Nonparametric Identification, Estimation, and Testing of Auction Models

9409569 Vuong拍卖是现实市场中许多价格形成的机制。对拍卖的进一步了解可以帮助设计更有效的定价机制，从而改善社会的资源配置。在当今的公共政策领域，拍卖设计问题在金融证券市场(拍卖美国国库券)、环境(拍卖排放限额)和电信(拍卖无线电/个人通信频谱)中具有真正的重要性。尽管拍卖的重要性和关于拍卖的大量理论研究，但由于拍卖模型的结构估计存在严重的计算困难，关于拍卖经济学的严谨实证研究很少。这个项目建立在最近由调查员在获得识别、估计和测试结构方法中的理论拍卖模型的计算容易的方法方面的突破。更具体地说，本项目的目标是解决基于真实数据的拍卖模型的结构估计的以下挑战：(1)以前的结构分析完全依赖于理论拍卖模型的结构元素的某些参数规范。这个项目开发了不需要这种参数信息的结构方法；以及(2)众所周知，基于参数分析的经济政策结论可能不会对结构要素的错误指定具有健壮性。该项目开发了一个两步非参数程序，它不受这个问题的影响，计算非常简单，对于拍卖数据集中典型的样本大小来说，表现得令人满意。

9409569 Vuong Auctions are the mechanisms by which many prices in real world markets are formed. Further understanding of auctions can help design more efficient pricing mechanisms and thereby improve society's allocation of resources. In the public policy arena today, auction design questions are of real importance in financial securities markets (auctions of U.S. Treasury bills), the environment (auctions of emission's allowances), and telecommunications (auctions for radio/personal communications spectrum). Despite the importance of auctions and the huge body of theoretical research on auctions, there has been little rigorous empirical research on the economics of auctions because of severe computational difficulties with the structural estimation of auction models. This project builds on recent breakthroughs by the investigator in obtaining computationally tractable methods of identifying, estimating, and testing theoretical auction models within a structural approach. More specifically, the objective of this project is to address the following challenges to structural estimation of auction models on real data: (1) Previous structural analyses have exclusively relied upon some parametric specification of the structural elements of the theoretical auction model. This project develops structural methods that do not require such parametric information; and (2) It is known that economic policy conclusions based on parametric analyses may not be robust to misspecification of the structural elements. The project develops a two-step nonparametric procedure that is not subject to this problem, is quite simple computationally and behaves satisfactorily for sample sizes typically found in auction data sets.