Computation and robustness in sealed-bid auctions.

摘要

The study of auctions has been an enduring theme in the economics and game theory literatures for several decades. Since there is a large body of theoretical work on auctions, and since auctions are naturally-occurring games which are used to organize many markets, auctions provide a setting for the analysis of questions of robustness of design and convergence to equilibrium which are central to the modern development of game theory.; Chapter 2 introduces an algorithm for the approximation of Bayes-Nash equilibrium in a broad class of auctions. In addition to its usefulness in computing equilibrium, analysis of the convergence behavior of the algorithm yields information on the strength of the incentives for bidders to undertake the computational time and costs needed to compute optimal bids.; The question of robustness of equilibrium predictions is taken up more directly in Chapters 3 and 4. Chapter 3 characterizes the set of ϵ-equilibria in independent private values auctions. It introduces an algorithm for locating ϵ-equilibria which most extremely deviate from Bayes-Nash equilibrium predictions for revenue and efficiency. The algorithm is suitable for more general auction environments. Chapter 4 examines the set of ϵ-equilibria in kth-price private-values auctions in the special case where bidders employ bidding strategies that are linear in their type. The analysis suggests a laboratory experimental design to evaluate bidder optimization incentives and the structure of bidding errors under different auction rules.