The Canonical Type Space for Interdependent Preferences

这笔奖金将资助两个研究项目。这两个项目都将与普林斯顿大学的Wolfgang Pesendorfer共同进行。行为经济学和实验经济学的最新研究强调了将心理因素纳入标准经济模型的重要性。这些因素与通常包含在经济模型中的因素的区别在于，前者不能作为后果描述的一部分加以考虑。第一个研究项目将创建和分析一个相互依赖偏好的模型，该模型将允许构建个人关心他人个性的经济模型。该模型将(i)促进对现有实验证据的分析，这些证据表明，个人分享奖励的意愿取决于与之分享奖励的人的个性，（ii）提出额外的应用，（iii）为Harsanyi的类型概念提供替代基础。初步结果包括分析利他主义、互惠主义和符合（或违反）社会标准的愿望的影响。已证明的两个定理建立了相互依赖偏好类型规范空间的存在唯一性，并表明该空间中的每种类型都可以被解释为由特征和“个性”组成的一对。未来的研究将集中在(i)开发包含相互依赖偏好和不对称信息的扩展，（ii）理解当前模型与通过信念模拟相互依赖偏好的模型之间的关系，以及（iii）应用。这个项目更广泛的影响将是确定影响行为的新因素，并发展更具描述性的社会互动模型。第二个项目是对选举竞争的分析。标准的唐斯分析得出的政治竞争模型支持这样的结论：即使有相当一部分选民不知情，候选人的政策选择和选举结果也会反映出中间选民的偏好。fromU.S证据。选举往往不能支持这一结论。到目前为止，这项研究已经确定，如果(i)一部分选民既不了解一位候选人的政策选择，也不了解偏好的分布，（ii）选民对一位或另一位候选人的个性有一点偏好，（iii）候选人知道偏好的分布，标准的唐氏结果是可以推翻的。在一个只有一个战略候选人的模型中，已经证明，一个最关心当选的候选人面对一个几乎完全不知情的选民，永远不会选择中间偏好的政策，而是在大多数人喜欢他的个性时当选。相比之下，一个主要关注政策的候选人总是选择中间偏好的政策，但永远不会当选。未来的研究将侧重于将分析扩展到多个“候选人”做出政策决策的模型，并将所得模型应用于拍卖理论和公司治理中的问题。这个项目的智力价值在于建立了党派政治（即候选人未能采用中间偏好政策）、聚合失败（即选民拒绝中间偏好政策）的可能性，并表明即使在代理人完全理性的情况下，看似天真的行为（基于人格偏好的投票）也可以维持在均衡状态。而且不太关心个性。这个项目的更广泛的影响是(i)评估工具投票模式的有效性和局限性，（ii）了解民主制度在信息汇总和确定社会期望的政策结果方面的有效性

This award will fund two research projects. Both of these projects will be undertaken jointly with Wolfgang Pesendorfer of Princeton University.Recent research in behavioral and experimental economics emphasizes the importance of incorporatingpsychological factors into standard economic models. What distinguishes such factors from those normally included in economic models is the fact that the former cannot be accounted for as a part of the description of consequences. The first research project will create and analyze a model of interdependent preferences that will allow for the construction of economic models in which individuals have concerns about the personalities of others. This model will (i) facilitate the analysis of existing experimental evidence showing that individuals' willingness to share rewards depends on the personality of the person with whom the rewards are to be shared, (ii) suggest additional applications, and (iii) provide alternative foundations for Harsanyi's notion of a type. Preliminary results have included analyzing the effects of altruism, reciprocity and a desire to conformto (or violate) social standards. The two theorems already proven establish the existence and uniqueness of the canonical space of interdependent preference types and show that each type in that space can be interpreted as a pair consisting of a characteristic and a "personality". Future research will focus on (i) developing extension that incorporate both interdependent preferences and asymmetric information, (ii) understanding the relationship between the current model and those that model interdependent preferences through beliefs, and (iii) applications. The broader impact of this project will be to identify new factors that influence behavior and to develop more descriptive models of social interaction.The second project is an analysis of electoral competition. Standard Downsian analysis yields models of political competition supporting the conclusion that candidates' policy choices and election outcomes will reflect median-voter preferences even if a significant fraction of the voters are uninformed. Evidence fromU.S. elections often fails to support this conclusion. So far, this research has established that standard Downsian outcomes can be overturned if (i) some fraction of the voters are uninformed of both the policy choice of one candidate and of the distribution of preferences, (ii) voters have some small preference for the personality of one or other candidate, and (iii) the candidate knows the distribution of preferences. In a model with one strategic candidate, it has been shown that a candidate mostly concerned with getting elected facing an almost fully uninformed electorate never chooses the median preferred policy but gets elected whenever the majority prefers his personality. In contrast, a candidate who is mostly concerned with policy, always chooses the median preferred policy but never gets elected. Future research will focus on extending the analysis to models where multiple "candidates" make policy decisions and to apply the resulting models to problems in auction theory and corporate governnance. The intellectual merit of this project is in establishing the possibility of partisan politics (i.e., the candidate's failure to adopt median preferred policies), aggregation failure (i.e., voters' rejection of median preferred policies), and in showing that seemingly naive behavior (voting based on personality preferences) can be sustained in equilibrium even when agents are fully rational. and care little about personalities.The broader impact of this project is to (i) assess the e.ectiveness and limitations of instrumentalvoting models and (ii) understand the e.ectiveness of democratic systems in informationaggregation and in determining socially desirable policy outcomes.1