Analysis of Auctions

拍卖一直是信息不对称市场中战略行为的理论和实证研究的一个富有成效的焦点。然而，尽管有丰富且不断增长的文献，拍卖理论忽略了现实世界拍卖市场潜在的重要特征，关于拍卖的基本经验问题仍然没有得到回答。这个项目用新的经验和理论分析解决了其中的一些缺陷。该项目的经验部分侧重于拍卖模型的“结构性”估计，其中一个理论模型被用来将观察到的出价映射为模型基元。结构估计提供了对数据的准确解释，对于政策分析可能特别有价值。然而，存在一个严重的危险：如果支持经验规范的理论模型与实际产生数据的经济过程不太匹配，结果可能会极具误导性。为了部分解决这一问题，提出了两种新的估计方法。第一个解决了拍卖在理论和实践中的一个众所周知的差异：事实上，英国的拍卖通常是以“口头递增投标”的形式(投标人自己喊出价格)而不是大多数理论工作中设想的“按钮拍卖”形式进行的。这种错配对于解读来自英国拍卖的数据具有重要意义。尽管一个解决方案是开发一个更具描述性的理论模型，但试图解释在口头拍卖的动态环境中提出的出价选择面临严峻挑战。本文提出的方法能够对口头递增投标拍卖中投标人估值的分布进行半参数估计，而不需要精确地说明生成观察到的投标的过程。这种方法仅依赖于经济学理论的微弱含义，对于生成数据的“真实模型”细节的变化是稳健的。这项技术应用于美国林业局木材拍卖的数据，用于解决有关储备价格的重要政策问题，并评估其相对于现有方法的方法贡献。第二种方法牺牲了假设的分布规范的灵活性，以便在基础理论模型中允许比以前在经验研究中使用的更一般的信息结构。特别是，该模型允许对象的投标人价值在事前不对称，投标人价值信号的精度不对称，以及私人和公共价值分量的分离信号。这些复杂性很可能存在于许多应用中，并对政策和理论评估具有重要影响。在对联邦近海矿业权拍卖的申请中，对每一种复杂情况的存在进行了测试。采矿量的辅助数据被用来构建经验规范的测试，并调查基本理论模型的灵活性(这里允许)和分布规范的灵活性(其他方法允许)之间的权衡。本项目的第二部分是对潜在买家可以在二级市场交易时最优销售程序的理论研究。大多数以拍卖方式出售的商品都存在这种转售机会，并可能对设计销售机制所面临的限制产生重大影响。尽管如此，几乎所有的拍卖文献都忽略了转售的可能性。研究了转售机会对卖家获取剩余的能力的影响，以及转售对最优销售机制下分配效率的影响。对于转售机会之后的标准拍卖，对保留价的使用、多件物品的捆绑、卖家在拍卖前的信息披露以及买家的信息获取都进行了调查。

Auctions have been a productive focus for theoretical and empirical studies of strategic behavior in markets with asymmetric information. Yet despite a rich and growing literature, auction theory has ignored potentially important features of real world auction markets, and fundamental empirical questions regarding auctions remain unanswered. This project addresses some of these shortcomings with new empirical and theoretical analysis. The empirical component of the project focuses on ``structural'' estimation of auction models, where a theoretical model is used to map observed bids into model primitives. Structural estimation provides a precise interpretation of the data and can be particularly valuable for policy analysis. However, there is a serious danger: if the theoretical model underlying the empirical specification is poorly matched to the economic processes actually generating the data, results can be wildly misleading. Two new estimation approaches are proposed to partially address this problem. The first addresses a well known discrepancy between auctions in theory and practice: the fact that English auctions are most often held in an `oral ascending bid'' format (where bidders call out prices on their own) rather than the ``button auction'' format envisioned in most theoretical work. This mismatch has important implications for the interpretation of data from English auctions. While one solution would be development of a more descriptive theoretical model, an attempt to explain choices of bids called out in the dynamic environment of an oral auction faces severe challenges. The approach proposed here enables semi-parametric estimation of the distributions of bidder valuations at oral ascending bid auctions without requiring a precise specification of the process generating the observed bids. Relying on only weak implications of economic theory, this approach is robust to variations in fine details of the `true model` generating the data. This technique is applied to data from U.S. Forest Service timber auctions, where it is used to address important policy questions regarding reserve prices and to assess its methodological contribution relative to existing approaches. The second approach sacrifices flexibility of the assumed distributional specification in order to allow a more general information structure in the underlying theoretical model than has previously been used in empirical studies. In particular, the model allows ex ante asymmetry in bidders' values for the object, asymmetry in the precision of bidders' signals of values, and separate signals of both private and common value components. These complications are likely to exist in many applications and have important implications for policy and evaluation of theory. Tests are performed for the presence of each of these complications in an application to Federal auctions of offshore mineral rights. Auxiliary data on quantities of minerals extracted are used to construct tests of the empirical specification and to investigate the tradeoff between flexibility of the underlying theoretical model, as permitted here, and flexibility of the distributional specification, as allowed by other approaches. The second component of this project is a theoretical investigation of optimal selling procedures when potential buyers can trade in a secondary market. Such resale opportunities exist for most goods sold by auction and can have significant impacts on the constraints faced in designing a selling mechanism. Nonetheless, nearly all of the auction literature has ignored the possibility of resale. Implications of resale opportunities for the ability of sellers to extract surplus are investigated, as is the impact of resale on the efficiency of allocations under optimal selling mechanisms. For standard auctions followed by a resale opportunity, the use of reserve prices, bundling of multiple objects, pre-auction information revelation by the seller, and information acquisition by buyers are all investigated.