Estimation of Bidder Valuations in an FCC Spectrum Auction

The Federal Communications Commission (FCC) uses auctions to allocate radio spectrum frequencies to wireless service providers. The innovation of the auction design is that it o¤ers many heterogeneous licenses simultaneously in one ascending auction. This paper develops an empirical model and procedure to estimate bidder valuations. Given that the complex nature of the auction does not admit formal modeling in a general setting, I do not explore a particular model of equilibrium bidding. Instead, I propose two revealed preference inequalities which should hold in any reasonable model of these auctions. The …rst inequality requires that a bidder never bids on a license at a bidding price above the expected marginal revenue of the license. The second inequality is that if a bidder bids on license A, but not on license B, the expected marginal surplus from winning license A at the bidding price is greater than that of license B. I employ an estimation strategy that generates a map from the observed bidding behavior to a set of distributions of bidder valuations consistent with these behavioral assumptions. A part of the strategy uses an estimator developed by Pakes, Porter, Ho and Ishii (2006). An advantage of this estimator is that it can accommodate a ‡exible speci…cation of unobserved heterogeneity. It allows bidder-license speci…c values to capture private information. I apply the empirical model to an auction held in 2006. Using the estimated distribution of bidder valuations, I estimate bidder markups in order to gauge the level of competition in this auction. The estimated bidder markups are large: the median for local bidders such as rural telephone companies is 26%, whereas it is 31% for global bidders such as nationwide carriers. This suggests that there were large distortionary e¤ects of informational rents in the auction. I am grateful to my advisor Patrick Bajari for his continuous support and guidance. I also wish to thank Amil Petrin, Kyoo Il Kim, Minjung Park, Tom Holmes and Bob Town, as well as participants of the Applied Micro seminar at the University of Minnesota for their helpful comments. Of course, all remaining errors are my own. Email: [email protected]