DANIEL W.DREZNER 713 some types of concessions have very poor or expensive substitutes.Political assets are more heterogeneous and less substitutable than economic assets.Concessions that transfer information,institutional access,or asset-specific investments alter the local distribution of power in a way that other power resources cannot without serious political externalities.Second,the assumption of fungible power is not necessarily true in the short run.It may be possible to redirect power resources toward a particular target,but it cannot be done immediately,and there are associated opportunity costs with it.While states must keep the aggregate distribu- tion of power in mind,they will be more concerned about the local correlation of forces and policy levers. The importance of relative transfers depends on the expectation of future conflict.If the concession is significant but the number of expected disputes in the bilateral relationship is small,the relative distribution of payoffs will matter less. Previous game-theoretic models have highlighted the importance of conflict expec- tations.Robert Powell(1991)argues that relative gains concern is salient when a military attack is expected.The logic developed here includes instances beyond military force,but is consistent with Powell's argument.Duncan Snidal (1991) concludes that relative gains concern will be salient when the locus ofa foreign policy leader's attention is on the bilateral dispute.A high expected probability of future confrontations and threats from another state will lead to an enhanced concern for relative gains in bilateral dealings with that state. Conflict expectations will also affect states through their concern over their bargaining reputation.In a future conflict,foreign policy leaders will consider the history of prior bilateral negotiations in developing conjectures about the other states'behavior.For example,if a country has a history of acquiescing in the face of deadlock,the other country will have the incentive to act tough in crises,so as to obtain a better outcome.By contrast,if that state has a history of refusing to fold even if that action is costly,the other foreign policy leader will take that history into account in later rounds of negotiations. If a nation-state demonstrates that it prefers to incur a costly outcome,provided it hurts the other side even more,it can affect the outcomes of those future conflicts. Furthermore,reputations are interdependent;one state's boost in reputation can come at the expense of the other.In a crisis,whichever side backs down is perceived to have lost credibility.At the same time,the side that triumphed will find its credibility enhanced.Therefore,leaders must be concerned that backing down in the present will raise the other state's expectation of success in later episodes.States will care about the distribution of payoffs in the present because it will alter their perceived reputations in the other state. Reputation becomes more important as the expectation ofrepeated interactions, and the shadow of the future,increases.If players anticipate that a game will be played often,they will be more willing to sacrifice current gains for a larger payoff in the future.Countries will always have some incentive to be concerned about their reputation.Nevertheless,the incentive will be much stronger when they expect a greater number of potential conflicts.States have both tangible and abstract reasons to factor in conflict expectations when crafting their existing strategies. Casual empiricism suggests that the expectation of future conflict varies with the dyad.For example,during the Cold War,the United States clearly expected more zero-sum disputes with the Soviet Union than with West Germany.At present,India can anticipate a greater number of disagreements with Pakistan than with Bangla- desh.For the foreseeable future,the United States is more concerned with a possible North Korean nuclear program than a probable Israeli nuclear program.In each of the cases where more future conflict is expected,one or both states perceive a significant threat from the other.In relationships where threat perception is high, so is the expectation of future zero-sum conflicts
some types of concessions have very poor or expensive substitutes. Political assets are more heterogeneous and less substitutable than economic assets. Concessions that transfer information, institutional access, or asset-specific investments alter the local distribution of power in a way that other power resources cannot without serious political externalities. Second, the assumption of fungible power is not necessarily true in the short run. It may be possible to redirect power resources toward a particular target, but it cannot be done immediately, and there are associated opportunity costs with it. While states must keep the aggregate distribution of power in mind, they will be more concerned about the local correlation of forces and policy levers. The importance of relative transfers depends on the expectation of future conflict. If the concession is significant but the number of expected disputes in the bilateral relationship is small, the relative distribution of payoffs will matter less. Previous game-theoretic models have highlighted the importance of conflict expectations. Robert Powell (1991) argues that relative gains concern is salient when a military attack is expected. The logic developed here includes instances beyond military force, but is consistent with Powell's argument. Duncan Snidal (1991) concludes that relative gains concern will be salientwhen the locus of a foreign policy leader's attention is on the bilateral dispute. A high expected probability of future confrontations and threats from another state will lead to an enhanced concern for relative gains in bilateral dealings with that state. Conflict expectations will also affect states through their concern over their bargaining reputation. In a future conflict, foreign policy leaders will consider the history of prior bilateral negotiations in developing conjectures about the other states' behavior. For example, if a country has a history of acquiescing in the face of deadlock, the other country will have the incentive to act tough in crises, so as to obtain a better outcome. By contrast, if that state has a history of refusing to fold even if that action is costly, the other foreign policy leader will take that history into account in later rounds of negotiations. If a nation-state demonstrates that it prefers to incur a costly outcome, provided it hurts the other side even more, it can affect the outcomes of those future conflicts. Furthermore, reputations are interdependent; one state's boost in reputation can come at the expense of the other. In a crisis, whichever side backs down is perceived to have lost credibility. At the same time, the side that triumphed will find its credibility enhanced. Therefore, leaders must be concerned that backing down in the present will raise the other state's expectation of success in later episodes. States will care about the distribution of payoffs in the present because it will alter their perceived reputations in the other state. Reputation becomes more important as the expectation of repeated interactions, and the shadow of the future, increases. If players anticipate that a game will be played often, they will be more willing to sacrifice current gains for a larger payoff in the future. Countries will always have some incentive to be concerned about their reputation. Nevertheless, the incentive will be much stronger when they expect a greater number of potential conflicts. States have both tangible and abstract reasons to factor in conflict expectations when crafting their existing strategies. Casual empiricism suggests that the expectation of future conflict varies with the dyad. For example, during the Cold War, the United States clearly expected more zero-sum disputes with the Soviet Union than with West Germany. At present, India can anticipate a greater number of disagreements with Pakistan than with Bangladesh. For the foreseeable future, the United States is more concerned with a possible North Korean nuclear program than a probable Israeli nuclear program. In each of the cases where more future conflict is expected, one or both states perceive a significant threat from the other. In relationships where threat perception is high, so is the expectation of future zero-sum conflicts
714 Conflict Expectations A logical extension of this argument is that the concern for relative gains and reputation is inversely correlated with the degree of alignment.Allies will anticipate few disputes,and care less about relative gains and reputation.Adversaries will anticipate frequent threats of conflict,and therefore care a great deal about relative gains and reputation. It is often unclear what theorists mean when they talk about allies,adversaries, or alignment.My definitions are as follows:states are allies if they share a history of cooperation and mutual trust on security and other issues that is not disrupted by shifts in the international distribution of power.States are adversaries if they share a history of discord and conflict on various issues that is not disrupted by large shifts in the international distribution of power.By allies I do not mean states that temporarily join coalitions to fight a common enemy,such as the United States and the Soviet Union in World War II or the United States and Syria in the Gulf War. By adversaries I do not mean states that have highly public but ephemeral spats over a single issue,such as the dispute between France and the United States over the Uruguay round. In sum:the model will assume that states act as rational,unitary utility-maximiz- ers.National preferences are partially motivated by conflict expectations.There are two effects of this expectation.First,states will have some concern for relative gains, because concessions made in the present can be used against nation-states in the future.Second,countries are concerned that conceding in the present will damage their reputation in future interactions.Both of these concerns vary with the expec- tation of future conflict.Therefore,states will care more about relative gains and reputation effects when their perception of threat is palpable.Because states balance against their threats,conflict expectations will be greatest between adversaries. Disputes will also be anticipated between allies,but to a lesser degree. A Game-Theoretic Model of Economic Coercion There are two ways to formalize the conflict expectations model.The first is to assume states are egotistic utility-maximizers with an expectation of a future dispute where the outcome depends on relative capabilities.The second is to create a reduced-form utility function,incorporating the conflict expectations into the actors'utility functions in the form of a concern for relative gains.As Powell (1994:336)observes,there is no a priori answer as to which approach is better.For this article,the reduced-form model will be used for two reasons.First,this version of the conflict expectations model is easier to present and analyze.3 Second,the focus of this article is the effect of conflict expectations on economic statecraft,not on the causes of those expectations. All games must have players,strategies,and payoffs.The coercion game has two actors:Sender and Target,occasionally labeled S andTfor notational convenience.4 These actors are rational and are assumed to have full information about the possible strategies and payoffs.The game is played only once.5 The structure of the game can be seen in Figure 1.Sender moves first;it can elect to do nothing and end the game at Status Quo Ante(SQA),or it can choose to make 3 Versions of the model that are more explicit about the future implications of present concessions produce results that are substantially similar to the reduced-form game presented here.See Drezner,1998. 4 It could be argued that many high-profile coercion efforts do not have just one sender but many.An examination of these events shows that most sanctions episodes have one dominant sender who persuades and cajoles other states into cooperation.For more on this see Martin,1992. 5A different variant of the game used two stages in an effort to incorporate the shadow of the future.The results are not fundamentally affected
'114 ConJlict Expectations A logical extension of this argument is that the concern for relative gains and reputation is inversely correlated with the degree of alignment. Allies will anticipate few disputes, and care less about relative gains and reputation. Adversaries will anticipate frequent threats of conflict, and therefore care a great deal about relative gains and reputation. It is often unclear what theorists mean when they talk about allies, adversaries, or alignment. My definitions are as follows: states are allies if they share a history of cooperation and mutual trust on security and other issues that is not disrupted by shifts in the international distribution of power. States are adversaries if they share a history of discord and conflict on various issues that is not disrupted by large shifts in the international distribution of power. By allies I do not mean states that temporarily join coalitions to fight a common enemy, such as the United States and the Soviet Union in World War I1 or the United States and Syria in the Gulf War. By adversaries I do not mean states that have highly public but ephemeral spats over a single issue, such as the dispute between France and the United States over the Uruguay round. In sum: the model will assume that states act as rational, unitary utility-maximizers. National preferences are partially motivated by conflict expectations. There are two effects of this expectation. First, states will have some concern for relative gains, because concessions made in the present can be used against nation-states in the future. Second, countries are concerned that conceding in the present will damage their reputation in future interactions. Both of these concerns vary with the expectation of future conflict. Therefore, states will care more about relative gains and reputation effects when their perception of threat is palpable. Because states balance against their threats, conflict expectations will be greatest between adversaries. Disputes will also be anticipated between allies, but to a lesser degree. A Game-Theoretic Model of Economic Coercion There are two ways to formalize the conflict expectations model. The first is $0 assume states are egotistic utility-maximizers with an expectation of a future dispute where the outcome depends on relative capabilities. The second is to create a reduced-form utility function, incorporating the conflict expectations into the actors' utility functions in the form of a concern for relative gains. As Powell (1994:336) observes, there is no a priori answer as to which approach is better. For this article, the reduced-form model will be used for two reasons. First, this version of the conflict expectations model is easier to present and analyze.3 Second, the focus of this article is the effect of conflict expectations on economic statecraft, not on the causes of those expectations. All games must have players, strategies, and payoffs. The coercion game has two actors: Sender and Target, occasionally labeled S and T for notational con~enience.~ These actors are rational and are assumed to have full information about the possible strategies and payoffs. The game is played only onceS5 The structure of the game can be seen in Figure 1.Sender moves first; it can elect to do nothing and end the game at Status Quo Ante (SQA), or it can choose to make 3 Versions of the model that are more explicit about the future implications of present concessions produce results that are substantially similar to the reduced-form game presented here. See Drezner, 1998. 4 It could be argued that many high-profile coercion efforts do not have just one sender but many. An examination of these events shows that most sanctions episodes have one dominant sender who persuades and cajoles other states into cooperation. For more on this see Martin, 1992. 5 A different variant of the game used hvo stages in an effort to incorporate the shadow of the future. The results are not fundamentally affected
DANIEL W.DREZNER 715 a demand(D),attached to a threat of economic coercion.6 It might represent a shift in Target's policy on an international issue,a shift in one of Target's policies toward Sender,or a change in one of its domestic policies that has international ramifica- tions.The demand is an action that hurts the target regime and benefits the sender regime.Note that the sender can calibrate the size of its demand;this decision is endogenous to the game. If Sender chooses to make a demand,then Target must decide between backing down and standing firm.If it chooses to back down,the outcome is Acquiescence (AQ);Target agrees to Sender's demand.Ifit chooses to stand firm,however,Sender has the last move.It could choose to back down and accept the Status Quo(SQ),or carry out the threat of disrupting economic exchange,which produces an outcome of Deadlock(DL).The Deadlock outcome means that Sender disrupts some bilateral economic exchange.This could include suspending aid,imposing trade barriers, freezing financial assets,or reducing investment flows.This action would obviously be painful to Target,and presumably painful to Sender.The opportunity costs of such an action would be the scarce resources needed to compensate for the interrupted exchange.Small opportunity costs imply that the costs of substitution are low (i.e.,low levels of asset-specific investment).The Deadlock outcome is essentially a stalemate;Sender and Target both incur costs,but Target makes no concessions. Figure I also shows the payoffs for each outcome.All of the payoffs are relative to the Status Quo outcome.Sender and Target payoffs in the Status Quo are normalized to zero.Status Quo Ante produces the same outcome,plus a small increase(o)for both actors.The difference between SQ and SQA is that Sender's threat temporarily freezes the bilateral relationship,preventing further increases of trade or investment.Even if the threat is not carried out,the crisis is sufficient to cast a pall on the economic relationship.Acquiescence also delivers the Status Quo payoffs,but there is also a direct transfer;Sender gains and Target loses the demand D.7 Finally,the Deadlock outcome disrupts the bilateral relationship.Both actors suffer costs from the loss ofeconomic exchange.Sender and Target receive penalties of -c(s)and -c(t),respectively. If the sender and target countries cared only about their short-run absolute benefits,backwards induction produces a unique outcome.Sender,choosing be- tween backing down or standing firm,will always back down;it will prefer the benefits of continued economic ties to the costs of economic disruption.That is to say,0>-c(s).Moving backwards,Target,at its decision node,knows that Sender will back down.Therefore,its strategic choice is between acquiescing to Sender's demands,or standing firm and reaching a Status Quo outcome.Since Target must concede in an AQ outcome,it will always prefer the Status Quo payoff and elect to stand firm.Finally,Sender must decide between doing nothing and arriving at Status Quo Ante,or making a threat and reaching the Status Quo.Because Sender always prefers the additional benefit from the SQA payoff,it will do nothing. Therefore,if one assumes that Sender and Target care only about their own payoffs, the unique equilibrium of this game is Status Quo Ante.Because the sender will not prefer to carry out a costly threat,it will opt to do nothing. If one assumes that only immediate absolute gains matter,the game is,frankly, boring.Even if Target suffers much greater costs than Sender,Sender cannot credibly threaten coercion,because it will always prefer to back down and incur fewer 6 Another variant could be that Sender implements sanctions and threatens to keep them in place unless Target meets its demands. 7The results do not change appreciably if Sender and Target place different values on the demand
a demand (D), attached to a threat of economic coercion.6 It might represent a shift in Target's policy on an international issue, a shift in one of Target's policies toward Sender, or a change in one of its domestic policies that has international ramifications. The demand is an action that hurts the target regime and benefits the sender regime. Note that the sender can calibrate the size of its demand; this decision is endogenous to the game. If Sender chooses to make a demand, then Target must decide between backing down and standing firm. If it chooses to back down, the outcome is Acquiescence (AQ); Target agrees to Sender's demand. If it chooses to stand firm, however, Sender has the last move. It could choose to back down and accept the Status Quo (SQ), or carry out the threat of disrupting economic exchange, which produces an outcome of Deadlock (DL). The Deadlock outcome means that Sender disrupts some bilateral economic exchange. This could include suspending aid, imposing trade barriers, freezing financial assets, or reducing investment flows. This action would obviously be painful to Target, and presumably painful to Sender. The opportunity costs of such an action would be the scarce resources needed to compensate for the interrupted exchange. Small opportunity costs imply that the costs of substitution are low (i.e., low levels of asset-specific investment). The Deadlock outcome is essentially a stalemate; Sender and Target both incur costs, but Target makes no concessions. Figure 1 also shows the payoffs for each outcome. All of the payoffs are relative to the Status Quo outcome. Sender and Target payoffs in the Status Quo are normalized to zero. Status Quo Ante produces the same outcome, plus a small increase (a)for both actors. The difference between SQ and SQA is that Sender's threat temporarily freezes the bilateral relationship, preventing further increases of trade or investment. Even if the threat is not carried out, the crisis is sufficient to cast a pall on the economic relationship. Acquiescence also delivers the Status Quo pa offs, but there is also a direct transfer; Sender gains and Target loses the demand D.4' Finally, the Deadlock outcome disrupts the bilateral relationship. Both actors suffer costs from the loss of economic exchange. Sender and Target receive penalties of -c(s) and -c(t), respectively. If the sender and target countries cared only about their short-run absolute benefits, backwards induction produces a unique outcome. Sender, cfioosing between backing down or standing firm, will always back down; it will prefer the benefits of continued economic ties to the costs of economic disruption. That is to say, 0 > -c(s). Moving backwards, Target, at its decision node, knows that Sender will back down. Therefore, its strategic choice is between acquiescing to Sender's demands, or standing firm and reaching a Status Quo outcome. Since Target must concede in an AQ outcome, it will always prefer the Status Quo payoff and elect to stand firm. Finally, Sender must decide between doing nothing and arriving at Status Quo Ante, or making a threat and reaching the Status Quo. Because Sender always prefers the additional benefit from the SQA payoff, it will do nothing. Therefore, if one assumes that Sender and Target care only about their own payoffs, the unique equilibrium of this game is Status Quo Ante. Because the sender will not prefer to carry out a costly threat, it will opt to do nothing. If,one assumes that only immediate absolute gains matter, the game is, frankly, boring. Even if Target suffers much greater costs than Sender, Sender cannot credibly threaten coercion, because it will always prefer to back down and incur fewer 6 Another variant could be that Sender implements sanctions and threatens to keep them in place unless Target meets its demands. 7 The results do not change appreciably if Sender and Target place different values on the demand