Introduction

The literature on the principal-agent problem has been concerned with the nature of optimal contracts when there is difficulty in monitoring all aspects of an agent's performance. For the most part, this literature has considered models in which payment based directly on a measure of output provides an enforceable contract. For a recent example, see Grossman and Hart (1983). This requires that information about an agent's output be symmetric so that both parties know what payment is due under the contract and, at least in principle, could get that payment enforced legally if necessary. The output measure does not have to be perfect as long as neither party is in a position to impart a systematic bias to it. However, in many employment situations, information about an agent's output is asymmetric in the sense that the agent cannot verify the principal's observation of it.