¹ Fearon, James, “Rationalist Explanations for War,” InternationalOrganization 49 (Summer 1995)Google Scholar; Powell, Robert, In the Shadow of Power: States and Strategies in International Politics (Princeton: Princeton University Press, 1999)Google Scholar.

² For the seminal study on information transmission, see Crawford, Vincent P. and Sobel, Joel, “Strategic Information Transmission,” Econometrica 50 (November 1982)CrossRefGoogle Scholar; for an example that parallels mediation, see Gibbons, Robert, “Learning in Equilibrium Models of Arbitration,” American Economic Review 78 (December 1988)Google Scholar.

³ Kydd, Andrew, “Which Side Are You On? Bias, Credibility, and Mediation,” AmericanJournalof Political Science 47 (October 2003)Google Scholar.

⁴ Walt, Stephen M., “Rigor or Rigor Mortis? Rational Choice and Security Studies,” International Security 23 (Spring 1999)CrossRefGoogle Scholar.

⁵ Fearon (fn. 1) and Powell (fn. 1) identify commitment problems along with asymmetric information as a root cause of war. Walter empirically evaluates the commitment problem argument in the context of conflict management. See Walter, Barbara, “The Critical Barrier to Civil War Settlement,” International Organization 51 (Summer 1997)CrossRefGoogle Scholar; and idem, Committing to Peace: The Successful Settlement of Civil Wars (Princeton: Princeton University Press, 2002)Google Scholar. For a comprehensive treatment of commitment problems, see Powell, Robert, “War as a Commitment Problem,” International Organization 60 (Winter 2006)CrossRefGoogle Scholar.

⁶ Fearon (fn. 1); Powell (fn. 1). Uncertainty about the other party's resolve stems from private information about a number of important issues, including their war-fighting capabilities, their attitudes toward risk, the degree to which they value the issue in question, their patience, and their willingness to incur various costs. This study assumes that uncertainty is the result of private information concerning the disputants' costs for fighting, which is also a measure of resolve.

⁷ Bercovitch, Jacob, “International Mediation: A Study of Incidence, Strategies, and Conditions of Successful Outcomes,” Cooperation and Conflict 21 (September 1986)CrossRefGoogle Scholar; Bercovitch, Jacob, Anagnoson, Theodore, and Wille, Donnette L., “Some Conceptual Issues and Empirical Trends in the Study of Successful Mediation in International Relations” Journal of Peace Research 28 (February 1991)CrossRefGoogle Scholar; Touval, Saadia and Zartman, I. William, eds., International Mediation in Theory and Practice (Boulder, Colo.: Westview Press, 1985)Google Scholar; Zartman, I. William, Ripefor Resolution: Conflict and Intervention in Africa, 2nd ed. (New York: Oxford University Press, 1989)Google Scholar.

⁸ Dixon, William J., “Third-Party Techniques for Preventing Conflict Escalation and Promoting Peaceful Settlements,” International Organization 50 (Autumn 1996)CrossRefGoogle Scholar; Regan, Patrick M. and Stam, Allan C., “In the Nick of Time: Conflict Management, Mediation Timing, and the Duration of Interstate Disputes,” International Studies Quarterly 44 (June 2000)CrossRefGoogle Scholar. Cf. Beardesley, Kyle C., Quinn, David M., Biswas, Bidisha, and Wilkenfield, Jonathan, “Mediation Style and Crisis Outcomes,” Journal of Conflict Resolution 50 (February 2006)Google Scholar. Beardesley et al. represent one of the few efforts to develop and specify more carefully the theories from which mediation hypotheses are derived.

⁹ Regan, Patrick M., “Choosing to Intervene: Outside Interventions in Internal Conflicts,” Journal of Politics 60 (August 1998)CrossRefGoogle Scholar; Greig, Michael J., “Stepping into the Fray: When Do Mediators Mediate?” American Journal of Political Science 49 (April 2004)Google Scholar; Terris, Lesley G. and Maoz, Zeev, “Rational Mediation: A Theory and a Test,” Journal of Peace Research 42 (September 2005)CrossRefGoogle Scholar.

¹⁰ Princen, Thomas, Intermediaries in International Conflict (Princeton: Princeton University Press, 1992), 26Google Scholar.

¹¹ Ibid., 26. Bercovitch also comes to the conclusion that mediators may help disputants “achieve some convergence of expectations by reducing distortion, ignorance, misperception, or unrealistic perceptions”; see Bercovitch (fn. 8), 98. For the seminal work on misperception and war, see Jervis, Robert, Perception and Misperception in International Politics (Princeton: Princeton University Press, 1976)Google Scholar. Note that other parts of Princen's analysis (fn. 10) do fall squarely under a rationalist approach. For example, see the discussion of bargaining ranges and the use of mediator leverage (chap. 3). There is also an interesting discussion of how information can be “pooled” to create a bargaining chip that mediators may use as leverage (pp. 41–43).

¹² Experimental economists have also explored how self-serving biases may prevent information from being processed in a way that allows beliefs to update in a manner consistent with Bayesian updating. For one example, see Babcock, Linda and Loewenstein, George, “Explaining Bargaining Impasse: The Role of Self-Serving Biases,” Journal of Economic Perspectives 11 (Winter 1997)CrossRefGoogle Scholar.

¹³ Kydd (fn. 3), 598, emphasis added. Cf. Andrew Kydd, “When Can Mediators Build Trust?” American Political Science Review (forthcoming, August 2006). Here Kydd argues that mediator reputation and repeated play may provide alternative explanations for why impartial mediators may prove effective.

¹⁴ This model builds on Fearon (fn. 1) and an earlier, draft version of Kydd (fn. 3).

¹⁵ This assumption is consistent with Fearon (fn. 1) and Powell (fn. 1). The key issue under examination here is how mediation impacts crisis bargaining in the shadow of war. Thus, it makes sense to bracket the process of war analytically. Some of the insights developed in this analysis may nevertheless extend to bargaining during war, but this issue would benefit from additional analytic and empirical treatment. For a formal model that examines two players that bargain while fighting, see Powell, Robert, “Bargaining and Fighting while Learning,” American Journal of Political Science 48 (April 2004)CrossRefGoogle Scholar. See also Slantchev, Branislav L., “The Power to Hurt: Costly Conflict with Completely Informed States,” American Political Science Review 47 (March 2003)Google Scholar.

¹⁶ This is consistent with a wide range of models in international relations. See, for example, Fearon (fn. 1) and Powell (fn. 1).

¹⁷ A high-resolve defender has low costs for fighting. If bargaining breaks down, it is less costly for this type of player to employ force. Conversely, a defender with high resolve has low costs because military force is less costly to employ. We therefore know that d ^L > d ^H.

¹⁸ For example, a demand where x > x ^L will be rejected by all adversaries. A demand where x < x ^H will also not occur because it is unnecessarily low—even a high-resolve player would accept x ^H. Intermediate range offers (x ^*H < x < x ^*L) also do not make sense. An offer in this range will be rejected by an adversary with high resolve, but it will exceed the minimum demand necessary for appeasing an adversary with low resolve. Thus, Player C will want to demand only x ^H or x ^L.

¹⁹ This decision is in keeping with Kydd (fn. 3) and other previous efforts at producing a formal model of the effects of information sharing. For other models that treat the root source of information as exogenous to the model, see Crawford and Sobel (fn. 2); and Calvert, Randall L., “The Value of Biased Information: A Rational Choice Model of Political Advice,” Journal of Politics 47 (May 1985)CrossRefGoogle Scholar.

²⁰ The situation is quite different for more intrusive forms of third-party intervention. When third parties possess sticks, they can force their way into negotiations. Similarly, if they have carrots, they can buy their way into the negotiations by promising to finance or underwrite the settlement.

²¹ Princen (fn. 10). Economists have also examined how information may be revealed through arbitration. See, for example, Gibbons (fn. 2).

²² For the information to be deemed independent, all that is required is that the third party is receiving the information from some source other than Player C. Otherwise the third party would simply be repeating or playing back the same information to Player C.

²³ See fn. 19.

²⁴ This issue is especially interesting when one considers that the disputants will have strategic reasons to misrepresent information to the third party. A related question concerns the decision to mediate in the first place. For three recent studies that explain the reason why third parties decide to mediate, see Regan (fn. 9); Greig (fn. 9); and Terris and Maoz (fn. 9). Economists have also explored this question in relation to arbitration. For example, see Bloom, David E. and Cavanagh, Christopher L., “Negotiator Behavior under Arbitration,” American Economic Review 77 (May 1987)Google Scholar.

²⁵ To solve the model, these dimensions will be collapsed to create four ideal types. Two of the ideal types match the strictly neutral and strictly biased third parties examined in Kydd's study. The other two ideal types are new and help explain why impartial mediators may outperform biased ones.

²⁶ The exact point or range at which a third party is impartial is difficult to determine because there are a number of different ways that impartiality can be operationalized. Impartiality can mean (1) that β is exactly in the middle of the issue space (β =.5); (2) that it is somewhere between the two different negotiated settlements under consideration ((p + d ^H) < β <(p + d ^L)); or (3) that it is exactly between the two different negotiated settlements under consideration (β=p+((d ^H+d ^L))/2. For the following analysis, definition 2 is adopted because it is reasonable to assume that impartiality, given the structure of this game, means that the third party's ideal point is in between the two negotiated settlements that are under consideration. The problem with definition 1 is that an impartial third party is essentially the same as a third party that is biased in favor of Player D. Here a third party will always attempt to skew the settlement to the left of the bargaining range (toward x ^*H). Rather than capturing an essential feature of the environment, this definition is heavily influenced by the model's take-it-or-leave-it structure and the assumption that Player D comes in only one of two types. Similarly, definition 3 means that impartiality occurs only at one precise point on a continuum. Thus, being just.01 off this mark means that the third party is already starting to exhibit bias. For the following analysis, definition 2 is adopted because it is reasonable to assume that impartiality, given the structure of this model, means that the third party's ideal point is somewhere in between the two negotiated settlements that are under consideration. All points within this range are considered equally impartial.

²⁷ For our purposes, solving for a truth-telling equilibrium is sufficient; Kydd (fn. 3), 599, 608. See also Crawford and Sobel (fn. 2); and Calvert (fn. 19). Future studies of mediation and conflict management may benefit from introducing uncertainty about the third party's preferences and considering equilibriums for lying.

²⁸ This biased third party is essentially the same as the strictly biased third party examined by Kydd. A strictly biased third party has the same ideal point as the challenger. However, note that in this study, the third party does not share the identical utility function as either of the disputants (that is, the third party does not share the disputants' costs for fighting).

²⁹ Note that if λ=1, then the value of β does not matter. This type of third party may in theory be biased toward either player or impartial, but as his dominant concern is avoiding war, the specific preferences over different settlements does not matter.

³⁰ Fearon (fn. 1); Powell (fn. 1); Kydd (fn. 3).

³¹ Kydd (fn. 3) labels it hypothesis 1 (p. 607).

³² Kydd (fn. 3); Princen (fn. 10); see also the studies cited in fn. 8. For an alternative viewpoint, see Kuperman, Alan J., “The Other Lesson of Rwanda: Mediators Sometimes Do More Damage Than Good,” SAIS Review 16 (Winter-Spring 1996)CrossRefGoogle Scholar; idem, The Limits of Humanitarian Intervention: Genocide in Rwanda (Washington, D.C.: Brookings Institution Press, 2001)Google Scholar. For a critique of the application of moral hazard theory to third-party intervention, see Rauchhaus, Robert W., “Conflict Management and the Misapplication of Moral Hazard Theory,” in Crawford, Timothy W. and Kuperman, Alan J., Gambling on Humanitarian Intervention: Moral Hazard, Rebellion and Civil War (New York: Routledge, 2006)Google Scholar.

³³ On the futility of asking whether assumptions are true or false, see Friedman, Milton, “The Methodology of Positive Economics,” in Friedman, Milton, ed., Essays in Positive Economics (Chicago: University of Chicago Press, 1953)Google Scholar; Waltz, Kenneth N., Theory of International Politics (Reading: Addison-Wesley, 1979)Google Scholar.

³⁴ Dixon (fn. 8).

³⁵ Sherman, Frank L., “SHERFACS: A Cross-Paradigm, Hierarchical and Contextually Sensitive Conflict Management Data Set,” International Interactions 20 (January-March 1994)CrossRefGoogle Scholar. The data set is available for download at the University of Michigan's Inter-university Consortium for Political and Social Research; www.icpsr.umich.edu. Since SHERFACS is less well known than the Correlates of War (cow) project or International Crisis Behavior (ICB) data, a few words are in order. The SHERFACS data set subsumes and extends a number of other conflict-management data sets. The most recent study on which it builds is Alker, Hayward R. and Sherman, Frank L., “Collective Security-Seeking Practices since 1945,” in Frei, Daniel, ed., Managing International Crises (Beverly Hills, Calif: Sage, 1982)Google Scholar. Earlier studies along this research line include Haas, Ernst B., Why We Still Need the United Nations (Berkeley: Institute of International Studies, 1986)Google Scholar; Robert Lyle Butterworth, Managing Interstate Conflict, 1975–1979: Data with Synopses (Final Report to the Defense Advanced Research Projects Agency, 1980); Butterworth, Robert Lyle, with Scranton, Margaret E., Managing Interstate Conflict, 1945–74: Data with Synopses (Pittsburgh, Pa.: University Center for International Studies, 1976)Google Scholar; Haas, Ernst B., Butterworth, Robert Lyle, and Nye, Joseph S., Conflict Management by International Organizations (Morristown, N.J.: General Learning Press, 1972)Google Scholar; Nye, Joseph S., Peace in Parts (Boston: Little, Brown and Company, 1968)Google Scholar; Haas, Ernst B., Collective Security and the Future International System, Monograph Series in World Affairs, vol. 5, no. 1 (Denver, Colo.: University of Denver, 1968)Google Scholar. For another study that uses SHERFACS and offers a thorough discussion of its origins and evolution, see Eyerman, Joe and Hart, Robert A. Jr., “An Empirical Test of the Audience Cost Proposition: Democracy Speaks Louder than Words,” Journal of Conflict Resolution 40 (December 1996)CrossRefGoogle Scholar.

³⁶ SHERFACS breaks disputes down into discrete phases and allows analysts to track the varying intensity of disputes over time. Efforts to evaluate conflict management against most other data sets can run into problems because, in addition to the lack of information on competing conflict-managing activities, the data set provides only aggregate dispute characteristics. This makes it difficult to determine whether intervention is causally related to the intensity or duration of a dispute. Dixon (fn. 8) is one of the only studies to date to use SHERFACS' burdensome phase-related structure. To allow for comparison between the studies, the same coding and testing procedures are adopted wherever possible. Because the studies have different purposes, however, it is necessary to make a number of modifications, which are discussed in detail below.

³⁷ Dixon (fn. 8).

³⁸ Skjelsbask, Kjell, “Peaceful Settlement of Disputes by the United Nations and Other Intergovernmental Bodies,” Cooperation and Conflict 21 (September 1986)Google Scholar. Studies define and operationalize mediation in a myriad of ways. To complicate matters, the type of mediation examined in Kydd's study is not the same as the mediation examined in Dixon's study. To avoid confusing the different types of mediation and create consistency across these studies, some of the variables are renamed. Mediation in Dixon's study is relabeled Heavy Mediation. Dixon's communication variable, which is in keeping with what Kydd generally refers to as mediation, is relabeled here as Light Mediation.

³⁹ Dixon (fn. 8), 658; and Skjelsbiek (fn. 38), 148.

⁴⁰ In Dixon (fn. 8), light mediation (communication) consists of the provision of good offices and the enunciation of key issues. Heavy mediation, by contrast, involves three types of third-party activity: arbitration, conciliation, and traditional mediation. Having gained a more nuanced understanding of light mediation as a result of the formal model developed in the previous section, it is necessary to modify the light mediation variable to ensure that it is measuring the correct empirical phenomenon. To that end, n i this article conciliation is removed from heavy mediation and instead treated as a form of light mediation. This is done because conciliation is very limited in its scope. According to Merrills, conciliation includes the exchange and verification of information and the “examination of the dispute and attempts to define the terms of a settlement.” See Merrills, John G., International Dispute Settlement, 2nd ed. (Cambridge, U.K.: Grotius Press, 1991), 39Google Scholar. Heavy mediation, by contrast, includes activities such as “the drafting and promotion of plans for conflict resolution.” See Skjelsbask (fn. 38), 13. Clearly, this is not the point of conciliation, which is limited to the facilitation of communication and other efforts to define the terms of a potential settlement. As we will see, this reclassification does not significantly change the number of observations associated with light or heavy mediation. Moreover, since conciliation was the “lightest” form of heavy mediation in Dixon's study, reclassifying it should increase, rather than decrease, the substantive impact of heavy mediation, given that this lighter form of conflict management is not expected to have much impact. The other five conflict-management variables—coercive intervention, adjudication, observation, humanitarian aid, and public appeals—remain unchanged.

⁴¹ The table is derived from Sherman (fn. 35) and Dixon (fn. 8). Disputes can transition through any number of phases and may do so in any order, with two exceptions: first, if a dispute makes the transition to phase 3, it can never go back to phase 1 or phase 2. This rule makes sense given that once hostilities have taken place, it is impossible to be in phase 1 or phase 2, which, by definition, is a situation where force has never been employed. Similarly, once a dispute transitions to phase 6, it is by definition over and cannot return to another phase. Aside form these two rules, disputes may move between these phase types in any order. Disputes can start with phase 1, phase 2, or phase 3. Those that begin as phase 1 or phase 2 may move between phase 1 and phase 2 one or more times. Disputes may also move between phases 3, 4, and 5 in any order, any number of times. Lastly, disputes may end by reaching phase 6, or they may end in any phase without ever moving to phase 6.

⁴² Bloomfield, Lincoln P. and Leiss, Amelia C., Controlling Small Wars:A Strategyfor the 1970's (New York: Knopf, 1969)Google Scholar.

⁴³ Dixon (fn. 8), 661.

⁴⁴ Hass, Ernst B., “Collective Conflict Management: Evidence for a New World Order?” in Weiss, Thomas G., ed., Collective Security in a Changing World (Boulder, Colo.: Lynne Rienner, 1993)Google Scholar.

⁴⁵ Dixon (fn. 8), 674.

⁴⁶ Ibid., 669–70.

⁴⁷ The coding criterion in SHERFACS is consistent with Haas, Butterworth, and Nye (fn. 35). In many instances, it is not difficult for researchers to determine relative bias because a third party will provide visible economic, political, or military aid to one of the parties. Bias is often also easy to determine because of public statements or declaratory policy. However, in other instances, determining bias is more difficult and may rely on the third party's past relations with the disputants. Following criteria established by Haas, Butterworth, and Nye, SHERFACS ranks the relative bias or impartiality of third parties on a 5-point scale.

⁴⁸ Dixon (fn. 8), 661.

⁴⁹ Kuperman (fn. 32).

⁵⁰ Although it was necessary to include multiple conflict-management techniques for the statistical analysis, this study does offer a theory or formal model that explains or predicts how the other conflict-management techniques work. One can speculate that the relationship between adjudication and war is either spurious or causal. For example, one might conclude that the relationship is merely the result of a spurious correlation stemming from the fact that most attempts to go before the court are often last-ditch efforts when war is inevitable. Thus, it may make sense that these cases are associated with both higher rates of fighting and an increased likelihood of peaceful settlement. However, there is at least one explanation of why adjudication may cause fighting.Unlike mediation or other bargaining processes that are political in nature, legal decisions by international courts may produce decisions that do not reflect political realities. Thus, courts may make rulings that fully legitimate one side's claims and propose an unsustainable settlement. A report issued by the Carnegie Commission on Preventing Deadly Conflict inadvertently provides some support for this hypothesis. See Laudy, Mark, “The Vatican Mediation of the Beagle Channel Dispute: Crisis Intervention and Forum Building,” in Greenberg, Melanie C., Barton, John H., and McGuinness, Margaret E., eds., Words over War: Mediation and Arbitration to Prevent Deadly Conflict (Lanham, Md.: Rowman and Littlefield, 2000)Google Scholar.

⁵¹ With p=.5, the expected payoff for all types of third party equals.5. If the value of p is greater or less than.5, then the expected payoff will depend on the balance of power and the third party's relative bias.

⁵² As stated above, this analysis does not assume that the third party pays the same cost for fighting as the disputants. Third parties, when serving as mediators, do not normally use military force, nor are they the targets of military force. A future study may benefit from relaxing this assumption. If the third party's costs for fighting are included as a variable, then as the costs approach the disputant's costs for fighting (or perhaps even exceed them because of reputation or other costs), the “provocateur effect” will diminish and ultimately collapse.

⁵³ Future analyses might relax the assumption that p=.5 and examine how different distributions of power will influence the behavior of third parties with different degrees of bias.