Robust Models of CEO Turnover: New Evidence on Relative Performance Evaluation

Abstract

We examine the robustness of empirical models of CEO turnover and find that the procedure used to categorize turnover events and the method used to construct performance metrics can have large effects on inferences. Common modeling choices have serious shortcomings including (a) ignoring important information, (b) relying on information that may be systematically biased towards the hypothesis of interest, and (c) weighing extreme observations more heavily than the underlying theory would suggest. We identify a small set of robustness checks and model permutations that should span the reasonable set in many turnover modeling contexts. Using these checks, we show that the surprising recent evidence that CEO turnover depends on industry returns is non-robust and that there is no convincing evidence of an independent role of industry stock returns in predicting CEO turnover. We conclude that efficient learning with full relative performance evaluation is a reasonable description of turnover behavior. JEL Classification: G30; G31; G32