Full Revenue Extraction for Collusion-Proof Auctions

Abstract

In this thesis we study a problem in auction design on how to counteract bidder collusion while maximizing seller revenue. We provide a mechanism that is strongly collusion-proof when players' values are distributed on Uniform[0, 1], in that it not only guarantees the seller full revenue equal to the expected maximum of bidders' values, but also guarantees that each bidder has nonnegative ex ante utility when bidding truthfully. We then extend our mechanism to the case when players' values come from an arbitrary distribution. The strong coalition-proofness of the mechanism is rooted in: - Bidder payments being not just a function of their values but also with a constant additive term - Bidder payments being reduced by the maximal bid in any subset of players not including that bidder, thus taking into account all possible coalitions that could harm the player.

We show how sacrificing ex interim and ex post individual rationality allows for the seller to extract the full or first best revenue equal to expectation of the maximum of player values.