Relative performance evaluation in organizations with information networks

In this paper, I study a principal-multiagent model where the principal adopts the relative performance evaluation (RPE) as the compensation strategy, and where 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, the effect of adding links among agents on the effort level is ambiguous, and depends on the third-order derivatives of the reward function