Print
E-MailDirk Jenter is Associate Professor of Finance at the London School of Economics & Political Science and Katharina Lewellen is Associate Professor of Business Administration at the Tuck School of Business at Dartmouth.
Replacing poorly performing CEOs is one of the key tasks of corporate boards and a potentially important source of CEO incentives. However, prior literature finds that forced CEO departures are rare, and that many CEOs remain in office in spite of poor performance. Our paper revisits this evidence and finds that performance-related turnovers are much more common than previously thought. We diverge from prior studies that classify turnovers into forced and voluntary and instead introduce the concept of performance-induced turnover, which we define as turnover that would not have occurred had performance been “good”.
To operationalize this, we estimate the turnover probability at high levels of performance and assume that any turnovers in excess of this probability are caused by performance being worse. Using this approach, we estimate that close to half of all CEO turnovers are performance induced, many more than the 20% of turnovers the prior literature classifies as forced. The key reason for this discrepancy is simple: turnovers classified as “voluntary” by prior studies are substantially more likely after poor performance, suggesting that many are in fact performance induced. This is especially true for turnovers of CEOs aged 60 or older, which standard algorithms classify as automatically voluntary.
Shifting focus from forced to performance-induced turnover changes our understanding of governance dynamics. For example, prior studies find that forced turnovers decline significantly with CEO tenure and attribute this decline to increasing entrenchment, possibly due to CEOs’ cooption of boards. In contrast, we find that the rate of performance-induced turnover is stable through much of CEO tenure and declines only for the most seasoned CEOs. In a similar vein, we find that performance-induced turnover is high also for founder CEOs, who are usually considered to be entrenched. This suggests that entrenchment is limited as boards are willing to replace even long-tenured and founder CEOs who underperform.
A practical advantage of our approach is that it side-steps the need to categorize turnovers and thus avoids biases due to misclassification. The concept of performance-induced turnover is also interesting in its own right: what matters for future firm performance is that bad firm-CEO matches are dissolved. Whether a departure is initiated by the board or the CEO is of second-order importance. It is worthwhile to note, however, that forced and performance-induced turnover, while conceptually distinct, are correlated: in our sample, turnovers classified algorithmically as “forced” have an 82% probability of being performance induced, compared to only 41% of “voluntary” turnovers.
We present two applications of our framework. As the first one, we test predictions of a standard Bayesian learning model of CEO turnover. In this model, the board learns CEO quality (or the quality of the CEO-firm match) by observing performance signals over time, and it dismisses the CEO once the quality estimate falls below a threshold. Our tests reject the baseline version of this model in which the CEO’s true quality is stable through time. We find that boards rely much more strongly on recent performance signals and ignore signals from the more distant past. In addition, learning about CEO quality appears slow as boards frequently dismiss CEOs even after observing performance for many years. These patterns are consistent with a version of the Bayesian learning model in which CEO quality is subject to frequent and sizable shocks, or alternatively, with models outside of the standard learning framework.
As our second application, we examine performance-induced turnover around corporate events that typically coincide with high levels of CEO turnover: corporate scandals, activist campaigns, and sell-offs by institutional investors. We find that all three events are associated with more forced and more performance-induced turnovers, but no increase in turnovers unrelated to performance. This points towards boards revising downwards their assessments of CEOs around these events, or alternatively, to boards facing increased pressure to fire CEOs who underperform.
In sum, our results underscore the importance of firm performance in CEO turnover and suggest that the threat of a performance-driven turnover plays an important role in incentivizing CEOs.
The complete paper is available for download here.
One Comment
Anyone who has spent considerable time in a large number of board rooms will know that many executive departures that are characterized externally as voluntary are in fact performance-based (including CEOs). This is done for a variety of reasons (face-saving, smooth transition, morale and retention of others, etc.). The approach taken by this study is interesting and gets at a portion of these performance-based departures. However, there are a number of performance-based departures that are not easily measured by externally reported factors such as financial performance and KPIs (e.g., poor leadership styles, disruptive presence, inability to scale further, engagement level, etc.).