Abstract This paper studies the efficiency of collusion between supervisors and supervisees. Building on Tirole (1986)'s results that deterring collusion with infinitely risk averse supervisors is impossible, while it is costless to do so under risk neutrality, we develop here a theory of...