This paper re-examines executive incentive compensation, using a principal-agent model in which the principal is downside risk averse, or prudent (as a number of empirical facts, regulations and scholarly works suggest it should be), instead of risk neutral (as it has been commonly assumed so far in the literature). We f

ind that optimal incentive pay should then be ‘approximately concave’ in performance (in a precise sense), the approximation being closer the more prudent the principal is relative to the agent. This means that an executive should face higher-powered incentives while in the bad states, but be given somewhat weaker incentives when things are going well. Such a statement runs counter to current evidence that incentive compensation packages often put substantial weight on convex devices (such as stock options). We show that this disparity can be justified under certain limited liability and taxation regimes. Some empirical research directions (investigating the composition of incentive pay, notably) and public policy implications are briefly discussed.