Estimating the sensitivity of CEO pay to accounting-based performance : gross versus net measures

In the empirical estimation of the relation between CEO pay and both firm and peer performance, researchers typically include conventional accounting-based measures that reflect firm performance net of executive pay expense. We analytically show that when firms evaluate CEO performance relative to peers, the coefficients on such net accounting measures are biased either upwards or downwards in the regression estimate of CEO incentives, relative to the Board of Directors’ choice of CEO incentives. Intuitively, net accounting measures include executive compensation, and the additional covariances bias the regression coefficients. In a cross-sectional estimation of CEO pay-for-performance sensitivity, we document evidence of, on average, an attenuation bias in the regression coefficients on focal-firm net performance and peer net performance. This evidence may help explain puzzling findings in prior work of both low CEO pay-for-performance sensitivity and a lack of widespread use of relative performance evaluation (RPE). Our results imply that in CEO pay regressions, a researcher can remove estimation bias in both CEO pay-for-performance sensitivity and RPE detection by using gross rather than net accounting measures, i.e., by adding back executive pay to net accounting measures for both the focal firm and peers in measuring performance