Biased performance evaluation in a model of career concerns: incentives versus ex-post optimality

Abstract

I study a career concerns model in which the principal receives information about the agent’s performance from an intermediary (evaluator). I show that, in general, a biased evaluator is ex-ante optimal for the principal. The ex-ante optimal bias solves the tradeoff between ex-post optimality of the principal’s decisions about the agent and incentive provision. It is “anti-agent” (“pro-agent”) when the agent has an a priori high value (low value) for the principal. It increases with the strength of the agent’s career concerns and decreases with the degree of uncertainty about his ability. Delegating decisions to the evaluator dominates communication with her when ex-ante optimality calls for a sufficiently large bias.