Mechanism Design in Hidden Action and Hidden Information: Richness and Pure-VCG (Revised version of F-386)

Abstract

We investigate general mechanism design problems in which agents can take hidden actions that influence state distribution. Their action choices exert significant externality effects on their valuation functions through this influence. We characterize all mechanisms that resolve the hidden action problem (i.e., that induce a targeted action profile). A variety of action choices shrinks the set of mechanisms that induce the targeted action profile, leading to the equivalence properties in the ex-post term with respect to payoffs, payments, and revenues. When the agents can take unilateral deviations to change the state distribution in various directions (i.e., when the action profile satisfies richness), pure-VCG mechanisms-the simplest form of canonical VCG mechanism, which is implemented via open-bid descending procedures that determine the losers’ compensation-are the only mechanisms that induce an efficient action profile. Contrariwise, the popular pivot mechanism, implemented by ascending auctions that determine the winner’s payment, generally fails to induce any efficient action profile.