Estimating State Preferences in International Crises:Promise and Limitations in Fully Structural Estimation1PRELIMINARY AND INCOMPLETE,DO NOT CITE WITHOUT PERMISSION.Jeffrey B. Lewis2Kenneth A. Schultz3Department of Political Science Department of Political ScienceUCLA Stanford UniversityFebruary 11, 20051Earlier versions of this paper was presented the Midwest Political Science Association Meetings,Chicago IL, April 2004, and at the American Political Science Association Annual Meeting, ChicagoIL, Sept. 2004. We gratefully acknowledge the support of the National Science Foundation (NSF-SES-0241647). Thanks to Cesar Zucco for research assistance.2email: [email protected]: [email protected] IntroductionA central problem in the study of international conflict is to understand the factors thatinfluence states’ incentives and behavior in crisis situations involving the threat or use offorce. What political, military, and economic variables influence how states (or, perhapsmore accurately, their leaders) assess the attractiveness of waging war? Under what condi-tions will the targets of threats choose to yield or resist? What factors make it likely thatleaders will implement the threats they make? There is no shortage of hypotheses regardingthese questions in the IR literature. Recent interest, for example, has centered around suchquestions as• whether regime type affects the propensity to win wars (e.g., Lake 1992; Reiter andStam 2002), the costs of war (e.g., Russett 1993; Bueno de Mesquita and Siverson1995), and the costs of backing down from a threat (e.g., Fearon 1994; Schultz 1999,2001b);• how territorial interests influence the stakes and outcomes of international crises (e.g.,Huth 1996; Vasquez 1993; Diehl 1991);• how alliances and strategic interests influence the the willingness to engage in crises orwars (e.g., Gowa 1999);• whether economic interdependence and/or development makes war less attractive (e.g.,Russett and Oneal 2001; Mousseau 2000);• and how historical rivalries influence states’ willingness to wage war or make concessions(e.g., Crescenzi and Enterline 2001; Crescenzi, Kathman, and Long 2004).A challenge that arises in testing such hypotheses is that the outcomes of internationalcrises are the product of interdependent choices: what a state decides to do depends not onlyon its own preferences but also on its beliefs and expectations about the preferences of theopponent. Moreover, these decisions are typically made under a great deal of uncertainty.In such settings, it is generally not straightforward to infer actors’ underlying preferencesEstimating State Preferences 2from their observable choices (Frieden 1999). Thus, understanding how factors like regimetype, power, and alliances affect states’ preferences over war and peace calls for an empiricaltechnique that is sensitive to dynamics of strategic interaction (Signorino 1999, 2003; Smith1999).In this paper, we provide some preliminary results using such a technique. The methodthat we employ involves making inferences from revealed preferences through the use of afully structural strategic model. That is, we use observational data on state choices in crisesto infer the distribution of preferences over the possible outcomes and the effects of covariateson those preferences. The empirical estimator is derived directly from the equilibrium of anextensive form crisis bargaining game with incomplete information. Many of the details andproperties of this estimator were discussed in an earlier paper (Lewis and Schultz 2003). Here,we apply this technique to a new data set on international crises in the period 1919-1939.The data combine events from two widely used data sets—the International Crisis Behavior(ICB) and Militarized Interstate Dispute (MID) projects. Using the cases identified in thesesources, we coded the outcome of each incident according to the four possible outcomes ofthe theoretical model.While the small size of the current data set prevents an extensive test of all the relevanthypotheses in the IR literature, here we provide some preliminary results on the effects ofseveral variables of interest. Most of the results we obtain are sensible, suggesting thatthere is promise in pursuing this line of inquiry. Of particular note are some interestingeffects associated with regime type; in particular, democracy seems to both increase thecosts associated with backing down from a threat and to decrease the attractiveness of war.Such a multi-faceted effect is not only important in its own right, but it demonstrates theability of our method to uncover complex relationships that might be hard to tease out usingstandard methods.That said, we pursue this research with a certain amount of caution and skepticism. As wehope to make clear later, fully structural empirical models have limitations that researcherswill have to appreciate before embracing them. In particular, this approach requires theanalyst to put enormous faith in the underlying theoretical model, a leap that many maybe uncomfortable taking. Nonetheless, the field has shown recent interest in this kind ofEstimating State Preferences 3A Crisis Bargaining Game½¼¾»AChallenge Not challenge########cccccccc}½¼¾»SQ(SQA, SQB)½¼¾»BNot resist Resist########cccccccc}½¼¾»ACQ(ACQA, ACQB)½¼¾»AFight Not fight########cccccccc}½¼¾»SF(SFA, SFB)}½¼¾»BD(BDA, BDB)Figure 1: Simple crisis bargaining game.analysis (esp. Signorino 1999, 2003; Smith 1999), and it is worthwhile to push this frontierforward to see where it leads us.This paper proceeds as follows. The first section sketches the method we employ byshowing how we derived an empirical estimator from an extensive form crisis bargaininggame. Section 2 then discusses some the of the pros and cons of using this approach toestimating state preferences in international crises. Section 3 discusses the data used in thisanalysis. Section 4 then presents our preliminary estimation results.Estimating State Preferences 42 From Theoretical Model to Empirical Estimator2.1 Theoretical ModelThe theoretical model underlying this analysis is a simple crisis bargaining game with in-complete information. Figure 1 depicts the extensive form.Two states, A and B, are assumed to have some dispute. The game begins with a decisionby A whether or not to challenge B by issuing a demand to alter the status quo in A’s favor.A challenge is assumed to involve an explicit threat to use