Optimal Dynamic Relative Performance Evaluation

The theoretical prediction of a negative coefficient on positively correlated peer performance underlies much of the empirical literature on relative performance evaluation. This prediction is commonly obtained from the special case of a single period setting where the variance-covariance matrix of the available performance measures is exogenously restricted to be independent of the evaluee's action. Using the dynamic approach of Holmström and Milgrom (1987), I study the properties of contracts that optimally condition an agent's compensation both on his own performance and on how well he fares relative to a peer (group) when these restrictions are not imposed. I show that if the covariance is non-zero, the optimal contract is linear in own and peer performance as well as the correlation between own and peer performance. In contrast, and in line with the preponderance of the empirical evidence, in its simplest form the model predicts that the expected coefficient on peer performance is exactly zero.