Private Information and Performance Measurement Bias with Multiple Agents

This paper studies the effect of performance measurement error and bias on the principal's choice of whether to appoint a supervisor who signals private, pre-decision, productivity information to a subordinate. Without a supervisor, both agents are privately informed and relative performance evaluation is optimal, and an increase in the conservative bias increases information rents only if the bias is insignificant. With a supervisor, the subordinate learns about productivity by observing the supervisor's effort, reducing the supervisor's incentives to shirk. Thus, joint performance evaluation can be optimal in compensating the supervisor, and an increase in the conservative bias reduces the supervisor's rents. The principal prefers a supervisor if the decision-making aspect of the agents' information is insignificant, implying that effort incentives are paramount. Further, with a less accurate performance measure, the principal's preference for a supervisor increases but only if the level of the conservative bias is small