material gain to punish those who violate social norms of fairness. That these behaviors are stronger in nations with a greater extent of market interactions is puzzling. Understanding why these cross-country correlations occur will cast further doubt on the idea that market-based economies necessarily promote “universal venality.”
Sustaining Social Order in Liberal and Other Societies
The difference between the cooperating and free-riding subject pools in this cross-cultural study—between, on the one hand, Boston and Zurich, and, on the other, Athens and Muscat—lies in the use of punishment and the response to being punished. In the experiment without the punishment option, subjects in Samara, Dnipropetrovs’k, and Muscat contributed more than those in Boston, Nottingham, and Zurich. The reason these subject pools did less well in the punishment version of the game is that a significant amount of punishment was directed not only at shirkers but also at high contributors. This may have been done in retaliation for punishment received in earlier rounds by subjects who believed (correctly) that the high contributors were doing most of the punishing (figure 5.4). The authors termed this practice—punishment of those contributing the same or more than the subject—“antisocial punishment.” Other experiments have found the same patterns.
The extent of antisocial punishment was significantly and inversely correlated with the societal measures mentioned above: rule of law (r = −0.53), democracy (r = −0.59) individualism (r = −0.63), and social equality (r = −0.72). In the five highest-contributing subject pools (Boston, Copenhagen, St. Gallen, Zurich, and Nottingham), shirkers who were punished responded by significantly increasing their contributions in subsequent periods. In only one of the five lowest-contributing pools (Athens, Riyadh, Muscat, Dnipropetrovs’k, and Samara) did shirkers respond positively to punishment. In the other four, their response was not significantly different from zero.
Figure 5.4. Antisocial punishment in a Public Goods game To the right of 0 are punishments imposed on those who contributed the same or more than the punisher (antisocial punishment). (Data from Herrmann, Thoni, and Gaechter 2008b.)
A plausible explanation for these differing uses of punishment and the responses of its targets is that punishment works only if it is regarded as legitimate and conveys the signal that the target has violated widely held norms. It appears that the punishment of free riders, even by complete strangers, is legitimate and evokes shame, not anger, in Boston and Copenhagen, but is seen differently in Muscat and Samara.
The results of an experimental exploration of the effect of legitimacy on the efficacy of punishment by Archan Ertan, Talbott Page, and Louis Putterman are consistent with this interpretation. Before playing the Public Goods game, each group of experimental subjects in Providence, Rhode Island, was invited to deliberate and then vote on whether peer punishment should be allowed, and whether it should be restricted in any way. This is very unlike the highly individualistic setup of the more common Public Goods with Punishment game, in which the punishment decision is made individually and simultaneously. But the novel communication aspect of the experimental protocol adopted by Ertan and his coauthors matches real-world practices of cooperation and norm enforcement. As is clear from ethnographic and other studies, persuasion, gossip, and ridicule play an important part in sustaining social norms; the punishment of transgressors is rarely carried out by an individual acting alone.29
Here is what the experimenters found: “No group ever allowed punishment of high contributors, most groups eventually voted to allow punishment of low contributors, and the result was both high contributions and high efficiency levels.”30 Apparently, majority determination of the punishment system made the punishment of shirkers not just an incentive but also a signal of group norms.
This result suggests an explanation for the contrasting levels of cooperation sustained by peer punishment in experiments with subject pools from liberal and other societies. Consider the structure of what anthropologists call a “lineage-segmented” society. Lineages are the fundamental social unit, composed of families sharing a (perhaps quite distant) common ancestor. In these societies, families mitigate risks by providing and seeking help during times of need. In addition to risk pooling and redistribution, lineages are responsible for the moral instruction and behavior of their members and for the rectification of any transgressions toward either members or nonmembers, including punishment and compensation where appropriate.31 Punishment by a nonmember for a member’s misbehavior may itself be considered a transgression requiring rectification or inviting retaliation. Ernst Gellner’s description of pastoralists as “a system of mutually trusting kinsmen” is an example. These lineages are “strong, self-policing, self-defending, politically participating groups … They defend themselves by means of indiscriminate retaliation against the group of any aggressor. Hence they also police themselves and their own members, for they do not wish to provoke retaliation.”32
In liberal societies, by contrast, the tasks of moral instruction and maintenance of order are routinely entrusted to individuals who are unrelated and, at least initially, unknown to those whom they teach, police, or judge. In an inversion of the moral code of lineage-segmented societies, the legitimacy of these teachers, police, and court officers is based on their anonymity and lack of relationship to those they interact with. This legitimacy is enhanced by uniforms, academic degrees, and official titles acquired (at least ideally) through a process of fair competition, not family favor.
Perhaps this explains why, when Boston subjects who contributed less than the average in the Public Goods game were punished, they substantially increased their contributions, while under the same conditions, subjects in Dnipropetrovs’k actually reduced theirs (though not by a significant amount). While the incentive to contribute more was no doubt salient in both cases, the message that the incentive sent may have differed. Boston subjects may have read the fine as disapproval by fellow citizens, while those in Dnipropetrovs’k may have seen it as an insult.
My hypothesis is that the different ways that order is maintained in liberal and lineage-based societies are part of the explanation of the cross-cultural differences observed in the experiments. It has yet to be tested empirically, but if it were borne out, it would direct attention not to the cultural consequences of markets but rather to liberal political, judicial, and other nonmarket institutions as the key to liberal civic culture. This differs from the usual explanation of the civic culture of liberal societies, namely, the doux commerce hypothesis, which credits the exchange process itself.
While living in England (1726-29), Voltaire was astounded that at the London Stock Exchange, “the Jew, the Mohammedan, the Christian deal with one another as if they were of the same religion, and give the name infidel only to those who go bankrupt … The Presbyterian trusts the Anabaptist and the Anglican takes the Quaker at his word”; and “upon leaving this peaceful and free assembly, some withdraw to the synagogue, … others retire to their churches … others to have a drink … and everyone is happy.”33 Perhaps he was observing “the civilizing effect of markets,” which the Wall Street Journal celebrated with the help of our experiments.
Understanding why people in the more market-exposed societies in our study made more generous offers in the game, and were more likely to reject low offers, requires two pieces of information. First, in many of the populations we studied, interactions with strangers are often fraught with danger. But this is less true where regular market exchanges occur, for the simple reason that some of the strangers one meets in a market provide opportunities for mutual gain. Second, our experimental subjects played anonymously, which may have cued them to play in a way that would have been appropriate toward strangers. A plausible explanation of the more generous and fair-minded experimental behavior seen in the more market-oriented societies, then, is that people learn from their market experiences that fair dealing with strangers is often profitable. Maybe the members of the London Stock Exchange who so impressed Voltaire had learned something similar.
This possible explanation for the civic values exhibited by market societies suggests that there may be something to the key hypothesis of doux commerce thinkers such as Voltaire and Montesquieu. Adam Smith comes closest to providing a causal mechanism that would explain why markets might foster a robust civic culture. He contrasts the probity of merchants with the untrustworthiness of ambassadors: “When a person makes perhaps twenty contracts in a day, he cannot gain so much by endeavoring to impose on his neighbors, as the very appearance of a cheat would make him lose. When people seldom deal with one another, we find that they are somewhat disposed to cheat, because they can gain more by a smart trick than they can lose by the injury which it does their character.”34
Smith is describing a reputation-based variant of a large class of game-theoretic models: in cases where contracts are incomplete or promises unenforceable, frequent, repeated exchanges with known individuals allow for retaliation against opportunistic behavior. The availability of punishment provides incentives that induce otherwise self-regarding individuals to adopt norms such as honesty and diligence toward their partners, thereby underwriting exchanges whose the mutual benefits would otherwise be compromised by malfeasance.35
If Smith is right, markets with a restricted number of people, in which exchanges take place repeatedly over a long period, might promote honest dealing. And just as Kohn showed that lessons learned at work about authority and independence are generalized to childrearing values and other realms, social norms associated with markets might become more generally diffused. Perhaps this explains why, when the anthropologist showed up with an odd game to play and real money on the table, the subjects from more market-exposed populations were more concerned about fairness and were more generous to their partners compared with other groups.
Smith’s argument that repeated interactions with many individuals all known to one another may promote honest dealing makes sense. But it does not explain why markets should provide a more favorable setting for this dynamic than institutions such as families, states, or teams of people who regularly work together. In these non-market settings, the number of people with whom one interacts is smaller than in most markets, and the repetition of interactions much greater. So Smith’s reasoning should apply with greater force outside the market. I return to Smith’s honest merchant in the next chapter. For now, let me say that I do not think Smith’s version of the doux commerce hypothesis (or any of its variants) adequately accounts for the civic-minded citizenry of many of the highly market-oriented societies. Instead, I think that the explanation has everything to do with nonmarket aspects of liberal social orders.
A Liberal Civic Culture
Here is my proposed resolution of the puzzle of the robust civic cultures in societies in which markets play a major role. I will first explain the reasoning (extending my joint work with Ugo Pagano and his joint work with Massimo D’Antoni), and then present some evidence.36
Liberal states have neither the information nor the coercive reach to eliminate opportunism and malfeasance. But they can and do protect citizens from worst-case outcomes such as personal injury, loss of property, and other calamities. The result, writes Norbert Elias, is a “civilizing process” based on the fact that “the threat which one person represents for another is subject to stricter control”; as a result, “everyday life is freer of sudden reversals of fortune [and] physical violence is confined to the barracks.”37 This attenuation of calamity is accomplished through the rule of law, occupational and other forms of mobility allowing people an exit option when faced with cataclysmic loss, and, more recently, social insurance.
By reducing risk, these aspects of liberal society become substitutes for the kinds of familial and parochial ties on which lineage segments and other traditional identities rest. Because these ties become less valuable, they are less likely to be sought and maintained. The result is a cultural environment favorable to the evolution of universal norms, which apply to strangers as well as to the clan. In addition, de facto insurance against worst-case outcomes may free people to act on their social preferences by ensuring that those who conform to moral norms of generosity or cooperation will not be exploited by their self-interested fellow citizens.
While this risk-reduction aspect of liberal society affects the entire panoply of social interactions, I will illustrate it by showing how it would promote trust in a market exchange. (A game-theoretic model of this argument appears in appendix 4.) Consider a population composed of a large number of people who interact in pairs in an exchange in which they may either behave opportunistically (for example, by attempting to steal the other’s goods) or trade goods for their mutual benefit. Call these strategies “defect” and “cooperate.” Suppose a defector takes the goods of a cooperator but runs the risk of getting a beating for his behavior; in this case, cooperating is a best response to being paired with a known cooperator. Defecting is always the best response to a defector, because the defector can fully exploit the undefended cooperator. Though mutual cooperation maximizes total payoffs (and also the individual payoffs for both individuals), a trader paired with an unknown stranger would defect in the absence of a reasonable assurance that the stranger is a cooperator.
Put yourself in the shoes of a trader facing an unknown potential trading partner. How confident would you need to be that your exchange partner is trustworthy (that is, will not defect) in order for you to cooperate? This “minimum degree of confidence required to trust the other” will depend on the consequences of inadvertently cooperating with a defector. If being exploited by a defector inflicted serious costs on you as a hapless cooperator, you would cooperate only if you were virtually certain that he was trustworthy. If, on the other hand, the worst that could happen to you as a naïve cooperator was not that bad, you would be much more willing to take a chance on trusting an unknown person.
The rule of law and other aspects of the liberal state make the consequences of mistakenly trusting a defector much less dire. As a result, the rule of law lowers the bar for how much you would have to know about your partner before trusting him. Thus, the rule of law could promote the spread of trusting expectations and hence of trusting behavior in a population. John Rawls provides a complementary argument: “when it is dangerous to stick to the rules when others are not,” “public institutions” may penalize defectors, thereby reducing their numbers. This lowers the probability that a cooperator will be exploited by a defector, and so minimizes the would-be cooperator’s motivation to preemptively defect as a strategy to reduce risk.38
Markets, of course, are part of this story. In the example above, the occasion for a trusting relationship between buyer and seller would not have arisen without the possibility of mutual gain through exchange. The synergistic effects of markets and the rule of law in favoring the evolution of trust among strangers may account both for Voltaire’s observations of cooperation among men of different religions on the London Stock Exchange, and for the surprising results of our cross-cultural experiments. A summary of the causal model I have just described is in figure 5.5.
The other way that the rule of law may support social preferences is by protecting those who conform to social norms from exploitation by defectors. This protective effect may occur not only because defectors are fewer under the rule of law (Rawls’s argument) but also because the knowledge that defectors will be punished for their transgressions may reduce the appeal of preemptive defection by someone who would otherwise prefer to cooperate.
Figure 5.5. Markets, liberalism, and civic preferences Arrows indicate causal effects, with (+) and (−) indicating, respectively, “enhances” and “undermines.” The (+) on the arrow from markets to civic preferences recognizes the force of the doux commerce arguments that markets may facilitate cooperative relationships with strangers (as suggested by the cross-cultural experiments).
This second crowding-in effect is evident in an experiment conducted by the social psychologists Mizuhu Shinada and Toshio Yamagishi among students at Hokkaido University. They cooperated more in a Public Goods game when they were assured that others (but not they, themselves) would be punished if they did not contribute sufficiently, even though the punishment had no effect on the subject’s own material incentives to defect or contribute.39
The subjects apparently wanted to be cooperative, but wished even more to avoid being exploited by defectors. Iris Bohnet and her coauthors call this motive “betrayal aversion.”40 The assurance that defectors would be punished by a third party reduced a subject’s fear that a defector would gain at her expense. Similar synergies occur in natural settings: social norms support the observance of traffic regulations, but these may unravel in the absence of state-imposed sanctions for flagrant violations. These or similar synergies may account for the few experiments in which material incentives and moral motives appeared to be complements rather than substitutes, the former enhancing the salience of the latter.
In addition, the emergence of the rule of law appears to have been associated with a parallel shift from trust in kin and other particular individuals to generalized trust, consistent with Yamagishi’s “emancipation theory of trust.”41Guido Tabellini, for example, showed that generalized (rather than familial) trust appears to thrive in countries with a long history of liberal political institutions.42 In a large sample of immigrants to Europe, the strong inverse association between measures of political participation, such as signing petitions or participating in demonstrations or boycotts, and the extent of one’s obligation to respect and care for one’s children and parents was also consistent with this view.43
The substitution of generalized trust for familial or parochial norms appears to have been at work during the expansion of the eleventh-century Mediterranean trading system, in which familial, communal, and other parochial systems of so-called collectivist contract enforcement were eclipsed by more universalistic, state-based, individualist systems.44 These are some of the reasons why market-based societies may exhibit high levels universalism in the definition and application of social norms.
Markets may have assisted the “civilizing process” in other ways. The spread of markets often contributed to the emergence of national states bound by the rule of law, and if my argument is correct, this dynamic favored the evolution of generalized trust. In addition, expansion of markets favored the proliferation of more universal social norms by promoting the national systems of schooling-by-strangers that Gellner termed exo-socialization.45 Gellner writes that markets can regulate the division of labor at the national level only if parochial traditional cultures are replaced by more universal values consistent with the extensive interaction with strangers in market environments. The resulting national standardization of language and culture facilitated occupational and geographic mobility, rendering individuals’ income-earning assets less specific to place and craft. The result was to complement the other literal and de facto forms of insurance provided by liberal institutions.
I have suggested that the puzzle of vibrant civic cultures in many market-based societies may be resolved by paying attention to how geographic and occupational mobility, the rule of law, and other aspects of liberal societies preserve social order by sustaining civic virtues. If I am right, then the kinds of incentives and constraints that people face in a liberal democratic and market-based society sometimes lead to a kind of crowding in of social preferences rather than the crowding out more commonly seen in experiments.
This result was just what Aristotle’s Legislator is hoping to accomplish. He does not seek to design some incentive-compatible mechanism that would allow him to set aside concerns about public morality. Instead, he wants to develop public policies that would allow incentives and constraints to work synergistically rather than at cross-purposes with peoples’ ethical and other-regarding dispositions.
It is time, thinks the Legislator, to visit a few economics faculties to see what help they can offer on this score. Recalling some unpleasant rebuffs from economists in the past, he will be surprised by the warm welcome he now receives.