Stopping agents from ``cheating'' (Q1117129)

Optimal incentive compatible mechanisms may have multiple equilibria. \textit{J. S. Demski} and \textit{D. Sappington} [ibid. 33, 152-171 (1984; Zbl 0541.90001)] have shown in a principal/agent model where two agents are interrelated through their production function that there exists an equilibrium, i.e. pair of optimal incentive contracts exists which forms a Bayes-Nash equilibrium. However, there is another pair of equilibrium strategies whose outcomes are Pareto-superior from the viewpoint of the agents. The latter equilibrium is such that the principal is strictly worse off. In order to avoid this state, the authors propose to strengthen the incentive constraints of one agent to make his output choices a dominant strategy for him. This method is, of course, costly for the principal. The present paper proposes a costless method of stopping the agents from realizing the equilibrium which is undesirable for the principal. The authors design a mechanism which is such that the principal offers one agent a range of extra output options. By choosing one of these options the agent can send signals to the principal. The signals inform the principal about the extent to which the other agent cheats from the perspective of the first agent. The proposed mechanism is displayed as a payment scheme in matrix form. These payments have the property that a unique Bayes-Nash equilibrium is implemented. The proof of this proposition is done in nine steps.