Evaluating Panel Regression Estimators in Corporate Finance: Evidence from CEO Pay
43 Pages Posted: 22 May 2018
Date Written: April 25, 2018
The use of panel data in corporate finance is ubiquitous to estimate the impact of choices made by managers and/or shareholders on firm value. This study evaluates the properties of five regression estimators when one uses panel data. Specifically, we analyze three existing estimators (namely, the pooled OLS (POLS) estimator, the standard random-effects (RE) estimator, and the standard firm fixed-effects (FE) estimator), and introduce two additional estimators (efficient firm fixed-effects (EFE) estimator, and the efficient correlated random-effects (ECRE) estimator). Theoretically and in simulations we find that the EFE and ECRE estimators are consistent and have the lowest uncertainty, the FE estimator is consistent but has high uncertainty, and the POLS and RE estimators are inconsistent. However, when we use panel data to examine the relationship between CEO pay and firm value, we find the EFE estimator is the most appropriate. For applied research, we suggest a J-test to check whether the assumptions of the ECRE estimator hold. If they do, one can use either the ECRE or the EFE estimators. If they do not, one should use the EFE estimator.
Suggested Citation: Suggested Citation