Economic Coercion 647 for sanctions imposition.Nevertheless,these models agree with the bargaining "folk theorem":in those situations when sanctions are most likely to work,they are least likely to be imposed. Daniel Drezner combines issue indivisibility and the inability to credibly com- mit to explain the imposition of sanctions.6 He presents a complete information model in which the target will make concessions if the sender prefers a deadlock outcome of sanctions imposition to backing down.In the basic version of the model, one should only observe threats or very brief imposition of sanctions.In a refined version of the model,sanctions imposition can be an equilibrium outcome,pro- vided two conditions are met:the demand is indivisible and expectations of future conflict are high.Under conditions of high conflict expectations,both the sender and target fear that any concessions made in the present will leave them in a weak- ened bargaining position in future conflicts,making credible commitments more difficult to achieve.Empirically,the model predicts that when sanctions are actu- ally imposed,the outcome is a sustained deadlock between adversaries.The cases of economic coercion that generate concessions will end at the threat stage and are thus more difficult to observe. Three modeling efforts combine imperfect information and issue indivisibility to explain the rational imposition of sanctions.Alistair Smith,as well as T.Clifton Morgan and Anne Miers,develop one-sided incomplete information models of eco- nomic coercion that lead to similar empirical predictions.17 In both models,the sender does not know whether the target prefers to stand firm or prefers to acqui- esce to the sender's demands rather than suffer the cost of sanctions.The models differ in that Morgan and Miers assume a discrete one-shot game,whereas Smith uses a continuous time approach.The predicted outcomes are similar.For Morgan and Miers,the sender's lack of information about the target's resolve,and the tar- get's incentive to signal a high degree of resolve,can lead to the imposition of sanctions.Morgan and Miers'results predict that the probability of a successful use of economic coercion is greater at the threat stage than at the implementation stage.They conclude:"there are severe selection bias problems with empirical studies that focus only on those cases in which sanctions were applied...sanc- tions strategies may be far more successful than one would conclude from looking only at these cases."18 Smith comes to a similar conclusion-if the target con- cedes,it will do so at the threat stage.He observes:"the length of sanctions will be short.In fact,one may never actually see the sanctions at all.Particularly if it is costly to back down in the face of sanctions,B [the target]may preempt sanc- tions and unilaterally change its policy."These conclusions and empirical pre- dictions are consistent with Drezner's model. 16.Drezner 1999. 17.See Smith 1996;and Morgan and Miers 1999. 18.Morgan and Miers 1999,16. 19.Smith1996,240
Economic Coercion 647 for sanctions imposition. Nevertheless, these models agree with the bargaining "folk theorem": in those situations when sanctions are most likely to work, they are least likely to be imposed. Daniel Drezner combines issue indivisibility and the inability to credibly commit to explain the imposition of sanctions.16 He presents a complete information model in which the target will make concessions if the sender prefers a deadlock outcome of sanctions imposition to backing down. In the basic version of the model, one should only observe threats or very brief imposition of sanctions. In a refined version of the model, sanctions imposition can be an equilibrium outcome, provided two conditions are met: the demand is indivisible and expectations of future conflict are high. Under conditions of high conflict expectations, both the sender and target fear that any concessions made in the present will leave them in a weakened bargaining position in future conflicts, making credible commitments more difficult to achieve. Empirically, the model predicts that when sanctions are actually imposed, the outcome is a sustained deadlock between adversaries. The cases of economic coercion that generate concessions will end at the threat stage and are thus more difficult to observe. Three modeling efforts combine imperfect information and issue indivisibility to explain the rational imposition of sanctions. Alistair Smith, as well as T. Clifton Morgan and Anne Miers, develop one-sided incomplete information models of economic coercion that lead to similar empirical predictions.'' In both models, the sender does not know whether the target prefers to stand firm or prefers to acquiesce to the sender's demands rather than suffer the cost of sanctions. The models differ in that Morgan and Miers assume a discrete one-shot game, whereas Smith uses a continuous time approach. The predicted outcomes are similar. For Morgan and Miers, the sender's lack of information about the target's resolve, and the target's incentive to signal a high degree of resolve, can lead to the imposition of sanctions. Morgan and Miers' results predict that the probability of a successful use of economic coercion is greater at the threat stage than at the implementation stage. They conclude: "there are severe selection bias problems with empirical studies that focus only on those cases in which sanctions were applied . .. sanctions strategies may be far more successful than one would conclude from looking only at these ca~es."'~ Smith comes to a similar conclusion-if the target concedes, it will do so at the threat stage. He observes: "the length of sanctions will be short. In fact, one may never actually see the sanctions at all. Particularly if it is costly to back down in the face of sanctions, B [the target] may preempt sanctions and unilaterally change its policy." l9 These conclusions and empirical predictions are consistent with Drezner's model. 16. Drezner 1999. 17. See Smith 1996; and Morgan and Miers 1999. 18. Morgan and Miers 1999, 16. 19. Smith 1996, 240
648 International Organization Dean Lacy and Emerson Niou create a model with incomplete information on both sides:the sender does not know how resolute the target state is,and the tar- get does not know whether the sender state is resolute.Similar to Smith and Mor- gan and Miers,these authors also assume issue indivisibility.Lacy and Niou's conclusions are identical to these other approaches: Empirical studies that examine cases only in which sanctions were imposed systematically omit a class of cases that represent successful sanctions,though the sanctions were threatened but not imposed.Examining cases of only im- posed sanctions generates a serious selection bias in empirical research on sanctions.20 To reiterate,these models provide different explanations for why one should ob- serve the imposition of sanctions,but provide the same explanation for why most successful uses of economic coercion should end before sanctions are imposed.A target that prefers conceding to deadlock and believes that the sender will carry out its threat will acquiesce before imposition to avoid incurring the cost of sanc- tions.Because all of these models rely on the same game structure,they share a similar empirical prediction:sanctions should yield more concessions at the threat stage than at the implementation stage.The robustness of this prediction to the different assumptions about the distribution of information is quite striking. This prediction also stands in marked contrast to alternative theories of economic sanctions.The assumption that sanctions are generally ineffective has given greater purchase to approaches that stress domestic or symbolic reasons for employing sanctions.Kim Richard Nossal argues that economic statecraft is used according to the logic of appropriateness,as a form of punishment rather than an attempt at coercion.21 Scholars emphasizing domestic politics argue that sanctions are im- posed even if the sanctioning government expects them to fail,to satiate public pressure for action in a crisis or to direct benefits towards rent-seeking coali- tions.22 These theories assume that for sanctions to have any utility to the sender, they must be imposed. Testing for Selection Bias To determine the validity of the selection effect argument,it is necessary to focus on events when sanctions are threatened but not imposed.23 The existence of a 20.Lacy and Niou 2000,18-19. 21.Nossal 1989. 22.See Drury 1998;Kaempfer and Lowenberg 1988 and 1992;and Hiscox 2000. 23.Nooruddin 2002 tests for selection bias in the Hufbauer,Schott,and Elliott data using a cen- sored probit model,but his methodology has two significant flaws.First,his model of sanctions impo- sition does not have a threat stage.Second,his dependent variable conflates sanctions success with the duration of imposition,which leaves the empirical results extremely sensitive to outlier cases in which sanctions have been imposed indefinitely,such as the U.S.embargoes against Cuba or North Korea
648 International Organization Dean Lacy and Emerson Niou create a model with incomplete information on both sides: the sender does not know how resolute the target state is, and the target does not know whether the sender state is resolute. Similar to Smith and Morgan and Miers, these authors also assume issue indivisibility. Lacy and Niou's conclusions are identical to these other approaches: Empirical studies that examine cases only in which sanctions were imposed systematically omit a class of cases that represent successful sanctions, though the sanctions were threatened but not imposed. Examining cases of only imposed sanctions generates a serious selection bias in empirical research on sanctions.20 To reiterate, these models provide different explanations for why one should observe the imposition of sanctions, but provide the same explanation for why most successful uses of economic coercion should end before sanctions are imposed. A target that prefers conceding to deadlock and believes that the sender will carry out its threat will acquiesce before imposition to avoid incurring the cost of sanctions. Because all of these models rely on the same game structure, they share a similar empirical prediction: sanctions should yield more concessions at the threat stage than at the implementation stage. The robustness of this prediction to the different assumptions about the distribution of information is quite striking. This prediction also stands in marked contrast to alternative theories of economic sanctions. The assumption that sanctions are generally ineffective has given greater purchase to approaches that stress domestic or symbolic reasons for employing sanctions. Kim Richard Nossal argues that economic statecraft is used according to the logic of appropriateness, as a form of punishment rather than an attempt at c~ercion.~' Scholars emphasizing domestic politics argue that sanctions are imposed even if the sanctioning government expects them to fail, to satiate public pressure for action in a crisis or to direct benefits towards rent-seeking coalition~.~~ These theories assume that for sanctions to have any utility to the sender, they must be imposed. Testing for Selection Bias To determine the validity of the selection effect argument, it is necessary to focus on events when sanctions are threatened but not imposed.23 The existence of a 20. Lacy and Niou 2000, 18-19. 21. Nossal 1989. 22. See Drury 1998; Kaempfer and Lowenberg 1988 and 1992; and Hiscox 2000. 23. Nooruddin 2002 tests for selection bias in the Hufbauer, Schott, and Elliott data using a censored probit model, but his methodology has two significant flaws. First, his model of sanctions imposition does not have a threat stage. Second, his dependent variable conflates sanctions success with the duration of imposition, which leaves the empirical results extremely sensitive to outlier cases in which sanctions have been imposed indefinitely, such as the U.S. embargoes against Cuba or North Korea