Earnings Distortion and CEO Compensation

This paper tests multitask agency theory in which the agent has multi-dimensional tasks and both an undistorted and a distorted performance measure are available for contracting. Accounting earnings is modeled as a distorted performance measure and stock returns as an undistorted measure. First, I demonstrate that the marginal products of CEOs' earnings management actions with respect to accounting earnings, in general, are not equal to the marginal products of their earnings management actions with respect to stock returns. In doing so, I provide evidence in support of the hypothesis that accounting earnings is a distorted performance measure and CEOs' earnings management actions are a major source of the distortion. This distortion measure is then used to investigate cross-sectional variation in the relative sensitivity of CEO compensation to accounting earnings. I find that the weight placed on accounting earnings, relative to stock returns, in CEO compensation decreases as earnings distortion increases. Overall, the results provide evidence consistent with multitask agency theory, which predicts that optimal compensation contracts put less weight on the distorted performance measure as its distortion increases in order to improve the efficiency of CEOs' effort allocation across tasks.