Optimal incentives and the time dimension of performance measurement

In many occupations, the consequences of an agent’s actions become known only over time. Should a principal then compensate an agent based on early but noisy information about performance, or later but more accurate information, or both? To answer this question, I study a two-period model in which a principal can pay a risk-averse agent based on both true output, which is realized with delay, and an early signal of output. The agent can borrow against future income, and can commit to a long-term contract. I show that under very general conditions the optimal wage contract depends on the early signal as well as on output even if the signal is merely a noisy version of output; that is, if the signal is uninformative of effort, given output. Specifically, for given output levels, better signals are on average associated with higher wages. Thus, an important characteristic of any performance measure is the time at which it is generated, which expands the range of signals useful for contracting well beyond that implied by the classic Informativeness Principle. The results also shed light on the use of forward-looking performance measures such as stock returns in managerial incentive contracts. JEL codes: D86, D91, M52