Copyright (c) The University of Chicago

Eric A. Posner.* 1998. “Symbols, Signals, and Social Norms in Politics and the Law.” The Journal of Legal Studies 27 (June): 765-98.

 

 

* Professor of Law, University of Pennsylvania. My thanks to Matthew Adler, Douglas Baird, Steve Coate, Jason Johnston, Timur Kuran, Richard McAdams, Geoffrey Miller, Eric Rasmusen, and participants at this conference and at a talk at the American Law and Economics Association annual meeting, Toronto, 1997. Karen Schoen provided valuable research assistance.

SUMMARY:   ... The lone protestor who burns an American flag on the steps of the courthouse also ignites a political firestorm, and when the smoke finally clears, it reveals a political and legal system in great disarray: political leaders condemning the perpetrator in the strongest terms, prosecutors struggling to bring him to justice, legislators scrambling to outlaw flag desecration, commentators wringing their hands beneath the looming shadow of constitutional amendment. ... For the purpose of the example, I will refer in a stylized way to the signal of "saluting the flag" on appropriate occasions, but this behavior, which might seem too cheap to serve as a signal, should be taken as a synecdoche for the patriotic behaviors mentioned above, and it will be placed in its proper context in succeeding sections of this article. ... The receiver believes that anyone who salutes the flag is a cooperator and anyone who fails to salute the flag is a cheater. ... The cooperator will not salute the flag ($ 6 > $ 6 - $ 3); nor will the cheater ($ 2 > $ 2 - $ 3). ... Suppose also that if someone does not salute the flag, the receiver believes that he is a cheater. ... Instead, flag-waving becomes a hollow ritual, and this may drive cooperators over time to abandon this signal--and the norm entrepreneur to create new ones (this is an example of symbol transformation). ... The Hermeneutic Effect: The Creation of Social Meaning and the Problem of Reification. When the law changes a separating equilibrium into an active pooling equilibrium, people stop associating the signal with cooperativeness or patriotism and instead associate the signal with not-being-a-cheater. ...  

 

ABSTRACT

This article uses a signaling model to explain the role of symbols in people's behavior and beliefs, with special attention to legal manipulation of symbols. It is argued that certain actions become symbolic because they have the proper cost structure and because they are, for historical or psychological reasons, focal. The government can in theory use standard legal instruments (which mainly affect the cost of the signal) to change equilibrium behavior and belief. The use of the law in this way is likely to have unpredictable effects because of multiple equilibria and of the sensitivity of behavior to parameters, but it occurs frequently because lobbying and other actions that influence lawmaking can become signals themselves, and the law is simply an equilibrium outcome. The analysis is used to discuss flag desecration, censorship, voting, and antidiscrimination laws.

TEXT-1:  [*765]  I. INTRODUCTION

SYMBOLS dominate American politics and permeate the law, but they are poorly understood. The lone protestor who burns an American flag on the steps of the courthouse also ignites a political firestorm, and when the smoke finally clears, it reveals a political and legal system in great disarray: political leaders condemning the perpetrator in the strongest terms, prosecutors struggling to bring him to justice, legislators scrambling to outlaw flag desecration, commentators wringing their hands beneath the looming shadow of constitutional amendment. By any standard, this is an extraordinary uproar. And yet the burning of the American flag harms no one in any way. In a country where the dominant political view holds that the state's main purpose is to prevent coercive harm, the attention lavished on the question whether a person should be allowed to burn an American flag--an action that does not coerce or harm anyone--can only be called an embarrassment.  [*766]  True, people are offended at the desecration of a symbol. But why is the American flag a symbol? And why do symbols matter?

That they do matter is beyond dispute. A little research reveals a multitude of controversies over symbols both in this country and abroad. Examples from the United States include the placement of creches and other religious symbols on public property, the display of the Confederate flag in certain southern universities, n1 the refusal by some citizens to use social security numbers and license plates with patriotic slogans on them, marches by Nazis and Klan members, "hate speech" on university campuses, and presidential apologies to victims of harsh government policies. Overseas, examples include controversies over the respect owed to the deutsche mark, the kangaroo, and the Emperor of Japan. n2 The first response of citizens to the fall of the communist regimes in Eastern Europe was not the privatization of industry and the reform of the secret police, but the destruction of statues, the renaming of streets, and the redesign of national flags.

Many of these examples seem trivial, but the influence of symbols on substantive policies and legal change cannot be doubted. The politician commits political suicide who supports a "wrong" (in his supporters' eyes) modification of laws relating to abortion, affirmative action, Social Security, gun control, and other highly charged policies, no matter how trivial the proposed modification. When voters interpret support for a political program as a symbol of a politician's commitment to their values and interests, political compromise becomes difficult, even impossible. n3

What accounts for the prominence of the flag desecration issue and other issues of pure symbolism? One theory is that the violation of symbols causes "psychic harm," another is that the violation of symbols conflicts with "social norms." But these theories do not explain why one behavior has symbolic resonance that causes psychic harm or violates social norms, while some other behavior does not. Acknowledgment of a right to be free of any psychic harm, moreover, would throw all traditional liberal rights--to speech, religion, privacy, conscience--into doubt. To provide a foundation for a legal order that protects people from some psychic harms and not others, one must explain how such harms arise.

This article uses a signaling model to explain why symbols matter. Symbols  [*767]  matter because a person's manifested attitude toward symbols tells others something about that person's character. People rely heavily on this information when deciding whether to engage in cooperative behavior in all realms of life. Indeed, because symbols matter so much, people's efforts to show respect for them lead to significant forms of conformity that can be described as "social norms." When symbols change, some people obtain advantages in forming cooperative relationships while other people lose advantages they had. Because changes in symbols can thus result in material loss for some people, these people resist when the government or other people challenge a particular system of symbols. Moreover, because cooperative behavior can be highly sensitive to symbols, there are great incentives for the government to regulate symbols, and dangers, too.

II. ANALYSIS

A. Games of Cooperation

In ordinary life people engage in symbolic behavior all the time. They shake hands, applaud in theaters, salute the flag, wear stylish clothes, exchange wedding rings, bow, present gifts, observe diplomatic protocol, and show deference to superiors. In every case, the symbolic behavior is intended as a signal that the agent has a characteristic that the agent wants the receiver of the signal to believe that the agent has, but that the receiver cannot directly observe. When the symbolic action serves no obvious private, "substantive" interest, as is the case of the actions listed above, everyone understands its symbolic nature. One does not applaud a performance because the act of bringing one's hands together is pleasurable but because applause signals to the performer and the other audience members that one enjoyed the show. Much behavior, however, has both symbolic and substantive purposes. One might discriminate against members of a minority group both because one dislikes them and because one wants to show others that one dislikes members of the minority group. One might obtain an education both because one believes that an education will improve one's skills and because one wants to show employers that one is smart enough to be able to earn a degree. One might participate in a patriotic parade both because one enjoys the spectacle and because one wants to show one's neighbors that one has patriotic feelings.

To capture these ideas and their implications for the legal system, I use a signaling model. The model will be referred to as the "cooperation game." Cooperation refers to any kind of cooperative relationship that can be modeled as a repeat prisoner's dilemma, including business, family, and social relationships. I should mention that in the space I have I can no more than  [*768]  sketch out the way the argument would proceed. A more detailed analysis must await future work.

In the cooperation game citizens of the high type ("cooperators") care about future gains relative to present gains and earn more through joint action with others than they earn by acting alone. In contrast, citizens of the low type ("cheaters") care more about the present and have adequate private opportunities for gain. Both types would cheat in a one-shot prisoner's dilemma, but in a repeated version of this game that has no definite ending and appropriate payoffs, the cooperators would never cheat while the cheaters would cheat on the first move. The reason is that the cooperators care enough about future payoffs that they would want the game to continue, whereas the cheaters value the immediate payoffs more than the (highly discounted) stream of future payoffs.

Somewhat artificially, we imagine that the two types ("senders") seek to cooperate with members of another group ("receivers"). The receivers cannot distinguish the senders by type. If the receivers cooperate with anyone who approaches them, they will sometimes gain and sometimes lose. The cooperators do not cheat them on the first round, so the receivers know that they are cooperators, and the two players then can cooperate indefinitely and obtain mutual gains. The cheaters cheat the receivers on the first round, and the receivers then decline to cooperate further, but they already have sustained a one-time loss. Depending on the fraction of cheaters in the population and the difference between the payoffs from cooperation and from cheating, receivers might be willing to cooperate with anyone, but they would refuse to cooperate with everyone whenever the expected losses exceed the expected gains.

To avoid the possibility that receivers will not cooperate with them, high types try to provide credible evidence of their type by sending a signal to the receivers. A signal is any costly action that, if successful, reveals the type of the sender. A signal can distinguish a cooperator and a cheater only if the cooperator can afford to issue the signal and the cheater cannot. Examples of signals discussed in the literature include gift giving, advertising, the acquisition of education that has no value, and consumption of luxury goods or fashionable items. n4 The person who engages in these actions shows others that he expects to earn high returns from cooperation, for if he did not, he would not be able to afford the signal. It should be emphasized  [*769]  that signals may be ambiguous: gift giving, for example, may reflect a person's generosity or altruism rather than his discount rate. n5 And it should be emphasized that signals that have the wrong cost structure--that are too cheap or too costly--fail to distinguish the cooperaters and the cheaters, leading to equilibria in which everyone sends the signal or everyone abandons it. Cooperators send signals in order to reveal their type; to the extent that the signals have the wrong cost structure, cheaters will mimic the signals in order to avoid revealing to the receiver that they belong to the low type. If the cooperators distinguish themselves from cheaters, receivers will cooperate with the cooperators; if they fail to do so, the receivers will not cooperate with them (or if the payoffs are appropriate, they will cooperate with cooperators and cheaters to an equal extent).

Several kinds of signals have received little attention in the literature but are of great significance. When a person shows respect for the national flag, performs civic obligations like voting and jury duty, and avoids people who are considered enemies of the state but are in fact desirable cooperative partners in the short term, this person engages in actions that are costly. If receivers interpret these actions as a means of revealing one's type, and if the actions are cheap enough that cooperators can recover their costs through payoffs from cooperation, but costly enough that cheaters cannot recover their costs by defecting in the first round, the actions may serve as signals. Why these actions serve as signals, rather than any other costly action (such as sitting on a pillar for one month), will be discussed subsequently. When focusing on these kinds of signals, as opposed to gift giving, advertising, and the like, I will sometimes refer to the cooperation game as the "patriotism game," although the games are formally identical.

To give the reader a more precise sense of the influence of signaling incentives on behavior, I will just sketch the main points, using a simple numerical example. n6 For the purpose of the example, I will refer in a stylized way to the signal of "saluting the flag" on appropriate occasions, but this behavior, which might seem too cheap to serve as a signal, should be taken as a synecdoche for the patriotic behaviors mentioned above, and it will be placed in its proper context in succeeding sections of this article.

Senders are either cooperators or cheaters. The probability that a sender is a cooperator is 0.9. In the first move Nature determines a sender's type.  [*770]  The sender learns his type, then chooses either to salute or not to salute. The receiver initially does not know the senders' types but does know the distribution of types. The receiver observes whether the sender salutes, then decides either to cooperate with the sender or reject him. If the receiver rejects the sender, everyone receivers $ 0. If the receiver cooperates, and the sender is a cheater, then the sender cheats the receiver and obtains a gain ($ 2), and the receiver incurs a loss (-$ 2). If the receiver cooperates, and the sender is a cooperator, then the receiver and the cooperator embark on a long-term relationship of mutual cooperation from which they both receive a stream of relatively small payments, though large in the aggregate ($ 6 for each). Technically, the same stream of payments is available to the cheater and the cooperator, but because the cheater and the cooperator have different discount rates, the cheater values it at less than $ 2 (the defection payoff) while the cooperator values it at $ 6. It costs both types of senders $ 3 to salute the flag and $ 0 not to salute the flag.

A separating equilibrium occurs under the following conditions. The receiver believes that anyone who salutes the flag is a cooperator and anyone who fails to salute the flag is a cheater. Therefore, the receiver cooperates with anyone who salutes the flag (expecting a payoff of $ 6 > $ 0) and rejects anyone who fails to salute the flag ($ 0 > -$ 2). The cooperator receives $ 3 from saluting the flag ($ 6 - $ 3), which exceeds his payoff from failing to salute the flag ($ 0). The cheater receives $ 0 from failing to salute the flag, which exceeds the payoff from saluting the flag ($ 2 - $ 3 = -$ 1). Therefore, neither sender deviates from his strategy. The receiver has no reason to revise his assumption that all flag saluters are cooperators, because in equilibrium this is true. The equilibrium is a perfect Bayesian equilibrium.

Another perfect Bayesian equilibrium is "passive pooling," by which I mean a pooling equilibrium in which no one issues the signal. Suppose no one salutes the flag, and the receiver correctly estimates the proportion of types in the population. The receiver cooperates with everyone, since the gain (.9($ 6) + .1(-$ 2) = $ 5.2) exceeds the gain from rejecting everyone ($ 0). The cooperator will not salute the flag ($ 6 > $ 6 - $ 3); nor will the cheater ($ 2 > $ 2 - $ 3). It does not matter what the receiver's beliefs are off the equilibrium path: even if he thinks that only cooperators salute the flag, no sender gains by doing so. Accordingly, passive pooling is a perfect Bayesian equilibrium.

"Active pooling," in which everyone issues the signal, is not a perfect Bayesian equilibrium: the cheater does better by deviating and failing to salute ($ 0) than by saluting ($ 2 - $ 3 = -$ 1). But we can create an active pooling equilibrium by adjusting the numbers. Suppose that the cheater gains $ 4 from cooperation, and suppose that the receiver believes that the  [*771]  probability that a person who salutes is a cooperator is 90 percent. Suppose also that if someone does not salute the flag, the receiver believes that he is a cheater. If everyone salutes the flag, the receiver will cooperate rather than reject (.9($ 6) + .1(-$ 2) = $ 5.2 > 0). The cooperator will not deviate: the gain from signaling ($ 6 - $ 3 = $ 3) exceeds the gain from not signaling ($ 0). The cheater will not deviate ($ 4 - $ 3 = $ 1 > $ 0).

There are many other possible equilibria, including equilibria in which people of one type randomize among signals. To keep the discussion simple, I avoid these issues. n7 The example suffices for my main purpose, which is to analyze the effect of legal interventions on symbolic behavior. For now I focus on three kinds of legal interventions. (A fourth is introduced in the next section.)

Changes in the Cost of Signaling. Consider again the original example. Suppose that the cost of saluting the flag falls to $ 1 for both types of sender. The cheater now gains $ 1 from saluting the flag ($ 2 - $ 1 = $ 1 > $ 0), and so the separating strategies are no longer a perfect Bayesian equilibrium. Passive and active pooling equilibria are now possible: passive, because when both types fail to salute, the receiver nevertheless cooperates ($ 5.2 > $ 0), so neither type gains enough by distinguishing himself as a cooperator to justify the cost of the signal ($ 6 > $ 5, $ 2 > $ 1); active, because when both types salute and the receiver believes that all nonsignalers are cheaters, neither type gains by failing to salute, being perceived as a cheater, and being rejected by the receiver ($ 5 > $ 0, $ 1 > $ 0).

Changes in the Senders' and the Receiver's Gains from Cooperation. We saw in our discussion of the active pooling equilibrium that an increase in the cheater's payoff from cooperation from $ 2 to $ 4 will sustain an active pooling equilibrium. In general, the higher the senders' gains from cooperation (relative to being rejected), the greater are their incentives to signal, to avoid being thought a cheater and rejected. Similarly, if the  [*772]  receiver's payoff from cooperating with a cheater is reduced by a sufficient amount, passive and active pooling can no longer be equilibria.

Changes in the Receiver's Beliefs about the Ratio of Types. Suppose there is a passive pooling equilibrium in which cooperation occurs. We saw that this is possible with the numbers used in the original example, and the result continues to hold if the receiver's payoff from cooperating with the cheater is even lower than in the original example, say, -$ 6. Suppose that a shock causes receivers to believe that 60 percent of all senders are cheaters, rather than 10 percent. In response, receivers will refuse to cooperate when types are not distinguishable (.4($ 6) + .6(-$ 6) = -$ 1.2 < 0). Now cooperators will deviate and salute the flag ($ 6 - $ 3 > 0). Cheaters will not deviate ($ 0 > $ 2 - $ 3). As a result, a separating equilibrium is formed. Receivers will not in the future correct their beliefs: since they cooperate only with those who salute the flag, and all those who salute the flag are cooperators, they will not learn the correct ratio.

B. The Norm Entrepreneur

The cooperation game requires that the signal be costly, but nothing about the game dictates the form of the signal. As long as an action is both actually and apparently costly, it can serve as a signal that the sender belongs to the high type. As noted above, gift giving, advertising, the purchase of fashions, and other forms of conspicuous consumption can serve as signals in the cooperation game. But signals can be ambiguous. Giving a gift or attending a parade does not reveal a person's discount rate if he or she enjoys these activities. The question is, then, how do certain behaviors take on symbolic value?

One answer draws on the idea of focal points. n8 Historical coincidence, physical qualities, and other attributes of a behavior can cause people to associate it with certain qualities of character. For example, April 19 has arisen, purely by chance, as a date on which an opponent of the government can distinguish himself by blowing up a building or engaging in other forms of protest. The coincidental occurrence of several significant conflicts between the government and right-wing opponents on that date has given it salience.

Other signals are fabricated. Martin Luther King Day, for example, was created by the federal government; it did not arise spontaneously. Once it was created, people could signal their loyalty to the civil rights movement by making gestures of respect on that day, in a publicly visible way.

Truly fabricated signals are rare. Usually, when the government or private  [*773]  individuals succeed in establishing certain actions as signals, they do so by drawing the public's attention to one of several conflicting focal points. To show respect for civil rights, should one have a holiday or some other celebration? Should it be in honor of King or another leader? Should it be on the date of King's birth or death or on the date of one of his accomplishments? History supplies a number of focal points; authoritative individuals enable citizens to coordinate around one.

To analyze these points more formally, we embed the cooperation or patriotism game in a larger game. Prior to the first move of the cooperation game, a "norm entrepreneur" n9 announces that a particular action will be a signal. The norm entrepreneur states that a particular action, for example, voting, will be understood as a signal of cooperativeness. n10 The cooperation game is modified in the following way. Each sender chooses among all possible signals. The sender might choose the signal recommended by the norm entrepreneur or might choose another signal. After the cooperation game is played, the norm entrepreneur receives a payoff that is a function of the number of people who issue the signal that he recommends. In some contexts it makes sense to say that the norm entrepreneur receives payoffs if he eliminates an active pooling equilibrium; in other contexts it makes sense to say that he receives payoffs if he transforms a passive pooling equilibrium into a separating equilibrium. The norm entrepreneur does not, ex ante, have full information about the payoffs of the parties in the cooperation game, so he cannot invariably choose the signal that maximizes his payoff. Because the game repeats itself, however, the norm entrepreneur can invent new signals in future rounds if earlier signals fail to yield high payoffs. n11

In choosing among signals senders have only a rough idea of which signals will allow them to distinguish themselves from cheaters and which signals will not. They must also worry about whether the receiver will recognize the signal as an effort to distinguish oneself, as opposed to simply a pleasurable form of behavior. This is a coordination problem: as long as everyone, or most people, believe that certain behaviors serve as signals, and as long as these behaviors actually have the right cost structure, then  [*774]  these behaviors can act as signals. Senders will rely on custom (for example, gift giving on holidays), but also on meanings provided by the Zeitgeist (for example, anticommunism), and the suggestions of authoritative norm entrepreneurs narrow down the pool. As long as enough senders follow the suggestion of a person, he will become an authentic norm entrepreneur. (An important condition is salience: that is why even politically ignorant celebrities can make influential political and cultural announcements, and the descendants of royalty, no matter how undistinguished their abilities, could always find a following.)

The significance of the NE game, as I will call it, is that it shows that if a signal fails to create a separating equilibrium or an active pooling equilibrium, the norm entrepreneur has an incentive to create a new signal, in the hope that it will be more successful. But it also should be emphasized that signaling can arise spontanteously around random events that become focal (like April 19), with norm entrepreneurs playing no role. In either case, we will refer to the creation of new symbols to replace old symbols that fail to produce the appropriate equilibrium, as symbol transformations.

More realistically, many people--including politicians, academics, novelists, journalists, and other cultural players--compete to be a successful norm entrepreneur in the first round of the NE game. They propose signals; cooperators use the signal that seems most likely to create a separating equilibrium. There will be conflict and confusion, but pooling around a single signal may emerge, with benefits redounding to the norm entrepreneur(s) who proposed it.

We can now add a fourth alternative to the three methods for influencing signaling equilibria. When the state or a private norm entrepreneur successfully publicizes a new signal, people may substitute from an old signal to the new signal, and the new signal may change the equilibrium. n12

C. Why Patriotism?

Unlike signals such as gift giving and advertising, patriotic signals appeal to national values. But the differences between these actions are superficial. Actions can be interpreted as a signal of one's desire to cooperate when they have some connection to the well-being of the receiver. When the pool of players consists of people in a social or business relationship, signals typically are actions that appeal to the pleasures and interests of the members of the group, which can be quite local and idiosyncratic (barbecues, fishing expeditions, ritual fasting and feasting, exchanges of gifts, and so  [*775]  on). When the pool of players consists of a nation, the actions must be thinner--they must not cause signficant offense or cost to most of the people in the country. In such circumstances, one finds two kinds of symbolic behavior: "pure" symbols, such as saluting the flag, whose benefits are entirely nonmaterial; and "impure" symbols, such as voting and discriminating against enemies of the country, which produce public goods that benefit all or most citizens. The latter are more focal, in the sense that we all recognize that a person who sacrifices himself for the country is a patriot; the former require a concerted effort to create and disseminate a tradition that everyone understands.

To explain why demonstrations of patriotism can serve as signals, it helps to use a stylized example. The McCarthy era arose suddenly after the successful explosion by the Soviet Union of a hydrogen bomb and the exposure of Soviet spies in the American government, events that heightened fears about the security of the United States and provoked concern about national cohesiveness. Even in ordinary business and social relationships far removed from international events, citizens will think that any person who engages in an action that weakens American security, however remotely, may be a "cheater." After all, cheaters prefer short-term gains to long-term payoffs, and the decline of a country is usually a slow and insidious thing. But it is not necessary to assume that people believe that any person who is a remote threat to the overall security of the United States cannot be trusted to deliver widgets on time or to help construct a new fence. It is enough to assume that some people will draw attention to their own loyalty by ostentatiously avoiding (or ostracizing) people who engage in actions that may be interpreted as unpatriotic, no matter how small and unthreatening these actions. If successful, anyone who wants to avoid being thought a cheater will pool around the actions, including ostracism of people thought to be cheaters, resulting in conformity to supposedly patriotic norms.

As the Soviet threat increased, then, people felt they could distinguish themselves from cheaters by refusing to criticize the government (or at least the hawks in the government), engaging in ostentatious patriotic displays, casting suspicion on immigrants and others whose connection to the United States was attenuated, and criticizing people who criticized American values and especially those who supported the communist ideology of the Soviet Union.

Notice that the significance of world events was twofold. First, they supplied the focal point, thus suggesting the form of the signal. If the Soviet Union is the threat, and the Soviet Union is controlled by communists, then maybe American communists are also a threat. Avoiding American communists thus reveals one's commitment to cooperation with noncommunist  [*776]  citizens. Second, world events may have enhanced the perceived need to avoid cheaters. In times of peace, entering a relationship with a cheater may not be so dangerous: they may cheat you, but little is at stake anyway. In times of war, entering a relationship with a cheater is dangerous: in cheating you, they make you vulnerable to heightened scarcity in the wartime economy if not also to hostile powers. McCarthy, a classic norm entrepreneur, exploited these fears. He did not fabricate the association between communism and subversion, but he did provide theories and "evidence" to strengthen people's suspicions. Before the McCarthy era, avoiding communists was not a powerful signal of loyalty. During the McCarthy era, it became such a powerful signal of loyalty that a separating equilibrium was created. Only after some years did this equilibrium collapse, when it became believed that the Soviet threat was exaggerated, or that the threat from cheaters was exaggerated, or that the cost of erroneously ostracizing people who were not cheaters was too high. As a result, it was not worthwhile to incur the cost of sending a patriotic signal, since one would obtain similar gains whether one did or not, and the separating equilibrium converted to a pooling equilibrium in which no one (or few people) sent patriotic messages and no one (or few people) inferred from the failure to send such a message that a particular person would be an unreliable partner. n13

D. Assumptions and Qualifications

Senders and Receivers. One inelegant aspect of the model is the division of the population into receivers and senders, when in fact everyone is both a receiver and a sender. The rationale is that when deciding whether to signal, a person takes the rest of the population as given. The receiver can be thought of as the average person a sender expects to deal with, given that in any functional society most of the population consists of cooperators and a small fraction of the population consists of cheaters. The artificiality of these assumptions could be avoided by using a model in which everyone sends and receives signals, n14 but such a model would produce unnecessary complexity.

Cheaters. A person might gain little from cooperation because he has a high discount rate; but he also might gain little from cooperation because he does not desire that which others can offer or because the others demand as a signal of cooperativeness an action that this person finds unusually costly. Thus, the principled communist or liberal might refuse to show respect  [*777]  for the flag, or to engage in other supposedly patriotic actions, because he or she believes that the truly patriotic person is one who points out the deficiencies in the existing political system. The difference between a real cheater and, say, a principled liberal is crucial: the former does not care about the future while the latter cares about the future but believes that the required symbolic behavior is repugnant. But the two types are lumped together by the cooperator who is persuaded of the value of the signal. We return to this problem--the crudeness of signals--subsequently. n15

The Reason for Signaling: The Payoff from Private Cooperation. The cooperation and patriotism models assume that citizens issue signals, even self-consciously political signals, in order to induce other citizens to cooperate with them. An alternative purpose for issuing political signals is to reveal one's political preferences to the government and other private citizens, in the hope of influencing the decisions of the former and the voting of the latter. n16 A third purpose for issuing political signals is to reveal one's costs to potential competitors. n17 These motives are probably important, but for simplicity I ignore them.

Note also that it is better to think of receivers cooperating more or less with citizens, rather than making an all or nothing choice.

Discrete versus Continuous Types and Actions. An alternative model might assume that people fall along a range of continuous types, with a few people being supercooperators and a few people being supercheaters, and everyone else clustering around a moderate position. It might also assume that people could send a range of patriotic or cooperative signals, from the very cheap to the very expensive. A model that made these assumptions rather than assuming dichotomous types and actions would probably not result in substantially different conclusions, for our purposes, though the model has not yet been well developed in the literature. A particularly nice aspect of such a model, however, is that it might show how continuous preferences result in discontinuous behavior, a result that is artificially generated in dichotomous models. The result is more persuasive also, as it suggests that if people do not fully conform, their behavior will still reflect the  [*778]  influence of the conformity of others. n18 This result is consistent with observed behavior; its significance will be discussed subsequently. n19

III. STATE REGULATION OF SYMBOLIC BEHAVIOR

When the state announces some prohibition, for example, a prohibition on the burning of American flags, we can distinguish two kinds of effect that the prohibition may produce. First, the law has an effect on behavior: it increases or decreases the amount of flag burning. This effect will be called the behavioral effect of the law. Second, the law might change people's understanding of the behavior it influences. For example, those who continue to burn flags after the law is enacted may be thought to be unpatriotic, whereas prior to the law they might have been thought less unpatriotic or just odd; or, conversely, those who continue to burn flags may be admired as civil libertarians, whereas prior to the law they might have been thought to be unpatriotic. The law's effect on people's beliefs about the kind of person who engages in a particular action will be called that law's hermeneutic effect.

The cooperation game shows the various ways in which a law's behavioral and hermeneutic effects are generated. We mentioned the four methods earlier. First, the law can modify the cost of sending a signal. Second, the law can modify the payoffs from cooperation. Third, the law can modify receivers' beliefs about the proportion of types in the population. Fourth, the law can modify the norm entrepreneur's payoff from constructing a signal or the law can construct a signal itself. These four effects in combination may produce a change in the equilibrium: the behavior in this new equilibrium represents the law's behavioral effect; the beliefs in this equilibrium represent the law's hermeneutic effect.

The following examples illustrate these phenomena. No effort is made to be systematic, however; the examples focus on different aspects of the models. Section IV draws together the threads.

A. Flags

One way to signal one's cooperativeness is to show respect for the flag. I will call "saluting the flag" any of a range of actions, including actually saluting the flag when the opportunity arises, displaying the flag, especially when done consistently with the complicated rituals, and so on. I will call "denigrating the flag" any gesture perceived as disrespect to the flag, from  [*779]  ignoring the flag when one should show respect for it, to burning it on the steps of the courthouse. Saluting a flag is costly. Although cooperators and cheaters incur the same cost in saluting the flag, cooperators earn higher returns when flag saluting functions as a signal, because receivers do not cut off relations when they discover the cooperators' type, whereas they do end relations with the cheaters. The cooperation game shows why enthusiasm for the flag waxes and wanes. In times of crisis the cost of being ostracized is so great that no one would risk the punishment that might result from deviation from active pooling. In times of security because receivers lose little from cooperating with cheaters, they will cooperate with anyone, so cheaters do not bother to salute the flag. But if receivers will cooperate with people who do not salute the flag, cooperators will not incur the cost of saluting the flag. In times of tension separating may occur, as only the cooperators find it profitable to salute the flag.

Does it make sense for receivers to believe that people who fail to show respect for the flag are not good cooperative partners? Yes, because of the association with soldiers who fight and die for the flag in a just war. Soldiers are the preeminent cooperators, sacrificing themselves for the sake of the nation. No, because soldiers do not fight for the flag (but, say, for the King), and the flag is (say) simply a means of identifying property of the country; or no, because the flag is associated with a particular government, not the nation, and that government is despised. Thus, in many countries the idea of a sacred flag is ridiculous, as is the idea of burning it, and indeed few national flags are burned outside the United States. The flag has become in the United States a focal point of cooperativeness at the national level (that is, patriotism) as a result of purely contingent circumstances and traditions that do not exist in most other countries. n20

Any of the equilibria mentioned above may be desirable or not, because saluting the flag is an imperfect signal of propensity to engage in cooperation. Consider the separating equilibrium. Some people will show respect for the flag in order to obtain cooperative gains but still cheat when the time is right. People who believe that patriotism, rightly understood, requires nonparticipation in patriotic displays will decline to show respect for the flag because of the intrinsic costliness of doing so, and the gains from cooperating with them will be lost. As long as error is low enough that receivers gain by cooperating only with those who salute the flag, the equilibrium will be sustained. Notice also that when everyone salutes the flag to avoid being rejected, but it is generally believed that not everyone is a patriot,  [*780]  people will realize that some flag-wavers must be cheaters and that flag-waving is not a reliable signal of patriotism. Instead, flag-waving becomes a hollow ritual, and this may drive cooperators over time to abandon this signal--and the norm entrepreneur to create new ones (this is an example of symbol transformation).

The state, considered exogenously, can influence the flag-waving game in a variety of ways. First, the state can modify the cost of saluting the flag. Consider a law that not only illegalizes the burning of flags but requires people to show respect for the flag in various ways. (A law that simply banned the burning of flags would not impose any cost on the vast majority of people; its importance will be discussed in Section IVC.) Suppose that before the law is passed, a separating equilibrium exists. Cooperators show respect for flags; cheaters denigrate or even burn them. The law decreases the citizens' cost of saluting the flag by increasing the cost of substitutes, denigrating or ignoring the flag. Most plausibly, this change would create an active pooling equilibrium. Now that it costs so much not to send the signal, cheaters are forced to mimic the cooperators and send the signal.

However, the law could have a different effect on behavior. When the cost of not sending the signal rises, the receivers may anticipate that everyone--cooperators and cheaters alike--will show respect for the flag. This being the case, the receivers no longer can rely on respect for the flag as a reliable indicator of a person's type. If the receiver's loss from cooperating with a cheater is high enough, the receiver will refuse to cooperate with anyone when both types can afford to salute the flag. But if the senders anticipate this reaction, they will not bother to incur the cost of showing respect for the flag, even though the law has reduced this cost. Why incur this cost, if the receiver is not going to respond by cooperating? Accordingly, the law results in a passive pooling equilibrium, in which no one shows respect for the flag and no one cooperates with anyone else.

An additional complication arises when members of a deviant subcommunity show their commitment to each other by desecrating the flag. This action serves as an effective commitment mechanism because those who publicly desecrate the flag and are subsequently ostracized by members of the dominant community reduce the value of their opportunities outside their group, increase their incentive not to free ride on the group, and thus enhance their trustworthiness within the group. Thus, a law that punishes flag desecration could result in more, not less, flag desecration, and in the greater effectiveness of the antisocial activities of deviant groups.

Predicting the law's behavioral effect is impossible: it could increase the amount of respect shown for the flag or reduce it. The defender of flag desecration laws, however, might argue that the purpose of the law is to change  [*781]  beliefs, not behavior--to instill in people feelings of respect for the flag. But the hermeneutic effect of the law is likewise impossible to predict. Respect for the flag increases when people increasingly believe that only cooperators show respect for the flag. If the status quo equilibrium is passive pooling, and if the law creates a separating equilibrium, then the law might cause people to abandon their belief that flag saluters are odd or idiosyncratic people and accept the belief that they are cooperators and patriots. But if the status quo equilibrium is a separating equilibrium, and the purpose of the law is to enhance people's respect for the flag, the law may well fail. If it produces a passive pooling equilibrium, people will no longer associate saluting the flag with any character type. If the law produces an active pooling equilibrium, people will believe that everyone who waves a flag may be a cooperator or a cheater. We term this phenomenon reification. The law in this case ambiguates the meaning of the symbol, rather than increasing respect for the symbol. If the law's purpose is to enhance respect for the flag, then the law is self-defeating.

The complexity of predicting the effect of a flag-burning law on behavior and beliefs should be evident. One should doubt, then, the claim that a law against flag burning would have any predictable effect that would be socially desirable. Then why is there so much support in the United States for a law against flag burning? This question is answered in Section IVC. n21

B. Self-Censorship

Self-censorship, like respect for the flag, may emerge as a signal of cooperativeness. In the struggle to find ways of distinguishing themselves as cooperators, people accuse critics of the government of being cheaters, while drawing attention to their own support of the government, by implication a signal of their patriotism. The average person may like to criticize the government or feel an obligation to do so; thus, not doing so will usually be a cost. But cooperators recover their cost through cooperation with receivers; cheaters may not. The association between self-censorship and patriotism arises frequently in countries, no doubt in part because of traditions of deference to authority, and no doubt in part because of the threat posed by internecine conflict to national survival in times of war. Norm entrepreneurs frequently exploit these circumstances.

The difference between self-censorship and respect for the flag is just a difference in the cost structure of the action. Some people might find self-censorship more costly, others might find respect for the flag more costly.  [*782]  In any event, whether one action, the other, or both actions emerge as signals of cooperativeness depends on the various costs faced by the different types and the extent to which each signal is made focal by tradition and other circumstances.

The problem with self-censorship equilibria is that they may reduce the well-being of the population. This point is explored at length by Timur Kuran in a recent book, n22 so I will be brief. When the reputational costs are high enough, cheaters and serious-minded cooperators, who care deeply about the state but disagree with prevailing views, may avoid publicly criticizing the government, in which case valuable information is lost. n23 During the McCarthy era, when the cost of criticizing the government (or, at any rate, McCarthy and the policies he supported) became extremely high, many people--cooperators and cheaters alike--stopped criticizing the government, and in this equilibrium not only was socially valuable patriotic criticism lost, so was the valuable exposure of cheaters.

One of the most striking aspects of McCarthyism was that this campaign resulted from McCarthy's entrepreneurial modification of focal points, not from changes in the law. But we can also analyze the effect of a law, such as a censorship law. Imagine that a passive pooling equilibrium exists: no one censors himself. The state enacts a censorship law. If the law provides for the correct level of sanctions, it creates a separating equilibrium. The law has two effects one might care about: it reveals cheaters (the internal effect) and it discourages criticism of the government (the external effect). This is characteristic of all the laws under consideration. The pro-flag law reveals cheaters and it encourages respect for the flag. The censorship law may be more effective than the pro-flag law at exposing cheaters, but it may also cause more harm than the pro-flag law, because speech has valuable external effects that showing respect for the flag, for the most part, lacks.

The state can also produce or maintain separating equilibria or active pooling equilibria by influencing the receivers' beliefs about the fraction of cheaters in the population. Suppose that receivers believe that almost all people are cooperators and cooperate with everyone, so no one sends the signal of not criticizing the government. The government now warns the receivers that a lot of people are in fact cheaters. If the government is persuasive, the receivers may refuse to accept new cooperative partners (or, more realistically, they take precautionary measures that lower the gains to the sender); in response, the cooperators distinguish themselves as cooperators  [*783]  by declining to criticize the government. If cheaters would not be able to recover their costs, they will not mimic the cooperators, and a separating equilibrium is produced. Because receivers learn the types of only the senders they cooperate with, they will have no grounds for later revising their beliefs despite their inaccuracy.

Propaganda can be analyzed in another way. In the NE game the state can take the role of norm entrepreneur by issuing propaganda. By issuing propaganda the state creates an opportunity for people to signal their patriotism or subversion. The signal is self-censorship--resisting one's impulse to disagree with the propaganda. The more baldly untruthful the propaganda, the more clearly does a person signal his patriotism by declining to disagree with it. Official lies are in this way like a flag or a national holiday: they provide the opportunity for demonstrating one's loyalty to the government.

C. Voting and Other Forms of Civic Participation

Rational choice theorists have not produced a satisfactory explanation of why people vote. Given the vanishingly small chance of influencing the outcome of an election and the relatively high and certain cost of taking time from work and standing in line at the voting booth, one would expect people never to vote. Positing a "taste" for voting turns the problem into a tautology; other efforts to play with utility functions have not been satisfactory. n24 One might just assume a social norm in favor of voting, and work from there, n25 but this strategy transforms the question, Why do people vote? to the question, Why does a social norm require people to vote?

An adequate theory of voting behavior would take us too far afield, but we can sketch its outlines. The signal is the act of voting in the voting booth, not the vote in favor of one person or another. Voting is observable, even if the casting of the vote is not: one's friends, associates, and family members know that one voted, because one took time off work or simply told them that one did when duplicity would risk one's reputation or because one is seen waiting in line. The act of voting is costly both for the cooperator and for the cheater, but the cooperator recovers his costs through repeated rounds of cooperation with the receiver, while the cheater does not. In the separating equilibrium the voter is believed to be a cooperator, and the nonvoter is believed to be a cheater. This explanation turns the voting paradox on its head: voting functions as a signal precisely because the  [*784]  costs exceed the material gains. If voting were profitable, as would be the case if people were paid to vote or severely punished for not voting, or if everyone derived "expressive utility" from voting, then cooperators and cheaters would all vote. The motive for voting, on this theory, is not to satisfy some taste for voting, or for expressing one's views, or even for helping one's country--although all these motives would strengthen the patriot's incentive to vote--but to obtain cooperative returns from other private actors. Patriotism is relevant only as a theory of the origin of the psychological association of voting and cooperation.

One would expect a separating equilibrium in which the more cooperative people vote and the less cooperative decline to vote. In the United States many people disapprove of those who do not vote and admit it. This reaction is inexplicable under the theory that posits a taste for voting, whereas it follows from the signaling model: the disapproval expresses the receiver's view that you may not be a trustworthy person. Times of national emergency could stimulate a pooling equilibrium in which everyone (or nearly everyone) votes. The main reason for voting at such times is not that your single vote now matters (it still does not) but that it now matters if people think you are a cheater.

I do not have space to survey the evidence for and against the signaling theory of voting, but let me briefly mention a few pieces of evidence. First, the fact that voting increases with wealth and education n26 supports the signaling theory but contradicts the taste theory: an action can serve as a signal only if it is costly, and the cost of voting increases with one's opportunity costs. Second, voting increases with membership activity in organizations like parent-teacher associations, charitable organization, neighborhood organization, business organizations, interest groups, and unions n27 --all organizations that require cooperative behavior from their members. Third, respondents' frequent exaggeration of their voting behavior to pollsters suggests an embarrassment about admitting failure to vote, embarrassment that may express their concern about their reputation. n28

 [*785]  Some countries penalize nonvoters with small fines; Italy posts the names of nonvoters in a public area. All of these countries enjoy a high voting rate. n29 But while both kinds of law effectively reduce the relative cost of the signal, voting, the first method relies on the legal bureaucracy whereas the latter exploits social sanctions that exist independently of the law.

Hand-wringing about low voter turnout is often scoffed at, because it could mean that people are relatively satisfied with the political status quo. The problem, however, is that because of peace and prosperity, it matters less if one is believed to be a cheater, so people do not bother to vote as a signal of patriotism. If, as a result, voter turnout falls, there will be undersupplied information about the preferences of citizens, leading to mistaken actions by politicians. However, equilibria in which many or all people vote are not necessarily desirable, either. People who vote solely for reputational reasons will not take their vote seriously, failing to inform themselves about the various candidates and voting instead on the basis of the mellifluousness or familiarity of the candidates' names.

D. Discrimination on the Basis of Race, Ethnicity, and Religion

Discrimination against members of a minority group can serve as a signal of cooperativeness or "patriotism." Because members of the out-group have skills and other desirable qualities, refusal to deal with them commercially or socially is costly for a member of the in-group. As long as other members of the in-group recognize that discriminatory action constitutes self-sacrifice for the sake of their group, a separating or active pooling equilibrium can emerge. n30

But why would people think that discrimination against minorities is a signal of cooperativeness? Sometimes, the answer is simply that the members of the outcast group have social or ancestral connections with a group that is unambiguously a threat to the majority group. Examples include Americans of Japanese origin after the bombing of Pearl Harbor, and people of Serbian ancestry living in Croatia or Bosnia. Sometimes, the answer is that the members of the outcast group compete with members of the majority group for scarce resources. If black migration threatens the power of whites in the labor market, then whites can demonstrates their loyalty to each other by discriminating against blacks. Sometimes, a minority group  [*786]  prospers while the majority does poorly, and the source of the minority group's prosperity is not understood but can be loosely connected to the majority's problems. Jewish dominance of the credit market and other important markets in European history gave Jews political power out of proportion to their numbers, a result that was used to justify theories of their harmful influence. These phenomena are self-reinforcing: when members of an out-group do well in a market or in a society, they become suspect, and members of the larger society discriminate against them to show their loyalty to each other, but this makes it even more important for out-group members to rely on each other, enhancing their trust and their mutual gains, leading to further discrimination against them by the insiders. n31

Discrimination against groups for purely self-interested motives is often rationalized. A common rationalization that supports the use of discrimination is a theory of racial or ethnic superiority, and such theories are endlessly supplied by norm entrepreneurs. That these theories are transparently false (it cannot be the case that the Serbs are superior to the Croats, as the Serbs believe, and that the Croats are superior to the Serbs, as the Croats believe, or that the Aryans are superior to the Slavs, as the Nazis believed, and yet that Slavs are superior to everyone else, and so on) shows that the forces that drive people into signaling equilibria are more powerful than the desire for truth. Just as one shows one's patriotism in a totalitarian country by endorsing its obviously wrong propaganda (the more obviously wrong, the more effective the signal), one shows one's patriotism in a fascist country by participating in collective discrimination against minorities even when one has warm feelings toward them or the minorities are defined in an obviously arbitrary way. n32 The signaling theory thus emphasizes the arbitrariness of state ideologies, be they ethnic, religious, or nationalistic. What counts as a signal depends on the importance of cooperation for individuals (which rises and falls with fluctuations in economic and political circumstances), the historical accidents that create associations between certain kinds of behavior and certain kinds of people, and the relative costs of the actions that are made salient by these associations. This is in contrast with views that people with certain characteristics, such as ethnicity and race, "naturally" belong together under one government. Indeed, the signaling  [*787]  theory shows how ethnicity and race are inventions that respond to the demand for criteria that facilitate cooperation against potential threats. n33

The social cost of this behavior is obvious, but it is worth mentioning that, unlike the examples of flag burning, self-censorship, and voting, the signal is an injury to a third party who is not a player in the game. The members of the minority are not the "cheaters"; those who fail to discriminate against them are the "cheaters." Although any theory that justifies the view that people who fail to discriminate against a certain group are cheaters will almost certainly conclude that members of that group are cheaters as well, special venom is directed to the insider who breaks ranks and treats outsiders with respect. n34

It is unnecessary to repeat the analysis of equilibria, and instead I will quickly mention examples of each of the four kinds of laws described above. First, antidiscrimination laws raise the cost of sending the signal, thus possibly reducing the amount of discrimination. Second, because whites were required to ostracize a white who treated blacks as equals in the South n35 (this was a social norm, not a law), a receiver's payoff from cooperating with a white person who cooperated with blacks would decline by the expected sanction to the receiver for violating the norm of ostracism. Third, the use of propaganda to convince citizens of an internal threat from people not belonging to the minority (that is, the cheaters) could enhance incentives to engage in discrimination as a way of showing people that one is not such a cheater. Fourth, the state can, as a norm entrepreneur, attempt to establish the signal in the first place--for example, endorsing or rejecting doctrines of racial superiority and placing religious symbols on public property.

The signaling theory helps explain the rapidity with which norms of discrimination change. Jews had been assimilated in German society decades before the rise of Hitler; a few years later, they were outcasts. Blacks had been outcasts in American society centuries before the civil rights movement; within a generation, the most obvious forms of discrimination had vanished. It is possible that these changes were due to fundamental changes in attitude, but the rapidity of the change, and the widespread feeling, especially among minorities, that the changes have been shallow, suggest the influence of signaling equilibria.

 [*788]  More should be said about the state's role as a norm entrepreneur. Official pronouncements play an important role, because officials enjoy the attention of the nation and thus can cheaply create focal points--for example, Hitler spinning out theories of Jewish influence. The effectiveness of politicians in this way accounts for the heavily symbolic content of so much political behavior. Officeholders and candidates for office must endlessly shake hands with, march in parades with, and attend the ceremonies of people who belong to powerful ethnic organizations, because once a politician associates himself with a minority group, attempts by him later to exploit circumstances and blame national problems on that group will lack credibility. If the group were so evil, the politician's earlier association with it would count against him.

The norm entrepreneur recognizes that a threat to the social order is presented when the minority becomes invisible: laws defining the minority can prevent the signal from becoming too costly. The yellow star in Nazi Germany and antimiscegenation laws in the United States minimize the cost of signaling by eliminating ambiguity about whom one should discriminate against. Jim Crow laws enabled whites to signal racial discrimination, by giving them an opportunity, never taken, to publicly display their rejection of racial norms. These laws do not so much enforce segregation as constantly remind everyone that everyone else is sending the signal. Otherwise, bad memory, the potential gains from cooperating with minorities, and the ambiguity of behavior would over time diminish the effectiveness of the signal.

When any of these factors eliminates the effectiveness of a signal at identifying cheaters, while the felt need to identify cheaters remains, the norm entrepreneur will search out new signals. n36 The most striking examples of the resulting symbol transformation come, again, from the history of anti-Semitism in Europe. In Spain in the fourteenth century and again in Germany and elsewhere in the nineteenth century, discrimination on the basis of the Jewish religion lost its effectiveness at causing separation. In Spain the reason was that persecution drove Jews to emigrate or convert to Christianity; in Germany the reason was that theological grounds for discrimination lost their plausibility as the influence of religion declined generally. But the great need in both countries for national unity provoked a demand for a method for distinguishing the loyal from the disloyal, and this demand stimulated the development of a new theory. The new theory would be based on race, not religion. People who used to practice Jewish rituals or whose parents practiced Jewish rituals were classified as racial Jews. This  [*789]  approach dealt with the problems created by demographic and theological trends by redefining the body of people who could be the subject of discrimination. Norm entrepreneurs promoted racial theories to supplement (in Spain) or replace (in Germany) the religious theories. n37

IV. SOME THEMES

A. Positive Implications

As noted earlier, the laws and official pronouncements of the state can influence equilibria in four ways. First, a law typically influences the relative costs of sending a signal. An antidiscrimination law increases the cost of signaling patriotism through discrimination; a censorship law reduces the cost of signaling patriotism by not criticizing the government. Second, a law may influence the cooperative payoffs for sender and receiver. An example would be a law that taxes employers who hire critics of the government or people who violate widespread norms of racial exclusion. Third, a law may influence the receiver's beliefs about the proportion of types in the population. Propaganda produces this effect through persuasion; other laws, such as laws that deport or imprison people who belong to the wrong type, change the receiver's beliefs by changing the actual proportion of types. Fourth, a law may influence or crowd out efforts by norm entrepreneurs to construct new signals; laws and official acts create new signals or refine existing ones by providing opportunities to send the signal or by increasing its visibility.

In predicting the effect of a law, one must take account of several complications. First, the effect of a law depends on the status quo equilibrium. A censorship law imposed on an active pooling equilibrium may have no effect on behavior; but imposed on other equilibria, it may cause a significant change in behavior. Second, the effect of a law even on a given equilibrium may be unpredictable: as we saw in Section IIIA, a law that reduces the cost of a signal may produce an active pooling equilibrium, but it also may produce a passive pooling equilibrium. Third, a single law may influence equilibrium behavior through more than one of the four routes mentioned in the prior paragraph. A law against flag burning may both reduce  [*790]  the relative cost of flag-waving and increase the cost of flag-waving (for cheaters) by more firmly associating it with patriotic activity. The effects can thus be offsetting, but they can also be reinforcing. Fourth, a law designed to eliminate undesirable signaling equilibria in the dominant community may have perverse effects on behavior in deviant communities. This is the paradox illustrated most vividly by laws that punish discrimination and flag desecration. By increasing the cost of these activities, the law makes them viable commitment mechanisms in deviant or alienated groups, resulting in an increase in the undesirable activity rather than a reduction, at least among the members of those groups.

Two other complications are discontinuity and symbol transformation. To understand the problem of discontinuity, imagine an active pooling equilibrium in which everyone discriminates against members of a minority group. Assume that "tastes" for discrimination are distributed uniformly: many people would like not to discriminate but suppress their inclinations to avoid being ostracized. A law is enacted that prohibits discrimination against minorities. If the sanction is small, it is unlikely to affect behavior: the cost of ostracism exceeds the cost of the sanction. As the sanction is gradually increased, discrimination does not change by much. As long as the reputational sanction exceeds the legal sanction, behavior is unaffected except for those with the most extreme preferences. But at some threshold, when the legal sanction exceeds the reputational sanction, the amount of discrimination will decline discontinuously. The reason is that once people with the stronger preferences against discrimination deviate from the active pooling equilibrium in sufficient numbers, the reputational sanction disappears, and all the people who would cooperate with the members of the minority group but for the existence of the reputational sanction will stop discriminating. n38 This phenomenon means that a law that has a small sanction (or even no sanction, like an announcement by an official) may have a disproportionate influence on behavior, and a law that has a large sanction may have little or no influence on behavior. Social norms are robust, but they are also delicate. n39

Symbol transformations occur because exogenous changes cause old signals to fail, eliminating the ability of receivers to distinguish cooperators and cheaters and thus giving cooperators and norm entrepreneurs an incentive to discover new signals to replace the old signals. We have already discussed the substitution from discrimination on the basis of religion to discrimination on the basis of race, the latter becoming a signal of loyalty  [*791]  when the former lost its effectiveness. Another example, which is presented in stylized form, comes from the McCarthy period. At roughly the same time McCarthy was making the connection between communism and subversion, others were asserting a connection between homosexuality and subversion. Thanks to the widely read Kinsey report and the social and demographic dislocations caused by World War II, it suddenly became evident that a lot of people engaged in homosexual behavior, just as a lot of people supported communism. Because refusing to cooperate with identifiable homosexuals is costly, discrimination could serve as a signal of cooperativeness. But for this signal to function properly, it was necessary for people to believe that discrimination expressed a desire to signal cooperativeness rather than a moral conviction, prejudice, or taste. Norm entrepreneurs had to draw the connection between homosexuality and subversion, and this was done in a variety of ways--by appealing to traditional moral and religious antipathy to homosexual behavior while claiming the importance of unity for national security and by claiming that homosexuals corrupted youth on whose "manliness" the nation relied. n40 But because homosexuals were no more a threat than anyone else, and because it was therefore not much more costly for a real cheater to discriminate against a homosexual than it was for a patriot, something like an active pooling equilibrium resulted. Sweeping legal and nonlegal attacks on homosexuality paralleled the mostly unofficial attacks on communists. n41

Symbol transformation occurs when one signal (self-censorship) fails to expose cheaters, perhaps because of exogenous changes in payoffs, resulting in the substitution to or the addition of another signal (discrimination against homosexuals). The irony is that if any American citizens posed a threat to the United States in the 1950s, it is more likely that communists and other political critics did than homosexuals, and yet discrimination against homosexuals proved to be a far more powerful equilibrium than discrimination against communists. The reason is probably that this country's traditions prior to the 1960s supported political freedom and religious conformity much more than sexual freedom, so self-censorship was a less reliable indication of patriotism--indeed, could be interpreted as a failure of patriotism--than discrimination against homosexuals.

B. Normative Implications

When a law generates a new signaling equilibrium, one evaluates the law by comparing the new equilibrium with the old equilibrium. In doing so,  [*792]  one should take account of both the behavioral and hermeneutic effects of the law.

1. The Behavioral Effect. It is well known that signaling equilibria cannot be ranked according to their contribution to social welfare without information about the costs and benefits of each kind of equilibrium. A separating equilibrium might at first appear to be superior to a pooling equilibrium because information is revealed, and information is often valuable; but it is possible that the gains from information disclosure in a given separating equilibrium are less than the costs of signaling. It is also possible that an active pooling equilibrium is superior to the separating equilibrium, since revelation of information may reduce incentives to generate value. n42 Because one cannot generalize about the costs and benefits of each kind of equilibrium with any accuracy, I have resisted making conclusions about the overall welfare gains from converting one equilibrium into another.

Still, we can draw three useful generalizations. First, there is a tension between the motive that causes a person to take an action and that action's contribution to a public good; and there is a tension between two kinds of public goods produced by an action--the revelation of information, on the one hand, and the production of some other public good, on the other. For example, a person may vote or engage in self-censorship from the motive of signaling his cooperativeness or patriotism. Voting, happily, produces an external public good (revelation of political preferences); self-censorship produces an external public bad concealment of political preferences). Both kinds of action produce an internal public good, namely, the exposure of cheaters. But even here the cost of signaling may exceed the benefits of that information. There is nothing intrinsically socially beneficial about signaling: sometimes it is beneficial, sometimes it is not.

Second, signals are crude, and even equilibria that might be considered broadly socially beneficial result in some undesirable behavior. Some politically uninformed people vote to avoid reputational sanctions. It would be more desirable if they followed their nonreputational preferences and did not vote. When self-censorship equilibria arise, political stability is established, and this may be important in times of war, but cooperators with valid criticisms of the government will be silenced along with the cheaters.

Third, when signaling equilibria are beneficial, we can expect them to occur too rarely. The reason is that a signal is a public good. (This is not true in standard signaling models, but they do not account for the ambiguity of actions that may serve as signals.) To be established, everyone has to recognize that certain actions are cheaper for some types than for others;  [*793]  but because the benefits of the signal are enjoyed by everyone, whereas the costs of making this connection are born by a few (the norm entrepreneurs, the gossips), there will be too few kinds of signals that are available for use. n43 For similar reasons, harmful equilibria occur more rarely than they would if signals were not produced through collective efforts.

2. The Hermeneutic Effect: The Creation of Social Meaning and the Problem of Reification. When the law changes a separating equilibrium into an active pooling equilibrium, people stop associating the signal with cooperativeness or patriotism and instead associate the signal with not-being-a-cheater. If we care about the "social meaning" of the action that serves as a signal, this result may be undesirable. This is the difference between putting an American flag on your house in an American suburb and raising it in a hostile foreign country, or the difference between criticism of the authorities in a democracy and criticism of the authorities in a dictatorship, or the difference between wearing a pink triangle on a college campus in the 1990s and doing so in the 1970s or early 1980s. Because everyone or almost everyone issues the signal, it no longer distinguishes some people from others. The "social meaning" of an action, which can be defined as the belief that the average receiver has about the type of person who engages in that action, n44 has become reified.

When the law changes a separating equilibrium into a passive pooling equilibrium, the signal disappears. Veterans of wars complain that people do not take the flag as seriously now as they used to. Then, the signal, though reified, was not meaningless: few people failed to salute the flag, but those who did fail to salute the flag were necessarily cheaters, so those who did not were at least possibly cooperators. Now, in many circles saluting the flag is almost meaningless. The observer assumes that the person who salutes the flag is old-fashioned or even a bit dotty--a person with strange tastes. When so few people issue a signal that the observer no longer associates that signal with a particular type, the social meaning has been destroyed.

When the state converts a passive pooling equilibrium into a separating equilibrium, it produces social meaning. An action that previously had little significance now has a great deal. Saluting a piece of cloth displaying stars and stripes had no significance prior to the Revolutionary War; later, it would have a great deal. Discrimination against those who engage in homosexual behavior is seen initially as a reaction to a practice believed to be  [*794]  immoral; after state action it is seen as an expression of patriotism. A social meaning is created. n45

But for all the emphasis of the role of the state, social meanings can emerge and disappear spontaneously, and often in the face of state efforts to regulate them. To see why, suppose at time 1 there is a separating equilibrium, and cooperators salute the flag just to show that they are cooperators. At time 2 the state enacts a law that punishes people who fail to salute the flag. At time 3 there is an active pooling equilibrium, created by the cheaters' desire to avoid the punishment. But at time 3 saluting the flag is no longer a reliable signal of patriotism. If many people salute the flag just to avoid the sanction, then those who salute the flag are not necessarily cooperators; the salute becomes reified. In the patriotism game, both types continue to salute the flag, the cooperators fearing that if they fail to salute they will be mistaken for cheaters. But over time many people, cooperators or not, will begin to recognize saluting the flag as the empty ritual that it has become. Saluting the flag becomes an embarrassment, because everyone knows that people salute the flag just to avoid legal punishment. Some will conceal their embarrassment behind a mask of irony, but under such conditions the meaning of the salute may eventually flip, becoming instead a signal of fear of legal punishment rather than a signal of patriotism. n46 The person who salutes is slavishly obedient, fearful to offend the authorities or other people; the person who declines to salute has integrity and independence--has an authentic cooperative nature. Failing to salute the flag becomes a better signal of patriotism than saluting the flag, and I believe that in some circles this is the case.

The politicization of behavior occurs with the creation of a law that requires people to engage in some behavior in which previously they had engaged voluntarily. People already salute the flag or pray at ceremonies; then a law is created that requires exactly the same behavior. At first sight, one would expect the law not to affect behavior, perhaps even to intensify it. But the law may flip the signal, so that the sender fears that others will  [*795]  think that he engages in the behavior to comply with the law, rather than to express patriotism or religious fervor. The result may or may not be that people stop engaging in the behavior--that depends on the size of the sanctions and other parameters. The important conclusion for present purposes is that politicization destroys important social meanings by legally compelling behavior that derives its meaning in part from the fact that it is not required by law. This argument is an analogy to the argument that the commodification of goods and services through the market destroys social meanings when a behavior derives its meaning from the fact that it is given freely. n47

C. Endogenizing the State

Several recent articles argue that the state has an important role in managing norms and social meanings. Norms, social meanings, and similar phenomena create externalities that cannot be bargained over. Because the market should thus be expected to undersupply socially desirable norms and social meanings, the state should either supply them itself or encourage their supply through legal incentives. n48

To see a problem with this argument, suppose that everyone engages in self-censorship in order to avoid being labeled a cheater. This active pooling equilibrium does not serve the interest of most people, but every citizen is afraid to deviate. Now we might say that here is a collective action problem that the government could solve: for example, by subsidizing the publication of newspapers and journals, giving politicians free air time, enacting special legal immunities against charges of libel and slander, and granting other privileges and subsidies. The question is why we should expect the government to engage in any of these actions. The problem is not just that government officials may enjoy the lack of criticism. The problem is that if the government officials proposed these laws, their support for them would be taken as a signal that they are cheaters. Citizens will not lobby the government to enact these laws, because they fear that their lobbying--a violation of the self-censorship norm--would be taken as a signal that they are cheaters. We cannot expect the government to change socially undesirable social meanings when these meanings are sufficiently powerful.

These considerations return us to the question raised at the start of this article: why do people seek laws against flag burning when flag burning  [*796]  causes no "real" harm? One answer is that citizens signal their patriotism not only by engaging in patriotic activity but by lobbying or at least passively showing support for laws that punish people who engage in unpatriotic activity. The second follows from the first as a matter of logic, and since the signal, political support, is still relatively cheap, an active pooling equilibrium can result. Once support for a law against flag burning is taken as a signal of one's patriotism, elected government officials cannot afford the political consequences of opposing such a law, nor can legal academics who hope for appointment to an important government position. n49

A problem with government-led norm entrepreneurship is that it produces wasteful competition for government resources. Competing groups treat the government as an instrument for conveying their symbols. An important example concerns the placement of religious symbols on public property. The problem is that observers understand that the symbol represents the successful lobbying efforts of one religious group and thus that this religious group has significant political influence. This inference causes receivers to assume that members of the religious group exist in great numbers or that they have great power--in either case, they are likely to be attractive (or unavoidable) cooperative partners. Thus, attempts by members of the majority to use discrimination against members of this group as a signal of loyalty or patriotism will fail. It is thus in the interest of religious groups to compete for influence over the government's decision to use religious symbols. Constitutional restrictions on establishment deter wasteful competition among religious groups over the use of the government as an instrument for recruiting members through its endorsement of their beliefs. n50 The free exercise clause and equal protection rules, to some extent,  [*797]  protect members of minority religions against the kind of discrimination that would stimulate the pursuit of political power in the first place.

V. CONCLUSION: SYMBOLS AND SOCIAL NORMS

Because the term, "social norm," is used in ordinary language to refer to many different kinds of behavior, understanding social norms requires the use of different models in different contexts. n51 In this article, I have argued that an important class of social norms arises from signaling games in which people choose actions that signal loyalty to states and communities. Because people often engage in a particular behavior only in order to show that they are loyal, that behavior has the peculiarly empty quality of a symbol: people take little or no pleasure from the behavior, but engage in it for the sake of reputation. Notice that the social norm is endogenous to the model: the social norm describes the behavior that arises in equilibrium. It is not that X punishes Y for violating a social norm; rather, X (and many other people) avoids Y because Y's behavior reveals to X that association with Y will not serve X's interests. Although in common speech we say that Y's behavior violates a social norm, the punishment is endogenous, not imposed by an external force. n52

The article contributes to the literature on law and social norms in three ways. First, it incorporates symbolic behavior in the general analysis. Second, it shows the ways in which a signaling model sheds light on the relationship between laws and social norms. The article argues that signaling is crucial to understanding social norms. The main alternative, the repeat game model, can explain how cooperation is possible, but it does not explain how patterned or norm-influenced behavior arises. Most other discussions of social norms in the legal literature differ in two important ways: by making exogenous assumptions about why people cooperate, and by endogenizing preferences. n53 For reasons that should be clear, I claim that this approach is backward. One can make a great deal of progress in explaining  [*798]  law and behavior without resorting to arguments about preferences, which are notoriously difficult to analyze. Third, the article shows several ways in which social norms can be pathological or inefficient. n54

The latter point suggests, of course, the possibility of desirable state intervention, but we have also seen several reasons for skepticism about proposals to involve the state in self-conscious efforts to "regulate" social meaning. First, government officials do not stand outside the signaling game. They, like citizens, are prisoners of symbols when the symbols are sufficiently powerful. Truman and Eisenhower were powerless to resist McCarthy at his height, because any effort to criticize McCarthy would have been interpreted as a signal of the presidents' pusillanimity, even sympathy, toward America's enemies. Second, the results of government efforts to change or sustain symbols, whether through legal devices or official exhortation, are inherently unpredictable. Thus, government efforts to change signals can backfire, leading to a strengthening of symbols that the government sought to change (a possible example is Clinton's efforts to increase acceptance of gays in the military), or to reification of the desired symbol (a possible effect of a law banning flag desecration). And when government efforts, whether deliberately or not, destroy or reify existing symbols, norm entrepreneurs will propose new symbols that may have worse effects than the old ones. When searching for examples of successful government-led efforts to exploit symbolic behavior, one repeatedly finds the most vivid triumphs in the histories of fascist and totalitarian states--not attractive models for the United States, a country in which the most striking successes in norm entrepreneurship have been achieved by nongovernmental movements such as civil rights, feminism, and religious evangelism.

 

FOOTNOTES: n1 Kevin Sack, Symbols of Old South Feed a New Bitterness, N.Y. Times, February 8, 1997, at 1.

n2 Nathaniel C. Nash, The New Symbol of Germany: Faith in a Scrap of Paper, N.Y. Times, June 25, 1995; Philip Shenon, Battling over a National Symbol: It's on the Menu, N.Y. Times, July 10, 1995; Nicholas D. Kristof, Japan's State Symbols: Now You See Them . . ., N.Y. Times, November 12, 1995.

n3 A detailed discussion of this problem can be found in Michael J. Piore, Beyond Individualism, ch. 2 (1995).

n4 A. M. Spence, Market Signaling (1974); Colin Camerer, Gifts as Economic Signals and Social Symbols, 94 Am. J. Soc. S180 (1988); Benjamin Klein & Keith B. Leffler, The Role of Market Forces in Assuring Contractual Performance, 89 J. Pol. Econ. 615 (1981); Wolfgang Pesendorfer, Design Innovation and Fashion Cycles, 85 Am. Econ. Rev. 771 (1995); Laurie Simon Bagwell & B. Douglas Bernheim, Veblen Effects in a Theory of Conspicuous Consumption, 86 Am. Econ. Rev. 349 (1996).

n5 See Eric A. Posner, Altruism, Status, and Trust in the Law of Gifts and Gratuitous Promises, 1997 Wis. L. Rev. 657.

n6 The model resembles advertising models. See, for example, Klein & Leffler, supra note 4; Paul Milgrom & John Roberts, Price and Advertising Signals of Product Quality, 94 J. Pol. Econ. 796 (1986). Also compare the model of symbols in Jack L. Carr & Janet Landa, The Economics of Symbols, Clan Names, and Religions, 12 J. Legal Stud. 135 (1983).

n7 The complications arise from the difficulty of arriving at satisfactory equilibrium concepts. As an illustration, one could argue that the active pooling equilibrium is not plausible. The cooperator would prefer to deviate if the receiver would cooperate with those who deviate, because the cooperator would then save the cost of the signal (earning $ 6 rather than $ 3). The receiver might in fact believe that anyone who deviates would be 90 percent likely a cooperator, because the cooperator and the cheater both do better by deviating than by signaling, assuming the receiver would in fact cooperate with someone who deviates, and given that both would do better, the receiver has no reason to believe that only a cheater would deviate. If the receiver did believe that anyone who deviates is 90 percent likely a cooperator, he would cooperate; therefore, the cooperator would, anticipating this, deviate. For a discussion of the equilibrium refinement used here, see Douglas G. Baird, Robert H. Gertner, & Randal C. Picker, Game Theory and the Law 255 (1994). If the cost from an unsuccessful deviation is high enough, however, the cooperator might find unacceptable the risk that the receiver would mistakenly reject him.

n8 See Thomas Schelling, The Strategy of Conflict 57 (1960).

n9 The term is from Cass R. Sunstein, Social Roles and Social Norms, 96 Colum. L. Rev. 903, 909 (1996).

n10 This is cheap talk; see Joseph Farrell & Matthew Rabin, Cheap Talk, 10 J. Econ. Persp. 103 (1996).

n11 This approach is loosely related to Pesendorfer's model, in which a designer manufactures goods that will be used by consumers to signal status; the designer's payoff results just from the fact that many consumers will want to buy the goods, and some consumers will pay a premium for them, because their exclusivity allows them to signal status. See Pesendorfer, supra note 4.

n12 This is analytically the same as changing the cost of the signal; only the interpretation differs.

n13 A brief overview of the McCarthy era can be found in McCarthyism (Thomas C. Reeves ed., 3d ed. 1989).

n14 See, for example, Pesendorfer, supra note 4; Camerer, supra note 4.

n15 A further confusion arises because in a subcommunity people signal their loyalty to each other by engaging in actions that violate the dominant norms. Thus, a person may burn a flag in order to show other critics of the government that he is deeply committed and a worthy ally.

n16 See Susanne Lohmann, Information Aggregation through Costly Political Action, 84 Am. Econ. Rev. 518 (1994); Eric Rasmusen, Lobbying When the Decisionmaker Can Acquire Independent Information, 77 Pub. Choice 899 (1993).

n17 Glazer and Konrad argue that firms lobby for protective regulations as a way of signaling to a potential entrant that they have low marginal costs, thus deterring entry. Amihai Glazer & Kai A. Konrad, Strategic Lobbying by Potential Industry Entrants, 7 Econ. Pol. 167 (1995).

n18 See B. Douglas Bernheim, A Theory of Conformity, 102 J. Pol. & Econ. 841 (1994).

n19 See Section IVA.

n20 Another example comes from Turkey, where flag-waving and burning have recently become important political symbols. See Gunter Seufert, The Sacred Aura of the Turkish Flag, 16 New Persp. on Turkey 53 (1997).

n21 For a different analysis of flag desecration issues, see Eric Rasmusen, The Economics of Desecration: Flag Burning and Related Activities, 27 J. Legal Stud. 245 (1988).

n22 See Timur Kuran, Private Truths, Public Lies (1995).

n23 When the relevant community is not the nation, but a smaller community, such as the university or the small town, a similar result obtains. See, for example, Glenn C. Loury, Self-Censorship in Public Discourse, 6 Rationality & Soc. 428 (1994).

n24 See Donald P. Green & Ian Shapiro, Pathologies of Rational Choice Theory 70 (1994).

n25 See, for example, Richard L. Hasen, Voting without Law? 144 U. Pa. L. Rev. 2135 (1996).

n26 See Jan E. Leighley & Jonathan Nagler, Individual and Systematic Influences on Turnout: Who Votes? 1984, 54 J. Pol. 718 (1992); Garey C. Durden & Patricia Gaynor, The Rational Behavior Theory of Voting Participation: Evidence from the 1970 and 1982 Elections, 53 Pub. Choice 231 (1987).

n27 See Stephen Knack, Civic Norms, Social Sanctions and Voter Turnout, 4 Rationality & Soc. 133, 143 (1992). In one poll 41 percent of regular voters agreed with the following reason for voting: "My friends and relatives almost always vote and I'd feel uncomfortable telling them I hadn't voted." Id. at 137.

n28 Stanley Presser & Michael Traugott, Little White Lies and Social Science Models, 56 Pub. Opinion Q. 77 (1992); see also Hasen, supra note 25, at 2160-61.

n29 Hasen, supra note 25, at 2169-71.

n30 The theory does not assume a "taste for discrimination" (Gary S. Becker, The Economics of Discrimination (2d ed. 1973)), or that physical traits are used as a proxy for unobservable characteristics (Kenneth J. Arrow, The Theory of Discrimination, in Discrimination in Labor Markets (Orley Ashenfelter & Albert Rees eds. 1973)).

n31 See Eric A. Posner, The Regulation of Groups: The Influence of Legal and Nonlegal Sanctions on Collective Action, 63 U. Chi. L. Rev. 133 (1996).

n32 Compare Richard H. McAdams, Cooperation and Conflict: The Economics of Group Status Production and Race Discrimination, 108 Harv. L. Rev. 1003, 1039-40 (1995). He relies on an assumption that people seek status, that is, they derive utility from being wealthier than other people. The signaling theory does not make this assumption, instead relying on the usual assumptions about preferences. Status, in this model, is endogenous: one has status if one has the reputation for being a desirable cooperative partner.

n33 The theory is thus consistent with the view that nations are "imagined communities"; see Benedict Anderson, Imagined Communities (1983); and historiography that emphasizes the contingency of the nation-state, for example, Eric J. Hobsbawm, Nations and Nationalism since 1780 (1990).

n34 See McAdams, supra note 32, at 1039-40; Timur Kuran, Ethnic Norms and Their Transformation through Reputational Cascades, in this issue, at 623.

n35 See McAdams, supra note 32.

n36 The politicization of race by political entrepreneurs is discussed in Jennifer Roback, Racism as Rent Seeking, 27 Econ. Inquiry 661 (1989).

n37 See B. Netanyahu, The Origins of the Inquisition in Fifteenth Century Spain 1052-54 (1995). Netanyahu puts more emphasis on the interest of the majority in seizing economic and political power from the Jews (in Spain, the conversos), than on the importance of cooperating on a national level, but he does elsewhere identify the demand for national unity in Spain as an important cause of the persecution of the conversos. Id. at 1004. Spain united around its opposition to the Jews; when religious Jews vanished, racial Jews had to be invented.

n38 Compare Bernheim, supra note 18; Kuran, supra note 34.

n39 See Sunstein, supra note 9, at 909.

n40 John D'Emilio, Sexual Politics, Sexual Communities: The Making of a Homosexual Minority in the United States, 1940-1970 (1983).

n41 Id.

n42 See Andreu Mas-Colell, Michael D. Whinston, & Jerry R. Green, Microeconomic Theory 455-57 (1995).

n43 See Lawrence Lessig, The Regulation of Social Meaning, 62 U. Chi. L. Rev. 943 (1995).

n44 Compare id.

n45 The creation of social meanings through spontaneous behavior on the part of the masses, but guided and exploited by norm entrepreneurs, helps explain the phenomenon of "invented traditions." The paradoxical connotation arises because old behaviors provide focal points which people use to signal their devotion to a new political entity. Wearing a kilt is distinctively Scottish in the sense that people in no other nations wear a kilt; but it only becomes a symbol of Scottish nationalism after later changes in social conditions made possible a Scottish "community," through which cooperative gains could be obtained. Then, norm entrepreneurs create further refinements of the signal by, for example, elaborating (entirely fictionally) on the connection between tartans and clans. See The Invention of Tradition (Eric J. Hobsbawm & Terence Ranger eds. 1983), especially Trevor-Roper's chapter on Scottish traditions.

n46 Compare Kuran, supra note 22.

n47 Compare Margaret Jane Radin, Market-Inalienability, 100 Harv. L. Rev. 1849 (1987).

n48 Lessig, supra note 43; Sunstein, supra note 9; Richard H. Pildes, The Unintended Cultural Consequences of Public Policy: A Comment on the Symposium, 89 Mich. L. Rev. 936 (1991).

n49 These observations may be used to lay the foundation of a more general theory of ideology, which would state the conditions under which people endorse (even in great numbers) political positions that are contrary to their interests. The germ of such a theory can be found in Kuran, supra note 22; and Loury, supra note 23; and see also Timur Kuran, Preference Falsification, Policy Continuity and Collective Conservatism, 97 Econ. J. 642 (1987). But this must be the subject of future research. A brief discussion can be found in Eric A. Posner, The Strategic Basis of Principled Behavior: A Critique of the Incommensurability Thesis, 146 U. Pa. L. Rev. (1998), in press.

n50 For a general discussion, see Eric A. Posner, The Regulation of Religious Groups, 2 Legal Theory 33 (1996). Objections to certain forms of "expressive" conduct by the government are thus derived from basic fears that such conduct will lead to changes in beliefs of private citizens that will reduce one's ability to enter into cooperative relationships. A person opposes legal endorsements of the views of a particular group, even if this law does not cause a direct injury to him, because the legal endorsement may stimulate a shift in attitudes that will injure him. This motivation more likely lies behind objections to race-based gerrymandering than general moral objections to "value reductionism." See Richard H. Pildes & Richard G. Niemi, Expressive Harms, "Bizarre Districts," and Voting Rights: Evaluating Election-District Appearances after Shaw v. Reno, 92 Mich. L. Rev. 483 (1993).

n51 For a discussion of different possible models, see Eric A. Posner, Efficient Norms, in The New Palgrave Dictionary of Economics and the Law (Peter Newman ed. 1998), in press.

n52 Thus, one avoids the second-order prisoner's dilemma (see Jon Elster, The Cement of Society: A Study of Social Order, 132-33 (1989)) that results from the theory, rejected here, that social norms are enforced by purposive collective action.

n53 For example, Sunstein, supra note 9; Lessig, supra note 43. See also Robert Cooter, Structural Adjudication and the New Law Merchant: A Model of Decentralized Law, 14 Int'l Rev. L. & Econ. 215 (1994); and Robert Cooter, Normative Failure Theory of Law, 82 Cornell L. Rev. 947 (1997), which combine an evolutionary theory of behavior and a theory of internalization of norms. Another approach derives social norms from the assumption that people desire status for its own sake. See Richard H. McAdams, The Origin, Development, and Regulation of Norms, 96 Mich. L. Rev. 338 (1997); see also Bernheim, supra note 18.

n54 Compare Eric A. Posner, Law, Economics, and Inefficient Norms, 144 U. Pa. L. Rev. 1697 (1996); Posner, supra note 51; but compare Robert Ellickson, Order without Law: How Neighbors Settle Disputes (1991).