Monitoring and Pay for Long-Run Performance

We develop a dynamic principal-agent model in which conditioning future pay- for-performance on monitoring signals is a perfect substitute for contemporaneous pay-for-performance in providing incentives. Average pay-for-performance is higher when monitoring is less efficient, because the conditioning is a less effective substitute in that case. Monitoring efficiency has a greater effect on pay-for-performance when negative signals have accumulated. Using changes in the availability of direct fights for board directors to a firm's headquarters as an exogenous shock to monitoring, we present new empirical evidence on the monitoring-compensation linkage that supports the model’s predictions.