Reduction of outcome variance: Optimality and incentives

It is well known that monitoring information can be used to reduce moral hazard problems. This paper analyzes a model in which the agent can affect the precision (or quality) of the information observed by the principal. In particular, the agent is assumed to be capable of supplying two types of effort: productive effort and effort that reduces the outcome variance. Variance reduction improves the principal's ability to use the outcome to infer the agent's productive effort. However, if the agent's variance-reducing actions are unobservable, a moral hazard problem related to those actions arises. The principal then faces a trade-off between obtaining monitoring benefits from variance reduction and the costs of motivating the agent to choose the desired level of outcome variance. Moreover, the incentives that the principal sets up to motivate the agent to reduce the variance also affect the agent's incentives to increase the outcome. The paper establishes conditions under which the principal motivates the agent to expend nonzero effort in reducing the variance. The incentives for variance reduction depend on the agent's degree of risk aversion and on the feasibility of nonmonotonic contracts. The results explain a variety of observed contracts such as bonus plans, bonus plans with an upper bound, and stock options.