Deriving The Fundamentals of a Case of Collusion in Equilibrium.

In three-tier principal-agent models, there is foten concern about the risk of collusion between the supervisor hired by the principal to deal with the agent, and the agent. A collusion-proof contract is often optimal, altough the literature has produced some situations in which collusion is optimal in equilibrium. One of these is in focus in this paper: if the principal does not know whether the supervisor is corrupt or not, collusion in equilibrium arises if the probability that he is corrupt is sufficiently low. This result is quite intuitive and is not contested as such. However, in the literature, it relies on the assumption that the principal cannot offer a self-selection scheme to the supervisor. This paper lifts this assumption and shows that collusion in equilibrium may arise anyway, provided that the honest supervisor needs incentives to reveal his type.