Empirical Analysis of Auctions

由于拍卖是主导现代经济理论的那种博弈论分析最令人信服地捕捉到的市场制度之一，拍卖数据为评估建模方法和评估理论预测的战略行为在实践中的重要性提供了重要机会。拍卖在公共资源的分配中也发挥着越来越重要的作用，因此有效分配和创收的目标为理解拍卖提供了强有力的政策动机。该项目由四个部分组成，通过对拍卖市场的实证研究和开发用于分析拍卖的新统计工具来促进这种理解。第一个部分解决了在卖方使用保留价的拍卖中对需求的估计。通过利用卖方估值异质性导致的保留价格的变化，人们可以始终如一地估计标准拍卖中需求特征的完整分布——这是广泛的政策模拟所需的模型原始。这是在没有参数分布假设的情况下完成的。第二个部分涉及从英国拍卖中的出价推断，其中竞标者可能不对称或估值相关。这也允许投标人（而不是计量经济学家）观察到一个共同因素改变了他们的私人价值——这是应用程序中的常见情况。第三部分开发了首价拍卖中常见值的非参数测试。利用拍卖计量经济学的最新进展，我们展示了如何使用投标人数量的变化来测试获胜者诅咒的存在，从而测试共同的价值观。第四个也是最后一个项目研究了一个在线拍卖市场，其中许多买家和卖家交易同一件物品。它讨论了拍卖规则、卖方声誉和其他拍卖特征对投标人参与和投标的影响。本文记录了套利行为，并研究了欺诈卖方行为的普遍性和预测因素。

Because auctions are among the market institutions most convincingly captured by the kind of game theoretic analysis that dominates modern economic theory, auction data provide important opportunities to evaluate modeling approaches and assess the importance in practice of strategic behavior predicted by theory. Auctions also play an increasingly important role in the allocation of public resources, so goals of efficient allocation and revenue generation provide strong policy motivations for understanding auctions. This project consists of four components that contribute to this understanding through empirical studies of auction markets and development of new statistical tools for analysis of auctions. The first component addresses estimation of demand at auctions in which sellers use reserve prices. By exploiting variation in the reserve prices that results from heterogeneity in seller valuations, one can consistently estimate the full distribution characterizing demand at standard auctions---the model primitive needed for a wide range of policy simulations. This is done without parametric distributional assumptions.The second component addresses inference from bids at English auctions in which bidders may be asymmetric or have correlated valuations. This allows also environments in which bidders, but not the econometrician, observe a common factor shifting their private values---a common situation in applications. The third component develops nonparametric tests for common values at first-price auctions. Exploiting recent advancements in the econometrics of auctions, we show how one can use variation in the number of bidders to test for the presence of the winner's curse and, therefore, common values. The fourth and final project studies an online auction market with many buyers and sellers trading a homogeneous object. It addresses the effects of auction rules, seller reputation and other auction characteristics on bidder participation and bids. Arbitrage behavior is documented, and the prevalence and predictors of fraudulent seller behavior are studied.