Risk Averse Supervisors and the Efficiency of Collusion

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 collusion based on a trade-off between the risk premia required by (less extreme) risk attitudes and incentives. This allows us to link the efficiency of collusion to the supervisor's risk aversion and to various parameters characterizing the economic environment in which collusion may take place. We are then able to derive implications for the design of organizations, like determining how the number of tasks/agents per supervisor or the level of competition may impact on the cost of collusion, studying the impact of vertical integration on those same costs, or characterizing the role of uncertainty on side-contracting.