Printer Friendly

Modeling the forms of international cooperation: distribution versus information.

Two pairs of matched problems obstruct international cooperation. One pair, sanctioning and monitoring problems, plagues the enforcement of cooperative arrangements. Sanctions may be needed to penalize those who renege on cooperative arrangements; what is the optimal way to penalize defectors? Monitoring is needed to determine whether defection has occurred; are sanctions needed? These problems exacerbate one another. Applying the proper sanctioning strategy is problematic when compliance is difficult to monitor, and monitoring is more difficult when sanctions and uncooperative behavior are similar.(1)

This pair of problems is well-known, but a second pair--problems of distribution and information--also impedes international cooperation.(2) Distributional and informational problems arise when the actors select precisely how they will cooperate from a set of possible solutions.(3) These problems precede sanctioning and monitoring problems: if the actors cannot agree on how they will cooperate, there is nothing to enforce. A distributional problem arises when the actors have different preferences over the solutions. An informational problem occurs when the actors are uncertain of the value of the available solutions and can benefit by sharing their knowledge. These two problems confound one another when combined. Distributional interests prevent the honest sharing of information.

Any issue that is a candidate for international cooperation presents a mixture of all four problems; sanctioning, monitoring, distribution, and information. Negotiations to cooperate address all four issues. This article focuses on the interaction of distributional and informational problems by themselves. I am not arguing that sanctioning and monitoring problems are unimportant for understanding international cooperation. Instead, I focus on distribution and information for analytical purposes. I examine why actors adopt different arrangements to address varied combinations of distributional and informational problems.

Actors bargain in an effort to resolve these problems. Successful negotiations require the parties to share some expectations about the solution to the problem over which they are bargaining. Actors adopt standardized negotiating procedures to assist the formation of such expectations. International regimes provide forums for negotiations and are believed to assist international cooperation by facilitating the convergence of actor expectations.(4) The institutions, norms, and practices of a regime assist the actors in anticipating one another's actions in a negotiation. The actors can then choose actions that make mutually desirable solutions more likely. Commonly accepted forms of negotiation can assist coordination of actions between actors. This article examines how such forms of cooperation can deal with the problems of distribution and information.

The model elaborates the classic two-by-two game of coordination known as the "battle of the sexes." The players in that game try to coordinate on one of two solutions. Both players are better off if they cooperate, but they disagree about which solution is preferable. That game represents the essence of a distributional problem. It is also the simplest model of the bargaining problem. The battle of the sexes game does not capture all the problems that face international cooperation. I do not use it as the sole model of international cooperation. Instead, international cooperation requires solving distributional problems, and battle of the sexes is the simplest representation of a distributional problem.

The model adds an informational problem to the distributional problem in battle of the sexes. The players are uncertain about which solution is preferable, and they hold private information about that uncertainty. The players both are better off in some situations if they share their private information honestly.

I examine how certain forms of formal communication between the players can increase the chance of cooperation between them.(5) The players must agree on the form and understanding of messages for communication to increase the chance of cooperation. These forms of communication help coordinate actions although they have no coercive power over any of the players. They alter the players' expectations about one another's actions by creating the opportunity to exchange meaningful messages. Successful communication requires both a forum for the exchange of messages and a shared interpretation of those messages.

In the model, the equilibria of the game provide a set of descriptions of stable behavior under a successful form of cooperation.(6) An equilibrium specifies what messages the players send to one another about the game, how they interpret those messages, and what actions they take after interpreting the messages. A regime fails when its members have incentives to deviate from the behavior that guides their convergent expectations. Because equilibria are self-enforcing, they are candidates for persistent forms of cooperation and model the incentives and behavior in such cooperation. The model has multiple equilibria. The relative importance of distribution and information influences the choice among the equilibria.

The model demonstrates five points. First and foremost, the combination of distributional and informational problems impedes the solution of either. Actors can solve either of these problems in isolation, but the combination of the two prevents a complete solution to both. Second, leadership solutions to coordination problems always exist, but leadership here is very different from hegemonic provision of public goods. Third, actors can cooperate in the face of anarchy even without a shadow of the future. Fourth, diffuse reciprocal strategies arise naturally in coordination problems through the actors' pursuit of their own interest.(7) Fifth, norms and institutions are intertwined within a successful regime. The model also predicts how the forms of cooperation vary with the relative importance of distributional and informational problems.

The model provides a way to begin to formalize how regimes help actors share information and knowledge to create shared understandings of their situation. In this way, it suggests the view of regimes as mechanisms of communication associated with Ernst Haas and with John Ruggie.(8) Their arguments generally have led to interpretive approaches to understanding regimes and the associated claim that rational models cannot explain the role of such shared understandings.(9) In this model, effective communication requires shared understandings between rational actors. This article suggests then that we can use rational models to explain the role of such shared understandings. Further, the approach here allows us to connect such understandings directly to the actors' choices and knowledge. We can see how behavior differs across disparate shared understandings.

The article first examines an example of continuing international cooperation, the International Telegraph Union (ITU), and then discusses existing formal models of international cooperation. It presents the limited information model and the different forms of cooperation within it, and then compares them to one another. Some examples illustrate how the combination of distributional and informational problems have been addressed. The article discusses the literature on cooperation in the light of the results of the model, then concludes by explaining the limits of the model.

The International Telegraph Union: an example of continuing international cooperation

The development of electric telegraphy in the mid-nineteenth century created the need for common standards for international messages. At first, bilateral agreements regulated messages across pairs of nations. As international traffic that crossed several borders grew, there was a clear need for multilateral standards. In 1865, Emperor Napoleon Ill of France convened the Paris Telegraph Conference, which led to the creation of the ITU. Numerous conferences were held following the Paris conference to update the regulations for international telegraphic messages. These conferences changed the regulations in response to changes in the technology and economics of telegraphy.

For example, the Saint Petersburg conference of 1875 considered the question of the unit size for billing telegrams. The existing unit size was twenty words; therefore, charges were assessed for every twenty words (rounded up) in a message. Extra-European members wished to reduce the unit size to single words in order to encourage more traffic. Charges on extra-European traffic were quite high to cover the high fixed cost of laying both undersea cable and cable in less developed areas. The unit size of twenty words was prohibitive for most extra-European traffic. Billing by the word would reduce the cost of sending short messages, and so encourage additional extra-European traffic. Many European states objected to reducing the unit size because it would cut their revenues from international messages. European members used international traffic to subsidize their domestic service, including governmental traffic. They believed that the additional traffic that cheaper rates would encourage would not compensate for the revenue loss and the increased cost of additional traffic, including the crowding out of governmental traffic as a nonmonetary cost. The members settled on a twenty-word unit size for European traffic and a one-word unit size for extra-European traffic at Saint Petersburg. However, the argument of the extra-European members proved to be correct; the unit size for European messages was reduced to one word in 1885.(10)

The ITU served as a forum for its members to resolve a host of issues facing international telegraphy. For example, in 1895 Germany proposed simplifying rates, billing, and accounting to encourage additional traffic and cut administrative costs. The proposed reform would have increased the share of transit charges that Germany collected by paying just the first transit nation in an international message. The regulation of acceptable codes for shortening messages was addressed, and an official codebook eventually was adopted. Codes were commonly used in the United States, who was not a member of the ITU because it had not nationalized its telegraph services. As cross-Atlantic traffic grew, arrangements were needed to regulate and encourage this traffic, and a common code was a central facet of the agreement.(11)

The issues the ITU addressed had four common characteristics. First, several solutions are always possible for each issue. The unit-size problem could have been solved many ways. The choice was not just between one-word and twenty-word units. The solution adopted in 1875 combined both sizes. Second, the actors are all better off if they coordinate on one solution than if they adopt different solutions. International telegraphy required international agreement on standards. Telegrams could not be sent if the sending agency did not know what rate to charge or if the receiving agency could not interpret the code of the message. All were better off adopting the same standard than leaving the ITU. Third, actors are often uncertain about which solution they prefer, and each possesses some information about which might be preferable., The consequences of the unit size on traffic and revenues was unclear in 1875. The European unit was changed to one word in 1885 because experience proved that the additional traffic generated enough revenue to cover both the loss of revenue on shorter messages and the cost of the added traffic. But that result was not obvious in 1875. Fourth, actors frequently have divergent preferences over which solution they prefer, given the constraint of their uncertainty. European members opposed the one-word unit because they feared it would reduce their revenues; extra-European members favored it because they wanted the additional traffic it would generate. The ITU solved a string of issues with these four characteristics over time. New problems emerged that required new agreements.(12) The unit size, accounting standards, and acceptable codes were just three of the standards needed for international telegraphy. The ITU provided a regular forum for negotiating agreements on standards.

These characteristics are not unique to the problems addressed by the ITU. Consider two other examples. One comprises the means by which the major money banks coordinate their actions when they organize a debt rescheduling.(13) Each debt rescheduling requires a new agreement about how much money is required and how nonperforming loans will be handled. Many possible refinancing packages are possible, but the banks need to coordinate on one plan. No one can know whether a particular rescheduling program will succeed. Even with this uncertainty, the banks disagree about the amount of "new money" required and how nonperforming loans should be treated. Issues of enforcing the agreement arise only after those two issues are resolved.

The problem of the pollution of the Mediterranean Sea, addressed by the Mediterranean Action Plan (Med Plan), also exhibits the four characteristics described above.(14) New agreements on different types of pollution have been negotiated as the Med Plan regime has developed. The different economic costs of various emission control levels of pollutants create disagreements about which level of pollution control should be established. The economic and environmental consequences of different controls cannot be known for certain before international regulations are adopted. Several actors, particularly Algeria, have reversed their positions on the Med Plan as the consequences of Mediterranean Sea pollution have become clearer. A dilemma of continuing cooperation occurs when actors face a string of problems with the four characteristics above.

These four characteristics address the problems of reaching an agreement. Problems of enforcement still exist even after an agreement is reached. However, there is nothing to enforce without an agreement.

All such international issues entail two problems on which the actors must cooperate to solve. The first is the problem of distribution: which solution will actors adopt in the face of their divergent preferences over the possible solutions? The second is problem of information: is there a solution that is best for all of them and if so, which solution? Because each actor has some, but not perfect, information on the value of different solutions, all actors have a joint incentive to pool their information to determine whether they agree that one solution is preferable to another. These two problems work against one another. Distributional problems create incentives to misrepresent one's private information in the hope of gaining what is likely to be a more favorable solution; yet, the actors require accurate communication messages to solve an informational problem.

The limits of other formal models of international cooperation

Formal models have spawned two major approaches to understanding international regimes. However, both focus on the sanctioning problem (and to a much lesser extent, the monitoring problem) at the expense of problems of distribution and information. The first draws on the public goods literature.(15) In this view, the provision of international order is seen as a public good. In the absence of coercion and selective incentives, public goods are underprovided unless one actor chooses to bear the burden of providing the public good alone. Hegemons provide international regimes as a means to establish order, which they and other states find beneficial.

This view generally has slighted an important distributional question in the provision of public goods.(16) Actors typically disagree about which form of provision of public goods is best. Hegemons desire a regime that produces the form of order that they prefer. Other actors may prefer an alternative regime, creating a distribution problem. Those other actors then have an incentive to misrepresent their interest in cooperating under the regime advanced by the hegemon.

Regimes also are seen as a means of communicating information to enhance the efficacy of reciprocal strategies of sanctioning. This view begins with Robert Axelrod's presentation of the value of the "tit-for-tat" strategy in open-ended iterated prisoners' dilemmas.(17) The tit-for-tat strategy is a simple reciprocal strategy of playing whatever move the opponent played in the prior round. Axelrod, drawing on the logic of the Folk theorem, shows that tit-for-tat forms a Nash equilibrium with itself that produces a higher payoff for both players than the strategy "defect on every round."(18)

If this result characterizes the world, institutions and regimes are unnecessary because the simple application of tit-for-tat leads to cooperative outcomes. But implementing this strategy requires knowledge of the other player's prior move. If noise in monitoring leads the players to misjudge the other's prior move, tit-for-tat quickly loses its desirable characteristics. Robert Axelrod and Robert Keohane suggest that institutions and regimes can enhance the chance of cooperation by reducing noise, which renders reciprocal strategies more effective at enforcing cooperation and avoiding mistaken feuds.(19) Axelrod and Keohane do not, however, analyze how institutions promote accurate communication. If a regime just publicizes information collected from its members, members may have an incentive to dissemble in their reports. The consequence is an informational problem.

Solving monitoring and sanctioning problems is important for the maintenance of regimes.(20) But some solutions to these problems create informational and distributional problems. Sanctioning creates a distributional problem if it is not known which actor of many must carry out a costly punishment. Different forms of sanction impose different patterns of costs across actors, leading to disagreement about which form of sanction should be invoked. An actor's compliance with monitoring requests from a regime can be threatened by incentives to distort both its own reports and its interpretation of others' reports. Such incentives to dissemble create an informational problem if it is not clear what actions constitute defections from the regime. Consequently, regimes may be unable to provide accurate information to their members.

The use of the prisoners' dilemma as the underlying model assumes away problems of distribution and information in favor of monitoring and sanctioning. There is only one way to cooperate in prisoners' dilemma; there are many ways to cooperate in the real world. To produce cooperation when distributional problems exist, actors must agree on how they will cooperate. If individual actors are uncertain about which solution is best, sharing their information about the value of candidate solutions can assist in achieving an outcome all would prefer. If problems of both distribution and information exist, an actor may be motivated to misrepresent its information about the value of solutions in the hope of producing an outcome it believes is better for itself. The distribution problem interferes with solving the information problem. A stylized model can help us understand these problems of continuing cooperation in the abstract and how regular forms of cooperation could address them.

A model of coordination under uncertainty

Regimes help actors solve dilemmas of continuing cooperation. I model those dilemmas with a variation of the classic two-person game, battle of the sexes. Battle of the Sexes, also known as a coordination game, is the two-by-two game shown in Figure 1. Its name is based on a story described by R. Duncan Luce and Howard Raiffa:

A man, player 1, and a woman, player 2, each have two choices for an

evening's entertainment. Each can either go to a prize fight [strategy A] or

to a ballet [strategy B]. Following the usual cultural stereotype, the man

much prefers the fight and the woman the ballet; however, to both it is

more important that they go out together than that each see the preferred

entertainment.(21)

Both players are better off playing the same move, but they have conflicting preferences over the efficient outcomes.(22) Player 1 prefers they both play move A and player 2, move B. The magnitude of a (a > 1) gives the intensity of this distributional disagreement. The greater a is, the more strongly each prefers that they coordinate on its preferred move. I refer to "a" as the degree of distributional conflict. The degree of distributional conflict reflects the risks the actors are willing to accept to gain their preferred outcome. As a increases, the actors become more interested in securing their preferred solution at the expense of guaranteeing that they agree on some solution. The degree of distributional conflict does not reflect the relative capabilities of the actors in any fashion.

Battle of the sexes has three Nash equilibria, two in pure strategies: (A;A) and (B;B); and one in mixed strategies: [a/(a + 1)A, 1/(a + 1)B]; [1/(a + 1) A, a/(a + 1)B]. The mixed-strategy equilibrium is read as (player 1's mixed strategy; player 2's mixed strategy). I separate the moves within each player's strategy with commas and players' strategies from one another with semicolons. The probabilities of each move in a mixed strategy are given within brackets. For example, Player 1 plays A with probability a/(a + 1) in this equilibrium. In equilibrium, no player has an incentive to change its move unilaterally. Each equilibrium forms a stable and self-enforcing combination of strategies. The pure strategy equilibria guarantee that the players coordinate their strategies with one player receiving a payoff of a and the other player a payoff of 1. In the mixed-strategy equilibrium, the players choose the same move with a probability of 2a/[(a + 1).sup.2], and each has an expected payoff of a/(a + 1).

Actors negotiate to coordinate their policies when they cooperate. Battle of the sexes captures the motivations underlying negotiations and is the simplest model of the negotiation problem. Imagine a buyer and seller negotiating over the sale of a house. There are only two prices available, high and low, and both parties are willing to trade at either price. Both the buyer and the seller are better off agreeing on a price than not, but they disagree about which price is best. The buyer prefers the low price and the seller, the high price. The moves in battle of the sexes correspond to the negotiating positions of the parties; the negotiations break down if the players adopt different positions.(23) The two different possible agreements correspond to the two pure strategy equilibria of battle of the sexes.

Battle of the sexes does not reflect all the aspects of a negotiation or a problem of international cooperation. But the limitations of battle of the sexes as a model of bargaining parallel the limitations of prisoners' dilemma as a model of defection from agreements. There is a wide range of possible agreements in a negotiation, and there are many different ways to defect from an agreement. Battle of the sexes allows only two possible agreements, and prisoners' dilemma has only one way to defect. Both in negotiations and after defection, either side can revise its position, and these revisions can go on indefinitely. In both of the simplified models, the players have only one move. They cannot revise their positions in response to the other side's position. I use battle of the sexes because it is the simplest representation of a distributional problem.

For many international issues, states may be somewhat uncertain about the value of outcomes. This is the problem of information. To represent such uncertainty in the model, I add the possibility that the players are playing another game in which both prefer the same move. Neither player knows for certain which game they are playing, but both have some private information about it. The players' situation parallels that of actors in international politics described by Keohane: "perceptions of self-interest depend both on the actors' expectations of the likely consequences that will follow from particular actions and on their fundamental values."(24) Sometimes, the players do not know the consequences of their actions in the model, but they do have some expectations about those consequences. Further, they can disagree about the possible consequences of any move. The resulting game played under circumstances of limited information captures the conditions of a dilemma of continuing cooperation.

The players are playing one of three two-by-two games. The first is battle of the sexes. The second and third games are shown in Figure 2. The players in the latter two games still face a coordination problem but now prefer to coordinate on the same move. These games also have three Nash equilibria, two in pure strategies, (A;A) and (B;B), and one in mixed strategies, wherein each player plays the preferred move with probability 1/(a + 1) and the other move with probability a/(a + 1). I refer to these games as "both prefer move A" and "both prefer move B" as given in Figure 2 and the two together as the "both-prefer" games.(25) These both-prefer games provide a simple model of situations in which the parties may need to negotiate to coordinate their actions, even though they do not disagree about what outcomes are preferable.(26)

To model uncertainty about the state of affairs, a round of the game begins with a random draw, not observed by either player, that determines which of the three games they are playing. I refer to the game actually selected in a particular round as "the game being played." The probability of each of the three games varies with a parameter p. Battle of the sexes occurs with probability 1 - p, and each of the both-prefer games with probability 1/2 p. The parameter p reflects the degree of similarity in the players' payoffs across the two strategies. When p = 0, both players know they are playing battle of the sexes and so know that they disagree about which move is preferable. When p = 1, they know they are playing either of the other two games and so have identical preferences over the two moves. I assume the players know the value of p but not the outcome of the random draw. They have a common belief about how likely it is that they share preferences over the strategies. I refer to p as the "probability of identical interest." The nature of the issue about which the actors are trying to cooperate determines this probability.

After the game they are playing has been selected, each player receives private information about that game in the form of a "signal." Each player receives its signal privately; each can observe only the signal it receives. Signals can be thought of as each players' current knowledge about the value of the two possible solutions. Such knowledge could be the result of research on the consequences of different policies or existing beliefs about the efficacy of policies. The players consult their own experts on policy in the area (e.g., scientists or economists) to learn what solution is better before entering negotiations on which solution to adopt. The experts provide a recommendation that one type of solution, (A;A) or (B;B), is better for the player. Each player uses its expert's recommendation to judge which solution is better for it. However, the experts are not always correct. The players discount the experts' recommendations in light of their accuracy. Unlike the literature on signaling games, I make a distinction between "signals," which give private information about the game being played, and "messages," which the players can send to one another after receiving their signals. This distinction is made purely because a player's type--the signal it receives--does not always allow it to know its preferred outcome.(27)

In this model, two signals are possible, signal 1 and signal 2. If battle of the sexes is the game being played, each has a one-half chance of receiving signal 1 and a one-half chance of receiving signal 2. If both prefer move A (B) is the game being played, both players receive the same signal (either signal 1 or signal 2).(28)

These signals create different expectations between the players about the game they are playing. If player 1 receives a signal of 1, it knows it is playing either battle of the sexes or both prefer move A. It knows then that it prefers coordinating on (A;A) to (B;B) for that round of the game because (A;A) produces payoff a for player 1 and (B;B), payoff 1 in both games. But if player 1 receives a signal of 2, it does not know whether it prefers (A;A) to (B;B). It could be playing either battle of the sexes or both prefer move B. Similarly, player 2 prefers (B;B) to (A;A) if it receives a signal of 2 and cannot tell if it receives a signal of 1. I refer to signal 1 for player 1 and signal 2 for player 2 as "favorable signals." Signal 2 for player 1 and signal 1 for player 2 are "unfavorable signals."

This uncertainty about payoffs creates a mutual interest in sharing information. If one player has received signal 1 and the other signal 2 and if the players share that information honestly, they would know that they must be playing battle of the sexes. If they have both received the same signal, they cannot be certain they are playing the corresponding both-prefer game, but their confidence that they are rises. Because the experts are not always correct, the players can benefit by sharing their recommendations.

The model captures the combination of distributional and informational problems that complicate international cooperation. Regardless of which game they are playing, both players are better off if they can agree to coordinate on one solution. But coordination is easier if they are playing one of the both-prefer games because they prefer the same solution. In contrast, when they are playing battle of the sexes they are better off if they coordinate, but they do not agree on which solution is better. The possibility that they are playing battle of the sexes creates a distributional problem, while the fact that neither player can determine which game they are playing creates an informational problem. If they share their knowledge about the game being played, their beliefs about which outcome they prefer may change. I use "knowledge" here in the way Ernst Haas does in his approach to regimes--it is a guide to the solution that is better for the actors.(29)

Even if the players accurately report their private information to each other, there is still some chance that they are playing battle of the sexes. They cannot eliminate the distributional problem. Both players always possess some incentive, then, to dissemble about their private information. If the game being played is battle of the sexes, such dissembling might convince the other player to coordinate on their preferred outcome. Increasing the degree of distributional conflict (a) heightens the intensity of these problems, since the risk each side will accept to achieve its preferred outcome rises. This magnifies the distributional problem. The importance of solving the informational problem also rises because the value of the jointly preferred solution is greater relative to the less-preferred solution. Increasing the probability of identical interest (p) raises the importance of the informational problem relative to the distributional problem. As p increases, it becomes more likely that the game being played is a both-prefer game. The players would like to share information to increase their confidence that they are playing a both-prefer game.

Regimes assist coordination of action by structuring communication among actors.(30) I model communication by adding a move whereby the players can inform one another about the signal they have received. I refer to the reports the players exchange as "messages" and formalize them as message 1 and message 2. Actors send messages after they have received their signals but before choosing their move for that round. The exchange of messages corresponds to the exchange of views in negotiations. The opportunity to exchange messages gives the players a chance to negotiate which solution they will adopt. The form of those negotiations affects the chance that the players will succeed in adopting a solution. When the players fail to coordinate, the negotiations have failed. As we will see later, there are different ways for the players to use the messages to coordinate their actions. These different uses correspond roughly to different forms of bargaining.

To represent the repeated aspect of international cooperation, the model consists of a series of "rounds." Each round begins with a new random draw of the game being played. The players receive their signals based on the game they are playing in the round. They select their "moves," either move A or move B, simultaneously, after receiving their signals. Their moves and the game being played are revealed and payoffs rewarded. The game can repeat beginning with a new draw for a known finite number of times. Any communication between the players takes place after they have received their signals about the game being played but before they select their moves. Figure 3 presents the sequence of a round of the model. Each round of the model can be thought of as a particular agreement that the players could make on an issue of the moment. Multiple rounds model a string of possible agreements the players could reach.

I focus on stationary strategies of the model rather than on conditional punishment strategies. A conditional punishment strategy selects moves based in part on the prior play of the game, as in tit-for-tat. A stationary strategy does not condition its moves on play in prior rounds of the game. I focus on stationary strategies for several reasons. First, the well-known results of tit-for-tat in the iterated prisoners' dilemma have drawn much attention to the efficacy of punishment strategies. I show that stationary strategies can also be effective in producing cooperation between rational actors with divergent interests. Second, punishment strategies require indefinite iteration (either infinite with discounted payoffs or random chances of ending the game), which may be a tenuous assumption in some cases. The strategies for producing cooperation that I discuss here work for finite plays of the model as well as for indefinitely repeated plays. Third, the Folk theorem shows that large ranges of behavior can be supported in equilibrium using punishment strategies. Any outcome where both players receive at least as much as their minmax value in the stage game can be supported by a punishment strategy provided that discount factors are not too high.(31) It is difficult to derive testable hypotheses from models based on the Folk theorem because many types of behavior are possible in such models.(32)

The following section presents the equilibria of the model. These equilibria give stable arrangements for sending and understanding messages and selecting moves based on those messages. Neither player has an incentive to defect from an equilibrium once both understand they are playing that equilibrium. The set of equilibria are the candidates for regimes that could persist over time. A variety of forms of cooperation--stable arrangements that assist coordination--are possible. Different values of the degree of distributional conflict and probability of identical interest make different regimes preferable.

The equilibria of the model

The model has four different types of equilibrium.(33) The four equilibria are distinguished by how the actors use the information in their messages. The first type of equilibrium, a "babbling equilibrium," lacks any effective regime. In a babbling equilibrium, each player always sends the same message regardless of the signal it has received. The actors adopt equivalent actions and receive equal expected payoffs, making this equilibrium symmetric.(34) The players' messages convey no meaning whatsoever (hence, the name). Neither player can learn anything about the other player's upcoming move from these messages.

The players choose their moves in a babbling equilibrium based on the signals they have received. When the probability of identical interest (p) is high, the actors follow the lead of the signal they receive; they choose move A if they receive signal I and move B if they receive signal 2.(35) When it is unlikely they have identical interests- formally when p < (a - 1)1(3a - 1)--both players follow the lead of a favorable signal (e.g., player 1 plays move A if it receives signal 1) but mix moves between A and B when they receive an unfavorable signal. The probability of cooperation and the players' expected payoffs rise with the probability of identical interests in the babbling equilibrium. When they cannot have identical interests (p = 0), both players know they are playing battle of the sexes and play its mixed-strategy equilibrium. When they must have identical interests (p = 1), the signal each receives suffices to coordinate their actions on the superior move.

The babbling equilibrium is analogous to the absence of a working regime. Without some mechanism to create expectations about what the other player will do, both players must proceed on their own. Even without structured communication, the actors in the model often coordinate their actions for their mutual benefit. Nonetheless, success in achieving cooperative outcomes is not necessarily evidence of successful cooperation. To understand what evidence would allow us to determine when actors cooperate, we must compare the babbling equilibrium to equilibria with meaningful communication.

A functioning regime elicits messages that allow the players to form expectations about one another's upcoming move. The second type of equilibrium, the "pure coordination equilibrium," addresses only the distributional problem. The players make no use of their private information and communicate purely to coordinate their moves. Each actor sends one of the two messages. If the messages match, they both play move A; if they do not match, they both play B. Using these messages, coordination is always achieved, and both players expect the same payoff [(a + 1)/2] from a round of the model. Each player has an incentive to play the move suggested by the messages because it expects the other player also to play that move. Neither player has any preference over sending one message over another because the outcome on which they coordinate depends upon both players' messages.

The pure coordination equilibrium produces gains over the babbling equilibrium when the distributional problem looms larger than the informational problem (that is, when p is small); in other words, it focuses solely on the distributional problem at the expense of the informational problem. The pure coordination equilibrium ignores whatever information the signals convey about the game being played. It selects both solutions with equal probability, even when the actors are playing one of the both-prefer games. It can lead the actors to outcomes that are Pareto-suboptimal ex post facto--when they coordinate on the inferior outcome in a both-prefer game.

If the players know they are playing battle of the sexes, the pure coordination equilibrium solves the distributional problem of that game. The players always coordinate their moves. Thus, distributional issues are not sufficient to explain a lack of cooperation. The pure coordination equilibrium is analogous to accepting the principle of "splitting the difference" in all negotiations. Splitting the difference insures that neither party is cheated and avoids any chance that negotiations will fail. But that rule forgoes the opportunity to search for a solution that both actors might prefer to an equal division of the current issue, such as a deal that links new issues into the deal.

Regimes that address the informational problem require messages that convey information about what signals each player has seen. In the third type of equilibrium, a "communicative equilibrium," the players' messages correspond to the signals they have received.(36) Such messages, however, create an incentive for the players to dissemble. An actor can benefit sometimes by sending the message that it has received a favorable signal when in fact it has received an unfavorable signal. Players receiving messages discount them based on the possibility that the sender is lying about its signal. How the actors interpret messages and which move they then play varies with the degree of distributional conflict (a) and the probability of identical interest (p). These "interpretations" prescribe what move each player should play in response to each combination of messages. Some interpretations and the moves they recommend cannot be supported in equilibrium.

When it is likely that the players have identical interests (p > 1/2), an "honest communicative equilibrium" exists. Both players accurately report the signals they receive. If both players send message 1, then they both play move A. If both players send message 2, they play B. If one player sends message 1 and the other message 2, they follow the procedure used in the pure coordination equilibrium to determine which move to play. The honest communicative equilibrium always produces coordination and each player expects the same payoff [ap + 1/2(1 - p) (a + 1)]. If this equilibrium can be sustained, no other equilibrium produces a greater total payoff for both players.

When it is likely that the players do not have identical interests (p < 1/2), the players lose the incentive to report their signals honestly given the interpretation of the honest communicative equilibrium. Actors who receive unfavorable signals would prefer lying to sending an honest message if they believe the interpretation of the honest communicative equilibrium is in effect. When each actor expects the other to send a dishonest message, a "dishonest communicative equilibrium" is possible. The actors discount one another's messages of favorable signals in dishonest communicative equilibria. One interpretation that does so directs the actors to play A if both send message 1, to play B if both send message 2, to use the pure coordination equilibrium if player 1 sends message 2 and player 2 sends message 1, and to play the babbling equilibrium if player 1 sends message 1 and player 2 sends message 2. This equilibrium rewards both players with successful coordination when they both send honest messages of unfavorable signals. Both sides protect themselves against the possibility of dishonest messages by playing the babbling equilibrium when they both send messages of favorable signals. Playing the dishonest communicative equilibrium entails a chance that the players will fail to coordinate, creating a cost for dishonest signals; but it also produces a penalty for honest messages of favorable signals.

The likelihood of coordination and the payoff the players expect from a dishonest communicative equilibrium depends upon the probability of identical interest (p). The lower p is, the greater the players' incentive to send dishonest messages when unfavorable signals are received. Dishonest communicative equilibria are less likely to produce coordination as the distributional problem looms larger than the informational problem. Increasing the degree of distributional conflict (a) also dwindles the probability of cooperation by exacerbating the distributional problem. If 1/2 > p > (a - 1)1(3a + 1), players employ the babbling equilibrium, wherein strategies are dictated by the signals players have received. Players that receive unfavorable signals do not, however, always send honest messages, and the probability of a dishonest message rises with a. The probability of coordination under the dishonest communicative equilibrium is 1/2p + 1/2p + (1 - p) [a/(a + 1)2], and each player's expected payoff is ap + 1/4(1 - p) (a + 2). Both are higher in the dishonest communicative equilibrium than in the babbling equilibrium.

When the probability of identical interest is low--that is, when p (a - 1)/(3a + 1)--coordination becomes increasingly difficult to produce through a communicative equilibrium. In some cases, no communicative equilibrium is possible. Players receiving unfavorable signals always send dishonest messages in these cases, and the players find themselves in a babbling equilibrium. When players send the same message no matter what signals they have received, messages lose their ability to communicate information. Under these circumstances, the pure coordination equilibrium produces a higher expected payoff for both players than any possible communicative equilibrium. Thus, under these conditions actors prefer a pure coordination equilibrium to any communicative equilibrium. The problem of distribution overwhelms the problem of information.(37)

Negotiations occurring within communicative equilibria, both honest and dishonest, center on searching for the best possible deal for the players. When an honest communicative equilibrium can be supported, the likelihood of identical interest is sufficiently high that the players reveal their signals honestly. Dishonest communicative equilibria demonstrate the limits of negotiations as a mechanism to communicate information. The distributional problem is more significant here than under an honest communicative equilibrium. Both players have an incentive to misrepresent their signal to advance their interest if battle of the sexes is the game being played. Such incentives reduce the possibility of successful coordination because each player is less likely to take the other's message seriously. Each is more likely to interpret the other's negotiating stance to be self-interested rather than an indication of the true value of one of the solutions.

The equilibria considered so far are all symmetric; both players are treated identically. In the fourth type of equilibrium, a "leadership equilibrium," the players send their messages sequentially, giving them different roles. One player sends a message about its signal; the second player then tells the first which move it (the second) will play.(38) I refer to the latter player as the leader and the former as the follower. The follower always "babbles"--sends the same message regardless of the signal received--and so the leader cannot learn anything about the follower's private information from its message. The follower can never gain by sending a message that it has seen an unfavorable signal. If the leader has received a favorable signal, it always plays its preferred move regardless of the follower's message. If the leader has not received a favorable signal, the follower is always better off if it convinces the leader that it has received a favorable signal. The follower always says it has received a favorable signal, and the leader, knowing this, ignores its message. Once the leader has announced its chosen move, the dynamics of battle of the sexes and the both-prefer games drive both players to adopt the announced move. The players always succeed in coordinating their moves in a leadership equilibrium, but they cannot make use of the information in the follower's signal. Like the pure coordination equilibrium, a leadership equilibrium concentrates on the distributional problem at the expense of the informational problem.

The players' values for a leadership equilibrium depend upon the probability of identical interest. When it is unlikely that they have identical interests (p < 1/2), the leader directs the follower to play the leader's preferred move in battle of the sexes--e.g., (A;A) if player 1 is the leader--regardless of the signal it receives. The leader expects to receive a greater payoff than the follower: a - 1/2p (a - 1) versus 1 + 1/2p (a - 1). If it is likely that the players have identical interests (p > 1/2), the leader follows the signal it receives. If it receives a favorable signal, it orders the follower to play its preferred move in battle of the sexes; if it receives an unfavorable signal, it orders the follower to play the other move. Both players expect the same payoff in this case, namely, 1/2(1 - p + a + ap).

Leadership equilibria are inefficient because the follower's signal is ignored. This inefficiency always leads to losses for the leader and sometimes leads to losses for the follower. When p < 1/2, the players coordinate on the inferior move when they play the game wherein both prefer the follower's preferred move. If 1/3 < p < 1/2, the leader will change the move it orders if it knows the follower's signal. If p < 1/2, the leader will always order its preferred move in battle of the sexes if it knows when both players receive unfavorable signals. This loss to the leader is a transfer to the follower rather than a joint loss to the pair.

None of these four equilibria solves both problems at the same time. The pure coordination and leadership equilibria coordinate the players' actions while neglecting the informational problem. Communicative equilibria fail to coordinate actions sometimes. Thus, it can be seen that problems of distribution and information interfere with one another. A form of cooperation can solve one, but not both, of these problems.

Four different ways to address the joint problems of information and distribution have been presented. In what follows I discuss how these solutions reflect the logic of bargaining and communication in international cooperation.

Regimes as forums for bargaining and communication

The equilibria of the model parallel behavior under effective regimes. A regime provides a forum for its members to negotiate solutions to recurring problems of coordination. A successful regime provides incentives for its members to engage in communication that coordinates their behavior; it promotes exchanges of information about its members' collective problems. This section compares how the equilibria defined above, on the one hand, and regimes, on the other, coordinate actions.

Equilibria in the model assist the actors in achieving cooperative outcomes by providing mechanisms for meaningful communication. How does communication coordinate the players' actions? Optimal play of all three games requires anticipation of the other player's move. If a player knows which move the other player will make, it should match that move. The equilibria help the players form those expectations. If the players believe they are playing a both-prefer game and they both know the other has that belief, coordinating on the mutually preferred move is easy. Communicative equilibria nurture these beliefs by allowing the players to share the information in their signals about the nature of the game. If the players believe they are playing battle of the sexes, coordination on one of the two moves is in both players' interest. The pure coordination equilibrium does this by creating mutual expectations about the move to be played. Once the leader announces its move in a leadership equilibrium, both players want to play that move. All the interpretations in these different equilibria produce such mutual expectations without imposing costs on the players.

The equilibria also produce norms of behavior. The actors' behavior in any equilibrium, except the leadership equilibrium, matches the norm of diffuse reciprocity. Diffuse reciprocity occurs when the players anticipate that they will be compensated for short-term losses in the long term, so they cooperate even when cooperation entails a short-term loss. In the model, the equilibrium resolves how the players will coordinate. Each is willing to play the other's preferred move in a round if the equilibrium dictates that outcome. But a player's willingness to make these concessions is not based on the expectation of compensation in the next round. Unlike the play of tit-for-tat in the iterated prisoners' dilemma, reciprocal rewards from the equilibria here are diffused over many rounds. An actor may have to accept playing the other's preferred move for several rounds, but going into any one round it expects at least one-half chance that they will coordinate on its preferred move. Obeying the recommendation of an accepted interpretation is in both players' short-run interest.(39) They act as if they are obligated to follow the equilibrium, a crucial element of diffuse reciprocity according to Keohane.(40) In the long run, they expect their benefits to exceed any short-run relative losses.(41) Ex ante rationality in the model parallels the norm of diffuse reciprocity.

The players' expectations about one another's moves are analogous to norms within a regime. Strategies are complete plans for playing the model; they specify particular messages and moves for specific situations and histories of the model. Like norms, the players' cognizance of one another's equilibrium strategies allows them to determine when another player has deviated from suggested behavior. Norms then are generated by an equilibrium of the model. Some equilibria, like the honest communicative equilibrium, produce exacting norms, honesty in this case. Others can support only weak norms of behavior. Regimes produce norms that help the actors create expectations about one another's future behavior. Different regimes produce norms of varying strength, just as the different equilibria in the model do.

Institutions have no immediate parallel in the model. In successful regimes, institutions help the players implement the real-world analog of an interpretation. An interpretation in the model assists the coordination of the players' moves only if they both agree to accept that interpretation before a round begins. The form and interpretation of messages must be common knowledge for them to coordinate actions. Frequent changes in interpretations undermine that common knowledge. Institutions obstruct players from changing the interpretation from round to round to suit their immediate ends. Institutions increase the cost of changing the interpretation and help routinize the form and exchange of messages. Because the parameters (a and p) of the model are fixed, how the players change from one equilibrium to another as conditions vary cannot be addressed by this model.

Just as regimes require both norms and institutions, an equilibrium intertwines strategies and interpretations so that both are required.(42) Strategies and their interpretations each depend on the other. Without an interpretation, the strategies followed would not be optimal. Without the corresponding strategies, an interpretation produces misleading expectations.

To see how the logic of the model reflects some aspects of reality, let us return to the three examples of continuing international cooperation mentioned above. The ITU helped coordinate policy on international telegraphy by providing a regular forum for negotiating changes in regulations. Those negotiations addressed both the distributional and informational issues of such policies. However, merely arguing that efficiency could be improved by a change in policies was not always sufficient to bring about that change. The German proposal in 1895 to simplify the accounting and billing procedures was not adopted in part because of its distributional consequences. In that sense, the ITU reflects the dishonest communicative equilibrium. Other nations discounted the efficiency gains of the German proposal because of its obvious effect of increasing German revenues since Germany served as a transit carrier of telegrams. The distributional issue confounded the transmission of information through the negotiations.

How major money banks determine the funds they will require to reschedule a debt also displays the characteristics of a dishonest communicative equilibrium.(43) After it becomes apparent that a state's debts must be rescheduled, banks establish a creditor committee that collects information on the situation from each participating bank. This information is not easily available and is needed to determine how much "new money" is needed. The banks differ about this figure based on their own exposure in that country and their domestic regulatory structure, producing an incentive to dissemble when providing information to the creditor committee. The greater a bank's existing loans to a country, the greater the loss it will absorb in the rescheduling of those loans. Restructuring affects the profits of U.S. banks more than those of German banks because the former do not have so-called hidden reserves to cover their losses and so treat those loans as nonperforming. Strategic misrepresentation surfaces in the discussion of which loans should be included in the total of "old money." The more old money a bank has in a country, the more new money it will have to provide. Each bank has an incentive, then, to argue that some of its loans do not belong in the pool of old money. But if all member banks do so, the ability of the creditor committee to determine the magnitude of the problem and how much new money the banks should provide would be compromised. The banks then face a problem of both information and distribution. They need to know how much old money is at risk, but they disagree about how much new money is needed once they agree on the figure for old money. Member banks help solve the informational problem by treating all loans equally--with some notable exceptions like officially guaranteed loans and short-term trade credits. This general standard eliminates the problem of negotiating how the different classes of loans will be treated in each case. The banks can also anticipate their exposure at the beginning of a rescheduling and know that they will not be singled out to provide an excessive amount of new money. The system does extract accurate information from the member banks. Peter Haas's discussion of the Med Plan, a regime for marine pollution control, may provide an example of an honest communicative equilibrium.(44) The essential problem of controlling pollution of the Mediterranean is ascertaining its extent and consequences. Enforcement of the regime's mandates

dates has not been problematic. The Med Plan has used its resources to fund studies of the consequences of pollution and has collected accurate information on emissions by publishing only total figures, rather than potentially embarrassing national emission figures. The epistemic community of ecologists plays a large role in this regime. They are the source of the information needed to determine what should be done. The regular meetings sponsored by the Med Plan provide a setting for scientists to exchange research on the consequences of different pollutants. These exchanges lead to common views on what regulations are needed and how national policies should enact those regulations in the face of the economic costs of emission controls. Like an honest communicative equilibrium in the model, the Med Plan has concentrated on the exchange of information among actors.

There are also problems of enforcement in each of these three cases. My brief discussions here address just the problems of distribution and information, how they interact, and how they have been dealt with in the three cases of successful international cooperation described above. A complete discussion of these cases also would have to address how the problems of enforcement were handled.

Choosing among forms of cooperation

The players are always better off adopting an equilibrium other than the babbling equilibrium, even if they expect the model to last just one round. Unlike tit-for-tat in the iterated prisoners' dilemma, these tools of cooperation can succeed even if there are no future rounds of the model. Egoists can cooperate under anarchy even if there is no future.(45) The self-enforcing nature of the three games guarantees compliance once the players have agreed how they should cooperate.

But which equilibrium should the players adopt? Figure 4 presents a graphical comparison of the expected payoff to each player from each equilibrium as p, the probability of identical interest, varies. The choice between communicative and coordination equilibria reflects the twin problems of information and distribution. Both players prefer a communicative equilibrium to a pure coordination equilibrium when p > a / (3a - 2). Coordination equilibria are more attractive as the informational problem wanes in importance relative to the distributional problem (as p declines). Leadership equilibria pose a distributional problem. Leaders are always best off in a leadership equilibrium. Followers are better off in a pure coordination equilibrium when p < 1/2 and as well off in an honest communicative equilibrium when p < 1/2. Followers almost always prefer a communicative equilibrium to a leadership equilibrium (except when p is small).

When will the players adopt one equilibrium over the others? For middling values of p, that is when a/(3a - 2) < p > 1/2, both players prefer a dishonest communicative equilibrium. Although the leader does better in a leadership equilibrium, the follower can offer a side-payment to the leader to adopt the communicative equilibrium, yielding a better result for both players. For smaller values of p, either a pure coordination or a leadership equilibrium is likely. Both produce the same total payoff, differing only in their distribution of that payoff. For all but small values of p, the follower is better off in the babbling equilibrium than in a leadership equilibrium. A threat by the follower to refuse a leadership equilibrium may be sufficient to force the leader to accept a pure coordination equilibrium here. For large values of p, the honest communicative and leadership equilibria produce the same value for both players. In summary, communicative equilibria occur when the probability of identical interest is substantial, while lower values of p give rise to pure coordination equilibria. Leadership equilibria are possible either when p is small or large, but some symmetric equilibria can produce the benefits of leadership. Leadership equilibria cannot occur for middling values of p because the followers are better off with no regime than they are under such an equilibrium.

These choices among the equilibria lead directly to testable hypotheses about when we should expect different forms of cooperation. Issues differ in the relative importance of distributional problems to informational problems. Issue-areas in which information is the more prominent problem should produce cooperation like the communicative equilibrium. Negotiations focus first on exchanging information and seeking a consensus about what solutions are best. When distributional issues are more important than informational ones, pure coordination and leadership equilibria are more likely. Split-the-difference bargaining and other types of formulaic negotiations that attempt to guarantee fairness are more likely in those settings. These hypotheses, of course, address how cooperation deals with these two issues. Regimes could address enforcement differently, even if they use similar mechanisms to reach agreement.

But regimes are not always adopted; why not? First, the actors may expect a regime to produce asymmetrical benefits as the leadership equilibrium does. The follower would object to any regime that reduced its payoff below what it would receive without a regime. Even if the disadvantaged player benefitted from a biased regime, it might withdraw from that regime to try to force a redistribution of benefits. If the regime produces a Pareto-optimal outcome, any increase in the disadvantaged player's payoff must cause the other player's payoff to decrease. Expected asymmetry in the distribution of benefits could lead to an attempt to renegotiate the terms of the regime. If the negotiations were to fail, the regime would collapse.

Second, actors might see close ties between the establishment of a regime and the outcome of several rounds of the model. A regime is more difficult to establish when the actors can predict its distributive consequences. The mutual interest in regimes in the model depends on the players' inability to predict what outcome the regime will produce in any given round of the model. If the players can foresee the immediate consequences of a regime, the relative losers may be less interested in its adoption. Once again, they might hold out for a distribution of benefits of a regime more favorable to their immediate interests.(46)

Third, some situations pose problems beyond the problems of distribution and information analyzed here. It is widely believed that regimes have difficulties resolving security issues.(47) Security issues do reflect some of the features of the model: each side has a preferred solution to the issue, and generally both sides prefer either solution to not coordinating on a solution--war. It is unlikely that they have identical interests, however (p is close to 0). The pure coordination equilibrium should render both actors better off. However, there is one very important difference between most security issues and the model. In the model, both sides must cooperate for either to receive its most preferred outcome; in security issues, one actor may be able to secure its most preferred outcome by force.(48) Even when neither side believes it can use force successfully, it cannot exclude the possibility that the other believes it can win through force. The inability to exclude the option of force undermines both communication and coordination. It destroys both sides' incentive to follow the equilibrium.

Knowledge and cooperation

The equilibria require intersubjective knowledge that a particular equilibrium exists in order to facilitate communication and coordination. This mutual acknowledgment could be reached either through agreement or by the emergence of convention. Vincent Crawford and Hans Haller have shown that actors can develop methods of coordinating strategies in coordination games without focal points.(49) According to their definition of a coordination game, the "players have identical preferences over strategy combinations, with two or more (Nash equilibrium) combinations at which each player's strategy choice is a unique best reply if the other player correctly anticipates it, but not in general otherwise."(50) Actors facing a combined informational and distributional problem could then develop an interpretation of messages over time in the absence of an explicit agreement.

Viewed in this way, the model suggests the view of regimes as mechanisms of communication associated with Ernst Haas and with Ruggie.(51) The actors are uncertain about their preferences over strategies, and they use intersubjective understandings to assist their communication of relevant knowledge about the problem they face. As Ruggie argues, the "normative framework" of a regime, which corresponds to the interpretation and the expected strategies it produces in the model, is as essential to a successful regime as its "instruments" (i.e., rules and procedures).(52)

The idea of "knowledge" formalized here is quite limited. Policy coordination requires not only international agreement but also shared understandings about the consequences of policies. For example, international macroeconomic cooperation relies on economic theory that explains how policies affect economic conditions. Because economic theory changes and no one theory can be accepted as definitive and final, actors may disagree about the consequences of proposed policies. In a very uncertain world, enduring cooperation requires shared understandings about the consequences of policies or at least what consequences each actor expects from each possible policy. Otherwise, actors cannot form the expectations of one another's policies needed for successful coordination. In game theory, such shared understandings are called "common knowledge." Something is common knowledge if everyone knows it, if everyone knows that everyone knows it, and so on in an infinite regress. Regimes can be thought of as assisting the creation of common knowledge. In the model, the players' payoffs for the two forms of cooperation are not common knowledge. By exchanging private information, the players increase their common knowledge about their payoffs, which can increase the likelihood of coordination.

Hegemony versus leadership

The leadership equilibrium contrasts with the role that many identify with hegemony in the establishment and maintenance of regimes. There are two versions of hegemonic stability theory, one theorizing benevolent and the other self-interested hegemons. In the former view, the hegemon benevolently shoulders the burden of establishing a regime.(53) Free trade illustrates this view. Although all nations benefit from loosening trade restrictions, hegemons-like Great Britain in the nineteenth century and the United States in the twentieth century-prefer minimal restrictions. Other states, like the followers in the model, prefer some restrictions on trade to protect nascent industries. In the other view of hegemonic stability, hegemons are not benevolent; they simply establish the order they prefer.(54)

The idea of leadership in the model neatly combines both views. The follower does better in a leadership equilibrium than in the babbling equilibrium when the probability of identical interest is small or large (when p is close to 0 or > 1/2). These cases parallel the benevolent view of hegemons; both sides are better off under a leadership equilibrium. But like the version that views hegemons as self-interested, the leader always gets the outcome it believes is best when it issues its directives.(55) Both sides are generally better off with a leader, but leadership advances the leader's interests as it perceives them.

Leadership in the model does not require superior capabilities. The mutual recognition that one actor is a leader is sufficient by itself to sustain leadership. Leadership equilibria work because they create stable expectations about what the actors will do. An advantage in capabilities is irrelevant to successful leadership in the model. A leader in the model has power; it achieves the outcomes that it prefers by eliciting supportive actions from the follower. Then actors can have power in the model even without asymmetry in capabilities.

The observation that capabilities may not be necessary for leadership raises two questions for hegemonic stability theory. First, why would relative decline in the hegemon's capabilities cause a regime to fail? Because capabilities are not necessary for leadership, such equilibria should not fail as the leader declines. Keohane's question about how international cooperation is sustained as the hegemon's capabilities decline is irrelevant in the model.(56) Keohane's answer to his own question, however, is correct: an operating hegemonic equilibrium creates self-sustaining expectations. Regimes under leadership persist because they are in both actors' interest.

Second, how is leadership achieved? Here the model provides no clue. Advantages in capabilities, either economic or military, are beside the point. Capabilities do not affect either the probability of identical interest (p) or the degree of distributional conflict (a). All that is needed to establish a leadership equilibrium is a shared understanding that one actor is the leader. It may be that preponderance in capabilities furthers these expectations in the follower. In that case an argument--and even better a model--that shows how these expectations arise should be advanced.

These points mirror the observation that the qualifications for hegemony have never been made explicit.(57) Is it Size of economy, military power, or something else? Does relative or absolute size matter?(58) Without such a definition, we cannot test hegemonic stability theory. This conceptual vagueness leads to the confusion underlying the debate on whether U.S. hegemony has passed.(59) If we do not know what characterizes hegemony, we can hardly tell whether it has passed.

To be fair to hegemonic stability theory, many reasons why hegemony fosters regimes have been advanced. This model alone cannot address all these arguments. For example, hegemonic stability theory sees the sanctioning problem as more central than informational and distributional problems. James Alt, Randall Calvert, and Brian Humes focus on the sanctioning problem and produce a similar result to the model here.(60) Hegemony can be sustained in their model even as the hegemon becomes weaker (as the probability that punishment is costly to the hegemon, w, increases) if the allies begin with a sufficiently strong belief about the willingness of the hegemon to punish defection--[beta]/([alpha] t [beta]) in their model. But how strong such a belief must be depends upon the benefits to the follower of unchallenged defection (b in their model). Even if we focus on the sanctioning problem, it is not clear that a decline in the hegemon's capabilities leads to a weakening of regimes. Hegemonic stability theory requires explicit models that explain how hegemony fosters and supports regimes. The existing models of hegemonic stability do not provide a sufficient explanation.

The leadership equilibria provide a crude way to formalize the effects of hegemonic leadership apart from the effects of hegemonic capabilities. Leadership in the model reflects some characteristics of leadership in the literature. Oran Young discusses how three types of political leadership--structural, entrepreneurial, and intellectual--assist in the creation of regimes.(61) Although the model discussed here addresses how regimes operate rather than how they are established, leadership in the model parallels elements of entrepreneurial and intellectual leadership. Entrepreneurial leadership shapes issues so that all parties to a negotiation see benefits for themselves in a deal. Intellectual leadership provides a system of thought that allows all to see the benefits of an agreement. Both types of leadership rely on the ability to provide information that creates common interests by shaping how states perceive their interests. Once the follower knows which move the leader will play, playing that same move is always in its interest.

Leadership in the model also displays a simple idea of the "second and third faces of hegemony" discussed by Scott James and David Lake.(62) The first face is the direct use of sanctions; the second face is the use of policy to alter others' incentives internally; the third face is the use of ideology to alter others' perceptions of their interests. Like the second and third faces, leadership in the model provokes the follower to obey the leader by changing its interests. This formal understanding of leadership is rudimentary, but it provides a way to connect leadership clearly to the follower's choices.

Conclusion

This article has presented a way to formalize cooperation in the face of problems of distribution and information. The model examines situations in which the actors must balance these twin problems. Both actors face uncertainty about which move is preferable within a round of the model. Arrangements to share information can increase the probability of coordinating action by assisting the convergence of the actors' expectations about one another's moves. These arrangements require prior agreement between the actors about their specific form. Several different equilibria are possible; the most advantageous equilibrium depends upon the relative importance of the two problems.

The model demonstrates that first, leadership solutions to coordination problems always exist--but the idea of leadership here is very different from the concept of hegemonic provision of public goods. Second, institutions without any power of enforcement can help actors coordinate. They achieve cooperation in the face of anarchy even without a shadow of the future. Third, diffuse reciprocal strategies arise naturally in these equilibria through the actors' pursuit of their own interests. Fourth, norms and institutions are intertwined within a functioning regime. Fifth, communicative equilibria are more likely than coordination equilibria as the probability of identical interest increases. Finally, problems of distribution and information cannot be solved simultaneously.

Any stylized model has limitations. This model analyzes the problem of distribution and information isolated from problems of sanctioning and monitoring. This model does not address the latter problems at all. Real issues involve all four problems. Sanctioning and monitoring problems must also be addressed. The creation and implementation of regimes may be substantially more difficult than the model suggests because of these problems. The equilibria here are self-enforcing; actors follow the recommendations of the interpretation because it is in their immediate interest to do so. When defection can be profitable in the short run, sanctions may be needed to enforce cooperation. Effective sanctioning may not be possible if defection is difficult to detect.

Second, this model is highly stylized. It should not be taken as a model of any particular instance of cooperation. Instead, the model focuses on the interaction between distribution and information in the abstract. This abstraction helps us to see that distributional and informational problems cannot be solved simultaneously.

One stylization in the model is not an issue. The model is symmetric; the payoffs of the players are identical, and they face the same strategic problem. Real issues are not symmetrical. Introducing asymmetry in the payoffs does change the exact strategies used in each equilibrium, but it does not change the character of the equilibria. All four equilibria are possible with asymmetric payoffs.

International cooperation calls for some answer to the problems of distribution and information. Multiple solutions are possible, and actors have divergent preferences over those solutions. Because economic and political issues are complex, actors cannot always have clear preferences over which courses of action produce the most desirable outcomes. Communication can help to alleviate the collective problem of lack of knowledge of the consequences of different actions. Further, these two problems exacerbate one another. The distributional problem hinders solutions to the informational problem. Creating and implementing regimes that can assist cooperation depends on addressing them successfully.

Appendix

This appendix presents proofs of the various equilibria presented in the text. I denote an equilibrium as: (player 1's move if signal 1 is received, player 1's move if signal 2 is received; player 2's move if signal 1 is received, player 2's move if signal 2 is received).

Babbling equilibrium

I begin by calculating a player's beliefs about what game is being played after the player receives its signal. Because I am solving for a babbling equilibrium, I need only solve for one player's strategy, say player 1. (Player 2's strategy can be found by exchanging 1 and 2 and A and B in the following calculations.) Let "battle of the sexes is game being played" be abbreviated as BotS, "both prefer move A (B) is game being played" be abbreviated as BPMA (BPMB), and "signal 1 (2) received" be abbreviated as 1 (2), we have the following conditional probability from Bayes's theorem:

p(BotS[+ or -] = p(BotS)p(1 \ BotS) \ p(BotS)p(1\ BotS) + p(BPMA)p(1\BPMA)p(BPAM)p(BPMB)p(1\BPMA)

= 1/2 (1 - p)\ 1/2 (1 - p) + 1 (1/2 p) + (1/2 p) = 1 - p

Similarly, p(BPAM\1)= p, p(BPMB\1 = 0, p(BotS \2) = 1- p, p(BPMA\2) = 0, and P(BPMB 12) = p.

First, I find the conditions under which the players will follow their signals, that is, when (A,B;A,B) is an equilibrium. Any player receiving a favorable signal [i.e., player 1 (2) receives signal 1 (2)] always prefers playing its preferred strategy--A for player 1 and B for player 2. Let "player 2 plays A (B) " be abbreviated as 2A (2B) and "player 1's utility for the AA(BB) outcome" be abbreviated as [u.sub.1](AA) [[u.sub.1](BB)]. Given player 2's equilibrium strategy, p(2 A\BotS) = p(1\BotS) = 1/2, and p (2A\BPAM) = p(1\BPMA) = 1. Player I's optimal strategy given its signal and player 2's strategy is then calculated:

[u.sub.1](A\1) = p(BotS \1)p(2,4\BotS)[u.sub.1](AA\BotS)
                   + p(BPMA\1)p(2,4 BPMA)[u.sub.1](AA\BPMA)
                   + p(BPMB\1)p(2,4 BPMB)[u.sub.1](AA\BPMB)
                 = (1 - p)1/2(a) + p(1)(a) + (0)(0)(1) = 112a(1 + p)


[u.sub.1](B\1) =p(BotS\1p(2B\BotS)[u.sub.1](BB\BotS)
                   + p(BPMA\1)p(2B\BPMA)[u.sub.1](BB\BPMA)
                   + p(BPMB\1)p(2B\BPMB)[u.sub.2](BB\BPMB)
                = (1 - p)1/2(1) + (p)(0)(1) + (0)(1)(a) = 1/2(1 - p)


As [u.sub.1](A\1)[greater than or equal to][u.sub.1](B\1), player 1 will always play A if it recei values of p such that player 1 will play B if it receives signal 2 can be found:

[u.sub.1](A\12) = p(BotS\2)p(2A\BotS)[u.sub.1](AA\BotS)
                    + p(BPMA\2)p(2A\BPMA)[u.sub.1](AA\BPMA)
                    + p(BPMB\2)p(2A\BPMB)[u.sub.1](AA\BPMB)
                    + p(BOMB\2)p(2A\BPMB)[u.sub.1](AA\BPMB)
                  = (1 - p)1/2(a) + (0)(1)(a) + (p)(0)(1) =1/2a(1-p)
                 + p(BPMA\2)p(2B\BPMA)[u.sub.1](BB\BPMA)
                 + p(BPMB\2)p(2B\BPMB)[u.sub.1](BB\BPMB)
               = (1 - p)1/2(1) + (0)(0)(1) + (p)(1)(a) = 1/2(1 - p) + ap


Player 1 prefers playing move B to move A when [u.sub.1](B \2) [less than or equal to] [u.sub/1](A above values and solving for p, we arrive at p [less than or equal to] (a - 1)/(3a - 1).

For p < (a - 1)/(3a - 1), players who receive unfavorable signals play a mixed strategy. We can solve for this mixed strategy bY finding the probability q that 2 plays B when it receives signal 1 that makes player 1 indifferent between playing move A and move B when it receives signal 2: [u.sub.1](A\2) = (1 - p)(1/2q)(a) + p(0)(1) = 1/2aq(1 - p) [u.sub.1](B\2) = (1 - p)(1 - 1/2q)(1) + p(1)(a) = ap + (1 - p)(1 - 1/2q)

Setting the two expected utilities above equal and solving for q, we obtain q = 2(1 - p + ap)/ (a + 1)(1 - p)

The babbling equilibrium without communication when p + (a - 1)1(3a - 1) is {A, [qA, (1- q)B];[(1 - q)A,qB],B} for q above.

Pure coordination equilibrium

To show that the pure coordination equilibrium indeed is an equilibrium, begin with play after messages have been exchanged. Whenever one player expects the other to playa (or B), it should play A (or B) regardless of which game is being played. Because both players understand what strategy they are to play after the messages are revealed, their behavior at this stage is self-enforcing.

At the message stage of the game, both players send each message with a probability of 1/2, making each other indifferent between sending either of the messages. If one player sends a message with a probability greater than 1/2, the other can increase its payoff by sending the corresponding message (the matching message for player 1, the other message for player 2). Then the strategy pairing is an equilibrium.

Communicative equilibria

In the honest communicative equilibrium, each player's message reports accurately the signal it has received and the combination of the messages determines on which equilibrium players coordinate. If both send message 1 (or 2), they both play A (or B, respectively). If the messages do not match, they send a second set of messages using the pure coordination equilibrium above, producing (A;A) one-half of the time and (B;B) the other half. At the stage of playing moves, this equilibrium is self-enforcing. I determine the conditions under which the players send honest messages.

Both players report favorable signals accurately. The question of honesty lies with players receiving unfavorable signals. Let "player 1 sends message 1" be abbreviated as 1s1. Player 1's expected utility for sending each type of message, given that it has received a signal of 2, is thus calculated: [u.sub.1](1s\2) = p(BotS \2)[p(2s1\BotS)[u.sub.1](aa \BotS)p(2s2 \BotS)[u.sub.1](2s2\BotS)]

. p(bpm412)[p(2sl IBPMA)U,(AA IBPMA)

p(2s2\BPMA)[u.sub.1](1/2AA, 1/2BB\BPMA)]

. p (BPMB\2[p (2s1\BPMB)[u.sub.1](AA\BPMB)

+ p(2s2\BPMB)[u.sub.1](1/2AA, 1/2BB\BPMBA)]

= (1 - p){1/2(a) + 1/2 (a + 1)[} + 0[(1)(a) (0) 0)(1)] + p)[(0)(1) + (1)(a)]

= 1/2(a - 1)(1 - p)

[u.sub.1](1s2\2) = (1 - p){1/2[1/2(a + 1)] + 1/2(1)) (0)[(1)(a) + (0)(1)]+ (p)[(0)(1) + (1)(a)]

= 1/2p(a - 1)

Setting the two expected utilities above equal, p = 1/2. For p [less than or equal to] 1/2 then, t the honest communicative equilibrium form an equilibrium.

For p < 1/2, honest reporting of players receiving unfavorable signals cannot be supported in equilibrium. One interpretation that can be supported is described in the text. Message pairs of (1,1), (2,1), and (2,2) are treated as in the honest communicative equilibrium, but (1,2) directs the players to play the babbling equilibrium of the game without communication where the players follow their signals [i.e., A,B;A,B)]. Again the strategies are self-enforcing at the stage of playing moves, so only the signaling stage is checked. If r is the probability that player 2 sends message 2 when it receives signal 1, the expected values for player 1 of sending messages of 1 and 2 if signal 2 is received can then be calculated:

[u.sub.1](1s1\2) = 1/2(1 - p)(1 - r)a + 1/2(1 - p)r(0) + 1/2(1 - p)(1) + p(a)

= 1/2(1 - p)(1 + a - ar) + ap

[u.sub.1](1s2\2) = 1/2(1 - p)(1 - r)[1/2(a + 1)] + 1/2(1 - p)r(1) + 1/2(1 - p)(1) + p(a)

= 1/4(1 - p)(3 + a + r - ar) + ap

Setting the two expected utilities equal and solving for r, r = (a - 1)1(a + 1). Further, the babbling equilibrium without communication can only be supported if player I's updated probability that one of the both-prefer games is being played after receiving signal 1 and observing messages (1,2), p[BPMB 1 (1,2)], is greater than or equal to (a - 1)1(3a - 1). Otherwise, (A,B;AB) is not an equilibrium.

p[BPMB\(1, 2)] = p(2s2\BPMB)p(BPMB)/p(2s2 \BPMB)p(BPMB) + p(2.s2 \BotS)p(BotS)

(1)(p) + 1/2 + 1/2r)(1 - p) = p / 1/2(1- r)(1 - p)

Setting this equal to (a - 1)1(3a - 1), substituting for r, and solving for the original probability p, I obtain p [less than or equal to] (a - 1)1(3a + 1).

For p < (a - 1)1(3a + 1), a dishonest communicative equilibrium can be supported using the other babbling equilibrium without communication. However, the pure coordination equilibrium Pareto dominates that equilibrium, so I omit it.

Leadership equilibria

Once the leader announces its intended move, both players have an incentive to play that move. I concentrate then on showing first, which move the leader will announce and second, why the follower always sends the same message regardless of what signal it has received. Assume for convenience that player I is the leader. Then the leader prefers playing A if the game is BotS. If it receives signal 1 (a favorable signal), it knows it is playing either BotS or BPMA, implying it is always better off coordinating on A. If it receives signal 2, it knows it is playing either BotS or BPMB with probabilities p and 1 - p, respectively. The leader's expected utility for announcing moves A and B are as follows:

[u.sub.1](a 12) = (a)p(BotS \2) + (1)p(BPMB\2) = ap + 1 - p

[u.sub.1](B\2) = (1)p(BotS\2) + (a)P(BPMB\2) = p + a - ap

When p = 1/2, these values are equal; when p 1/2, player 1 prefers A; and when p > 1/2, player 1 prefers B.

To see that player 2 always sends the same message, note that player 1 announces move A if it receives signal 1 regardless of player 2's message. Then player 2's message can influence player 1 only when player 1 receives signal 2. But then the players are playing either BotS or BPMB and player 2 always prefers move B regardless of which signal it has received. Player 2 always sends the same message regardless of which signal it has received.

Earlier versions of this article were presented at the annual meeting of the American Political Science Association, San Francisco, 30 August-2 September 1992, and the annual meeting of the Public Choice Society, New Orleans, La., 20-22 March 1992. I thank Bruce Bueno de Mesquita, Randy Calvert, Jim Fearon, Hein Goemans, Robert Keohane, Jochen Lorentzen, and Robert Powell for their comments on this article. This article benefitted from seminar presentations at: th Politics and Organizations Seminar at Stanford University, Washington University, the University of Rochester, the University of California at Berkeley, the PIPES seminar at the University of Chicago, and at the University of Illinois at Urbana-Champaign. I retain all responsibility for all interpretations and any errors herein. (1.) For discussions of the interaction between sanctioning and monitoring problems, see Jonathan Bendor, "In Good Times and Bad: Reciprocity in an Uncertain World," American Journal of Political Science 31 (August 1987), pp. 531-58; and George W. Downs and David M. Rocke, Tacit Bargaining, Arms Races, and Arms Control (Ann Arbor: University of Michigan Press, 1990). (2.) My distinction between sanctioning and monitoring on one hand and distribution and information on the other parallels Stein's distinction between collaboration and coordination. Sanctioning and monitoring are problems of collaboration, and distribution and information are problems of coordination. See Arthur A. Stein, Why Nations Cooperate: Circumstance and Choice in International Relations (Ithaca, N.Y.: Cornell University Press, 1990). Lisa Martin adds problems of suasion and assurance to Stein's two problems in "Interests, Power, and Multilateralism," International Organization 46 (Autumn 1992), pp. 765-92. The model here addresses both of Martin's problems. Suasion is a special case of coordination, and assurance is captured in the both-prefer games described below. (3.) For discussions of some of the problems of reaching agreement on how to cooperate, see Stephen D. Krasner, "Global Communications and National Power: Life on the Pareto Frontier," World Politics 43 (April 1991), pp. 336-66; and James K. Sebenius, "Challenging Conventional Explanations of International Cooperation: Negotiation Analysis and the Case of Epistemic Communities," International Organization 46 (Winter 1992), pp. 323-365. (4.) See Robert O. Keohane, After Hegemony: Cooperation and Discord in the World Political Economy (Princeton, N.J.: Princeton University Press, 1984); and Stephen D. Krasner, "Structural Causes and Regime Consequences: Regimes as Intervening Variables," in Stephen D. Krasner, ed., International Regimes (Ithaca, N.Y.: Cornell University Press, 1982), pp. 1-21. (5.) The work here draws heavily on work in economics on "cheap talk"--cost-free, nonbinding messages. See Joseph Farrell, "Cheap Talk, Coordination, and Entry," Rand Journal of Economics 19 (Spring 1987), pp. 34-39; and Joseph Farrell and Robert Gibbons, "Cheap Talk Can Matter in Bargaining," Journal of Economic Theory 48 (June 1989), pp. 221-37. For a discussion of the signaling game framework and an analysis of how signals can aid cooperation, see Vincent P. Crawford and Joel Sobel, "Strategic Information Transmission," Econometrica 50 (November 1982), pp. 1431-51. For an analysis of how different forms of communication assist in coordination problems in which there is limited information, see Jeffrey S. Banks and Randall L. Calvert, "A Battle-of-the-Sexes Game with incomplete Information," Games and Economic Behavior, vol. 4, no. 2, 1992, pp. 347-72. (6.) Schotter develops the view that social institutions, of which regimes are an example, are equilibria of a game. See Andrew Schotter, The Economic Theory of Social Institutions (New York: Cambridge University Press, 1981). (7.) For a discussion of diffuse reciprocity, see Robert O. Keohane, "Reciprocity in International Relations," International Organization 40 (Winter 1986), pp, 1-27. (8.) See Ernst B. Haas, "Why Collaborate? Issue-Linkage and International Regimes," World Politics 32 (April 1980), pp. 357-405; Ernst B. Haas, "Words Can Hurt You; or, Who Said What to Whom About Regimes," in Krasner, International Regimes, pp. 23-59; and John Gerard Ruggie, "International Regimes, Transactions, and Change: Embedded Liberalism in the Postwar Economic Order," in ibid., pp. 195-231. (9.) See for example, Alexander Wendt, "Anarchy is What States Make of It: The Social Construction of Power Politics," International Organization 46 (Spring 1992), pp. 391-425. (10.) George A. Codding, Jr., The International Telecommunications Union: An Experiment in International Cooperation (Leiden: E.J. Brill, 1952), pp. 60-61. (11.) Ibid., pp. 61-75. (12.) If the actors must reach an agreement that will persist over time, agreement may be difficult in the face of distributional interests. Fearon shows that long "shadows of the future" make reaching such agreements more difficult. See James Fearon, "Cooperation and Bargaining Under Anarchy," paper presented at the annual meeting of the Public Choice Society, New Orleans, La., 19-21 March 1993. (13.) See pp. 206-14 of Charles Lipson, "Bankers' Dilemmas: Private Cooperation in Rescheduling Sovereign Debts," in Kenneth A. Oye, ed., Cooperation Under Anarchy (Princeton, N.J.: Princeton University Press, 1986), pp. 200-225. (14.) Peter M. Haas, "Do Regimes Matter? Epistemic Communities and Mediterranean Pollution Control," International Organization 43 (Summer 1989), pp. 377-403. (15.) Charles P. Kindleberger, The World in Depression, 1929-1939, 2d ed. (Berkeley: University of California Press, 1986), pp. 288-305. Also see Duncan Snidal, "The Limits of Hegemonic Stability Theory," International Organization 39 (Autumn 1985), pp. 579-614, for a critique of this view. (16.) Similarly, externalities create distributional problems. The Coase theorem states that such problems can be resolved by private negotiation if property rights are already defined, and the actors have perfect information and do not face transaction costs. But the adoption of a property rights regime creates a distributional problem. Keohane discusses the Coase theorem and international cooperation in After Hegemony, pp. 85-87. (17.) See Robert Axelrod, The Evolution of Cooperation (New York: Basic Books, 1984). Prisoners' dilemma is the two-by-two game with a dominant strategy equilibrium, typically labeled D for "defect," that is not Pareto-optimal. In single-shot play or iterated games with a fixed end point, D is the dominant strategy of the game on all rounds. The outcome produced by the other strategy, typically labeled C for "cooperate," improves both players' payoff over DD. (18.) Tit-for-tat is not a perfect equilibrium of the open-ended iterated prisoners' dilemma because it requires suboptimal behavior if one side defects; see Eric Rasmusen, Games and Information (Cambridge: Basil Blackwell, 1989), p. 91. Actors are better off if they do not carry ou the punishments dictated by tit-for-tat. (19.) Robert Axelrod and Robert O. Keohane, "Achieving Cooperation Under Anarchy: Strategies and Institutions," in Oye, Cooperation Underanarchy, pp. 226-54. (20.) James E. Alt, Randall L. Calvert, and Brian D. Humes, "Reputation and Hegemonic Stability: A Game-Theoretic Analysis," American Political Science Review 82 (June 1988), pp. 445-66. (21.) R. Duncan Luce and Howard Raiffa, Games and Decisions (New York: John Wiley, 1957), p. 91. Snidal and Stein each also argue that battle of the sexes exhibits elements of the problem of international cooperation that the prisoners' dilemma does not. See Duncan Snidal, "Coordination Versus Prisoners' Dilemma: Implications for International Cooperation and Regimes," American Political Science Review 79 (December 1985), pp. 923-42; and Arthur A. Stein, "Coordination and Collaboration: Regimes in an Anarchic World," in Krasner, International Regimes, pp. 115-40, respectively. (22.) I use the term "move" rather than "strategy" to describe each player's choice in a particular round. Strategies in game theory are complete plans to play the game. A strategy here specifies what moves should be played and what messages--if any--should be sent in any possible round. For an introduction to game theory, see James D. Morrow, Game Theory for Political Scientists (Princeton, N.J.: Princeton University Press, 1994). (23.) Alternatively, one can assume that the parties agree to a middle price halfway between the two prices if the buyer proposes a high price and the seller, a low price. This variation does not change the strategic logic of battle of the sexes. There are still two pure strategy equilibria, (A; and (B;B), and a symmetric mixed-strategy equilibrium. The probabilities of each move in the mixed-strategy equilibrium do change. (24.) Keohane, After Hegemony, p. 63. (25.) Both-prefer games are also known as "stag hunt" or an assurance game; see Martin, "Interests, Power, and Multilateralism," and Stein, Why Nations Cooperate. (26.) The model is symmetric because all three games are symmetric. Both players have the same payoff for the three possible outcomes: more preferred coordination, less preferred coordination, and no coordination. This symmetry has no effect on the equilibria of the game, as I discuss later. (27.) For a review of signaling games and their application to political science, see Jeffrey S. Banks, Signaling Games in Political Science (New York: Gordon and Breach, 1991). (28.) The symmetry of equal chances of each game being played and each signal does not influence the set of equilibria for the model. Changing these probabilities alters the mixed strategies in each equilibrium. (29.) E. Haas, "Why Collaborate?" pp. 367-70. (30.) Keohane adds that regimes assist cooperation by changing transaction costs to favor mutually beneficial trades. See After Hegemony, pp. 89-92. (31.) See James Friedman, "A Noncooperative Equilibrium for Supergames," Review of Economic Studies 38 (January 1971), pp. 1-12; and Drew Fudenberg and Eric Maskin, "The Folk Theorem in Repeated Games with Discounting or with Incomplete Information," Economettica 54 (May 1986), pp. 533-54. (32.) Robert Pahre, "A Model of Economic and Security Cooperation and the Case of Postwar Europe," University of Rochester, Rochester, N.Y., 1991. (33.) I examine perfect Bayesian equilibria of the model. Actors act rationally at each move given their beliefs in a perfect Bayesian equilibrium; see Morrow, Game Theory for Political Scientists, chap. 6. In this model, the actors' beliefs are the probability of the game they are playing given t information the actor has at that moment. Sequential rationality requires that the actors' messages and moves are optimal given their beliefs when those messages are sent and moves made and that their beliefs are calculated from their equilibrium strategies using Bayes's law whenever possible. (34.) The model also has asymmetric equilibria without communication. In such an asymmetric equilibrium, the actors always coordinate on one move--A or B--leading to different expected payoffs. Because the game is symmetric, there is no a priori way to distinguish the players. Harsany and Selten argue that in the absence of any way to distinguish the players in a symmetric game, only symmetric equilibria are plausible; see John C. Harsanyi and Reinhart Selten, A General Theory of Equilibrium Selection in Games (Cambridge, Mass: MIT Press, 1988), pp. 70-71. I ignore the asymmetric equilibria of the game. Without some mechanism to explain how the players share mutual expectations that they will always play one move, we cannot explain how the players can play an asymmetric equilibrium. I discuss later a regime with communication that produces asymmetric payoffs for the players. (35.) For specifications of all equilibria and proofs, please see the appendix. (36.) By assuming that messages honestly report information unless the players have incentives to dissemble, I follow Rabin's concept of "credible message" equilibrium; see Matthew Rabin, "Communication Between Rational Agents," Journal of Economic Theory 51 (June 1990), pp. 144-70. There are equilibria that mirror those discussed in the text, i.e., wherein the messages mean the opposite of the observed signal. As Rabin cogently argues, agents agree on forms of communication that can be used to refine the set of equilibria of a game. Equilibrium behavior depends not only on the incentives of the game but also on forms of communication between the players. (37.) The pure coordination equilibrium is also preferable to the dishonest communicative equilibrium when p < a/(3a - 2), which is greater than (a - 1)1(3a + 1).A later section discusses choosing among different possible equilibria. (38.) There are other ways to send messages sequentially that have the same effect. For example, the first player's message indicates what move the players will coordinate on unless the second player objects to that move in its response. In all cases, the second message provides the cue that the players use to coordinate. (39.) The self-enforcing feature of battle of the sexes contrasts with prisoners' dilemma, in which players always have a short-run incentive to defect. General reciprocal punishment strategies also produce behavior that looks like Keohane's diffuse reciprocity. See Bendor, "In Good Times and Bad," and Downs and Rocke, Tacit Bargaining, Anns Races, and Arms Control (40.) Keohane, "Reciprocity in International Relations," pp. 20-21. (41.) In the leadership regime, there is no reciprocity for the follower. The follower does the leader's bidding because following the leader's directives is in its short-run interest. Still, the leadership equilibrium does produce an equal expected distribution of the total payoff when p > 1/2. This equality results because the follower denies the leader the information that the leader needs to exploit the follower. (42.) Keohane, After Hegemony, p. 59. (43.) Lipson, "Bankers' Dilemmas." (44.) P. Haas, "Do Regimes Matter?" (45.) For the claim that the "shadow of the future" is necessary for cooperation among egoists, see Axelrod, The Evolution of Cooperation. (46.) Oran R. Young, "The Politics of international Regime Formation: Managing Natural Resources and the Environment," International Organization 43 (Summer 1989), pp. 349-75 and specifically pp. 361-62 and 367. (47.) Jervis argues that the Concert of Europe functioned as a regime during the period 1815-22 and that other examples of security regimes can be found. However, he concedes that security regimes are rare. See his essays, "Security Regimes," in Krasner, International Regimes, pp. 173-94; and "From Balance to Concert: A Study of International Security Cooperation," in Oye, Cooperation Under Anarchy, pp. 58-79. (48.) See James D. Morrow, "A continuous-Outcome Expected Utility Theory of War," Journal of Conflict Resolution 29 (September 1985), pp. 473-502; and James D. Morrow, "Capabilities, Uncertainty, and Resolve: A Limited Information Model of Crisis Bargaining," American Journal of Political Science 33 (November 1989), pp. 941-72. (49.) Vincent P. Crawford and Hans Haller, "Learning How to Cooperate: Optimal Play in Repeated Coordination Games," Econometrica 58 (May 1990), pp. 571-95. (50.) Ibid., p. 571. (51.) See E. Haas, "Why Collaborate?"; E. Haas, "Words Can Hurt You"; and Ruggie, "International Regimes, Transactions, and Change." (52.) Ruggie, "International Regimes, Transactions, and Change," p. 200. (53.) Kindleberger, The World in Depression, pp. 288-305. (54.) Stephen D. Krasner, "State Power and the Structure of International Trade," World Politics 28 (April 1976), pp. 317-47. (55.) We cannot say that the leader benefits more than the follower because that statement makes an interpersonal comparison of utilities. (56.) Keohane, After Hegemony. (57.) Timothy J. McKeown, "Hegemonic Stability Theory and Nineteenth Century Tariff Levels in Europe," International Organization 37 (Winter 1983), pp. 73-91. (58.) Snidal shows that the public goods argument for hegemony relies on absolute rather than relative size of economy. See Snidal, "The Limits of Hegemonic Stability Theory," p. 589. (59.) See Jacek Kugler and A.F.K. Organski, "The End of Hegemony?" International Interactions, vol. 15, no. 2, 1989, pp. 11 3-28; Bruce Russett, "The Mysterious Case of Vanishing Hegemony; or Is Mark Twain Really Dead?" International Organization 39 (Spring 1985), pp. 207-31; and Susan Strange, "The Persistent Myth of Lost Hegemony," International Organization 41 (Autumn 1987), pp. 551-74. All argue that the United States is still a hegemon. (60.) Alt, Calvert, and Humes, "Reputation and Hegemonic Stability." (61.) Oran R. Young, "Political Leadership and Regime Formation: On the Development of Institutions in International Society," International Organization 45 (Summer 1991), pp. 281-308. (62.) Scott C. James and David A. Lake, "The Second Face of Hegemony: Britain's Repeal of the Corn Laws and the American Walker Tariff of 1846," international Organization 43 (Winter 1989), pp. 1-29.
COPYRIGHT 1994 Cambridge University Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1994 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Morrow, James D.
Publication:International Organization
Date:Jun 22, 1994
Words:16698
Previous Article:A security regime among democracies: cooperation among Iroquois nations.
Next Article:Political learning by doing: Gorbachev as uncommitted thinker and motivated learner.
Topics:

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters