Robustly collusion-proof implementation

A contract with multiple agents may be susceptible to collusion. We show that agents' collusion imposes no cost in a large class of circumstances with risk neutral agents, including both uncorrelated and correlated types. In those circumstances, any payoff the principal can attain in the absence of collusion, including the second-best level, can be attained in the presence of collusion in a way robust to many aspects of collusion 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 an ex post incentive compatible and ex post individually rational mechanism. Our collusion-proof implementation also sheds light on the extent to which hierarchical delegation of contracts can optimally respond to collusion.