Although earlier chapters focused on descriptive claims about the relationship between law and social norms, normative judgments tumbled in now and then. This chapter and the next two chapters gather them together. The goals of these chapters are modest: not to convince you of the value of any particu-lar legal reform so much as to convince you that common normative judg-ments in legal analysis should take account of complexities of nonlegal regula-tion more often than they do. This chapter focuses on the efficiency of group and extra-group norms, and the relationship between norms, welfare systems, and distributive justice.
Moon Cakes
An article in the Wall Street Journal (C. Smith 1998) describes a ritual that, despite its foreign provenance, is intensely familiar. It concerns the ritual ex-change of the moon cake.
Years ago the quarter-pound cakes, most often filled with a paste of mashed lotus root, sugar and oil, were prized gifts, a rare rib-sticking treat that would keep well into the icy winter months when most people subsisted on cabbage. But the urban Chinese are richer now, and moon cakes have become more bother than bounty. Like Christmas fruitcakes in the U.S., they get passed from person to person until the festival ends—and the last one holding the cakes has to eat them or quietly throw them away.
The moon cake had once been a meaningful gift. When people dined on cabbage every day, a moon cake must have provided some welcome variety.
The puzzle is why people gave away their moon cakes, in return for more moon cakes, rather than consuming the moon cakes that they had bought or produced themselves. Transaction costs alone would have made such ex-changes inefficient (a problem solved currently by a thriving secondary mar-ket of moon cake coupons). The answer to the puzzle is that the gift of a moon cake was a signal that people conveyed to their friends, relatives, and as-sociates, to show that they were good cooperators. Like other non-monetary gifts, the moon cake was both costly to the donor and not so valuable to the donee that it would swamp the cooperative benefits of a relationship.
Moon cakes from famous shops such as Xinghua Lou get passed around the most. Ms. Li believes many boxes pass through five or six hands ahead of the “best before” date stamped on the side. One recent newspa-per article recounted the case of a man who was given the same box of moon cakes that he had given away weeks before.
Why is the signal a gift of moon cakes, rather than some other item? The answer is that people give each other moon cakes this year because people gave each other moon cakes last year. At any time t, people must act in confor-mity with expectations based on time t-1. If they do not, then other people would begin to doubt whether they intend to continue a relationship, just as Americans wonder whether the failure to send a Christmas or holiday card this year, after sending one last year, is meant to be a signal.
The origin of the moon cake exchange is shrouded in the mists of time—
the tradition is a thousand years old—but part of the explanation for the rit-ual is surely that the holiday during which moon cakes are conveyed takes place on the full moon closest to the autumnal equinox. And the holiday cele-brating the harvest occurred during the full moon presumably because that day could be readily identified year after year by farmers who witnessed the lo-cation of the sun and the phase of the moon but did not have written calen-dars. People could have given each other pieces of clothing, or other kinds of food, but the moon cake was suggested by the holiday, which itself was sug-gested by the salience of a moment during the harvest. The moon, and hence the moon cake, was a focal point around which people coordinated their sig-naling behavior.
Wang Yafang opens her first box of the season and cuts a guest a quarter cake . . . The 53-year-old housewife expects to receive more than a dozen boxes before the festival is over, though her family manages to chew through only one box each year.
When prosperity came, the moon cake lost its appeal. People did not want moon cakes, because they could purchase more delicious pastries at their lo-cal bakeries and anyway they now had a more varied diet than in the past when cabbage dominated their meals. Still, no one could deviate from the thousand-year-old equilibrium. Given that people expected moon cakes from their friends and associates, disappointment of this expectation could only be interpreted as an indication that the relationship was at an end.1To avoid this inference, people continued to give moon cakes to people who did not want them, and received in return moon cakes that they did not want either.
“I’ll send two boxes to my little brother’s family, two to each of my hus-band’s brothers, two to my parents . . .” she says, counting down how she will dispose of them all. “Somehow, we always end up with one box that we can’t give away,” she says with a sigh.
Although no one likes the taste of moon cakes, a tremendous quantity of them is manufactured (21 million by just one of many bakeries that make them) and vast sums are spent on them. The ritual may finally end, yet the conclusion is inescapable that it has persisted long beyond the point at which it ceased having any value, if it ever did have any value. The comic pathos of the situation is the reverse of the despair produced by dowry competitions in India, the difference being that trends in technology and demographics have made the signals too cheap rather than too expensive.
Social Norms and Efficiency
One finds in the literature two kinds of claims about the efficiency of social norms. First, many economists and law professors assume that social norms solve strategic dilemmas that would otherwise reduce overall well-being. An example is Frank and Cook’s (1997) argument that people overconsume posi-tional goods in a competition that leaves everyone or almost everyone worse off than he would be if the competition could be restricted. Frank and Cook suggest that disapproval of cosmetic surgery, conspicuous consumption, and similar behavior reflects welfare-enhancing social norms.
There are a number of difficulties with this claim. First, one must distin-guish between attitudes and behavior. An individual might envy or feel con-tempt for wealthy or beautiful people, but it does not follow that the accumu-lation of wealth or the purchase of cosmetic surgery violates a social norm.
More to the point, even members of the elite may recognize that the competi-tion over beauty and wealth may make everyone worse off, but if no one can break out of the competition, then it serves no purpose to say that the
behav-ior violates a social norm. It is thus useful to limit the use of the concept of so-cial norm to cases where people punish those who engage in the proscribed behavior—for example, by avoiding them—and there is little evidence that people in the United States try to sanction wealthy and attractive people in this way. Otherwise, one could not understand how people could condemn the moon cake exchange ritual while engaging in it. Yet people often feel trapped by social norms that they abhor.
Second, one must be careful about which people hold the attitudes under discussion. Certainly, there are social groups in which conspicuous consump-tion or the purchase of cosmetic surgery would provoke sancconsump-tions. A member of an old-order Amish community will be ostracized if he or she violates pro-hibitions against the display of ornaments. But I doubt that most Americans would take steps to sanction individuals who engage in conspicuous con-sumption or undergo cosmetic surgery.
Third, there remains a matter of interpretation. Suppose that Americans sanctioned individuals who undergo cosmetic surgery. They might sanction these people simply because they envy them and not because they realize that anyone who undergoes cosmetic surgery promotes a self-defeating competi-tion for beauty that reduces welfare in the aggregate because beauty is a posi-tional good. The former interpretation is more plausible. On the other side, one can imagine circumstances under which social norms against cosmetic surgery do arise, even if cosmetic surgery is not a positional good, and instead gives pleasure both to the patient and to most other people. I will say more on this possibility below.
Functionalism—the view that social practices and norms are efficient or adaptive in some way—is empirically false and methodologically sterile. To answer the question of whether a social norm is efficient, one must produce a theory about the supply of social norms as well as the demand. There are two main sources of supply, each corresponding to the two main kinds of norms—group norms and extra-group norms. I discuss each kind below.
Efficiency of Group Norms
Some scholars, notably Ellickson (1991), argue that social norms are efficient when they arise in close-knit groups. Ellickson advances this argument mainly as an empirical claim. Ellickson’s study of cattle ranchers and farmers in Shasta County, California shows that these people do not rely heavily on the law and do not even have a good idea of what the law is, yet they cooperate in impressive ways. Neighboring landowners repair fences, retrieve stray cattle, pay for damage done by their cattle to the property of others, keep their promises, and pay their debts. These findings resemble the findings of studies
in the literature on common pools (Ostrom 1990). This literature shows that individuals in a remarkable array of settings—Turkish fisheries, Alpine pas-tures, Japanese farming regions—cooperate in the production of collective goods in the absence of an effective legal regime. These studies decisively con-tradict the view that people will never cooperate unless so required by law.
As an aside, I should observe that the studies do not show that the observed cooperation is optimal. True, the farmers fix the fences rather than letting them fall to pieces, but optimal cooperation might require more than that.
The standard model of cooperation in teams shows that if each party’s indi-vidual interest in the collective good is strong enough, the parties will produce some of it, and cooperation will fail just on the margin. The parties repair fences but not as quickly as would a single owner of the fences. They return a neighbor’s stray cattle but not as rapidly as they would recover their own. Sim-ilar criticisms can be made of the studies of common pools (for details, see E.
Posner 1996a).
Let us turn to the theoretical reasons for believing that group norms are likely to be efficient or inefficient. Ellickson does not make the functionalist mistake of assuming that because cattle ranchers want efficient norms, they will produce such norms. To explain the supply of efficient norms, he appeals to the theory of repeated games. This theory shows that two people in a re-peated prisoner’s dilemma might engage in the optimal level of cooperation.
But as Ellickson acknowledges, the theory does not show that optimal cooper-ation is necessary. A farmer who shares a length of fence with another farmer might rationally or mistakenly make repairs too rarely; if the other farmer re-taliates by acting in the same way, a suboptimal equilibrium will result. Still, optimal cooperation between two people seems likely.
Extension of the two-person game to the n-person game, however, is fraught with difficulties. Suppose a rancher’s cattle stray onto the land of sev-eral neighbors who themselves do not own cattle. There is an optimal level of cooperation, but the form of cooperation may be complex and difficult to re-alize. Suppose that the optimal level of cooperation is achieved when the rancher expends x units of effort to restrain his cattle, each neighbor expends y units of effort to return his cattle, and the rancher compensates the neigh-bors either by making occasional cash payments or repairing fences or helping out in other ways. True, a neighbor might meet his obligations and expend y units of effort, but he might instead let the rancher’s cattle wander onto some-one else’s land. The rancher’s threat of retaliation might not be sufficient to enlist his cooperation, and the other neighbors might not be able to coordi-nate to punish the shirker. Everyone has an incentive to free-ride, resulting in a low level of cooperation or none at all. This is, of course, the collective ac-tion problem. Although game theorists have shown that in principle certain
equilibrium strategies could result in n-person cooperation, these strategies seem implausible (see Chapter 2.)
Let me now turn to the signaling model. To understand the efficiency implications of the signaling model, begin by imagining that information costs are zero. Because everyone knows the type of everyone else, and because signaling is costly, no one sends signals. Good types match up with other good types, and bad types match up with other bad types or not at all. After players match up, some amount of cooperation occurs within those matches.
Not even the assumption of perfect information guarantees that the optimal amount of cooperation would occur, but let us assume that such a result is plausible. In the repeated prisoner’s dilemma without information costs, each player knows that if both players choose a sufficiently cooperative strategy, like tit-for-tat, both will do better than if both choose a less cooperative strategy, such as cheat-but-then-play-tit-for-tat. The cooperative surplus—
which might consist of money profits (Chapter 9), the pleasures of friendship (Chapter 4), the generation of political influence (Chapter 7), the marital sur-plus (Chapter 5)—is the same thing as the internal collective good, a concept introduced in Chapter 2.
Information costs are not zero. When information costs cross a threshold, good types signal in order to distinguish themselves from bad types. De-pending on the relevant parameters, signaling may or may not produce sepa-rating equilibria, but the relevant point is that there is no reason to believe that any given equilibrium is likely to be more efficient than an alternative.
When people signal, they impose costs on third parties, so they do not have the proper incentives to engage in efficient signaling.
To see why, imagine a pooling equilibrium in which no one sends a signal but in which “receivers” (those who receive a signal) cooperate with everyone.
Receivers would cooperate as long as the expected gains from cooperating with the good types minus the expected losses from cooperating with the bad types exceeded the value of their alternative opportunities. One might observe in such an equilibrium some or a fair amount of cooperation but not necessar-ily optimal cooperation. Now suppose that exogenous changes supply the good types with a technology that enables them to distinguish themselves from the bad types by issuing differentially costly signals. If the good types start signaling, the receivers might stop cooperating with the bad types and cooperate solely with the good types. Because the receivers would devote more time to their relations with good types, it is possible that the gains from coop-eration would increase for the good types, justifying the good types’ invest-ment in the signals.
When the good types engage in signaling, then, they increase the value of
the internal collective good obtained by themselves and the receivers but they also decrease the value of the internal collective good obtained by bad types. If this latter decrease is small and the number of bad types is low, then there will be an efficiency gain. Otherwise, there will be an efficiency loss. And if the bad types are able to mimic the signal, and do so in order to avoid being iden-tified as bad types, in equilibrium signaling costs are incurred without any off-setting informational gains. Because of these problems, we cannot say at this level of abstraction whether the separating equilibrium produces greater social wealth than the pooling equilibrium does. The value of the internal collective goods in the new equilibrium may be greater or less than the amount pro-duced in the old equilibrium.
The emergence of signaling has a similarly indeterminate effect on what was called the external collective good. The signaling technology, whatever it happens to be, might cause the good types to signal their type by aiding others or hurting them. If people signal their type by voting, giving philanthropic gifts, complying with underenforced laws, or volunteering in soup kitchens, then it is possible that the separating equilibrium will have desirable attrib-utes. Whether in fact we like the behavior that emerges in that equilibrium depends on what people would otherwise do and the exact nature of the sig-nal. If they vote without informing themselves or substitute from support of worthwhile charities to support of less worthwhile but more visible charities, then the signaling behavior will be objectionable. If people signal their type by shunning minorities or by engaging in self-censorship, then the separating equilibrium is again inferior to the initial pooling equilibrium. If people sig-nal to each other by exchanging moon cakes, and everyone participates in this activity, then no information is revealed while resources are squandered. If farmers signal by ostentatiously repairing fences, then too much fence-repair-ing will occur.
When everyone signals, so that a pooling equilibrium exists, there is no rev-elation of information and thus no contribution to an internal collective good. If, at the same time, the signal does not contribute to an external collec-tive good, then the equilibrium is unambiguously bad, because the cost of sig-naling is incurred without the production of offsetting benefits. And if the signal actually injures third parties, as with the case of racial discrimination, then the pooling equilibrium is also bad. As noted above, however, other cases are ambiguous. When everyone signals but the signaling contributes to an ex-ternal collective good, the equilibrium is desirable, as long as the value of the external collective good exceeds the cost of the signaling. When a separating
When everyone signals, so that a pooling equilibrium exists, there is no rev-elation of information and thus no contribution to an internal collective good. If, at the same time, the signal does not contribute to an external collec-tive good, then the equilibrium is unambiguously bad, because the cost of sig-naling is incurred without the production of offsetting benefits. And if the signal actually injures third parties, as with the case of racial discrimination, then the pooling equilibrium is also bad. As noted above, however, other cases are ambiguous. When everyone signals but the signaling contributes to an ex-ternal collective good, the equilibrium is desirable, as long as the value of the external collective good exceeds the cost of the signaling. When a separating