Relative Performance Evaluation in Organizations with Information Networks

In this paper, I study a principal-multiagent model in which the principal adopts relative performance evaluation (RPE) as the compensation scheme, and agents are connected in an information network. The ability of each agent is private information, but adjacent agents in the information network can observe each other’s ability. Given RPE as the compensation scheme, and assuming that each agent is identical ex ante, the effect of adding links among agents on the effort level is ambiguous, and is determined by the second-(complementarity) and third-order (“risk” attitude) cross-partial derivatives of the compensation function. Complete information networks may not be optimal for maximizing the performance of the organization, whereas empty networks can be optimal in some cases