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...