Collusion-Proof Implementation of Optimal Mechanisms

A contract with multiple agents may be susceptible to collusion. We show that agents’ collusion can be prevented at no cost in a large class of circumstances with risk neutral agents — virtually all circumstances with uncorrelated types and a broad set of circumstances with correlated types. In those circumstances, an optimal outcome in the absence of collusion can be made collusion proof, in a way robust to the precise manner of coalition formation and its behavior. The collusion-proof implementation generalizes to a setting in which only a subset of agents may collude, provided that noncollusive agents’ incentives can be protected via ex post individually rational, dominant strategy implementation.