Evaluating the performance of CEOs is one of the most important tasks of corporate boards of directors. When deciding whether to retain or dismiss CEOs, boards should follow the informativeness principle developed by Holmström (1979) and include all valuable performance signals regarding the quality of the CEOs. Of course, boards should also ignore all uninformative performance signals. For example, CEOs should not be rewarded or punished for, in effect, getting lucky or unlucky.
In a recent working paper, I investigate whether boards violate the informativeness principle in firing CEOs by failing to ignore outcomes that are conditionally uninformative. If such outcomes are taken into account, “outcome bias” is present, which refers to the misattribution of good and bad luck to the actions of an agent (Brownback & Kuhn, 2019).
As uninformative performance outcomes, I use the information on whether a firm’s (relative) stock return in the previous fiscal year was barely positive or barely negative. Because neither firms nor CEOs have precise control over the (relative) stock returns, the outcome of either a barely positive or a barely negative (relative) return can be viewed as random and thus uninformative. In other words, given that a firm’s (relative) return is very close to zero, it is just a matter of good or bad luck whether the (relative) return turns out to be positive or negative.
As a measure of absolute return, I use total shareholder return (TSR), including reinvested dividends, at the end of a fiscal year. Total shareholder returns are salient and usually reported in firms’ annual reports. Thus, boards know whether their firms have increased or decreased in shareholder value during the last year. As a measure of relative TSR, I use the difference between a firm’s TSR and the S&P 500 index return. Because the S&P 500 index return is a salient benchmark, boards are expected to know whether their firm could outperform this index during a given year. Suppose boards are less likely to dismiss CEOs when their firms’ (relative) TSRs are barely positive than when they are barely negative. In that case, boards violate the informativeness principle because CEOs are rewarded (or punished) for good (or bad) luck.
My data consist of 1500 S&P firms and their CEOs between 1993 and 2018. Using the open-source dismissal database constructed by Gentry et al. (2021), I categorize an instance of CEO turnover as forced if the dismissal is related to job performance and exclude all other turnovers. The final sample has approximately 28,000 firm-year observations, and the mean forced CEO turnover rate is 2.9 percent.
Using a regression discontinuity design, my results show that CEOs of firms with barely positive TSRs in the previous fiscal year are approximately 1.9 percentage points less likely to be dismissed than CEOs of firms with barely negative returns. This effect is considerable, as it corresponds to a reduction in the probability of forced CEO turnover of approximately 50 percent. A similar effect can be found for returns relative to the S&P 500 index return: A firm’s board is less likely to dismiss its CEO if the firm barely outperformed the S&P 500 index than if it barely underperformed the S&P 500 index.
Because boards have increasingly employed relative return measures in executive compensation contracts, I expect the relative return to become more salient for CEO turnover decisions in recent years. Indeed, my results show that the tendency of boards to consider uninformative absolute return outcomes decreased between 1993 and 2018. In contrast, the tendency to consider uninformative relative return outcomes has increased during the same period. These findings suggest that, while the focus of boards shifted to relative performance evaluation metrics, boards nevertheless have continued to consider return outcomes that are uninformative about the quality of the CEO.
Overall, my paper demonstrates that boards violate the informativeness principle by dismissing CEOs of firms with barely positive (relative) returns less frequently than CEOs of firms with barely negative (relative) returns. These results have important implications. First, basing CEO turnover decisions partly on factors beyond the control of CEOs is costly, as it exposes them to an additional, unnecessary source of risk. Second, boards should pay greater attention to process-oriented evaluation strategies to improve the quality of CEO turnover decisions. Finally, the existence of outcome bias in corporate boards’ CEO dismissal decisions indicates that this phenomenon might be more widespread in organizations than previously anticipated.
Brownback, A., & Kuhn, M. A. (2019). Understanding outcome bias. Games and Economic Behavior, 117, 342-360.
Gentry, R. J., Harrison, J. S., Quigley, T. J., & Boivie, S. (2021). A database of CEO turnover and dismissal in S&P 1500 firms, 2000 – 2018. Strategic Management Journal, 42(5), 968-991.
Holmström, B. (1979). Moral hazard and observability. The Bell Journal of Economics, 10(1), 74–91.
This post comes to us from Raphael Flepp at the University of Zurich. It is based on his recent article, “Uninformative Performance Signals and Forced CEO Turnover,” available here.