Revisiting managerial “style”: The replicability and falsifiability of manager fixed effects for firm policies

Research Summary We attempt to replicate Bertrand and Schoar (2003), who used manager fixed effects to provide evidence for managerial “style” as measured by metrics such as leverage, R&D spending, and Tobin's Q. We find that the statistical and economic significance of manager fixed effects is generally lower in our results than in theirs. This discrepancy motivated us to conduct placebo tests in which we randomize managers' spells at their firms. The results with the randomized data are generally as strong as those with the real data, suggesting (a) that the apparent explanatory power of manager fixed effects may be largely a statistical artifact and (b) that managerial “style” may manifest itself in more complex ways than can be ascertained from examining financial statement metrics. Managerial Summary An influential body of research uses large datasets of firm financial information to study whether individual top managers have a “style” as measured by metrics such as leverage and R&D spending. We attempt to replicate a leading paper in this body of research but obtain quite different, usually weaker results. We also try repeating the analysis using artificial datasets where managers are randomly assigned to firms other than those at which they really worked. These randomized datasets produce results that are generally as strong as those with the real data. The implications are that the statistical methodology used in this body of research may be generating spurious results and that managerial “style” may be too complex to ascertain from examining financial statement metrics.