Threat-Based Implementation of Incentive Compatible Mechanisms

We investigate necessary and sufficient conditions for threat-based incentive mechanisms (TBIMs), an extension of the Mirrlees's schemes, to approximately eliminate moral hazard in a principal-agent relationship with both problems of moral hazard and adverse selection. Assuming normal distributions of output, a necessary condition is that the agent's type and effort do not affect the variance of the distribution. Sufficient conditions on the likelihood ratio of the distributions appear to be stronger than those used in the pure adverse-selection models, and have their particular implications.