Abstract
This paper presents a model of executive compensation in which the executive is risk-averse and has specific knowledge -knowledge about the optimal actions to take that is costly to transfer to the principal. The model generates predictions that are consistent with the available evidence and provides a rationale for a number of unresolved puzzles in executive compensation. Notably, we find that relative performance evaluation is optimal only if the quality of specific knowledge is low. We also show (1) why some common risk components are not filtered out of executives' pay, (2) why performance is more likely to be evaluated relative to aggregate market movements than relative to industry movements, and (3) why executives with higher perceived abilities are given stronger incentives. Finally, we demonstrate that the relation between risk and incentives may be positive or negative, depending on the quality of the executive's specific knowledge.