Strong Collusion-Proof Implementation

The robustly collusion-proof (RCP) mechanism proposed by Che and Kim (2006) enables a principal to attain any payoff that could be achieved without any collusion, even when agents collude. Although the RCP mechanism is robust to various collusive arrangements that agents may devise, it relies on agents not to form certain extreme beliefs following a rejection of a collusive side contract. This paper strengthens the collusion-proofness notion to be robust to such beliefs, as well as any other aspects of coalition formation and its behavior, and proposes a mechanism that implements virtually any non collusive payoff for the principal in this considerably strong collusion-proof sense. The key issue is to guarantee the participation of agents in an RCP mechanism. The proposed mechanism achieves this situation by adding an option that each agent can exercise to protect himself against possible hold-up by his collusive partners.