Learning about the Details in CEO Compensation

Abstract

I document the richness of CEO compensation packages and show that boards learn about the desirability of the many complex package features through observing how these features are associated with firm performance. I first capture the detailed features of plan-based awards for CEOs of the largest U.S. public firms in a vector with more than 1,300 elements. I then demonstrate the complexity of boards' decisions on adding and dropping the detailed features. I hypothesize that boards learn about the efficacy of complex features by observing their correlation with performance—both at their own firms and at other firms. To test these hypotheses, I measure the similarity between any two compensation packages using a metric that assigns a shorter distance to more similar packages. My results support my learning hypotheses: firms that perform well in the current year award similar packages to their CEOs in the following year, whereas firms that perform poorly significantly change their packages in the following year; moreover, firms adjust their own CEO compensation packages to be more similar to that of well-performing firms, and less similar to that of poorly performing firms. These results hold after controlling for the effects from compensation peer firms, compensation-consultant sharing firms, board interlocking firms, and product market peers. I further show that a focal firm experiences better performance when its CEO compensation package becomes more similar to those used by its well-performing compensation benchmark firms. This paper demonstrates the importance of capturing the multi-dimensional details of CEO plan-based awards and studying changes in compensation packages in a holistic manner.