Games and Meta-Games: Pricing Rules for Combinatorial Mechanisms

Abstract: In settings where full incentive-compatibility is not available, such as core-constraint combinatorial auctions and budget-balanced combinatorial exchanges, we may wish to design mechanisms that are as incentive-compatible as possible. This paper offers a new characterization of approximate incentive-compatibility by casting the pricing problem as a meta-game between the center and the participating agents. Through a suitable set of simplifications, we describe the equilibrium of this game as a variational problem. We use this to characterize the space of optimal prices, enabling closed-form solutions in restricted cases, and numerically-determined prices in the general case. We offer theory motivating this approach, and numerical experiments showing its application.