Improving Empirical Applications of the Expected Utility Theory of War

研究人员建议对区域和时间差异的三个可能来源进行检查，以使对国际冲突的预期效用预测与实际国际冲突相吻合。对理性选择理论在国际冲突中的复杂一般实施(基于布埃诺·德梅斯基塔和拉尔曼1992年的《国际互动游戏》)的测试表明，该理论的预测在时间和空间上的适合性存在系统性变化。所谓的“战争期望效用理论”的这种突如其来的变化是该理论没有预料到的，它对理性选择模型在国际关系中的应用、对国际冲突的研究以及对国际关系中的定量分析都具有重要的意义。这种变异的系统性及其与关于学习和地区差异的合理假设的明显契合性表明，这种变异不仅仅是测量不当或自然变异的结果。因此，应将这些差异解释为更准确和在经验上适用于国际冲突的关键理性模式的一种手段。任何理性选择模型对跨越时间和空间的相互作用的适合度的变化，至少可以用两种一般方法来解释。首先，一些学者会把注意力集中在博弈结构上，拒绝接受国际互动博弈是跨越时间的所有并元互动的合适理论模型的前提。其次，其他人会争辩说，测量和应用问题是在应用环境中测试和使用预期效用理论的关键，并且当前在关键预期效用概念的测量和操作中存在着特别的问题，这些问题对期望效用理论的测试具有系统性的扭曲效应。这两个可能的问题在这个特定的案例中都有相关性。首席研究员首先计划调查国际互动游戏的游戏结构的关键变化如何影响其预测，以及这些预测与国际冲突行为的契合度。研究人员将重点放在关键博弈假设上，即所有国家都可以选择对彼此使用武力，并开发了一种计算机模拟，旨在按顺序放松博弈假设，并重新评估由此产生的平衡与全球和地区行为模式的契合度。调查员还审查了当前主要预期公用事业概念的运作和实施中涉及的关键衡量问题。特别是，密切关注距离和绝对国家权力的影响，作为国际互动博弈预测在广泛的经验应用中为何不符合预期的关键决定因素。这项工作部分借鉴了在开发Eugene(预期效用和数据生成程序)期间获得的见解，Eugene(预期效用和数据生成程序)是用于生成数据以全面测试国际互动游戏的预测的软件包(Bennett和Stam 1997,1998)。

The researcher proposes an examination of three likely sources of regional and temporal variation in the fit of expected utility predictions of international conflict to actual international conflict. Tests of the msot sophisticated general implementation of rational choice theory to international conflict (based on Bueno de Mesquita and Lalman's 1992 "international interaction game") reveal systematic variations in the fit of the predictions of the theory across time and space. This unexpected variation in the fit of the so-called "expected utility theory of war" is not predicted by the theory, and it has importance implications for the application of rational choice models to international relations, to the study of international conflict, and to quantitative analysis in international relations. The systematic nature of the variation and its apparent fit to plausible hypotheses about learning and regional diffferences suggest that the variation is not solely the result of poor measurement or natural variation. As a result, these variations should be explained as a means of obtaining a more accurate and empirically applicable implementation of a key rational model of international conflict. Variation in the fit of any rational choice model to interactions across time and space could be explained by at least two general approaches. First, some scholars would focus attention on the game structure, refusing to accept the premise that the international interaction game is an appropriate theoretical model of dyadic interaction for all dyads across time. Second, others would argue that measurement and application issues are the keys to testing and using expected utility theory in an applied setting, and that there are particular problems in the current measurement and operationalization of key expected utility concepts that have systematic distorting effects on tests of expected utility theory. Both of these possible problems have relevance in this particular case. The principal investigator first plans to investigate how critical changes in the game structure of the international interaction game affect its predictions and the fit of those predictions to international conflict behavior. The investigator does this by focusing on the key game assumption that all states have the option to use force against one another, and also by developming a computer simulation intended to relax game assumptions sequentially and to reassess the fit of the resulting euilibria to global and regional behavior patterns. The investigator also examines critical measurement issues involved in the current operationalization and implementation of key expected utility concepts. In particular, close attention is given to the effects of distance and absolute state power as key determinants of why the predictions of the international interaction game do not fit as expected in broad empirical application. This work draws in part on insights gained during the development of EUGene (Expected Utility and Data Generation Program), the software package used to generate data to test fully the predictions of the international interaction game (Bennett and Stam 1997, 1998).