A Complete Characterization of Equilibria in Common Agency Screening Games

We characterize the complete set of equilibrium allocations to intrinsic common agency screening games as the set of solutions to self-generating optimization programs. This analysis is performed both for continuous and discrete two-type models. These programs, in turn, can be thought of as maximization problems faced by a fictional surrogate principal with a simple set of incentive constraints that embed the non-cooperative behavior of principals in the underlying game. For the case of continuous types, we provide a complete characterization of equilibrium outcomes for regular environments by relying on techniques developed elsewhere for aggregate games and mechanism design problems with delegation. Those equilibria may be non-differentiable and/or exhibit discontinuities. Among those allocations, we stress the role the maximal equilibrium exhibits a n-fold distortion due to the principals' non-cooperative behavior. It is the unique equilibrium which is implemented by a tariff satisfying a biconjugacy requirement inherited from duality in convex analysis. This maximal equilibrium may not be the most preferred equilibrium allocation from the principals' point of view. We perform a similar analysis in the case of a discrete two-type model. We select within a large set of equilibria by imposing the same requirement of biconjugacy on equilibrium tariffs. Those outcomes are limits of equilibria exhibiting much bunching in nearby continuous type models which fail to be regular and require the use of ironing procedures.