Introduction

This research paper examines a cornerstone topic in corporate governance: the role of boards of directors in hiring and dismissing CEOs. It addresses three primary questions: how boards evaluate CEO performance, whether their decisions are systematically biased, and what the consequences of these decisions are for the firm. The authors propose that directors rely on benchmarks drawn from their own personal networks of interlocked firms, using the performance of those firms as implicit comparators when assessing a CEO’s ability. This benchmarking can stem from rational learning or from salience-driven heuristics, where interlocked firms provide readily available reference points or mental shortcuts for decision-making.

Sample and variables

The study constructs a comprehensive sample of S&P 1500 firms with data sourced from RiskMetrics, CRSP, Compustat, and IBES. The final firm-level sample comprises 23,350 firm-year observations from 2,587 unique firms and 586 forced turnovers. A turnover is classified as forced following established methodology where departures of CEOs younger than 60 are considered forced if the press does not report reasons such as death or health, or if the departure was announced less than six months before occurring. The key variable of interest, %Worst, is defined as the fraction of independent directors for whom the focal firm is the worst performer among their other directorships. On average, firms are the worst performers for approximately 18 per cent of their independent directors.

Empirical design

To study how director-specific performance evaluation affects the likelihood of CEO dismissals, the authors estimate the likelihood of forced CEO turnover using both logit and linear probability models. The specifications include cubic polynomials of stock returns and ROA to ensure that the %Worst variable does not simply capture non-linear effects of absolute firm performance. The researchers also employ an alternative specification where the unit of observation is a director-firm-year triplet, allowing them to compare dismissal decisions across different firms within a single director’s portfolio in the same year.

Forced CEO turnovers

The main result shows that the likelihood of forced CEO turnover is significantly higher when a firm performs poorly relative to a director’s other interlocked firms. A one-standard deviation increase in %Worst relates to a 0.72-0.79 percentage point increase in the probability of CEO dismissal, representing a 29-32 per cent increase relative to the unconditional probability. This effect is roughly one-third as large as the firm’s own stock returns and 1.4 to 2.5 times larger than the industry performance effect. The findings are robust to various industry definitions and persist even when the firm outperforms its industry peers, as illustrated by the 2011 dismissal of AMD’s CEO.

Economic mechanisms

The study explores whether this effect operates through attention and monitoring effort. Directors are found to be 0.606 percentage points less likely to miss board meetings when the focal firm is their worst-performing directorship, representing a 51 per cent decrease relative to the sample mean. The relation between benchmarking and CEO turnover is concentrated among directors with formal monitoring responsibilities on audit, compensation, or nominating committees. Furthermore, benchmarking effects intensify when performance is harder to assess, such as in firms with high analyst forecast errors or for younger CEOs. This supports a heuristic-based view where directors use mental shortcuts during complex evaluations.

Firm performance following CEO turnovers

Dismissals driven by interlock benchmarking are followed by weaker operating and market performance. A one-standard deviation increase in %Worst is associated with a 0.012 decline in post-turnover ROA, equivalent to roughly 12.8 per cent of the standard deviation of ROA. Market reactions to turnover announcements are also more negative; the cumulative abnormal return (CAR) is –3.4 per cent for firms with high %Worst compared with –1.0 per cent for firms with low %Worst. These results suggest that interlock-benchmarked dismissals are not easily explained as value-enhancing responses to poor performance.

Labour market outcomes

The research examines the careers of dismissed CEOs using a matched-sample difference-in-differences framework. While dismissed CEOs generally experience a 58.7 per cent decline in directorships, this penalty is attenuated by roughly 25 per cent for CEOs fired by boards with high %Worst rankings. This suggests that the director labour market views these dismissals as less informative or negative signals regarding CEO ability when they are driven by interlocked-directorship benchmarks.

Conclusion

The paper provides compelling evidence that outside directors use implicit relative performance benchmarks from their other board seats to shape CEO turnover decisions. This heuristic-driven behaviour has real consequences for governance and shareholder value, often leading to post-dismissal underperformance. The study highlights a novel channel through which interlocking boards affect corporate governance beyond traditional measures of expertise or director busyness.