This research will study market design by a combination of theoretical and experimental approaches. The theoretical portion of the project aims to integrate multi-item auction theory, particularly the theory of package auctions, with matching theory to develop practical means to solve harder resource allocation problems than is allowed by current mechanisms. In package auctions, participants bid for bundles consisting of multiple items. For example, the city of London uses a package auction to select bus companies to service combinations of bus routes.
Intellectual Merit: Auction theory and matching theory both have a long intellectual history dating to the early 1960s. Seminal contributions by William Vickrey in auction theory and by Gale and Shapley in matching theory laid much of the foundations for current thinking. The Gale-Shapley algorithm has served as a basis for several high-profile applications in recent years, including the match of doctors to hospital residency programs and the assignment of children to schools in New York and Boston, but the generalized Vickrey auction for heterogeneous goods appears never to have been applied. Its failure arises from certain limitations of the auction, especially its potential to yield very low revenues and its manipulability not to reported values but to shill bidders, coalitions, and sellers who may have an incentive disqualify some buyers. The theory will develop alternative designs that maintain as many of the desirable features of Vickrey auctions as possible while ensuring adequate revenues and immunity to the other identified manipulations. The experiments will test the practicality of related dynamic mechanisms, which are reported to have been used with great success in certain laboratory settings but which have not been systematically analyzed.
Broader Impacts: Auctions with package bidding have the potential to result in dramatic improvements in resource allocations and increases in revenues compared to traditional designs in many high-value settings. In the context of asset sales, package auctions have the advantage of attracting bids by bidders interested in either large or small packages, which is an especially important advantage when each set of bidders is small and neither alone would be enough to generate sufficiently competitive prices. In these settings, theory and experiment both indicate that these auctions can achieve efficiency when more traditional kinds of auctions cannot.
