Dynamic Contracts with Random Monitoring

In environments where a principal contracts with many agents who each execute numerous independent tasks, it is often infeasible to evaluate an agent'?s performance on all tasks. Incentives under moral hazard are instead provided by monitoring only a subset of randomly selected tasks. We characterize optimal dynamic contracts implemented with this type of random monitoring technology. We consider a stochastic environment where the agent?'s cost of effort varies over time, and analyze situations where this cost is public or private information. In an optimal contract, the terms the agent is promised when monitoring reveals compliance are as good as when no monitoring is performed, and for some cost types are better. These latter types receive a monitoring reward. We also elicit the dynamics of contract parameters over time. As time passes and the agent becomes richer, the monitoring reward decreases as the threat of forgoing the promised stream of future compensation provides sufficient incentives for compliance.