Dynamics of Performance Measurement Systems

We present a model of how organizations manage performance measures when gaming is revealed over time. The incentive designer does not know when it selects a performance measure whether it will communicate the right behavior. Only over time does the principal find out the agent's responses and then uses this additional information to update and finetune the incentive system. Using data from a government organization, we test the model's main prediction that the correlation between the performance measure and the true goal of the organization should change after the performance measure is included in the incentive system and we find some evidence consistent with this hypothesis.