Journal of Financial Economics 163 (2025) 103971

Contents lists available at ScienceDirect

Journal of Financial Economics

journal homepage: www.elsevier.com/locate/finec

CEO turnover and director reputation✩

Felix von Meyerinck a , Jonas Romer b , Markus Schmid c ,∗
a
University of Zurich, Department of Finance, Plattenstrasse 14, 8032 Zurich, Switzerland
b
University of St. Gallen, School of Finance, Unterer Graben 21, CH-9000 St. Gallen, Switzerland
c
University of St. Gallen, School of Finance, Swiss Finance Institute (SFI), and European Corporate Governance Institute (ECGI), Unterer Graben 21, CH-9000 St.
Gallen, Switzerland

ARTICLE INFO ABSTRACT

Dataset link: Replication kit (Code and pseudo d This paper analyzes the reputational effects of forced CEO turnovers on outside directors. We find that directors
ata) interlocked to a forced CEO turnover experience large and persistent increases in withheld votes at subsequent
re-elections relative to non-turnover-interlocked directors. Directors are not penalized for an involvement in
JEL classification:
G32 a turnover per se but for forced CEO turnovers that are related to governance failures by the board. Our
G34 results challenge the widespread view that forcing out a CEO can generally be understood as a sign of a
well-functioning corporate governance.
Keywords:
CEO turnover
Director elections
Director reputation
CEO succession
Shareholder voting

✩ Toni Whited was the editor for this article. We thank Toni Whited, an anonymous referee, Reena Aggarwal (AFA discussant), Yakov Amihud, Katsiaryna
Bardos (EFA discussant), Patrick Bolton, Pang-Li Chen (SWFA discussant), Lauren Cohen, Cláudia Custódio, Cesare Fracassi, Nickolay Gantchev, Marc Goergen
(WCGI discussant), Michael Halling, Harald Hau, Sheng Huang (Paris December Finance Meeting discussant), Sofia Johan, Oğ Karakaş (CICF discussant), Jonathan
Karpoff, Elisabeth Kempf, Peter Limbach, Alberto Manconi, Kasper Meisner Nielsen, Junnatun Naym (FMA discussant), Per Östberg, Yihui Pan (MFA discussant),
Vincenzo Pezone (ECGC discussant), Vesa Pursiainen, Luc Renneboog, Zacharias Sautner, Laura Starks, Roberto Tubaldi (SFI Research Days discussant), Michael
Ungeheuer, Milos Vulanovic (CFD discussant), Alexander Wagner, Tracy Wang (FIRS discussant), Michael Wittry, Ziwei Zhao, seminar participants at the University
of St.Gallen, the University of Konstanz, IESEG School of Management, the University of Tübingen, the University of Basel, the University of Liechtenstein, and
conference participants at the 2021 Midwest Finance Association (MFA) meeting, 2021 Southwestern Finance Association (SWFA) meeting, 2021 Swiss Society
for Financial Market Research (SGF) conference, 2021 Eastern Finance Association (EFA) meeting, the 2021 SFI Research Days, the 2021 International Young
Finance Scholars’ (IYFS) conference, the 2021 Financial Management Association (FMA) meeting, the 2022 American Finance Association (AFA) meeting, the 2022
China International Conference in Finance (CICF), the 2022 Corporate Finance Day (CFD), 2022 Workshop on Corporate Governance and Investment (WCGI),
the 2022 Paris December Finance Meetings, the 2023 Financial Intermediation Research Society (FIRS) conference, and the 2023 Erasmus Corporate Governance
Conference (ECGC) for helpful comments and suggestions. Part of the paper was written while Felix von Meyerinck was visiting Tilburg University and while
Jonas Romer was visiting the University of Texas at Austin.
∗ Corresponding author.
E-mail addresses: felix.vonmeyerinck@df.uzh.ch (F. von Meyerinck), jonas.romer@unisg.ch (J. Romer), markus.schmid@unisg.ch (M. Schmid).

https://doi.org/10.1016/j.jfineco.2024.103971
Received 2 December 2023; Received in revised form 5 November 2024; Accepted 5 November 2024
Available online 23 November 2024
0304-405X/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

1. Introduction actions and that directors respond to changes in withheld votes (e.g.,
Ertimur et al., 2012; Brochet and Srinivasan, 2014; Aggarwal et al.,
Monitoring and, if necessary, firing a CEO is one of the most 2019; Liu et al., 2020; Erel et al., 2021).3
important decisions made by corporate boards (e.g., Fama, 1980; Fama To implement our identification strategy, we estimate general-
and Jensen, 1983; Jenter and Kanaan, 2015; Jenter and Lewellen, ized difference-in-differences regressions on a sample of turnover-
2021). However, evidence on the reputational consequences of forced interlocked director re-elections and a control sample of non-turnover-
CEO turnovers on involved directors is scarce. Moreover, the existing interlocked director re-elections. Specifically, we regress the change in
literature has conflicting views on the direction of such reputational the share of withheld votes between two consecutive director elections
consequences, which reflect conflicting views on the corporate gover- on a treatment dummy variable indicating whether a director was
nance signal transmitted by forced CEO turnovers. On the one hand, involved in a forced CEO turnover at another firm between the two
firing a poorly performing CEO may be a sign of effective monitoring by election dates as well as firm-level and director-level control variables.
the board and thus indicate a well-functioning corporate governance. We first-difference outcome and control variables at the director-firm
This view represents the predominant position taken in the extant level. This approach ensures that time-invariant director characteristics,
empirical finance literature.1 Under this view, directors who force out a such as talent, and firm characteristics, such as corporate culture, do
CEO are expected to gain reputation. On the other hand, the need to fire not influence our estimates.4 Moreover, we include industry-year fixed
the CEO may indicate a governance failure (e.g., Jensen, 1993; Marcel effects. These fixed effects control for time trends, industry-specific
et al., 2017). Indeed, a better board may have replaced the CEO before trends, and unobserved time-varying industry shocks in withheld votes,
negative performance consequences became observable or by ensuring ensuring that we compare changes in vote outcomes between turnover-
a less disruptive transition to a new CEO. This view is supported by interlocked and non-turnover-interlocked directors within the same
the case-based study of Mace (1971) and theoretical work.2 Under this industry and year. In further tests, we augment our baseline regression
alternative view, directors involved in a forced turnover are expected with either firm or director fixed effects. These fixed effects additionally
to lose reputation. In this paper, we set out to empirically investigate remove time-invariant firm-specific and director-specific effects from
whether directors gain or lose reputation from forcing out a CEO. the director re-election outcomes, allowing us to compare turnover-
Unambiguous evidence on the direction of reputational effects derived interlocked directors with other non-turnover-interlocked directors at
from such an analysis allows us to paint a clearer picture of the the same firm or with the same director absent a forced CEO turnover.
corporate governance signal transmitted by a forced CEO turnover. Our results show that directors involved in a forced CEO turnover
A test of the reputational effect of forced CEO turnovers on in- experience a significant increase in withheld votes at their subsequent
volved directors is subject to two major empirical challenges. First, re-election at interlocked firms compared to directors not interlocked
turnover decisions are endogenous and often related to company per- to a forced turnover in the same industry and year. Director- and
formance (Fee et al., 2013). Second, widely used measures of director company-level control variables and different sets of fixed effects leave
reputation, such as gains and losses of board memberships, are subject our results largely unchanged, suggesting that controlling for various
to endogenous selection by directors (Levit and Malenko, 2016). To time-variant and time-invariant characteristics does not affect our find-
address the first major endogeneity concern, we study directors with
ings. Hence, forced turnovers do not appear to systematically coincide
multiple directorships who force out a CEO at one firm (which we refer
with other changes in director and firm characteristics at turnover-
to as ‘‘turnover firm’’) and observe director-level reputation measures
interlocked firms, providing support for the conjecture that forced
at the other firms on whose boards these directors sit (‘‘interlocked
turnovers represent exogenous shocks on interlocked directors’ vote
firms’’). These interlocked firms, and the directors sitting on their
outcomes. Overall, our finding of an increase in withheld votes asso-
boards, are largely unaffected by the characteristics of the turnover
ciated with involvement in a forced CEO turnover is consistent with
firm, including factors that led to the forced turnover. Hence, this
directors suffering a reputational loss, which challenges the predomi-
setting allows us to isolate the impact of a forced CEO turnover on
nant view that forcing out a CEO is a sign of well-functioning corporate
director reputation from company-level factors at the turnover firm.
governance at the board level.
To address the second major endogeneity concern, we employ the
In economic terms, we find that turnover-interlocked outside di-
change in the percentage of withheld votes (defined as the sum of votes
rectors experience a 1.20 percentage point increase in withheld votes.
withheld and votes against, divided by the total number of votes cast)
While this increase may appear modest, it is important to note that vote
in director re-elections as our primary measure of director reputation.
support for directors is generally high, with the average of withheld
Voting results reflect shareholders’ and not directors’ decisions. Hence,
votes in our sample amounting to 6.1%. Hence, an increase of 1.20
withheld votes constitute a direct measure of shareholder satisfaction
percentage points represents a sizeable increase of 19.6% over the
regarding individual directors that is not subject to endogenous se-
sample mean. Moreover, recent research shows that modest increases in
lection by directors. By using director vote outcomes as a measure
withheld votes often result in negative consequences for directors (e.g.,
of director reputation, we rely on a growing stream of literature that
Aggarwal et al., 2019). To put the economic magnitude of the doc-
shows that investors actively use withheld votes to evaluate directors’
umented effect into perspective, we conduct two additional analyses.
First, we compare the increase in withheld votes resulting from a
forced CEO turnover at an interlocked firm to the increase in withheld
1
See, for example, Weisbach (1988), Farrell and Whidbee (2000), Huson votes resulting from other corporate events at an interlocked firm. We
et al. (2001), Faleye (2007), Guo and Masulis (2015), Kempf et al. (2017),
Dasgupta et al. (2018), Cai et al. (2021), and Jenter and Lewellen (2021).
2
In Dow (2013)’s model, directors choose not to fire a bad CEO because
they do not want to reveal that they made a mistake in hiring her in the 3
For instance, Aggarwal et al. (2019) show that an increase in withheld
first place. Aghamolla and Hashimoto (2021) show that while aggressive votes in uncontested director elections leads to higher director turnover,
boards facilitate truthful communication between the CEO and the board, committee demotions, and reduced opportunities in the director labor market.
4
they tend to dismiss talented managers, resulting in costs to shareholders. The As we discuss below, our setting is akin to a staggered difference-in-
model of Adams and Ferreira (2007) suggests that it may be beneficial for differences setting in levels that includes two-way fixed effects (subject and
shareholders to elect a board that is friendly toward the CEO and does not time) but can easily accommodate repeated treatments (e.g., Heider and
monitor her too closely and/or impose a significant threat of replacement. Ljungqvist, 2015). As we show below, our setting is unlikely to suffer from the
Hence, a forced turnover may signal an aggressive board to shareholders, one ‘‘bad comparison’’ problem recently discussed in the literature (e.g., Callaway
that may eventually turn out to be detrimental to shareholder value. and Sant’Anna, 2021; Sun and Abraham, 2021; Baker et al., 2022).

2
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

consider four alternative events that have been shown to hurt direc- forced CEO turnover, rather than for the turnover. To distinguish be-
tor reputation, including restatements (Srinivasan, 2005), class action tween these two possible interpretations, we conduct a propensity score
lawsuits (Fich and Shivdasani, 2007), poison pill adoptions (Johnson matching analysis. In this analysis, we match firms with a forced CEO
et al., 2023), and bankruptcy filings (Gow et al., 2018). The increase in turnover to firms without a forced turnover but the same propensity to
withheld votes associated with firing a CEO is larger than the increase force out a CEO. We estimate the propensity of a forced CEO turnover
in withheld votes resulting from the other reputational events and is using the model of Peters and Wagner (2014) that includes various
only exceeded by a bankruptcy filing, arguably the most detrimen- company performance metrics and other firm, CEO, governance, and
tal event for corporate shareholders. Still, the increase in withheld industry characteristics. Our results show that turnover-interlocked
votes following a forced CEO turnover amounts to approximately half directors receive significantly more withheld votes than directors in-
the increase resulting from a bankruptcy. Thus, these results further terlocked to matched firms without a forced turnover but otherwise
support our assertion that the documented increase in withheld votes similar characteristics. The economic magnitude of this negative vote
following forced CEO turnovers is sizable. Second, we test whether effect is virtually identical to our baseline estimates, suggesting that
forced turnovers have the potential to result in unusually high levels the performance of the turnover firms does not constitute an omitted
of withheld votes at interlocked firms, defined as at least 15% of variable that drives our results.
votes cast (Bach and Metzger, 2017). We find that the likelihood A related concern is that there is an unknown – and thus omitted
of experiencing very high levels of withheld votes at an interlocked – reason why, in two firms with equally bad performance, one fires
firm increases significantly following a forced CEO turnover, again the CEO and the other does not. This unknown reason, in turn, might
confirming that such turnovers have a sizeable impact on interlocked affect investors’ inference about director ability. For instance, if a man-
directors’ withheld votes. agement failure drives bad performance, this failure may cause both a
We also test whether reputational losses are confined to certain forced CEO turnover and a negative updating about director ability.
types of turnovers or depend on directors’ involvement in the turnover. In contrast, if bad performance is simply due to bad luck, this may
Results from these tests show that directors are not penalized for all neither result in a forced CEO turnover nor convey any signal about
turnovers, but for forced CEO turnovers that appear to be related to director ability. To address this concern, we extend the set of covariates
governance failures by the directors. Specifically, directors are penal- in our propensity score estimation by a newspaper-based sentiment
ized for reactive forced turnovers, firings during the most productive index. As mistakes of the leadership team can be expected to trigger
period of a CEO’s tenure, poor monitoring of the CEO, and the lack of more negative newspaper coverage than bad luck, accounting for news
a successor to the outgoing CEO. In contrast, directors are not penalized sentiment can be expected to produce a sample that is balanced in
for unforced turnovers or for hiring the wrong CEO in the first place. terms of management’s involvement in the negative events that even-
Importantly, we uncover no sub-sample for which we find significant tually lead to bad performance. Our results show that differences in
reputational gains from an involvement in a forced CEO turnover. newspaper sentiment, and thus a potential omitted variable related to
Taken together, these results suggest that a forced CEO turnover is not a management failures prior to the forced CEO turnover, are unlikely to
credible signal of a board’s monitoring ability. In contrast, they suggest drive our results.
that firing the CEO is often indicative of governance failure at the board An alternative interpretation of our results is director distraction.
level. Forced CEO turnovers may demand significant time and effort from
We also investigate which shareholders penalize directors for in- directors involved in the turnover. Turnover-interlocked directors may
volvement in forced CEO turnovers. Systematically penalizing directors thus divert their attention away from the interlocked firms (e.g.,
for their actions across all board mandates requires shareholders to Masulis and Zhang, 2019; Stein and Zhao, 2019), resulting in a negative
observe and evaluate directors’ actions and to observe interlocked di- assessment by shareholders. We conduct two tests to rule out this
rectorships. We hypothesize that institutional investors with significant alternative explanation. First, we make use of sudden deaths of CEOs.
ownership stakes in both the turnover and interlocked firms are most Sudden CEO deaths represent shocks to the time demand of directors
likely to fulfill these requirements. Our results confirm this conjecture: similar to forced CEO departures. However, sudden CEO deaths repre-
We find the negative vote effect to be concentrated in director re- sent CEO departures that are outside of directors’ control and thus not
elections in which institutional investors hold above-average ownership expected to affect director reputation. If distraction drives our results,
stakes in both the turnover and the interlocked firms. sudden CEO deaths are expected to trigger an increase in withheld
We conduct several tests to assess the internal validity of our main votes at interlocked firms similar to that of forced CEO departures. Our
result. The identifying assumption central to a causal interpretation of results show that sudden CEO deaths do not affect directors’ re-election
difference-in-differences estimates is that treated and control samples results at interlocked firms. Second, we find no evidence of a decline
follow parallel trends. We show that there is no significant difference in board meeting attendance of directors interlocked to a forced CEO
in the change in withheld votes between turnover-interlocked and non- turnover. Hence, we find no evidence supporting a distraction-based
turnover-interlocked directors before forced turnovers, supporting the explanation.
notion that the parallel trends assumption holds. We also show that In the final part of the paper, we follow prior research (Farrell and
the timing of the changes in withheld votes coincides with that of the Whidbee, 2000; Ellis et al., 2021) and study the effect of forced CEO
turnovers and that there is no subsequent reversal, suggesting that the turnovers on future directorships. We acknowledge that this analysis
reputational loss is indeed driven by the turnover and is persistent. In a may be subject to endogeneity problems. Directors may choose to leave
placebo test, we show that directors joining the board of the turnover some board seats voluntarily, for instance, due to reputational concerns
firm after a forced departure but before their next re-election at the connected to the past performance of the turnover firm, increased
interlocked firm do not experience an increase in withheld votes. This busyness as a result of the CEO succession, or disagreements over
finding again supports our previous results suggesting that directors the turnover (Farrell and Whidbee, 2000; Ertimur et al., 2012; Levit
lose reputation for an involvement in a forced turnover, not for sitting and Malenko, 2016; Fahlenbrach et al., 2017; Masulis and Zhang,
on the board of a firm that recently fired the CEO. 2019; Ellis et al., 2021). To analyze director labor market outcomes
A potential concern with the interpretation of our results is an empirically, we track our sample of turnover-interlocked directors and
omitted variables bias. Specifically, firms that experience poor stock their board seats over the years following the forced CEO turnovers
price performance are more likely to force out their CEO (e.g., Dasgupta and compare them to other directors sitting on the board of the same
et al., 2018; Fee et al., 2018). As poor firm performance can also be interlocked firms. At the extensive margin, we find that directors
linked to bad monitoring (e.g., Klein, 1998), investors might simply involved in a forced CEO turnover have a higher propensity to leave
penalize directors for the poor firm performance, which resulted in the the director labor market. At the intensive margin, we document that

3
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

directors lose board seats five years after the turnover, with initially such as (Aggarwal et al., 2019), suggest that even modest increases in
lost directorships, primarily at the turnover firm, being offset by newly withheld votes lead to higher director turnover, committee demotions,
acquired board seats in the subsequent four years. However, such newly and reduced opportunities in the labor market. We contribute to this
acquired board seats are at smaller firms. Overall, these results suggest literature by establishing that forced CEO turnovers significantly affect
that reputational losses from involvement in a forced CEO turnover vote outcomes. In fact, we show that forced turnovers have the second
adversely affect director labor market opportunities.5 largest effect on vote outcomes among a set of corporate events that
Our paper contributes to three strands of research. First, our paper have been shown to hurt director reputation and are only exceeded by
adds to the literature on forced CEO turnovers. Many studies resort bankruptcy filings, arguably the most detrimental event for corporate
to forced CEO turnovers as an outcome variable when analyzing the shareholders.
benefits of good corporate governance (e.g., Weisbach, 1988; Farrell Finally, our paper relates to the literature on board interlocks. Prior
and Whidbee, 2000; Kempf et al., 2017; Dasgupta et al., 2018). This research shows that such board interlocks can propagate firm poli-
choice is supported by the empirical observation that stock prices cies (Davis, 1991; Bizjak et al., 2009; Stuart and Yim, 2010; Bouwman,
typically react positively to forced CEO turnover announcements (e.g., 2011; Chiu et al., 2013; Brown and Drake, 2014; Gopalan et al., 2021;
Denis and Denis, 1995; Borokhovich et al., 1996; Huson et al., 2001, Zhang, 2021; Foroughi et al., 2022). We extend this literature by
2004). However, short-term event studies around the turnover an- showing that actions taken by directors at one firm affect shareholders’
nouncement may not capture shareholders’ assessment of the boards’ assessment of these directors at interlocked firms.6 Showing that direc-
performance in the monitoring and firing of a CEO as such turnovers tors are held accountable for actions taken at other firms is important
often follow periods of poor company performance. While shareholders because it enables us to provide direct evidence on the disciplining
may greet the eventual decision to fire a poorly performing CEO, their effect of the director labor market, as suggested by Fama (1980) and
assessment of the board’s willingness and ability to monitor and replace Fama and Jensen (1983).
the CEO may depend on how much value was destroyed before the
turnover decision was made. We add to this literature by measuring 2. Sample and data
investors’ assessment of forced CEO turnovers through directors’ vote
2.1. Sample selection
outcomes at directorships held at other firms. Our setting allows us to
separate idiosyncratic turnover firm characteristics as well as specific
To compile our sample of CEO turnovers, we first identify all CEO
aspects of these turnovers from shareholders’ assessment of forced
departures from S&P 1500 firms between January 2003 and December
turnover decisions. Our results show that forced CEO turnovers are
2017 in BoardEx. We then conduct extensive news searches in Factiva
associated with significant reputational losses for involved directors.
to determine the exact departure announcement date, the name of the
This finding contradicts the general presumption in the extant empir-
replacement, whether the replacement was announced jointly with the
ical finance literature that forced CEO turnovers are a credible signal
departure, and the circumstances of the departure. We drop departures
of boards’ monitoring ability and indicate well-functioning corporate
that result from the firm being acquired, the firm acquiring another
governance. Instead, our findings suggest that forced CEO turnovers
company, or the firm selling or spinning off parts of its business because
are often perceived as a signal of poor monitoring by the board and,
such events often indicate a strategic realignment of the firm. We
thus, may indicate a governance failure that becomes visible upon the
also drop CEO departures that result from proxy contests, government
announcement of a forced turnover.
interventions, and other types of active monitoring by parties other
The existing empirical literature on the consequences of forced CEO
than the board of directors. Additionally, we remove departures for
turnovers for involved directors is very limited and produces ambigu- which we cannot find sufficient board meeting data in BoardEx and
ous findings. Farrell and Whidbee (2000) find that outside directors ISS. This leaves us with a sample of 1773 CEO departures involving
have a higher likelihood of leaving the turnover firm following a CEO 1739 CEOs at 1266 turnover firms.
dismissal but directors remaining with the turnover firm have a higher We follow previous literature in classifying CEO turnovers as either
likelihood of gaining additional board seats. In contrast, Ellis et al. ‘‘forced’’ or ‘‘unforced’’ (e.g., Parrino, 1997; Parrino et al., 2003; Peters
(2021) investigate directors’ learning from forced turnovers and find and Wagner, 2014; Jenter and Kanaan, 2015). We classify a turnover as
that directors involved in forced CEO turnovers, while showing some forced if newspaper articles indicate that the CEO is fired, is forced out
learning, do not gain other board positions following CEO dismissals. of her position, or departs due to unspecified policy differences. Parrino
One potential reason for these ambiguous results is that both studies use (1997) argues that CEOs departing below the age of 60 should be
future directorships as an outcome variable, which are a noisy proxy for treated with special care. He classifies such turnovers as forced if
director reputation and subject to endogeneity concerns, as explained (i) articles do not indicate that the CEO left due to poor health or
above. We add to this sparse literature on the consequences for di- acceptance of another position (elsewhere or within the firm) or (ii)
rectors from firing a CEO by establishing an unambiguous negative articles report that the CEO is retiring but firms do not announce the
link between forced turnovers and shareholder satisfaction as measured retirement at least six months before the succession. Our classification
through director re-election outcomes at interlocked firms, an arguably procedure follows that of Parrino (1997) but, in an attempt to increase
cleaner and more granular measure of director reputation. Our results the precision of the forced turnover classification, adds one additional
also suggest that the reputational losses that we document dominate criterion dealing with retained positions of the outgoing CEO at the
the learning benefits found in Ellis et al. (2021). turnover firm: We classify a turnover as unforced if the CEO does not
Second, we contribute to the literature on director voting. This liter- leave the firm within one month after the announced departure date,
ature uses vote outcomes as a measure of shareholder satisfaction and which includes the termination of a board membership, but does not in-
director reputation (e.g., Ertimur et al., 2012; Brochet and Srinivasan, clude a consulting position. The reason is that, after applying (Parrino,
2014; Liu et al., 2020; Erel et al., 2021). Results of recent studies,

6
In a contemporaneous paper, Johnson et al. (2023) show that directors
5
In further tests, we show that the loss of the directorship at the turnover involved in the adoption of a poison pill experience a decrease in vote margins
firm, which may be associated with a decline in influence, network connec- and an increase in the probability of losing a board seat at the pill-adopting
tions, and prestige, does not drive the documented increase in withheld votes. firm. In some of their tests, they expand their analyses to interlocked firms
Hence, these results support a reputation-based explanation for the increase in and show that directors experience a decrease in vote margins across all their
withheld votes. directorships.

4
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

1997)’s original algorithm, we discover that CEOs who are forced out Table 1
Forced CEO turnover and interlocked director characteristics at turnover firms.
of their role remain in an executive or board position at the turnover
firm for an average of 1.49 years. Such a long lead-time between the Panel A: Forced turnover characteristics

departure announcement and the effective departure suggests that the Forced CEO turnovers Interlocked directorships
departure is not forced but consensual. The relatively short cut-off of Mean N Mean N
one month ensures that our algorithm captures forced CEO turnovers Successor announced (d) 0.451 206 0.448 607
only, and thus keeps the number of false positives low.7 This results in Performance-induced (d) 0.876 201 0.895 591
283 (16%) turnovers being classified as forced.8 Honeymoon stage (d) 0.320 206 0.308 607
Harvest stage (d) 0.655 206 0.685 607
Next, we identify all outside directors who serve on the boards of the Decline stage (d) 0.024 206 0.012 607
283 turnover firms at the forced CEO departure announcement date and
Panel B: Interlocked director characteristics at turnover firms
their interlocked directorships using BoardEx.9 We classify a director as
Interlocked directorships
turnover-interlocked if she serves as an outside director on the board of
another firm besides the turnover firm at the turnover announcement Mean N

date. Co-opted director (d) 0.504 607

Our main outcome variable is the change in the share of withheld Monitoring committee member (d) 0.796 604

votes in director re-elections. Withheld votes are defined as the sum This table reports descriptive statistics on forced CEO turnover characteristics (Panel
of votes withheld and votes against, divided by the total number of A) and interlocked director characteristics at the turnover firms (Panel B). A CEO
departure at an S&P1500 firm between 2003 and 2017 is classified as a forced turnover
votes cast (e.g., Aggarwal et al., 2019).10 We collect data on director if (i) newspaper articles indicate that the CEO is fired, is forced out of her position,
re-elections from ISS Voting Analytics, which encompasses shareholder or departs due to unspecified policy differences, (ii) the CEO does not leave to take
votes of Russell 3000 firms from January 2003 onward. Our sample over an executive position at another organization, and (iii) the CEO leaves within
ends in December 2017. We only consider regular director elections, one month after the departure is announced, which includes the termination of a
board membership but does not include a consulting position at the turnover firm.
which make up for 96.33% of director elections in ISS, and exclude,
Interlocked directors are directors who contemporaneously serve on the board of a firm
for instance, elections at special meetings and contested elections. that announces a forced CEO turnover and on the board of another firm that does not
Computing the change in withheld votes from one shareholder meeting fire the CEO. Interlocked directors appear in our sample if re-election vote data from
to the next requires two consecutive elections per director-firm pair.11 ISS, director data from BoardEx and ISS, stock price data from CRSP, accounting data
We match directors in ISS Voting Analytics and BoardEx. This from Compustat, and ownership data from Thomson Reuters are available. Definitions
and data sources of all variables are provided in Table A.1.
merged dataset allows us to classify director re-elections as either
treated or control observations. Specifically, the treatment group com-
prises re-elections of directors who were involved in a forced CEO
turnover at another firm between this re-election date and the pre- After applying these filters, the final sample includes 88,406 direc-
vious election date, while the control group comprises re-elections of tor re-elections of 18,693 individual outside directors at 3,269 firms.
directors who were not involved in a forced CEO turnover at another The treatment sample includes 607 director re-elections that concern
firm between this re-election date and the previous election date. We directors interlocked to 206 forced CEO turnovers. The remaining
drop director re-elections for which we cannot retrieve sufficient stock 87,799 director re-elections concern non-turnover-interlocked directors
price data from CRSP, accounting data from Compustat, institutional and thus are eligible for the control sample.
ownership data from Thomson Reuters, and board data from BoardEx
and ISS. We also remove re-elections at financial and utility firms (SIC 2.2. Descriptive statistics
codes 6000–6999 and 4900–4999, respectively). Finally, we drop the
first re-elections of all directors at the turnover firms following the Table 1 reports summary statistics on the 206 forced CEO turnovers
forced CEO turnovers and the first re-elections of the departing CEOs and 607 interlocked directors at the turnover firms. Panel A reports
at other firms following the turnovers.12 , 13 statistics on the forced turnovers. In 45% of turnover announcements
(or 45% of interlocked directorships), firms announce the appointment
of a full replacement to the outgoing CEO jointly with the departure.
In 88% of turnovers (or 90% of interlocked directorships), we de-
7
This additional criterion changes 128 CEO turnovers, classified as ‘‘forced’’ fine the turnover as performance-induced using the model of Jenter
using Parrino (1997)’s algorithm, to ‘‘unforced’’.
8 and Lewellen (2021).14 Following Hambrick and Fukutomi (1991)
Figure IA.1 in the Internet Appendix displays the distribution of turnovers
as well as forced turnovers over time. The number of (forced) turnovers is
and Brochet et al. (2021), we categorize forced CEO turnovers based
fairly evenly distributed across sample years. Hence, our results are unlikely
to be driven by market-wide spikes in forced CEO turnovers.
9
We exclude inside (or executive) directors at both the turnover firms and
12
our sample firms for two reasons. First, inside directors are primarily involved Table IA.1 in the Internet Appendix illustrates the reduction in CEO
in daily business decisions and may thus be punished for the performance that turnovers and turnover-interlocked directorships for each of the seven filters
leads to the turnover, while outside directors are responsible and punished applied to the initial sample of 283 forced CEO turnovers and 1,787 turnover-
for monitoring and, if necessary, firing the CEO (e.g., Fama, 1980; Fama interlocked directorships, resulting in the final sample of 206 forced CEO
and Jensen, 1983). Second, inside directors’ vote results may depend on the turnovers and 607 turnover-interlocked directorships.
13
perceived performance in their executive roles rather than in their role as There are only 238 (0.27%) director re-elections of previously fired
corporate directors. CEOs in the control sample used in our main analysis. To address concerns
10
In unreported robustness tests, we alternatively employ the vote margin, that including previously fired CEOs in the control group affects our results,
defined as the percent of votes for a director minus the percent against, minus we drop these re-elections from the control sample and replicate our main
the percent abstaining, and find very similar results. analysis. Results are virtually identical (as shown in Table IA.2 in the Internet
11
About one-fifth of re-elections in our sample take place at firms with Appendix).
14
a staggered board. We keep these observations in our sample but remove We classify a forced CEO turnover as performance-induced if the implied
director re-elections that are further apart than five years or closer than one probability from Jenter and Lewellen (2021)’s two-probit model is above
quarter from the previous election. In Section 4.1, we show that our baseline 50%. The results of the two-probit estimation are reported in Table IA.4 in
findings are not sensitive to dropping re-elections of directors at firms with the Internet Appendix. The percentage of performance-induced forced CEO
staggered boards. We also show that the reputational effect is independent of turnovers is consistent with Jenter and Lewellen (2021), who find 82% of
the time between the turnover announcement and the re-election date. forced CEO turnovers to be performance-induced (see their Table 5, Panel A).

5
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 2 Table 3
Turnover-interlocked director and firm characteristics. Non-turnover-interlocked director and firm characteristics.
Panel A: Turnover-interlocked director characteristics Panel A: Non-turnover-interlocked director characteristics
Mean Median SD N Mean Median SD N
Age (yrs) 63.534 64.739 6.904 607 Age (yrs) 63.003 63.819* 8.707 87,786
Female (d) 0.173 0.000 0.379 607 Female (d) 0.131*** 0.000*** 0.337 87,799
# of other board seats 3.142 3.000 1.918 607 # of other board seats 1.770*** 1.000*** 2.093 87,799
Panel B: Turnover-interlocked director characteristics at interlocked firms Panel B: Non-turnover-interlocked director characteristics at non-interlocked firms
Mean Median SD N Mean Median SD N
% votes withheld 7.116 2.687 10.887 607 % votes withheld 6.139** 2.458** 9.623 87,799
Tenure (yrs) 8.328 6.995 5.514 607 Tenure (yrs) 8.691 6.995 6.604 87,799
ISS withhold/against (d) 0.074 0.000 0.262 607 ISS withhold/against (d) 0.096* 0.000* 0.295 87,799
Panel C: Turnover-interlocked firm characteristics Panel C: Non-turnover-interlocked firm characteristics
Mean Median SD N Mean Median SD N
Total assets (millions) 14,662.478 2,532.490 56,107.593 607 Total assets (millions) 10,957.859** 1,422.700*** 43,764.939 87,799
Tobin’s Q 2.028 1.647 1.220 607 Tobin’s Q 2.035 1.615 1.316 87,799
ROA 0.138 0.133 0.110 607 ROA 0.103*** 0.128*** 0.176 87,799
BH return (m270,m21) 0.022 −0.007 0.407 607 BH return (m270,m21) 0.006 −0.046*** 0.718 87,799
Board size 9.778 10.000 2.180 607 Board size 9.223*** 9.000*** 2.359 87,799
% outside directors 0.852 0.875 0.076 607 % outside directors 0.836*** 0.857*** 0.089 87,799
% busy outside directors 0.342 0.333 0.189 607 % busy outside directors 0.254*** 0.250*** 0.193 87,799
Institutional ownership (%) 0.773 0.818 0.218 607 Institutional ownership (%) 0.701*** 0.794*** 0.281 87,799

This table reports descriptive statistics on turnover-interlocked director characteristics This table reports descriptive statistics on non-turnover-interlocked director character-
(Panel A), turnover-interlocked director characteristics at the interlocked firms (Panel istics (Panel A), non-turnover-interlocked director characteristics at the non-interlocked
B), and turnover-interlocked firm characteristics (Panel C). A CEO departure at an firms (Panel B), and non-turnover-interlocked firm characteristics (Panel C). The
S&P1500 firm between 2003 and 2017 is classified as a forced CEO turnover if (i) sample comprises director re-elections between 2003 and 2017 from ISS Voting
newspaper articles indicate that the CEO is fired, is forced out of her position, or Analytics, augmented with director data from BoardEx and ISS, stock price data from
departs due to unspecified policy differences, (ii) the CEO does not leave to take CRSP, accounting data from Compustat, and ownership data from Thomson Reuters.
over an executive position at another organization, and (iii) the CEO leaves within Definitions and data sources of all variables are provided in Table A.1. *, **, and ***,
one month after the departure is announced, which includes the termination of a indicate statistical significance at the 10%, 5%, and 1% level, respectively, of tests for
board membership but does not include a consulting position at the turnover firm. differences in means and medians between the turnover-interlocked director sample
Interlocked directors are directors who contemporaneously serve on the board of a firm reported in Table 2 and the non-turnover-interlocked director sample reported in this
that announces a forced CEO turnover and on the board of another firm that does not table.
fire the CEO. Interlocked directors appear in our sample if re-election vote data from
ISS, director data from BoardEx and ISS, stock price data from CRSP, accounting data
from Compustat, and ownership data from Thomson Reuters are available. Definitions
and data sources of all variables are provided in Table A.1. and control samples are statistically significant, as indicated by the
asterisks in Table 3, these differences in levels are unlikely to drive
our results for at least two reasons. First, we estimate our regressions
on the timing of the turnover within a CEO’s tenure. Around 32% of in first-differences at the director-firm level and saturate our models
departures (or 31% of interlocked directorships) occur during the first with different fixed effects (see Section 3.2). Second, we use propensity
three years of a CEO’s tenure (the ‘‘honeymoon period’’), 66% (or 69% score-matched control samples to address concerns that our results are
of interlocked directorships) during years three to 13 of tenure (the driven by selection rather than treatment. In these matched samples,
‘‘harvest stage’’), and 2% (or 1% of interlocked directorships) after all differences in director- and firm-level control variables between
more than 13 years of tenure (the ‘‘decline stage’’). treatment and control samples are statistically insignificant. Still, our
Panel B reports descriptive statistics on the 607 interlocked directors results remain qualitatively unchanged (see Section 4.2).15
at the turnover firms as of the date of the CEO departure announce-
ment. About 50% of turnover-interlocked directors are ‘‘co-opted’’, 3. The effect of forced CEO turnovers on interlocked directors’
i.e., were appointed under the departing CEO (Coles et al., 2014). vote support
Almost 80% are members of either the nominating or the compensation
committee, making them more responsible for monitoring CEOs (e.g., 3.1. Univariate results
Chhaochharia and Grinstein, 2009; Guo and Masulis, 2015).
Table 2 reports summary statistics on turnover-interlocked directors We first test for the reputational effect of forced CEO turnovers
and interlocked firms as of the date of the first re-election following the univariately. To this end, we compare the change in withheld votes
forced CEO turnover. This sample constitutes the treatment group in in consecutive re-elections of directors interlocked to a forced CEO
our analysis. Panel A reports summary statistics on the characteristics turnover with the change in withheld votes of directors not interlocked
of the interlocked directors. On average, directors are 64 years old, to a forced CEO turnover. We do so by regressing the change in with-
are female in 17% of all cases, and hold 3.1 additional board seats held votes on a treatment indicator, that is, a dummy variable equal to
besides the one at the interlocked firm but including the one at the one if a director was involved in a forced CEO turnover at another firm
turnover firm. Panel B reports interlocked director characteristics at the between the date of this election and the date of the previous election,
interlocked firms. On average, turnover-interlocked directors receive
7.1% withheld votes. ISS recommends withholding votes for 7.4% of
the interlocked directors. Panel C reports summary statistics on inter- 15
Note: To obtain treatment status, directors need to hold at least two board
locked firms. On average, they have 14.7 billion total assets, generate
seats at the CEO’s departure, resulting in a significantly higher number of
around 14% return on assets, and institutional investors own 77% of board seats compared to control directors. This additional condition for the
their shares. treatment sample may cause some of the documented differences across the
Table 3 replicates Table 2 for the non-turnover-interlocked directors treatment and control samples as shown in Table 3. In Section 4.1, we conduct
and the firms on whose board they serve. This sample constitutes the an additional robustness test, in which we require control directors to have at
control group in our analysis. While many differences across treatment least two board seats as well. Again, results remain virtually unchanged.

6
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Fig. 1. Forced CEO turnovers and withheld votes at interlocked firms. This figure plots coefficient estimates obtained from an ordinary least squares regression of the change
in withheld votes in director re-elections on a dummy variable set equal to one if a director was involved in a forced CEO turnover at another firm between the date of this
re-election and the date of the previous election, and zero otherwise. Each bar represents the coefficient estimate obtained from a separate regression that uses leads or lags of
up to three elections relative to the first re-election after the forced CEO turnover. The sample comprises director re-elections between 2003 and 2017 from ISS Voting Analytics,
augmented with director data from BoardEx and ISS, stock price data from CRSP, accounting data from Compustat, and ownership data from Thomson Reuters. 𝑡-statistics are
reported in parentheses and are based on standard errors clustered at the Fama–French 48 industry level. *, **, and ***, indicate statistical significance at the 10%, 5%, and 1%
level, respectively. Definitions and data sources of all variables are provided in Table A.1.

and zero otherwise. We estimate this regression repeatedly for three performance, or industry shocks that may affect the value of certain
re-elections before to three re-elections after the first re-election that director characteristics. To control for such confounding factors, we
follows the forced CEO turnover using a treatment dummy variable estimate the following difference-in-differences regression:
indicating a turnover-interlock with a lead or lag of up to three re- 𝛥𝑖𝑗 𝑣𝑖𝑗 𝑡 = 𝛽 𝑇𝑖𝑗 𝑡,𝑡−1 + 𝛿 𝛥𝑖𝑗 𝑋𝑖𝑡 + 𝜃 𝛥𝑖𝑗 𝑍𝑗 𝑡 + 𝛼𝑠𝑡 + 𝜀𝑖𝑗 𝑡 , (1)
elections. The coefficient estimates for the treatment dummy variable
obtained in these regressions are displayed in Fig. 1. Absent a forced where 𝑖, 𝑗, 𝑠, and 𝑡 index director, firm, industry, and years, respec-
turnover, differences in changes in withheld votes between turnover- tively. 𝛥𝑖𝑗 is the first-difference operator between two elections of
interlocked and non-turnover-interlocked directors are small, ranging director 𝑖 at firm 𝑗. 𝑣𝑖𝑗 𝑡 is withheld votes of director 𝑖 at firm 𝑗 in
from −0.60 to +0.28 percentage points, and are statistically insignifi- year 𝑡, in percent. 𝑇𝑖𝑗 𝑡,𝑡−1 is the treatment indicator, that is, a dummy
cant. However, following a forced CEO turnover, turnover-interlocked variable equal to one if director 𝑖 was involved in a forced CEO turnover
directors experience a significant increase in withheld votes of 1.36 at another firm between the date of her re-election at firm 𝑗 in year
percentage points relative to non-turnover-interlocked directors. This 𝑡 and the date of her previous election, 𝑡 − 1, at firm 𝑗, and zero
increase in withheld votes represents a 22.1% increase over the sample otherwise.17 𝑋𝑖𝑡 and 𝑍𝑗 𝑡 are time-varying director-level and firm-level
mean. Hence, these univariate results suggest that directors suffer a control variables. 𝛼𝑠𝑡 are interacted industry-year fixed effects. 𝜀𝑖𝑗 𝑡 is
substantial reputational loss from a forced CEO turnover.16 the error term. Standard errors are clustered at the Fama–French 48
industry level.18
By estimating the regressions in first-differences, we remove unob-
3.2. Multivariate results
served heterogeneity at the director-firm level. Such a specification is
similar to a regression in levels with director-firm fixed effects but,
The results in the previous section suggest that forced CEO turnovers
unlike a specification in levels, it can easily accommodate repeated
are associated with increases in withheld votes at interlocked firms.
However, these results could be confounded by coinciding changes in
director characteristics, such as changes in the number of outside board
seats, changes in interlocked firm characteristics, such as changes in 17
For most director re-elections, 𝑡 and 𝑡 − 1 refer to the year 𝑡 and 𝑡 − 1,
respectively. However, in one-fifth of our sample, firms do not re-elect all
directors each year but in a staggered manner. For staggered re-elections, 𝑡
refers to the election in year 𝑡, while 𝑡 − 1 refers to the previous shareholder
16
Figure IA.2 in the Internet Appendix displays changes in withheld votes meeting for which we observe election outcomes. In Section 4.1, we show that
around forced CEO turnovers for turnover-interlocked and non-turnover- our baseline findings are not sensitive to dropping re-elections of directors
interlocked directors separately. The figure shows that the increase in withheld at firms with staggered boards. We also show that the reputational effect
votes of turnover-interlocked versus non-turnover-interlocked directors around is independent of the time between the turnover announcement and the
the turnover is driven by an increase in withheld votes of turnover-interlocked re-election date.
18
directors and not a reduction in withheld votes of non-turnover-interlocked Our results remain similar when we cluster the standard errors at the firm
directors. or the director level.

7
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 4 firms in the same industry and year. In Column 2, we add firm fixed
Forced CEO turnovers and withheld votes at interlocked firms.
effects to the model. These fixed effects additionally absorb unobserved
Dependent variable: % votes withheld
time-invariant firm-specific heterogeneity. This specification effectively
(1) (2) (3) compares turnover-interlocked directors to non-turnover-interlocked
Forced CEO turnover (d) 1.204*** 0.878** 1.034** directors at the same firm. The results show that turnover-interlocked
(3.071) (2.553) (2.407) directors experience a significant increase in withheld votes of 0.88
ISS withhold/against (d) 18.214*** 18.462*** 18.644*** percentage points over non-turnover-interlocked directors at the same
(29.462) (29.645) (26.995)
firm. In Column 3, we augment our baseline regression with director
# of other board seats 0.128*** 0.133*** 0.172*** fixed effects. These fixed effects remove unobserved time-invariant
(3.474) (3.100) (3.866)
director characteristics and produce a within-person comparison of the
Total assets, log 0.562*** 0.514** 0.434*
turnover-interlocked directors to themselves absent an involvement in
(2.940) (2.326) (1.909)
a forced CEO turnover. Our results again remain robust in this most
Tobin’s Q −0.194*** −0.140* −0.151*
(−2.827) (−1.976) (−1.871)
conservative specification: Directors interlocked to a forced turnover
face a significant increase in withheld votes of 1.03 percentage points
ROA −1.915** −2.160** −1.816*
(−2.676) (−2.336) (−1.981) compared to the re-elections of the same directors absent a forced
BH return (m270,m21) −0.586*** −0.564*** −0.596***
turnover. Hence, these findings are consistent with the notion that
(−6.892) (−6.273) (−6.711) directors suffer a reputational loss across all board mandates following
Board size −0.096** −0.087 −0.103 a forced CEO turnover. Coefficient estimates on the control variables
(−2.081) (−1.653) (−1.632) are generally consistent with prior research (e.g., Cai et al., 2009). For
% outside directors 2.536** 2.044* 2.009 example, a change of ISS’s recommendation from elect to withhold or
(2.427) (1.840) (1.607) vote against increases withheld votes significantly.19 Similarly, direc-
% busy outside directors 0.261 0.533 0.193 tors who gain board memberships subsequently receive more withheld
(0.614) (0.987) (0.354) votes. On the other hand, improved firm performance or increases in
Institutional ownership (%) 0.049 −0.009 0.062 growth opportunities reduce withheld votes.20
(0.153) (−0.020) (0.131) To gauge the economic magnitude of the documented vote effects,
Year × Industry FE Yes Yes Yes we first compare the observed increases in withheld votes to the
Firm FE No Yes No
sample mean. Directors generally receive high vote support, with a
Director FE No No Yes
mean of only 6.1% of votes withheld (see Table 3). Hence, increases
Observations 88,406 88,406 88,406
in withheld votes of between 0.88 and 1.20 percentage points, as
Firms 3,269 3,269 3,269
Directors 18,693 18,693 18,693 reported in Table 4, represent a sizeable increase of between 14.3%
Turnover-interlocked directorships 607 607 607 and 19.6% over the mean. Indeed, recent research shows that even
Adjusted 𝑅2 0.333 0.347 0.354 modest increases in withheld votes often have negative consequences
This table reports results from ordinary least squares regressions of the change in for directors (e.g., Aggarwal et al., 2019). Second, we compare changes
withheld votes in director re-elections on a dummy variable set equal to one if a in withheld votes following forced CEO turnovers to changes in with-
director was involved in a forced CEO turnover at another firm between the date of held votes following other corporate incidents that have been shown to
this re-election and the date of the previous election, and zero otherwise, as well as
control variables and fixed effects. All regressions include interacted industry-year fixed
adversely affect director reputation, including restatements (Srinivasan,
effects. The regression reported in Column 2 additionally includes firm fixed effects. The 2005), class action lawsuits (Fich and Shivdasani, 2007), poison pill
regression reported in Column 3 replaces firm fixed effects with director fixed effects. adoptions (Johnson et al., 2023), and bankruptcy filings (Gow et al.,
The sample comprises director re-elections between 2003 and 2017 from ISS Voting
Analytics, augmented with director data from BoardEx and ISS, stock price data from
CRSP, accounting data from Compustat, and ownership data from Thomson Reuters.
All variables are in first-differences. Definitions and data sources of all variables are 19
provided in Table A.1. 𝑡-statistics are reported in parentheses and are based on standard
We acknowledge that ISS’ vote recommendations at turnover-interlocked
errors clustered at the Fama–French 48 industry level. *, **, and ***, indicate statistical firms may be affected by the forced CEO turnovers. Hence, the ISS
significance at the 10%, 5%, and 1% level, respectively. withhold/against control variable may represent a ‘‘proxy control vari-
able’’ (Angrist and Pischke, 2009), a covariate whose omission can result in an
omitted variables bias that outweighs the bias resulting from the inclusion of
a bad control variable. To gauge the sensitivity of our results to the inclusion
treatments (the possibility that a director is involved in multiple forced of the ISS withhold/against control variable, we re-estimate our main analysis
CEO turnovers over our sample period). Interacted industry-year fixed reported in Table 4 without this control variable. Results are reported in Table
effects remove unobserved industry effects, time trends, and industry IA.5 in the Internet Appendix and are very similar to those reported in Table 4,
shocks. These steps ensure that the model captures the structural with coefficients of interest becoming somewhat larger and statistically more
differences between the treatment and control samples discussed in significant. Hence, our conclusions are not affected by the inclusion of the ISS
Section 2.2. As time-varying director controls, 𝑋𝑖𝑡 , we include the withhold/against control variable.
20
ISS vote recommendation indicator variable and the number of addi- Recent work in econometrics (e.g., Callaway and Sant’Anna, 2021; Sun
tional outside board seats. Director characteristics such as age, tenure, and Abraham, 2021) and finance (e.g., Baker et al., 2022) argues that stan-
and gender are removed due to their time-invariant nature in a first- dard DiD regressions with staggered treatment timing may not provide valid
differences setting. The set of time-varying control variables at the estimates of a causal effect even under random assignment of treatment. While
firm level, 𝑍𝑗 𝑡 , is based on prior corporate governance and voting this literature has yet to converge on how to correct for such bias, it agrees
that the bias tends to be smaller the larger the share of never-treated units.
literature (e.g., Cai et al., 2009; Fischer et al., 2009; Matvos and
In our case, the share of never-treated directors among all sample directors
Ostrovsky, 2010; Aggarwal et al., 2019; Johnson et al., 2023), and
is 97.6%, suggesting that any bias, if present, is likely to be small. Still, to
includes total assets, Tobin’s Q, ROA, past buy-and-hold returns, board
mitigate concerns that such a bias affects our estimates, we rerun our baseline
size, the fraction of outside directors on the board, the fraction of busy regression from Table 4, first by completely excluding treated directors from
outside directors, and institutional ownership. the control group and, second, by excluding treated directors from the control
Results from estimating the difference-in-differences regression in group after the onset of treatment. Results are reported in Panels A and B of
Eq. (1) are reported in Column 1 of Table 4. Following a forced Table IA.6 in the Internet Appendix, respectively. Results are similar to those
CEO turnover, turnover-interlocked directors receive 1.20 percentage in our baseline regression, suggesting that such a bias is likely to be small and
points more withheld votes than non-turnover-interlocked directors at thus unlikely to affect our results.

8
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 5
Cross-sectional tests: Turnover types.
Dependent variable: % votes withheld
(1) (2) (3) (4) (5) (6)
Forced CEO turnover (d) 1.173*** 0.873** 1.038**
(3.001) (2.541) (2.414)
Unforced CEO turnover (d) 0.490*** 0.137 0.239
(3.848) (1.150) (1.304)
Forced CEO turnover, performance-induced (d) 1.392*** 1.044*** 1.263**
(3.249) (2.730) (2.691)
Forced CEO turnover, non-performance-induced (d) −0.312 −0.491 −0.685
(−0.461) (−0.621) (−0.600)
Unforced CEO turnover, performance-induced (d) 0.596* 0.212 0.225
(1.985) (0.617) (0.562)
Unforced CEO turnover, non-performance-induced (d) 0.375*** 0.036 0.117
(2.943) (0.296) (0.704)
Control variables Yes Yes Yes Yes Yes Yes
Year × Industry FE Yes Yes Yes Yes Yes Yes
Firm FE No Yes No No Yes No
Director FE No No Yes No No Yes
Observations 88,406 88,406 88,406 88,322 88,322 88,322
Firms 3,269 3,269 3,269 3,268 3,268 3,268
Directors 18,693 18,693 18,693 18,691 18,691 18,691
Interlocked directorships 3,868 3,868 3,868 3,784 3,784 3,784
Adjusted 𝑅2 0.334 0.347 0.354 0.334 0.347 0.354

This table reports results from ordinary least squares regressions of the change in withheld votes in director re-elections on a dummy variable set equal to one if a director was
involved in a certain type of CEO turnover at another firm between the date of this re-election and the date of the previous election, and zero otherwise, as well as control
variables and fixed effects. Regressions reported in Columns 1 to 3 include a dummy variable set equal to one if a director was involved in a forced turnover, zero otherwise, and
a dummy variable set equal to one if a director was involved in an unforced turnover, and zero otherwise. Regressions reported in Columns 4 to 6 include four dummy variables
indicating whether a director was involved in a performance-induced forced turnover, a non-performance-induced forced turnover, a performance-induced unforced turnover, and
a non-performance-induced unforced turnover, and zero otherwise, respectively. The sample comprises director re-elections between 2003 and 2017 from ISS Voting Analytics,
augmented with director data from BoardEx and ISS, stock price data from CRSP, accounting data from Compustat, and ownership data from Thomson Reuters. All variables are
in first-differences. Control variables are the same as in Table 4. Definitions and data sources of all variables are provided in Table A.1. 𝑡-statistics are reported in parentheses and
are based on standard errors clustered at the Fama–French 48 industry level. *, **, and ***, indicate statistical significance at the 10%, 5%, and 1% level, respectively.

2018). To this end, we replicate our baseline regression from Column our conjecture that forced CEO turnovers represent exogenous shocks
1 of Table 4 and replace the interlocked turnover dummy variable on interlocked directors’ vote outcomes.
with alternative treatment indicators that are equal to one for directors
interlocked to a firm that restates earnings, a firm that becomes subject 3.3. Are directors penalized for governance failures?
to a securities litigation, a firm that adopts a poison pill, or a firm that
files for bankruptcy, and zero otherwise. Results are reported in Table In this section, we test whether reputational losses are confined to
IA.7 in the Internet Appendix. We find that the increase in withheld certain types of CEO turnovers or whether they depend on directors’
votes resulting from a forced CEO turnover (1.2 percentage points) involvement in the turnover. Such cross-sectional tests allow us to shed
is approximately 2.5 times larger than the vote effect of restating
light on the driving forces behind the documented reputational loss.
earnings (0.5 percentage points), 1.5 times larger than the vote effect of
While replacing a CEO is not expected to generally reflect negatively on
securities litigation or a poison pill adoption (both about 0.9 percentage
involved directors, the need to fire the CEO may indicate a governance
points), and half the vote effect of a bankruptcy filing (2.4 percentage
failure (e.g., Jensen, 1993; Marcel et al., 2017). For example, we would
points). Hence, the increase in withheld votes associated with firing a
expect directors to be held accountable for CEO turnovers that are
CEO is larger than the increase in withheld votes resulting from other
reactive and happen only after significant performance declines or
reputational events and is only exceeded by the increase in withheld
poorly prepared CEO turnovers that have disruptive effects on corpo-
votes resulting from a bankruptcy filing, arguably the most detrimental
rate leadership. More generally, we would expect that directors suffer
event for corporate shareholders. Finally, we investigate whether forced
reputational losses for a failure to monitor the outgoing CEO and take
CEO turnovers have the potential to lead to unusually high levels of
timely and appropriate action to replace a poorly performing CEO.
withheld votes at interlocked firms, defined as at least 15% of votes
cast (Bach and Metzger, 2017). The results in Table IA.8 in the Internet Directors may also suffer reputational losses for hiring the wrong CEO
Appendix show that directors are between 3.6% and 4.0% more likely in the first place.
to experience such high levels of withheld votes following a forced In our first test, we analyze whether directors are punished for firing
turnover at an interlocked firm, which is 44.2% to 49.1% higher than a CEO and not for a CEO turnover more generally. While directors take
the sample mean. Thus, forced CEO turnovers often result in unusually an active role in forced CEO turnovers, which have a high potential
high levels of withheld votes at interlocked firms. for disruption, unforced CEO turnovers are often the result of a CEO
In summary, these results suggest that directors suffer an economi- retiring or taking over responsibilities on the board of directors. Un-
cally large and statistically significant reputational loss across all board forced turnovers are thus expected to be smoother and less disruptive
mandates following a forced CEO turnover. These findings contrast and consequently to be associated with smaller reputational losses
with the presumption that forced CEO turnovers credibly signal a than forced turnovers. To compare re-election outcomes of directors
board’s monitoring ability and generally indicate good corporate gov- interlocked to forced and unforced turnovers, we augment the baseline
ernance. Moreover, our results survive when we control for a host of regressions from Table 4 with a dummy variable set equal to one if
observable and unobservable firm and director characteristics, suggest- a director was involved in a CEO turnover that we do not classify as
ing that forced CEO turnovers do not systematically coincide with other forced at another firm between the date of this re-election and the
changes in director and firm characteristics. Thus, the findings support date of the previous election, and zero otherwise. Column 1 of Table 5

9
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 6
Cross-sectional tests: Turnover and director characteristics.
Dependent variable: % votes withheld
(1) (2) (3) (4) (5)
Forced CEO turnover, honeymoon stage (d) 0.197
(0.314)
Forced CEO turnover, harvest stage (d) 1.679***
(4.288)
Forced CEO turnover, decline stage (d) −0.121
(−0.049)
Forced CEO turnover (d) 1.204***
(3.068)
Hired CEO and left before forced turnover (d) −0.002
(−0.003)
Forced CEO turnover, co-opted director (d) 1.670***
(3.373)
Forced CEO turnover, non-co-opted dir. (d) 0.720
(1.566)
Forced CEO turnover, monitoring comm. (d) 1.566***
(3.748)
Forced CEO turnover, non-mon. comm. (d) −0.246
(−0.348)
Forced CEO turnover, successor announced (d) 0.837
(1.539)
Forced CEO turnover, successor not ann. (d) 1.507***
(3.172)
Control variables Yes Yes Yes Yes Yes
Year × Industry FE Yes Yes Yes Yes Yes
Observations 88,406 88,406 88,406 88,403 88,406
Firms 3,269 3,269 3,269 3,269 3,269
Directors 18,693 18,693 18,693 18,693 18,693
Turnover-interlocked directorships 607 607 607 604 607
Adjusted 𝑅2 0.334 0.333 0.333 0.334 0.333

This table reports results from ordinary least squares regressions of the change in withheld votes in director re-elections on a dummy variable set equal to one if a director was
involved in a forced CEO turnover at another firm between the date of this re-election and the date of the previous election, and zero otherwise, as well as control variables and
interacted industry-year fixed effects. In the regression reported in Column 1, the dummy variable is split based on whether the turnover takes place during the first three years
(Honeymoon stage), between the third and 13th year (Harvest stage), or after the 13th year of the departing CEO’s tenure (Decline stage). The regression reported in Column 2
includes an indicator variable set equal to one if a director was involved in a forced turnover at another firm and an indicator variable set equal to one if a director was involved
in hiring the fired CEO but left the interlocked firm’s board. In the regression reported in Column 3, the dummy variable is split based on whether or not the turnover-interlocked
director was appointed to the turnover firm’s board during the departing CEO’s tenure (co-opted). In the regression reported in Column 4, the dummy variable is split based
on whether or not a full CEO replacement is announced with the departure. In the regression reported in Column 5, the dummy variable is split based on whether or not the
turnover-interlocked director was a member of the nominating or compensation committee of the turnover firm. The sample comprises director re-elections between 2003 and 2017
from ISS Voting Analytics, augmented with director data from BoardEx and ISS, stock price data from CRSP, accounting data from Compustat, and ownership data from Thomson
Reuters. All variables are in first-differences. Control variables are the same as in Table 4. Definitions and data sources of all variables are provided in Table A.1. 𝑡-statistics are
reported in parentheses and are based on standard errors clustered at the Fama–French 48 industry level. *, **, and ***, indicate statistical significance at the 10%, 5%, and 1%
level, respectively.

reports results from regressions with interacted year and industry fixed turnovers, classified using the procedure proposed by Jenter and Lewellen
effects. While we find positive and statistically significant increases (2021), as a proxy for late, reactive turnovers as they follow significant
in withheld votes for both forced and unforced turnovers, the vote performance declines. We then replicate our baseline regressions with
effect for unforced turnovers is about 60% smaller than the effect for four dummy variables indicating interlocks to forced and unforced
forced turnovers. Columns 2 and 3 report results from regressions that turnovers that are either performance-induced or non-performance-
additionally include firm and director fixed effects, respectively. The induced, and zero otherwise. In the specification with interacted year
coefficients on the unforced turnover dummy variable turn statistically and industry fixed effects, reported in Column 4 of Table 5, we find
insignificant and become economically smaller, while the coefficients positive and significant increases in withheld votes associated with
on the forced turnover dummy variable remain statistically significant performance-induced forced turnovers and insignificant reductions in
and economically sizeable. These results suggest that directors are withheld votes for non-performance-induced forced turnovers. Un-
penalized for forced turnovers, in which they took an active role and forced turnovers are associated with significant increases in withheld
which have a higher potential to be disruptive, and less so for unforced votes that are economically much smaller than those of performance-
turnovers. induced forced turnovers. When adding firm or director fixed effects,
Second, we test whether reputational losses are larger for reactive as in Columns 5 and 6, we continue to find statistically significant
turnovers than for timely, proactive turnovers. Ertugrul and Krish- and economically sizeable increases in withheld votes for performance-
nan (2011) conjecture that non-performance-induced turnovers are induced forced turnovers, while the effects of the three other turnover
indicative of a proactive board that fires an underperforming CEO types are statistically insignificant. Hence, reputational losses from CEO
before she can cause harm to firm value, while performance-induced turnovers are confined to performance-induced forced turnovers, which
forced turnovers are indicative of a reactive board that only reacts are indicative of reactive boards that act too late when damage is
to poor performance. Hence, we expect the largest reputational losses already done.
to result from forced turnovers that are performance-induced. Fol- An alternative way to measure the board’s reactiveness is to con-
lowing Ertugrul and Krishnan (2011), we use performance-induced sider the timing of the forced departure within a CEO’s tenure. The

10
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

CEO tenure literature documents an inverted U-shaped relation be- forced turnover-interlock dummy variable remains economically and
tween CEO tenure and firm value (e.g., Hambrick and Fukutomi, 1991; statistically significant, the coefficient on the dummy variable capturing
Henderson et al., 2006; Brochet et al., 2021). Brochet et al. (2021) involvement in the CEO’s hiring is economically small and statistically
show that in a large sample of public US companies, firm value peaks insignificant. This result suggests that directors are penalized for their
at about 13 years of CEO tenure on average. During the first three involvement in the CEO’s firing but not for their involvement in the
years of tenure, the ‘‘honeymoon stage’’, a newly appointed CEO gets hiring of the later fired CEO.
to know the company and starts implementing changes. The board Next, we test whether directors are penalized for neglecting their
can assess whether the new CEO meets expectations, with a proactive monitoring duties toward the CEO. As a first proxy for directors’
board firing a disappointing CEO to prevent future harm (Ertugrul monitoring intensity and quality, we use the co-option measure of Coles
and Krishnan, 2011). Hence, we do not expect significant reputational et al. (2014). They show that directors appointed after a CEO has
damages for directors involved in forced turnovers during this stage. resumed office (co-opted directors) are favorably inclined to this CEO
After a successful honeymoon stage, the CEO’s organizational changes and thus provide weaker monitoring. Hence, involvement in a forced
begin to bear fruit, and the CEO enters the ‘‘harvest stage’’. In the case turnover may reinforce shareholders’ assessment of co-opted directors’
of a poorly fitting CEO, who was still retained beyond the honeymoon limited willingness to act in shareholders’ best interests.23 We would
stage, negative performance consequences will become observable. thus expect that the reputational loss of a forced CEO turnover is
Hence, forced turnovers in the harvest stage may reflect a reactive larger for co-opted directors. To test this conjecture empirically, we
board, and we expect reputational damages for involved directors. split the forced interlocked turnover dummy into two dummy variables
Finally, after about 13 years of tenure, the positive effects start to depending on whether the director was appointed under the dismissed
be outweighed by the adverse effects of a deteriorating CEO-firm CEO. Results are reported in Column 3 of Table 6. Both coefficient
match and increased power and entrenchment, even if the CEO was estimates are positive, but the estimate for co-opted directors is about
an excellent match to the firm at appointment and remained so during twice the size of the estimate for non-co-opted directors and only the
the harvest stage. During this ‘‘decline stage’’, the CEOs’ performance former is statistically significant. This result confirms our conjecture
contribution turns negative. Turnovers during this stage are on average that directors are penalized for a failure to monitor the CEO properly.
value increasing (e.g., Brochet et al., 2021), while not necessarily Our second proxy variable for directors’ ability and willingness to
indicative of a reactive board, so we expect no adverse reputational monitor senior management is based on board committee membership.
effects for involved directors. We test this hypothesis by splitting the The board committees generally entrusted with monitoring the CEO are
forced turnover dummy variable into three dummy variables indicat- the nominating and compensation committees (e.g., Chhaochharia and
ing forced turnovers during these CEO life-cycle stages.21 Results are Grinstein, 2009; Guo and Masulis, 2015). Hence, we rerun our baseline
reported in Column 1 of Table 6. We find large and significant increases regression with two dummy variables, one is equal to one for turnover-
in withheld votes for forced turnovers during the harvest stage and involved directors who are members of either the nominating or the
insignificant changes in withheld votes, that are close to zero, for compensation committee at the turnover firm, and zero otherwise,
turnovers during the honeymoon and the decline stages. These results and the other variable is equal to one for turnover-involved directors
suggest that reputational losses are confined to forced turnovers during who are not members of these two committees, and zero otherwise.
the most productive period of a CEO’s tenure, that is turnovers that Results are reported in Column 4 of Table 6. We find large positive and
likely reflect a reactive board that took too long to recognize and statistically significant increases in withheld votes for members of the
correct a CEO-firm mismatch. nominating and compensation committees and small and statistically
We also test whether directors are not only penalized for firing the insignificant reductions for all other directors. This result suggests that
CEO but also for hiring the wrong CEO in the first place (e.g., Laux, directors suffer reputational losses from poor monitoring of the CEO.
2010). Specifically, firing a CEO may reveal new information on the Finally, we directly test whether reputational losses are larger for
CEO’s quality, the CEO-company match, and ultimately the CEO’s value directors involved in poorly prepared forced CEO turnovers. Dalton and
contribution to the company.22 However, shocks to the quality of a Dalton (2007) and Cvijanović et al. (2023) argue that the lack of an
CEO-firm match may result in a CEO firing, even though this CEO may heir apparent to the outgoing CEO signals the board’s unpreparedness
have been an excellent match at the time of appointment (e.g., Eisfeldt and results in more disruptive turnovers. Hence, director reputation is
and Kuhnen, 2013; Brochet et al., 2021). Thus, a CEO eventually fired expected to suffer more if no full replacement is announced jointly
for poor performance may still have been a reasonable choice at the with the CEO’s departure. To test this hypothesis, we replicate our
time of appointment, and her firing may thus not reflect negatively on baseline regression by replacing the forced CEO turnover interlock
directors involved in the CEO’s appointment. To separate reputational dummy variable with two dummy variables, one indicating involve-
effects from involvement in the CEO’s firing from potential reputational ment in a forced CEO departure with a full replacement announced
effects from involvement in the CEO’s hiring, we augment our baseline simultaneously and one without such an announcement. Results are
specification with a dummy variable equal to one if a director held a reported in Column 5 of Table 6. We find that directors involved in a
board seat at a turnover firm at the time of the fired CEO’s appointment forced CEO turnover with either an interim succession or no succession
but left the board before the turnover, and zero otherwise. Results announced experience an increase in withheld votes that is twice the
are reported in Column 2 of Table 6. While the coefficient on the size of the increase in withheld votes of directors involved in a forced
turnover with a full replacement announced. These results suggest that
shareholders penalize directors for poorly prepared, disruptive CEO
turnovers.
21
Note that this sample split corresponds to interaction terms between the
forced CEO turnover dummy variable and the CEO life-cycle stage dummies,
omitting stand-alone turnover characteristics. Stand-alone turnover charac-
23
teristics are omitted because they are only defined for turnover-interlocked Note that co-opted directors were hired after the CEO took office, while
directors but not for non-turnover-interlocked directors. non-co-opted directors were involved in the CEO’s hiring. However, the
22
Note that previous research shows that both initial uncertainty about previous test, reported in Column 2 of Table 6, shows that directors are not
a CEO’s fit to the company at appointment (e.g., Jovanovic, 1979; Allgood penalized for hiring the fired CEOs, suggesting that a test based on the co-
and Farrell, 2003; Ali and Zhang, 2015) as well as subsequent shocks to the option measure only captures shareholders’ reaction to weak monitoring and
CEO-firm match quality (e.g., Miller, 1991; Garrett and Pavan, 2012; Eisfeldt not to a potential involvement in the CEO’s hiring. Moreover, any reputational
and Kuhnen, 2013) make it difficult to assess a CEO’s quality at the time of loss from hiring would generate the opposite effect, so the documented penalty
appointment even ex-post. for weak monitoring of co-opted directors would be a conservative estimate.

11
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

The results in this section show that directors are not penalized Table 7
Cross-sectional tests: Common ownership in turnover and interlocked firms.
for all CEO turnovers but for forced turnovers that are related to
Dependent variable: % votes withheld
governance failures. Specifically, directors are penalized for reactive
forced turnovers, firings during the most productive period of a CEO’s (1) (2)
tenure, poor monitoring of the CEO, and the lack of a successor to the Forced CEO turnover ≥ mean common ownership (d) 1.442**
fired CEO. In contrast, directors are not penalized for hiring the wrong (2.678)

CEO in the first place. Importantly, our analysis uncovers no sub-sample Forced CEO turnover < mean common ownership (d) 0.900
(1.670)
for which we find significant reputational gains from involvement in a
turnover. Taken together, these results contradict the general presump- Forced CEO turnover ≥ median common ownership (d) 1.513**
(2.603)
tion in prior empirical finance research that a forced CEO turnover is a
Forced CEO turnover < median common ownership (d) 0.870*
credible signal of a board’s monitoring ability. In contrast, they suggest
(1.758)
that firing the CEO is indicative of governance failure at the board level.
Control variables Yes Yes
Year × Industry FE Yes Yes
3.4. Which shareholders vote against turnover-involved directors? Observations 88,390 88,390
Firms 3,269 3,269
In this section, we investigate which shareholders vote against Directors 18,693 18,693
directors involved in forced CEO turnovers. Systematically penalizing Turnover-interlocked directorships 591 591
Adjusted 𝑅2 0.334 0.334
directors across all board mandates requires shareholders to (i) observe
and evaluate directors’ actions, (ii) observe interlocked directorships, This table reports results from ordinary least squares regressions of the change in
withheld votes in director re-elections on a dummy variable set equal to one if a
and (iii) systematically vote against directors across all firms where director was involved in a forced CEO turnover at another firm between the date of
they hold board seats. Institutional investors with significant ownership this re-election and the date of the previous election, and zero otherwise, as well as
stakes in both the turnover and interlocked firms are most likely to control variables and interacted industry-year fixed effects. In the regression reported in
fulfill these requirements (e.g., He et al., 2019; Liu et al., 2020). Insti- Column 1 (2), the dummy variable is split into two variables depending on whether or
not common institutional ownership in the turnover and the interlocked firm is above
tutional investors are considered to be sophisticated (e.g., Gibson et al.,
the mean (median) common ownership. The sample comprises director re-elections
2004) and to impact the governance of their portfolio firms positively, between 2003 and 2017 from ISS Voting Analytics, augmented with director data from
independent of whether they follow an active (e.g., Brav et al., 2008) BoardEx and ISS, stock price data from CRSP, accounting data from Compustat, and
or passive investment approach (e.g., Appel et al., 2016).24 Hence, we ownership data from Thomson Reuters. All variables are in first-differences. Control
variables are the same as in Table 4. 𝑡-statistics are reported in parentheses and are
expect that turnover-interlocked directors receive more withheld votes
based on standard errors clustered at the Fama–French 48 industry level. *, **, and
if there is significant common ownership by institutional investors in ***, indicate statistical significance at the 10%, 5%, and 1% level, respectively.
both the interlocked and the turnover firm.
To test this conjecture empirically, we measure institutional in-
vestors’ common ownership in turnover and the turnover-interlocked
4. Internal validity
firms using Thomson Reuters’ institutional holdings (13F) database. We
identify all institutional investors that report holdings in the turnover
4.1. Parallel trends, treatment reversal, pseudo treatments, and selection
firms at the reporting date before the turnover announcements and
issues
holdings in the interlocked firms at the reporting date before the re-
election of turnover-interlocked directors. Common ownership is the
The identifying assumption central to any difference-in-differences
lower value of the fraction of shares outstanding held in the turnover
analysis is that treated and control observations share parallel trends
firm and the fraction of shares outstanding held in the interlocked firm.
before the onset of treatment. Specifically, turnover-interlocked direc-
We then sum up common ownership across investors at the director
tors and non-turnover-interlocked directors need to show insignificant
re-election level. Finally, we split the dummy variable indicating a
differences in their re-election results prior to forced CEO turnovers.
turnover-interlock into two dummy variables, one indicating involve-
Fig. 1 shows the differences in the change in withheld votes between
ment in a forced CEO turnover with common ownership between the
the turnover-interlocked directors and non-turnover-interlocked direc-
turnover and the interlocked firm above the sample mean (median) and
tors for a symmetric window covering seven re-elections around forced
one with common ownership below the sample mean (median).
CEO turnovers. In the three re-elections prior to forced CEO turnovers,
Results from re-estimating our baseline regression with the treat-
there are no significant differences in the changes in withheld votes
ment dummy split according to mean (median) common ownership are between turnover-interlocked and non-turnover-interlocked directors,
reported in Column 1 (2) of Table 7. We find that outside directors supporting the notion that the parallel trends assumption holds.
involved in a forced CEO turnover with above mean (median) common Fig. 1 further shows no evidence of a treatment reversal. Across
ownership experience an increase in withheld votes at an interlocked all three post-turnover re-elections, differences in the change in with-
firm that is 60% (74%) larger than the increase in withheld votes held votes between turnover-interlocked directors and non-turnover-
at an interlocked firm in response to turnovers with below mean interlocked directors are economically small and statistically insignif-
(median) common ownership. These results indicate that informed icant. To test for treatment reversal more formally, we rerun the
institutional investors engaged in both the turnover and the interlocked baseline regression from Column 1 of Table 4 with additional lags for
firms are responsible for penalizing directors for their involvement in one and two re-elections, respectively. Results are reported in Columns
the turnover.25 1 and 2 of Table 8. The coefficient on the treatment indicator remains
economically large and statistically significant in both regressions,

24
BlackRock, the world’s largest institutional investor based on Assets
Under Management, acted on various concerns by opposing 5,100 director re-
elections globally between July, 2019 and June 30, 2020. Notably, BlackRock which triggered a discussion around the resulting incentives for firms to
penalizes directors across all of their board mandates, for instance for holding compete (e.g., Azar et al., 2018; Lewellen and Lowry, 2021), might also
too many board seats. have a so-far unrevealed positive consequence: Substantial cross-shareholdings
25
These results suggest that the recent growth in institutional shareholdings facilitate monitoring of directors across firms, resulting in stronger incentives
and concentration of ownership in the portfolios of a few large asset managers, to perform.

12
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 8
Treatment reversal and pseudo treatments.
Dependent variable: % votes withheld
Baseline plus lag Baseline plus lags Pseudo 𝑡 − 1 Pseudo 𝑡 − 2 Joined after
(1) (2) (3) (4) (5)
Forced CEO turnover (d) 1.202*** 1.279*** 0.020 −0.390 1.202***
(3.052) (2.863) (0.054) (−0.835) (3.074)
Forced CEO turnover (d) 𝑡 − 1 0.097 0.373
(0.272) (1.066)
Forced CEO turnover (d) 𝑡 − 2 −0.325
(−0.506)
Joined after forced CEO turnover (d) 0.629
(0.583)
Control variables Yes Yes Yes Yes Yes
Year × Industry FE Yes Yes Yes Yes Yes
Observations 88,406 64,555 64,450 47,544 88,406
Firms 3,269 2,582 2,577 2,098 3,269
Directors 18,693 13,817 13,788 10,588 18,693
Turnover-interlocked directorships 607 424 420 315 607
Adjusted 𝑅2 0.333 0.360 0.418 0.431 0.333

This table reports results from ordinary least squares regressions of the change in withheld votes in director re-elections on a dummy variable set equal to one if a director was
involved in a forced CEO turnover at another firm between the date of this re-election and the date of the previous election, and zero otherwise, as well as control variables
and interacted industry-year fixed effects. In the regressions reported in Columns 1 and 2, the baseline regression reported in Column 1 of Table 4 is augmented with lags of the
dummy variable indicating a forced turnover at another firm. In the regressions reported in Columns 3 and 4, pseudo-treatment dummy variables are set equal to one for one
and two re-elections prior to the actual forced turnover, and zero otherwise, respectively. In the regression reported in Column 5, the baseline regression reported in Column 1 of
Table 4 is augmented with an indicator variable equal to one if the director joined the board after the departure of the CEO but before the date of her re-election. The sample
comprises director re-elections between 2003 and 2017 from ISS Voting Analytics, augmented with director data from BoardEx and ISS, stock price data from CRSP, accounting
data from Compustat, and ownership data from Thomson Reuters. All variables are in first-differences. Control variables are the same as in Table 4. Definitions and data sources
of all variables are provided in Table A.1. 𝑡-statistics are reported in parentheses and are based on standard errors clustered at the Fama–French 48 industry level. *, **, and ***,
indicate statistical significance at the 10%, 5%, and 1% level, respectively.

while the coefficients on the lagged treatment indicators are small in directors on average hold 1.4 board seats more than non-turnover-
magnitude and statistically insignificant. Hence, there is no evidence interlocked directors (see Tables 2 and 3). This selection bias may be
of a treatment reversal, implying that directors involved in a forced problematic because directors with board overlaps may fulfill different
CEO turnover experience a persistent increase in withheld votes at roles than directors without board overlaps (e.g., Geng et al., 2023).
interlocked firms. However, such a selection bias is unlikely to drive our results for
Additionally, we run placebo tests where treatment is set to 𝑡− 1 and two reasons. First, by first-differencing at the director-firm level and
𝑡 − 2, respectively. Hence, we look at re-elections that take place before including the change in the number of board seats as a covariate
the forced CEO turnovers. Columns 3 and 4 present the results. The in our regressions, we control for effects that are directly related to
coefficient on the pseudo-treatment indicator is small and statistically changes in the number of board seats as well as unobservable director
insignificant in both regressions, suggesting that increases in withheld characteristics correlated with the change in the number of board seats.
votes are due to forced turnovers at interlocked firms, as opposed to Second, our results hold when we add director fixed effects to our
alternative factors, such as deteriorating performance of the turnover regressions, which additionally control for potential selection effects at
firms preceding the turnovers. the director level. Nevertheless, to address the concern that selection
rather than treatment drives our results, we rerun our main analysis
In a related placebo test, we analyze re-elections of directors ap-
from Table 4 on a sample that excludes all non-turnover-interlocked
pointed to the board of the turnover firms after the forced CEO
directors who hold no additional board seats from the control sample.
turnovers. If directors are penalized for involvement in a forced CEO
The results are reported in Table IA.9 in the Internet Appendix and are
turnover, directors appointed shortly after the turnover should, in
virtually identical to the main results. Hence, a selection bias resulting
contrast to directors on the board at the time of the turnover, not
from our identification strategy does not drive our results.
be held accountable for the turnover and thus not face an increase
Finally, we analyze whether reputational effects are muted if in-
in withheld votes at their next re-election. To test this conjecture, we
terlocked directors are re-elected in a staggered manner. In staggered
augment the baseline specification from Column 1 of Table 4 with an
boards, the time gap between a CEO’s departure and the next director
indicator variable equal to one if a director joined the board of the re-election can be as long as five years. Such a long time lag may
turnover firm after the departure of the CEO but before the date of result in a muted reputational effect if shareholders tend to forget about
her next re-election at the interlocked firm. The results in Column 5 directors’ actions in the distant past. In Column 1 of Table IA.10 in the
of Table 8 show that the coefficient on the placebo dummy variable Internet Appendix, we report results obtained when we replicate the
is statistically insignificant while the coefficient on the treatment baseline regression from Column 1 of Table 4 but exclude firms with
dummy variable remains virtually unchanged compared to the baseline staggered boards. The estimate of the treatment effect is very similar to
regression. These results again suggest that it is an involvement in a the one obtained in our baseline regression. In Column 2, we split the
forced CEO turnover that results in a negative reputational effect. treatment indicator into two dummy variables depending on whether
A potential concern with our analysis is a selection bias that results the time between the date of the turnover announcement and the date
from our identification strategy. Specifically, obtaining treatment status of the next re-election is below or above one year, and zero otherwise,
requires interlocked directors to hold at least two outside directorships with time gaps above one year generally picking up staggered director
at the time of the turnover – one at the turnover firm and one at the re-elections. The coefficients on both variables are similar in magnitude
interlocked firm. In contrast, directors in the control group are not and remain statistically significant. In Column 3 (4), we split the
required to hold multiple board seats. As a result, turnover-interlocked treatment indicator into two dummy variables depending on whether

13
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

the time between the date of the turnover announcement and the date turnovers in our sample. The results of covariate balancing tests be-
of the next re-election is below or above 0.75 (1.25) years. The results tween the turnover and matched control firms are reported in Panel
are very similar to those in Column 2. Taken together, these results B of Table IA.13 in the Internet Appendix. The turnover sample and
suggest that variation in the time between elections of directors caused the matched control sample do not differ significantly in terms of any
by staggered elections does not bias our results. of the variables. Common support is illustrated in Figure IA.3 in the
In summary, this section shows that there is no significant differ- Internet Appendix. Panel B shows that the density distributions of the
ence in the change in withheld votes between turnover-interlocked propensity scores for the turnover and matched control samples align
and non-turnover-interlocked directors before forced turnovers, provid- closely across the full spectrum of propensity scores after matching.
ing empirical evidence in support of the parallel trends assumption. For each firm in our propensity score-matched sample, we identify
Moreover, the negative reputational effects are persistent, showing no all outside directors who also hold outside directorships at other firms
reversal in subsequent years. in our vote-share sample, following the methodology described in
Section 2. We identify 526 turnover-interlocked directorships held
4.2. Omitted variables and reverse causality by 383 individual directors at 411 firms, and 586 matched non-
turnover-interlocked directorships held by 418 individual directors at
A potential concern with the interpretation of our results is an omit- 453 matched firms.
ted variables bias. Specifically, firms that experience poor stock price Using this matched sample, we estimate regressions similar to our
performance are more likely to force out a CEO (e.g., Dasgupta et al., baseline regressions in Table 4. Because this propensity score-matched
2018; Fee et al., 2018). As poor firm performance can also be linked to sample is naturally cross-sectional, we estimate the regressions in levels
bad monitoring (e.g., Klein, 1998; Brick and Chidambaran, 2010), in- and not in changes. Hence, we add time-invariant director-level control
vestors might simply penalize directors interlocked to poorly perform- variables to these specifications. Table 9, Columns 1 to 3, present
ing firms. Under this alternative interpretation, causality would not the results. Column 1 reports results from a regression without fixed
run from forced turnovers to increases in withheld votes of turnover- effects. The regression reported in Column 2 additionally includes
interlocked directors. Instead, both forced CEO turnovers at the turnover industry and year fixed effects. In Column 3, we augment the regression
firms and increases in withheld votes of the turnover-interlocked di- from Column 1 with turnover event fixed effects. This most restrictive
rectors would be caused by poor firm performance of the turnover specification compares withheld votes between turnover-interlocked
firms. directors and matched non-turnover-interlocked directors within each
matched pair separately. Across all three columns, we find that direc-
To address concerns of an omitted variables bias, we conduct a
tors interlocked to firms that forced out their CEO receive significantly
propensity score matching analysis in which we match turnover firms
more withheld votes than directors interlocked to firms with the same
to non-turnover firms with the same propensity to force out the CEO.
propensity of a forced turnover but no actual turnover. The magnitude
We then compare withheld votes of directors interlocked to firms with
of the coefficients in Columns 1 and 2 is virtually identical to the
a forced turnover to withheld votes of directors interlocked to matched
magnitude in the baseline regression in Column 1 of Table 4. The
firms that do not replace their CEO. By doing so, we obtain balanced
coefficient in the most restrictive specification in Column 3 indicates
treatment and control samples comprising directors interlocked to firms
that interlocked directors on average receive 1.03 percentage points
that are similar in terms of the likelihood of firing the CEO and differ
more withheld votes (16.8% over the unconditional sample mean) than
only in the effective turnover decision. This setting thus allows us
non-turnover-interlocked directors serving on the board of firms with
to address the concern that factors that lead to turnover also drive
the same propensity to force out a CEO. These results suggest that the
increases in withheld votes of turnover-interlocked directors.
performance of the turnover firm is unlikely to constitute an omitted
We calculate propensity scores using the forced turnover likelihood
variable.28
model of Peters and Wagner (2014). To estimate this model, we con-
The forced CEO turnover likelihood model of Peters and Wagner
struct a turnover-firm panel that comprises each forced turnover in our
(2014) ensures that we obtain balanced treatment and control samples
sample and, for each turnover, all potential control firms within the
of directors interlocked to firms that are statistically indistinguishable
S&P 1500. We remove potential control firms that experience a forced
across a wide set of observable firm characteristics, including various
CEO turnover within five years before or after the turnover. This filter
metrics of company performance. However, there may still be an
ensures that the matched control firms do not include firms that have
unknown reason why, in two firms with equally bad performance, one
recently forced out or will soon force out their CEO. Additionally, we
fires the CEO and the other does not, and this reason might affect in-
remove potential control firms for which we find no outside directors
vestors’ inference about director ability. For example, bad performance
with interlocked directorships at other firms or no vote-share data
in one firm might be due to mistakes of the leadership team, leading to
at the interlocked firms. Our final sample for the propensity score both a forced turnover and negative updating about director ability. In
matching consists of 183 forced turnovers, 1,822 distinct potential contrast, bad performance in the other firm might be due to bad luck,
control firms, and 160,312 potential control firm-turnover pairs. A neither leading to a forced turnover nor conveying any signal about
covariate balancing test between the turnover firms and the potential director ability. To address this concern, we construct an alternative
control firms is presented in Panel A of Table IA.13 in the Internet
Appendix. Absent any matching, the two samples differ significantly
in most variables.
Results from estimating propensity scores are reported in Table 27
We improve match quality by imposing a maximum caliper width, i.e., a
IA.12 in the Internet Appendix.26 Within each turnover stratum, we maximum allowed distance between the turnover and control firms’ propensity
identify the control firm with the closest propensity score to the score (e.g., Austin, 2011a; Lechner and Strittmatter, 2019). We follow Austin
turnover firm.27 We find nearest neighbors for 181 of the 183 forced (2011b), who identifies an optimal caliper width of 0.2 times the pooled
standard deviation of the logit of propensity scores.
28
A concern with the use of the Peters and Wagner (2014) model for
estimating propensity scores is that it relies on performance measures com-
26
We follow the suggestion of Peters and Wagner (2014) to use a logistic puted over one year, which may be too short to capture the entire effect of
regression. Peters and Wagner (2014) use a linear probability model instead performance on forced turnovers. Hence, in a robustness test reported in Table
because they use the forced turnover likelihood model as a first stage in a IA.14 in the Internet Appendix, we replicate the analysis reported in Table 9
two-stage model and, to ensure consistency with the second stage, resort to a using performance measures computed over three years preceding the turnover
linear probability model in the first stage as well. when estimating propensity scores. The results remain very similar.

14
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 9
Propensity score matching analysis.
Dependent variable: % votes withheld
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Forced CEO turnover (d) 1.437** 1.433*** 1.026* 1.380** 1.387*** 0.856* 1.480** 1.243** 0.972*
(2.679) (2.953) (1.855) (2.436) (2.956) (1.772) (2.593) (2.225) (1.809)
ISS withhold/against (d) 19.520*** 19.334*** 19.707*** 19.092*** 19.234*** 19.472*** 21.822*** 22.316*** 22.199***
(10.431) (10.076) (10.252) (9.768) (9.620) (11.363) (11.650) (11.383) (14.852)
Female (d) −0.456 −0.301 −0.384 −0.783 −0.785 −0.075 −1.036 −0.541 −0.594
(−0.678) (−0.463) (−0.714) (−1.267) (−1.216) (−0.112) (−1.289) (−0.633) (−1.005)
# of other board seats −0.030 0.049 0.042 0.255* 0.301** 0.287** 0.270** 0.329*** 0.281**
(−0.190) (0.393) (0.235) (1.995) (2.241) (2.226) (2.494) (3.102) (2.127)
Age (yrs), log −0.642 −1.852 −1.983 1.958 0.578 1.104 3.016 2.939 1.636
(−0.235) (−0.684) (−0.884) (1.101) (0.246) (0.530) (1.409) (1.275) (0.614)
Tenure (yrs), log 1.490*** 1.301*** 1.786*** 0.882** 0.716* 0.570* 1.214*** 1.157*** 1.082***
(4.056) (3.820) (5.244) (2.171) (1.874) (1.791) (2.830) (2.922) (2.804)
Total assets, log 0.140 0.372 0.020 0.434 0.535** 0.382* 0.530* 0.527* 0.426
(0.552) (1.348) (0.080) (1.655) (2.159) (1.828) (1.749) (1.856) (1.561)
Tobin’s Q 0.045 0.160 0.058 −0.162 −0.082 −0.242 0.186 0.321* 0.158
(0.189) (0.670) (0.269) (−0.789) (−0.345) (−1.243) (1.138) (1.935) (0.678)
ROA −2.474 −4.112** −1.339 −2.196 −2.112 −1.535 −2.470 −3.574* −1.303
(−0.954) (−2.143) (−0.659) (−1.568) (−1.525) (−0.884) (−0.938) (−1.690) (−0.665)
BH return (m270,m21) −1.938*** −1.613*** −2.011*** −1.557*** −1.486*** −1.596*** −3.147*** −3.244*** −3.649***
(−3.144) (−2.952) (−3.201) (−3.399) (−3.173) (−2.863) (−4.358) (−3.896) (−4.556)
Board size −0.115 −0.046 −0.136 −0.162 −0.057 −0.167 −0.324** −0.192 −0.304
(−0.789) (−0.353) (−0.844) (−1.350) (−0.432) (−1.342) (−2.399) (−1.399) (−1.622)
% outside directors 6.880* −0.293 4.610 3.232 −0.517 1.450 7.033 3.466 1.719
(1.700) (−0.065) (1.254) (0.956) (−0.180) (0.446) (1.159) (0.578) (0.381)
% busy outside dirs −0.027 0.312 0.609 0.663 1.127 −0.657 0.688 0.918 0.213
(−0.021) (0.252) (0.403) (0.518) (0.929) (−0.490) (0.452) (0.630) (0.138)
Inst. ownership (%) 0.560 0.111 −0.497 −0.637 −0.312 −1.400 0.190 −0.400 −0.563
(0.486) (0.101) (−0.478) (−0.653) (−0.382) (−1.364) (0.168) (−0.320) (−0.530)
Year FE No Yes No No Yes No No Yes No
Industry FE No Yes No No Yes No No Yes No
Turnover FE No No Yes No No Yes No No Yes
Observations 1,112 1,112 1,112 1,072 1,072 1,072 935 935 935
Firms 745 745 745 730 730 730 646 646 646
Directors 762 762 762 720 720 720 619 619 619
Turnover-interl. d’ships 526 526 526 516 516 516 440 440 440
Adjusted 𝑅2 0.284 0.303 0.305 0.268 0.285 0.303 0.351 0.372 0.394

This table reports results from ordinary least squares regressions of withheld votes in director re-elections on a dummy variable set equal to one if a director was involved in a
forced CEO turnover at another firm between the date of this re-election and the date of the previous election, and zero otherwise, as well as control variables. The regressions
reported in Columns 2, 5, and 8 include year and interlocked firm industry fixed effects. The regressions reported in Columns 3, 6, and 9 include turnover event fixed effects.
The treatment sample comprises re-elections of directors involved in a forced CEO turnover at another firm. The control sample comprises re-elections of directors interlocked to a
sample of propensity score-matched firms with characteristics similar to the turnover firms. The sample in Columns 1 to 3 is based on propensity scores estimated using the forced
CEO turnover likelihood model of Peters and Wagner (2014). The sample in Columns 4 to 6 is based on propensity scores estimated using the forced CEO turnover likelihood
model of Peters and Wagner (2014) augmented with aggregate news sentiment prior to the CEO departure date, and the sample in Columns 7 to 9 is based on propensity scores
estimated using the forced CEO turnover likelihood model of Peters and Wagner (2014) augmented with the mean change in director re-election votes of all board members prior
to the CEO turnover. For details, see Section 4.2. Definitions and data sources of all variables are provided in Table A.1. 𝑡-statistics are reported in parentheses and are based on
standard errors clustered at the Fama–French 48 industry level (Columns 1, 2, 4, 5, 7, and 8) or at the turnover level (Columns 3, 6, and 9). *, **, and ***, indicate statistical
significance at the 10%, 5%, and 1% level, respectively.

matched control sample of firms using the CEO turnover likelihood between turnover and potential control firms absent matching shows
model of Peters and Wagner (2014) augmented with a newspaper-based that turnover firms show a significantly more negative news senti-
sentiment index. The rationale behind adding newspaper sentiment ment before the turnover (see Panel A of Table IA.15 in the Internet
is that mistakes of the leadership team can be expected to trigger Appendix), supporting the conjecture that Ravenpack’s AES captures
more negative newspaper coverage than bad luck. Matching on news negative news preceding turnover events. The results in Panel B of
sentiment thus yields a control sample that is balanced in terms of Table IA.15 in the Internet Appendix show that the propensity score
managerial involvement in negative events that lead to negative firm matching results in treatment and control samples that are balanced
performance, and eventually CEO turnover. We use Ravenpack’s ag- in terms of all observable characteristics, including news sentiment.
gregate event sentiment (AES) in the period preceding the turnover Propensity scores also align closely across the full spectrum of propen-
announcement as a proxy for firms’ news sentiment.29 A comparison sity scores after matching (see Figure IA.4 in the Internet Appendix).
Using the sample based on an augmented matching approach, we re-
estimate the regressions reported in Columns 1 to 3 of Table 9. The
results are reported in Columns 4 to 6 and are virtually identical
29
Ravenpack’s AES is a firm-level news sentiment score that ranges from to those obtained when constructing the control sample with Peters
zero to 100, with values below (above) 50 indicating negative (positive) and Wagner (2014)’s original CEO turnover likelihood model. Hence,
sentiment. The variable is constructed daily, using a 91-day rolling window, differences in (negative) newspaper sentiment, and thus an omitted
and considers all news sources from the Dow Jones universe (among others, variable related to management failures that lead to a forced CEO
Dow Jones Newswires, the Wall Street Journal, Barron’s, and MarketWatch). turnover, are unlikely to drive the observed increase in withheld votes.

15
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Another potential concern is reverse causality. Fos et al. (2018) find Table 10
Forced CEO turnovers and director distraction at interlocked firms.
that the CEO turnover-performance sensitivity increases before director
re-elections. Directors under threat of increased withheld votes might Dependent variable: % votes withheld Attends < 75%
of meetings (d)
feel inclined to fire their CEO to signal a well-functioning corporate
(1) (2) (3)
governance. Under such a scenario, the increase in withheld votes at
interlocked firms might reflect a general increase in withheld votes Forced CEO turnover (d) 1.211*** 1.209*** −0.001
(3.070) (3.067) (−0.238)
for the turnover-interlocked directors across all their board seats that
Sudden CEO death (d) −0.586
started before the directors decided to force out their CEO. Forced
(−1.005)
turnovers might then be the result of increases in withheld votes rather
Sudden CEO death (extended) (d) 0.768
than the other way around. To address this concern and to ensure
(0.870)
that causality runs from forced CEO turnovers to increases in withheld
Control variables Yes Yes Yes
votes, we again replicate our propensity score matching analysis and Year × Industry FE Yes Yes Yes
extend (Peters and Wagner, 2014)’s model by a variable that measures
Observations 88,372 88,363 51,451
the mean change in withheld votes across all director re-elections at the Firms 3,269 3,269 1,470
last meeting before the matching date versus the previous meeting. This Directors 18,692 18,692 10,382
matched sample is expected to be well-balanced in terms of fading vote Turnover-interlocked directorships 605 605 434
support for directors, which may eventually pressure them to fire the Death-interlocked directorships 28 39 –
Adjusted 𝑅2 0.333 0.333 0.036
CEO, and thus should address the reverse causality issue. The results in
Table IA.16 in the Internet Appendix indicate that turnover and control This table reports results from ordinary least squares regressions of the change in
withheld votes in director re-elections (Columns 1 and 2) or the change in board
firms differ in terms of most observable characteristics before matching meeting attendance (Column 3) on a dummy variable set equal to one if a director
(Panel A) but are statistically indistinguishable after matching (Panel was involved in a forced CEO turnover at another firm between the date of this re-
B). Most importantly, the matched sample is well-balanced in terms election and the date of the previous election, and zero otherwise, as well as control
of pressure at the turnover firm that stems from director re-elections variables and interacted industry-year fixed effects. The regression reported in Column
1 augments the baseline regression reported in Column 1 of Table 4 with a dummy
before the forced turnover. Propensity scores align closely across the
variable set equal to one if a director was involved in a sudden CEO death at another
full spectrum of propensity scores after matching (see Figure IA.5 in firm, and zero otherwise. In the regression reported in Column 2, the sample of
the Internet Appendix). The results from re-estimating Columns 1 to 3 unexpected CEO deaths is augmented with CEO sick leaves during which the CEO
of Table 9 using the matched sample that is based on this extended subsequently dies. The regression in Column 3 reports results of the baseline regression
reported in Column 1 of Table 4 with the outcome variable replaced by the change
propensity score model are reported in Columns 7 to 9 of Table 9.
in board meeting attendance. Meeting attendance is a dummy variable set equal to
Across all three columns, which include different fixed effects, we find one if a director attends less than 75% of all board meetings within a fiscal year at
that directors interlocked to firms that forced out their CEO receive the interlocked firm, and zero otherwise. The sample comprises director re-elections
significantly more withheld votes than directors interlocked to firms between 2003 and 2017 from ISS Voting Analytics, augmented with director data from
with the same propensity of a forced CEO turnover but no actual BoardEx and ISS, stock price data from CRSP, accounting data from Compustat, and
ownership data from Thomson Reuters. All variables are in first-differences. Control
turnover and similar pre-turnover election outcomes. Hence, reverse
variables are the same as in Table 4. Definitions and data sources of all variables are
causality is unlikely to drive our results. provided in Table A.1. 𝑡-statistics are reported in parentheses and are based on standard
In summary, the results in this section support our conjecture that errors clustered at the Fama–French 48 industry level. *, **, and ***, indicate statistical
neither poor performance, nor the (unobservable) reason behind the significance at the 10%, 5%, and 1% level, respectively.
turnover decision, nor negative re-election vote pressure at the turnover
firm drive our results. Instead, causality appears to run from forced CEO
turnovers to increases in withheld votes at turnover-interlocked firms.30 from the board of directors than sudden director deaths do. However,
sudden CEO deaths are, by definition, outside of the board’s control
4.3. Reputation versus distraction and thus are not expected to affect director reputation. Hence, if our
results are driven by director distraction, we would expect to observe
An alternative interpretation of the increase in withheld votes fol- an increase in withheld votes for directors serving on the board of
lowing a forced CEO turnover is director distraction. A forced turnover another firm whose CEO suddenly passes away. To test this empiri-
likely demands significant time and effort from directors involved in cally, we re-estimate the baseline regression in Column 1 of Table 4,
the process. Consequently, turnover-interlocked directors may divert augmented with a dummy variable equal to one if the director is
some of their limited attention from the interlocked firm to the turnover interlocked to another firm that experiences a sudden death of the
firm. Shareholders of the interlocked firm may penalize directors at the CEO, and zero otherwise. To construct this variable, we search for
subsequent re-election for a reduction in effort at the interlocked firm. CEO departures caused by sudden deaths as defined by Nguyen and
Such an explanation would be in line with Masulis and Zhang (2019) Nielsen (2010). To increase sample size, in an alternative specification,
and Stein and Zhao (2019), who associate increased director distraction we extend our measure of sudden CEO deaths to include cases where
from various sources, including CEO turnovers at interlocked firms, CEOs take health-related leaves of absence and die subsequently.31
with reduced monitoring efficiency. Results are reported in Columns 1 and 2 of Table 10. Consistent with
We conduct two tests of the director distraction hypothesis. The our reputation-based explanation, but inconsistent with a distraction-
first test explores sudden CEO deaths. Falato et al. (2014) show that based explanation, the coefficients on the dummy variables indicating
committee peers of suddenly deceased directors experience a workload an interlocked outside directorship to a firm experiencing the CEO’s
increase that negatively impacts the attention devoted to interlocked death are statistically insignificant in both columns.
firms. Sudden CEO deaths likely require even more attention and time In a second test of the director distraction hypothesis, we test
whether forced CEO turnovers affect board meeting attendance at inter-
locked firms. Masulis and Zhang (2019) show that directors suffering

30
These tests using matched samples also address the concern that our
baseline estimates are biased because of a size imbalance between treated and
31
control samples. Given that we continue to find results that are very similar to To ensure that these later deaths constitute a shock similar to sudden
our baseline regression, it seems unlikely that a sample size imbalance drives deaths, we restrict the extension to deaths that occur within 30 days of the
the result in our main analysis. sick leave announcement.

16
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

from external distractions typically attend fewer board meetings and seat over a one-year horizon and a 9.7 percentage point increase over
exhibit reduced board commitment, adversely affecting the perfor- a five-year horizon, respectively. The unconditional probabilities of
mance of the firms at which distracted directors hold board seats. If a losing a directorship in our sample are 13.1% and 46.0% over one
forced turnover distracts a director in a significant manner, we would and five years, respectively. Hence, the economic magnitude is sizeable:
expect to observe a decline in board meeting attendance at interlocked An involvement in a forced CEO turnover increases the probability of
firms. To test this conjecture, we re-estimate the baseline regression losing a directorship by 58% versus the sample mean over one year and
from Column 1 of Table 4 with the change in directors’ board meeting 21% over five years.
attendance at interlocked firms as an outcome variable. We obtain Next, we test whether the loss in directorships following a forced
board meeting attendance data from ISS, measured using a dummy CEO turnover is driven by a loss in the directorship at the turnover
variable set equal to one if a director attends less than 75% of all firm. To this end, we omit the directorship at the turnover firm when
board meetings within a fiscal year, and zero otherwise. Since ISS only computing our dependent variables and rerun the regressions from
provides these data for S&P1500 firms, our sample size is reduced to Columns 1 and 2. Results reported in Columns 3 and 4 show that the
51,451 directorships. The results are reported in Column 3 of Table 10. coefficients on the forced interlocked dummy variable turn economi-
The estimated coefficient is close to zero and statistically insignificant, cally and statistically insignificant in both columns. Hence, the overall
suggesting that there is no relationship between forced CEO turnovers loss in directorships is driven by a loss of directorships at the turnover
and directors’ board meeting attendance at interlocked firms. Hence, firm.32 , 33
these results are again inconsistent with director distraction driving our We also test whether directors can make up for the lost board seats
results. at the turnover firm by gaining board seats at other firms. To this
Overall, the results in this section show that exogenous shocks, end, we replace the dependent variable with dummy variables that are
which are confined to the workload of directors but do not reveal equal to one if a director gains at least one new directorship within
any information on directors’ monitoring capabilities, do not result in either one or five years following the turnover, respectively, and zero
increases in withheld votes. Together with the finding that forced CEO otherwise. The results are reported in Columns 5 and 6. They show
turnovers do not reduce board meeting attendance at the interlocked that turnover-interlocked directors are significantly more likely to gain
firms, these results suggest that our results are not caused by direc- new board seats five years after the forced turnover, but not within one
tor distraction but by shareholders reassessing individual directors’ year after the turnover. Five years after the turnover, the likelihood of
capabilities to monitor and, if necessary, fire CEOs. gaining a new board seat is 8.7 percentage points higher for turnover-
interlocked directors versus non-turnover-interlocked directors. This
5. The effect of forced CEO turnovers on directors’ labor market coefficient estimate is very similar in magnitude to that for directorship
opportunities losses reported in Column 2, suggesting that the gains in directorships
at other firms substitute for the losses of directorships at the turnover
So far, we have shown that forced CEO turnovers negatively affect firms. To explicitly test for such substitution effects, we again replicate
director reputation, as measured by changes in withheld votes at inter- the analysis and use a dummy variable that indicates a net loss of
locked firms. These reputational effects may extend into the director outside directorships as the dependent variable. The results reported
labor market and affect career prospects. Consistently, Aggarwal et al. in Columns 7 and 8 confirm our conjecture: On average, directors
(2019) show that directors who suffer an increase in withheld votes involved in a forced turnover have a 7% higher likelihood of losing a
in uncontested elections are more likely to sustain reduced labor mar- directorship in the first year after the turnover. However, this initial
ket opportunities. Similarly, Johnson et al. (2023) find that directors loss, driven by the loss of the directorship at the turnover firm, is
involved in the adoption of a poison pill suffer reputational damage offset by a higher likelihood of gaining directorships in subsequent
that impacts their career outlooks. Consequently, we expect that direc-
years. As a result, net losses of directorships five years after the forced
tors’ labor market opportunities are reduced following a forced CEO
turnover.
To test this conjecture empirically, we conduct analyses akin to
a difference-in-differences setting. Specifically, we analyze changes in 32
We also test whether the cross-sectional patterns on withheld votes
directors’ labor market opportunities between the forced CEO turnover described in Section 3.3 translate into the labor market and affect directors’
announcement date and up to five years after the turnover (first dif- losses of directorships. Results from cross-sectional regressions using the loss
ference) and compare these changes between turnover-interlocked di- of board seats as a dependent variable reported Table IA.17 in the Internet
rectors and non-turnover-interlocked directors (second difference). To Appendix are in line with those in Section 3.3.
33
construct the sample for this analysis, we use our director re-election Our finding that turnover-interlocked directors face a significant risk of
losing their directorship at the turnover firm may raise the concern that in-
sample and retain all outside directors of turnover-interlocked firms at
creases in withheld votes documented in Section 3 are driven by shareholders’
the CEO departure announcement date. Using BoardEx, we collect data
dissatisfaction with turnover-interlocked directors losing their board seat in the
on all outside directorships these directors hold at the CEO departure turnover firm, which may be accompanied by a decline in influence, network
date and all directorships they gain or lose within one and five years connections, and prestige. We conduct two tests to address this concern. First,
after the departure. In these analyses, we include turnover event fixed we drop turnover-interlocked directors who leave the board of the turnover
effects, controlling for all turnover-specific effects, as well as firm firm between the forced CEO turnover and the subsequent re-election from the
fixed effects, absorbing time-invariant characteristics of the interlocked sample and replicate our main analysis in Table 4 using this reduced sample.
firms. Results are reported in Table IA.18 in the Internet Appendix and are almost
We first analyze whether directors involved in a forced CEO identical to those reported in Table 4. Second, we re-estimate our regressions
turnover have a higher propensity for losing outside directorships from Columns 1 to 3 of Table 9 with a dummy variable set equal to one if
compared to other directors of the turnover-interlocked firms. Hence, a director leaves the turnover/matched control firm before the subsequent re-
election at the interlocked firm, and zero otherwise, and an interaction term
our first outcome variables are two dummy variables equal to one if
between this variable and the dummy variable for an interlock to a forced
a director loses at least one of her directorships within one or five
turnover. Results reported in Table IA.19 in the Internet Appendix show that
years following the forced turnover, respectively, and zero otherwise. neither the interaction term nor the standalone variable indicating the loss of
The positive and significant coefficients reported in Columns 1 and 2 a board seat at the turnover/matched control firm is statistically significant.
of Table 11 suggest that involvement in a forced turnover increases Results from both these tests suggest that the increase in withheld votes of
the likelihood of losing board seats. The obtained estimates indicate turnover-interlocked directors is independent of a loss of their board seat at
a 7.6 percentage point increase in the probability of losing a board the turnover firm.

17
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table 11
Forced CEO turnovers and gains and losses of directorships.
Dependent variable: Loss of directorship (d) Gain of directorship (d) Net loss of directorship (d)
w/ turnover firm w/o turnover firm
t = [0,1] t = [0,5] t = [0,1] t = [0,5] t = [0,1] t = [0,5] t = [0,1] t = [0,5]
(1) (2) (3) (4) (5) (6) (7) (8)
Forced CEO turnover (d) 0.076*** 0.097*** −0.000 −0.016 0.018 0.087*** 0.068*** 0.033
(2.925) (2.711) (−0.009) (−0.478) (0.776) (2.925) (2.875) (1.106)
# of other board seats 0.028*** 0.043*** 0.029*** 0.045*** 0.016*** 0.032*** 0.021*** 0.017***
(4.052) (3.696) (4.127) (3.769) (3.302) (5.059) (3.958) (3.099)
Female (d) −0.018 0.024 −0.014 0.019 0.001 0.018 −0.011 0.035*
(−1.147) (0.950) (−0.903) (0.806) (0.048) (0.654) (−0.758) (1.725)
Age (yrs), log 0.006 0.487*** −0.001 0.478*** −0.220*** −0.673*** 0.056 0.750***
(0.107) (5.484) (−0.022) (5.343) (−3.774) (−8.410) (1.001) (9.984)
Tenure (yrs), log 0.001 0.049*** 0.001 0.052*** −0.016 −0.067*** 0.000 0.074***
(0.134) (3.057) (0.128) (3.322) (−1.541) (−4.699) (0.012) (5.709)
Turnover FE Yes Yes Yes Yes Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes Yes Yes Yes Yes
Observations 4,013 3,300 4,013 3,300 4,013 3,300 4,013 3,300
Firms 465 404 465 404 465 404 465 404
Directors 2,885 2,425 2,885 2,425 2,885 2,425 2,885 2,425
Turnover-interlocked directorships 594 489 594 489 594 489 594 489
Adjusted 𝑅2 0.055 0.089 0.046 0.081 0.017 0.084 0.042 0.086

This table reports results from ordinary least squares regressions of variables capturing gains and losses of directorships within 𝑡 years after the turnover on a dummy variable set
equal to one if a director was involved in a forced CEO turnover between the date of this re-election and the date of the previous election, and zero otherwise, as well as director
control variables, turnover event fixed effects, and interlocked firm fixed effects. The dependent variable is a dummy variable set equal to one in case of a loss of directorships
(Columns 1–4), a gain in directorships (Columns 5–6), or a net loss in directorships (Columns 7–8), and zero otherwise. In the regressions reported in Columns 3 and 4, turnover
firms are omitted when constructing the dependent variable. The treatment sample comprises outside directorships of directors interlocked to forced CEO turnovers between 2003
and 2017. The control sample comprises outside directorships of directors not interlocked to a forced turnover but share a board seat with a director interlocked to a forced CEO
turnover at the turnover announcement. Definitions and data sources of all variables are provided in Table A.1. 𝑡-statistics are reported in parentheses and are based on standard
errors clustered at the turnover level. *, **, and ***, indicate statistical significance at the 10%, 5%, and 1% level, respectively.

Table 12
Forced CEO turnovers, changes in directorship quality, and exits from director labor market.
Dependent variable Difference in the sum of total assets over all Exit from labor market (d)
directorships
𝑡1 − 𝑡0 𝑡5 − 𝑡0 𝑡=1 𝑡=5
(1) (2) (3) (4)
Forced CEO turnover (d) −1.100*** −1.785*** 0.121*** 0.092***
(−3.504) (−3.298) (4.525) (4.773)
Director control variables Yes Yes Yes Yes
Turnover FE Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes
Observations 4,013 3,300 4,019 3,305
Firms 465 404 465 404
Directors 2,885 2,425 2,887 2,426
Turnover-interlocked directorships 594 489 594 489
Adjusted 𝑅2 0.026 0.102 0.115 0.057

This table reports results from ordinary least squares regressions of changes in total assets of all directorships (Columns 1 and 2) or a dummy variable indicating whether a
director holds zero board seats within 𝑡 years after the turnover (Columns 3 and 4) on a dummy variable set equal to one if a director was involved in a forced CEO turnover
between the date of this re-election and the date of the previous election, and zero otherwise, as well as director control variables, turnover event fixed effects, and interlocked
firm fixed effects. The treatment sample comprises outside directorships of directors interlocked to forced CEO turnovers between 2003 and 2017. The control sample comprises
outside directorships of directors not interlocked to a forced turnover but share a board seat with a director interlocked to a forced CEO turnover at the turnover announcement.
Director control variables are the same as in Table 11. Definitions and data sources of all variables are provided in Table A.1. 𝑡-statistics are reported in parentheses and are based
on standard errors clustered at the turnover level. *, **, and ***, indicate statistical significance at the 10%, 5%, and 1% level, respectively.

CEO turnover do not differ between turnover-interlocked directors and or five years after the turnover.34 To account for the skewness in
non-turnover-interlocked directors sitting on the boards of the same the distribution of these variables, we apply the inverse hyperbolic
firms. sine transformation (Burbidge et al., 1988).35 Results are reported in
A reputational loss may show not only in the number but also in Columns 1 and 2 of Table 12 and show that turnover-interlocked
the quality of the outside directorships. Board seats at larger firms
are associated with higher compensation (Ryan and Wiggins, 2004),
more power and prestige (Shivdasani, 1993; Adams and Ferreira, 34
In Table IA.20 in the Internet Appendix, we replicate this analysis using
2008; Masulis and Mobbs, 2014), and better networking opportuni- the change in market capitalization as the dependent variable and find very
ties (Yermack, 2004; Fich, 2005). Hence, we analyze the change in similar results.
the aggregate total assets represented by all firms at which a director 35
Note that we cannot use the natural logarithm of the change in total assets
holds outside board seats in the first year after the forced CEO turnover as these changes can take on negative values.

18
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table A.1
Variable definitions.
Panel A: Director characteristics
Variable Definition Source
% votes withheld Fraction of votes withheld and against the re-election of the director; ISS
(𝑣𝑜𝑡𝑒𝑑 𝑎𝑔 𝑎𝑖𝑛𝑠𝑡 + 𝑣𝑜𝑡𝑒𝑑 𝑎𝑏𝑠𝑡𝑎𝑖𝑛)∕(𝑣𝑜𝑡𝑒𝑑 𝑓 𝑜𝑟 + 𝑣𝑜𝑡𝑒𝑑 𝑎𝑔 𝑎𝑖𝑛𝑠𝑡 + 𝑣𝑜𝑡𝑒𝑑 𝑎𝑏𝑠𝑡𝑎𝑖𝑛)
≥15% votes withheld (d) Dummy variable equal to one if 15% or more of the votes cast in a director’s re-election are against ISS
her and were not against her at the previous election, and zero otherwise
Attends < 75% of meetings Dummy variable equal to one if the director attends less than 75% of all board meetings in a fiscal ISS
(d) year, zero otherwise
ISS withhold/against (d) Dummy variable equal to one if ISS recommends shareholders to withhold their votes or vote against ISS
the re-election of the director, zero otherwise
# of other board seats Number of other outside board seats held by the director BoardEx
Female (d) Dummy variable equal to one if the director is female, zero otherwise BoardEx/ISS
Age (yrs) Age of the director BoardEx/ISS
Tenure (yrs) Time in years since the director joined the board BoardEx/ISS
Monitoring committee (d) Dummy variable equal to one if the director is a member of the nominating or compensation BoardEx/ISS
committee of the turnover firm at the turnover announcement date, zero otherwise
Co-opted director (d) Dummy variable equal to one if the director became director of the turnover firm after the departing BoardEx/ISS
CEO was appointed, zero otherwise
≥ (<)𝑡 yrs until meeting Dummy variable equal to one if the time between turnover announcement date and the next ISS
date (d) re-election date is larger or equal to (below) 𝑡 years, zero otherwise
Leaves board (d) Dummy variable equal to one if the director leaves the board of the turnover firm until the next ISS
re-election, zero otherwise
Panel B: Firm characteristics
Variable Definition Source
Total assets Total assets in million USD; 𝐴𝑇 Compustat
Tobin’s Q Tobin’s Q (market value of assets to book value); (𝐴𝑇 + 𝐶 𝑆 𝐻 𝑂 ∗ 𝑃 𝑅𝐶 𝐶_𝐹 − 𝐶 𝐸 𝑄 − 𝑇 𝑋 𝐷𝐵)∕𝐴𝑇 . Compustat
Missing values in 𝑇 𝑋 𝐷𝐵 have been set to 0. Winsorized at the 1st and 99th percentiles
ROA Return on assets; 𝑐 𝑜𝑎𝑙𝑒𝑠𝑐 𝑒(𝑂𝐼 𝐵 𝐷𝑃 , 𝑆 𝐴𝐿𝐸 − 𝑋 𝑂𝑃 𝑅, 𝑅𝐸 𝑉 𝑇 − 𝑋 𝑂𝑃 𝑅)∕((𝐴𝑇 + 𝑙𝑎𝑔(𝐴𝑇 ))∕2). Winsorized at Compustat
the 1st and 99th percentiles
BH return (m270,m21) Buy-and-hold-return from 𝑡 − 270 to 𝑡 − 21 with 𝑡 = 0 being the director re-election date; adjusted for CRSP
equally weighted market return. Winsorized at the 1st and 99th percentiles
Board size Number of directors on the board Boardex
% outside directors The percentage of outside directors, as a fraction of board size BoardEx
% busy outside directors The percentage of outside directors with more than two board memberships, as a fraction of the BoardEx
number of outside directors
Institutional ownership (%) The percentage of shares owned by 13F institutions, as a fraction of shares outstanding Thomson Reuters
≥ (<) median (mean) Dummy variable equal to one if the common ownership between the interlocked firm and the Thomson Reuters
common ownership (d) turnover firm is at least equal to (below) the median (mean), zero otherwise
Panel C: Events at interlocked firms
Variable Definition Source
Forced CEO turnover (d) Dummy variable equal to one if a director was involved in a forced CEO turnover at an interlocked Hand-collected
firm, zero otherwise. A CEO departure is classified as forced if (i) newspaper articles indicate that the (Factiva/Boardex)
CEO is fired, is forced out of her position, or departs due to unspecified policy differences, (ii) the
CEO does not leave to take over an executive position at another organization, and (iii) the CEO
leaves within one month after the departure is announced, which includes the termination of a board
membership but does not include a consulting position at the turnover firm.
Unforced CEO turnover (d) Dummy variable equal to one if a director was involved in a CEO turnover not classified as forced at Hand-collected
an interlocked firm, zero otherwise (Factiva/Boardex)
Hired CEO and left before Dummy variable equal to one if the director hired the CEO who gets fired but left the turnover firm Hand-collected
forced CEO turnover (d) before the firing, zero otherwise (Factiva/Boardex)
Joined after forced CEO Dummy variable equal to one if a director joined the board of the turnover firm after the departure Hand-collected
turnover (d) of the CEO and before the date of her re-election at the interlocked firm, zero otherwise (Factiva/Boardex)
Sudden CEO death (d) Dummy variable equal to one if a director was involved in a CEO turnover at an interlocked firm Hand-collected
caused by the sudden death of the CEO, zero otherwise (Factiva/Boardex)
Sudden CEO death Dummy variable equal to one if a director was involved in a CEO turnover at an interlocked firm Hand-collected
(extended) (d) caused by the CEO’s sudden death, zero otherwise. We also consider cases where CEOs take (Factiva/Boardex)
health-related leaves of absence and die within 30 days
Restatement (d) Dummy variable equal to one if a director was involved in a restatement at an interlocked firm, zero Audit Analytics
otherwise
Class action lawsuit (d) Dummy variable equal to one if a director was involved in a class action lawsuit at an interlocked Securities Class
firm, zero otherwise Action
Clearinghouse
Poison pill (d) Dummy variable equal to one if a director was involved in the adoption of a poison pill at an Shark-Repellent
interlocked firm, zero otherwise
Bankruptcy (d) Dummy variable equal to one if a director was involved in a bankruptcy filing at an interlocked firm, Florida-UCLA-
zero otherwise LoPucki

(continued on next page)

directors experience a significant decline in aggregate total assets decline in total assets represented by the entire portfolio of firms at
represented by their board seats compared to non-turnover-interlocked which turnover-interlocked directors hold board seats.
directors. Hence, board seats lost, in particular at the turnover firm, An even more drastic labor market consequence than losses of
tend to be replaced by board seats at smaller firms, resulting in a directorships or declines in the quality of directorships is a complete

19
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Table A.1 (continued).

Panel D: CEO turnover characteristics
Variable Definition Source
Successor announced (d) Dummy variable equal to one if the company announces a full replacement with the departure of the Hand-collected
outgoing CEO, zero otherwise (Factiva/Boardex)
Performance-induced (d) Dummy variable equal to one if the CEO turnover is performance-induced as estimated using the See Table IA.3
two-probit model by Jenter and Lewellen (2021), zero otherwise
Honeymoon stage (d) Dummy variable equal to one if the outgoing CEO has a tenure of less than three years, zero BoardEx/ISS
otherwise
Harvest stage (d) Dummy variable equal to one if the outgoing CEO has a tenure of between three and 13 years, zero BoardEx/ISS
otherwise
Decline stage (d) Dummy variable equal to one if the outgoing CEO has a tenure of more than 13 years, zero otherwise BoardEx/ISS
Panel E: Director labor market outcomes
Variable Definition Source
Loss of directorship (d) Dummy variable equal to one if the director loses any of her outside directorships within 𝑡 years BoardEx
following the forced CEO turnover at an interlocked firm, zero otherwise
Gain of directorship (d) Dummy variable equal to one if the director obtains a new outside directorship within 𝑡 years BoardEx
following the forced CEO turnover at an interlocked firm, zero otherwise
Net loss of directorship (d) Dummy variable equal to one if the director loses more outside directorships than she gains within 𝑡 BoardEx
years following the forced CEO turnover at an interlocked firm, zero otherwise
Sum of total assets (market Sum of total assets (market capitalization) in million USD over all directorships a director holds; 𝐴𝑇 BoardEx, Compustat
cap.) over all directorships
Exit from labor market (d) Dummy variable equal to one if the director is alive and has no board memberships 𝑡 years following BoardEx
the forced CEO turnover at an interlocked firm, zero otherwise

This table reports variable definitions. Database mnemonics are provided in italics (if available).

exit from the labor market. In our final set of tests, we thus study This paper establishes novel evidence on the reputational effects of
whether directors involved in a forced turnover have a higher propen- forced CEO turnovers on directors involved in the turnover decision.
sity to leave the director labor market post-turnover. To this end, we Directors interlocked to a forced turnover experience an economically
construct a dummy variable equal to one if the total number of board large and statistically significant increase in withheld votes at sub-
seats held declines to zero and the director is still alive, and zero sequent director re-elections, suggesting that directors involved in a
otherwise. Results in Columns 3 and 4 show that the probability of forced turnover lose reputation. Additional tests show that directors are
exiting the director labor market increases significantly following a not penalized for involvement in a CEO turnover per se but for forced
forced turnover. Specifically, directors interlocked to a forced turnover CEO turnovers that are related to governance failures by the board.
are 12.1 (9.2) percentage points more likely to exit the director labor The results presented in this paper challenge the predominant view
market one (five) year(s) after a forced turnover relative to directors on that forcing out a CEO is a sign of well-functioning corporate gov-
the same board not interlocked to a turnover. These findings provide ernance at the board level. Our results support an alternative view:
additional evidence that involvement in a forced CEO turnover has Depending on the timing and circumstances of the turnover, forcing
adverse effects on director labor market opportunities. out a CEO can be perceived as a signal of failure in the monitoring,
In summary, results in this section provide some evidence consistent and in particular the firing of the CEO, and thus be detrimental to a
with directors’ labor market opportunities being reduced following a director’s reputation.
forced CEO turnover and thus support our main finding of a reputa-
tional loss resulting from involvement in a forced turnover. However, it CRediT authorship contribution statement
is important to note that all outcome variables used here may be subject
to endogeneity concerns, in particular, that directors may choose to Felix von Meyerinck: Writing – review & editing, Writing – original
either terminate some board seats or attempt to solicit new ones for draft, Methodology, Formal analysis, Conceptualization. Jonas Romer:
reasons unobservable to the researcher (Ertimur et al., 2012; Levit and Writing – review & editing, Writing – original draft, Methodology, Data
Malenko, 2016). curation, Formal analysis, Conceptualization. Markus Schmid: Writing
– review & editing, Writing – original draft, Supervision, Methodology,
6. Conclusion Conceptualization.

Monitoring and, if necessary, firing the CEO of a corporation is Declaration of competing interest
one of the primary tasks of the board of directors (e.g., Fama, 1980;
Fama and Jensen, 1983; Jenter and Kanaan, 2015; Jenter and Lewellen, The corresponding author serves as a member of the advisory board
2021). Hence, directors’ involvement in forced CEO turnovers can of Inrate AG, Switzerland, a proxy advisory firm and ESG rating agency.
be expected to be assessed by the market and to affect directors’ He has no other conflicts of interest to disclose.
reputation. However, research studying the consequences of forced The first and second author declares that they have no conflicts of
CEO turnovers on involved directors is scarce. Such analysis is em- interest to disclose.
pirically challenging as the turnover decision is endogenous and often
related to company performance (Fee et al., 2013). Moreover, the Appendix A
outcome variables commonly used to measure reputational effects for
directors, such as gains and losses in the number of board mem- See Table A.1.
berships, are subject to endogenous selection by directors (Levit and
Malenko, 2016). We overcome these challenges by using changes in Appendix B. Supplementary data
the percentage of withheld votes in director re-elections as our pri-
mary outcome variable and observing this outcome at interlocked Supplementary material related to this article can be found online
directorships. at https://doi.org/10.1016/j.jfineco.2024.103971.

20
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Data availability Eisfeldt, A.L., Kuhnen, C.M., 2013. CEO turnover in a competitive assignment
framework. J. Financ. Econ. 109 (2), 351–372.
Ellis, J., Guo, L., Mobbs, S., 2021. How does forced-CEO-turnover experience affect
Replication kit (Code and pseudo data).
directors? J. Financ. Quant. Anal. 56 (4), 1163–1191.
Erel, I., Stern, L.H., Tan, C., Weisbach, M.S., 2021. Selecting directors using machine
learning. Rev. Financ. Stud. 34 (7), 3226–3264.
References Ertimur, Y., Ferri, F., Maber, D.A., 2012. Reputation penalties for poor monitoring of
executive pay: Evidence from option backdating. J. Financ. Econ. 104 (1), 118–144.
Adams, R.B., Ferreira, D., 2007. A theory of friendly boards. J. Finance 62 (1), Ertugrul, M., Krishnan, K., 2011. Can CEO dismissals be proactive? J. Corp. Finance
217–250. 17 (1), 134–151.
Adams, R.B., Ferreira, D., 2008. Do director’s perform for pay? J. Account. Econ. 46 Fahlenbrach, R., Low, A., Stulz, R.M., 2017. Do independent director departures predict
(1), 154–171. future bad events? Rev. Financ. Stud. 30 (7), 2313–2358.
Aggarwal, R., Dahiya, S., Prabhala, N.R., 2019. The power of shareholder votes: Falato, A., Kadyrzhanova, D., Lel, U., 2014. Distracted directors: Does board busyness
Evidence from uncontested director elections. J. Financ. Econ. 133 (1), 134–153. hurt shareholder value? J. Financ. Econ. 113 (3), 404–426.
Aghamolla, C., Hashimoto, T., 2021. Aggressive boards and CEO turnover. J. Account. Faleye, O., 2007. Classified boards, firm value, and managerial entrenchment. J. Financ.
Res. 59 (2), 437–486. Econ. 83 (2), 501–529.
Ali, A., Zhang, W., 2015. CEO tenure and earnings management. J. Account. Econ. 59 Fama, E., 1980. Agency problems and the theory of the firm. J. Polit. Econ. 88 (2),
(1), 60–79. 288–307.
Allgood, S., Farrell, K., 2003. The match between CEO and firm. J. Bus. 76 (2), Fama, E.F., Jensen, M.C., 1983. Separation of ownership and control. J. Law Econ. 26
317–341. (2), 301–325.
Angrist, J.D., Pischke, J.-S., 2009. Mostly Harmless Econometrics: An Empiricist’s Farrell, K.A., Whidbee, D.A., 2000. The consequences of forced CEO succession for
Companion. Princeton University Press. outside directors. J. Bus. 73 (4), 597–627.
Appel, I.R., Gormley, T.A., Keim, D.B., 2016. Passive investors, not passive owners. J. Fee, C.E., Hadlock, C.J., Huang, J., Pierce, J.R., 2018. Robust models of CEO turnover:
Financ. Econ. 121 (1), 111–141. New evidence on relative performance evaluation. Rev. Corp. Finance Stud. 7 (1),
Austin, P.C., 2011a. An introduction to propensity score methods for reducing the 70–100.
effects of confounding in observational studies. Multivar. Behav. Res. 46 (3), Fee, C.E., Hadlock, C.J., Pierce, J.R., 2013. Managers with and without style: Evidence
399–424. using exogenous variation. Rev. Financ. Stud. 26 (3), 567–601.
Austin, P.C., 2011b. Optimal caliper widths for propensity-score matching when Fich, E.M., 2005. Are some outside directors better than others? Evidence from director
estimating differences in means and differences in proportions in observational appointments by Fortune 1000 firms. J. Bus. 78 (5), 1943–1971.
studies. Pharm. Statist. 10 (2), 150–161. Fich, E.M., Shivdasani, A., 2007. Financial fraud, director reputation, and shareholder
Azar, J., Schmalz, M.C., Tecu, I., 2018. Anticompetitive effects of common ownership. wealth. J. Financ. Econ. 86 (2), 306–336.
J. Finance 73 (4), 1513–1565. Fischer, P.E., Gramlich, J.D., Miller, B.P., White, H.D., 2009. Investor perceptions of
Bach, L., Metzger, D., 2017. How do shareholder proposals create value? Working board performance: Evidence from uncontested director elections. J. Account. Econ.
paper. Stockholm school of economics & Swedish house of finance. 48 (2–3), 172–189.
Baker, A.C., Larcker, D.F., Wang, C.C., 2022. How much should we trust staggered Foroughi, P., Marcus, A.J., Nguyen, V., Tehranian, H., 2022. Peer effects in corporate
difference-in-differences estimates? J. Financ. Econ. 144 (2), 370–395. governance practices: Evidence from universal demand laws. Rev. Financ. Stud. 35
Bizjak, J., Lemmon, M., Whitby, R., 2009. Option backdating and board interlocks. (1), 132–167.
Rev. Financ. Stud. 22 (11), 4821–4847. Fos, V., Li, K., Tsoutsoura, M., 2018. Do director elections matter? Rev. Financ. Stud.
Borokhovich, K.A., Parrino, R., Trapani, T., 1996. Outside directors and CEO selection. 31 (4), 1499–1531.
J. Financ. Quant. Anal. 31 (3), 337. Garrett, D.F., Pavan, A., 2012. Managerial turnover in a changing world. J. Polit. Econ.
Bouwman, C.H., 2011. Corporate governance propagation through overlapping 120 (5), 879–925.
directors. Rev. Financ. Stud. 24 (7), 2358–2394. Geng, H., Hau, H., Michaely, R., Nguyen, B., 2023. Does Board Overlap Promote
Brav, A., Jiang, W., Partnoy, F., Thomas, R., 2008. Hedge fund activism, corporate Coordination Between Firms? Working paper, University of Geneva.
governance, and firm performance. J. Finance 63 (4), 1729–1775. Gibson, S., Safieddine, A., Sonti, R., 2004. Smart investments by smart money: Evidence
Brick, I.E., Chidambaran, N.K., 2010. Board meetings, committee structure, and firm from seasoned equity offerings. J. Financ. Econ. 72 (3), 581–604.
value. J. Corp. Finance 16 (4), 533–553. Gopalan, R., Gormley, T.A., Kalda, A., 2021. It’s not so bad: Director bankruptcy
Brochet, F., Limbach, P., Schmid, M., Scholz-Daneshgari, M., 2021. CEO tenure and experience and corporate risk taking. J. Financ. Econ. 142, 261–292.
firm value. Acc. Rev. 96, 47–71. Gow, I.D., Wahid, A.S., Yu, G., 2018. Managing reputation: Evidence from biographies
Brochet, F., Srinivasan, S., 2014. Accountability of independent directors: Evidence from of corporate directors. J. Account. Econ. 66 (2–3), 448–469.
firms subject to securities litigation. J. Financ. Econ. 111 (2), 430–449. Guo, L., Masulis, R.W., 2015. Board structure and monitoring: New evidence from CEO
Brown, J.L., Drake, K.D., 2014. Network ties among low-tax firms. Acc. Rev. 89 (2), turnovers. Rev. Financ. Stud. 28 (10), 2770–2811.
483–510. Hambrick, D.C., Fukutomi, G.D., 1991. The seasons of a CEO’s tenure. Acad. Manag.
Burbidge, J.B., Magee, L., Robb, A.L., 1988. Alternative transformations to handle Rev. 16 (4), 719–742.
extreme values of the dependent variable. J. Amer. Statist. Assoc. 83 (401), He, J., Huang, J., Zhao, S., 2019. Internalizing governance externalities: The role of
123–127. institutional cross-ownership. J. Financ. Econ. 134 (2), 400–418.
Cai, J., Garner, J.L., Walkling, R.A., 2009. Electing directors. J. Finance 64 (5), Heider, F., Ljungqvist, A., 2015. As certain as debt and taxes: Estimating the tax
2389–2421. sensitivity of leverage from state tax changes. J. Financ. Econ. 118 (3), 684–712.
Cai, Y., Xu, J., Yang, J., 2021. Paying by donating: Corporate donations affiliated with Henderson, A.D., Miller, D., Hambrick, D.C., 2006. How quickly do CEOs become
independent directors. Rev. Financ. Stud. 34 (2), 618–660. obsolete? Industry dynamism, CEO tenure, and company performance. Strateg.
Callaway, B., Sant’Anna, P.H., 2021. Difference-in-differences with multiple time Manage. J. 27 (5), 447–460.
periods. J. Econometrics 225 (2), 200–230. Huson, M.R., Malatesta, P.H., Parrino, R., 2004. Managerial succession and firm
Chhaochharia, V., Grinstein, Y., 2009. CEO compensation and board structure. J. performance. J. Financ. Econ. 74 (2), 237–275.
Finance 64 (1), 231–261. Huson, M.R., Parrino, R., Starks, L.T., 2001. Internal monitoring mechanisms and CEO
Chiu, P.C., Teoh, S.H., Tian, F., 2013. Board interlocks and earnings management turnover: A long-term perspective. J. Finance 56 (6), 2265–2297.
contagion. Acc. Rev. 88 (3), 915–944. Jensen, M.C., 1993. The modern industrial revolution, exit, and the failure of internal
Coles, J.L., Daniel, N.D., Naveen, L., 2014. Co-opted boards. Rev. Financ. Stud. 27 (6), control systems. J. Finance 48 (3), 831–880.
1751–1796. Jenter, D., Kanaan, F., 2015. CEO turnover and relative performance evaluation. J.
Cvijanović, D., Gantchev, N., Li, R., 2023. CEO succession roulette. Manage. Sci. 69 Finance 70 (5), 2155–2184.
(10), 5794–5815. Jenter, D., Lewellen, K., 2021. Performance-induced CEO turnover. In: Denis, D. (Ed.),
Dalton, D.R., Dalton, C.M., 2007. CEO succession: Some finer - and perhaps provocative Rev. Financ. Stud. 34 (2), 569–617.
- points. J. Bus. Strateg. 28 (3), 6–8. Johnson, W.C., Karpoff, J.M., Wittry, M.D., 2023. The Consequences to Directors for De-
Dasgupta, S., Li, X., Wang, A.Y., 2018. Product market competition shocks, firm ploying Poison Pills. Working paper, Suffolk University, University of Washington,
performance, and forced CEO turnover. Rev. Financ. Stud. 31 (11), 4187–4231. Ohio State University.
Davis, G.F., 1991. Agents without principles? The spread of the poison pill through the Jovanovic, B., 1979. Job matching and the theory of turnover. J. Polit. Econ. 87 (5,
intercorporate network. Adm. Sci. Q. 36 (4), 583. Part 1), 972–990.
Denis, D.J., Denis, D.K., 1995. Performance changes following top management Kempf, E., Manconi, A., Spalt, O., 2017. Distracted shareholders and corporate actions.
dismissals. J. Finance 50 (4), 1029–1057. Rev. Financ. Stud. 30 (5), 1660–1695.
Dow, J., 2013. Boards, CEO entrenchment, and the cost of capital. J. Financ. Econ. Klein, A., 1998. Firm performance and board committee structure. J. Law Econ. 41 (1),
110 (3), 680–695. 275–304.

21
F. von Meyerinck et al. Journal of Financial Economics 163 (2025) 103971

Laux, V., 2010. Effects of litigation risk on board oversight and CEO incentive pay. Nguyen, B.D., Nielsen, K.M., 2010. The value of independent directors: Evidence from
Manage. Sci. 56 (6), 938–948. sudden deaths. J. Financ. Econ. 98 (3), 550–567.
Lechner, M., Strittmatter, A., 2019. Practical procedures to deal with common support Parrino, R., 1997. CEO turnover and outside succession A succession analysis. J. Financ.
problems in matching estimation. Econometric Rev. 38 (2), 193–207. Econ. 46 (2), 165–197.
Levit, D., Malenko, N., 2016. The labor market for directors and externalities in Parrino, R., Sias, R.W., Starks, L.T., 2003. Voting with their feet: Institutional ownership
corporate governance. J. Finance 71 (2), 775–808. changes around forced CEO turnover. J. Financ. Econ. 68 (1), 3–46.
Lewellen, K., Lowry, M., 2021. Does common ownership really increase firm Peters, F.S., Wagner, A.F., 2014. The executive turnover risk premium. J. Finance 69
coordination? J. Financ. Econ. 141 (1), 322–344. (4), 1529–1563.
Liu, C., Low, A., Masulis, R.W., Zhang, L., 2020. Monitoring the monitor: Dis- Ryan, H.E., Wiggins, R.A., 2004. Who is in whose pocket? Director compensation,
tracted institutional investors and board governance. Rev. Financ. Stud. 33 (10), board independence, and barriers to effective monitoring. J. Financ. Econ. 73 (3),
4489–4531. 497–524.
Mace, M.L., 1971. Directors: Myth and reality. Rutgers Law Review. Division of Shivdasani, A., 1993. Board composition, ownership structure, and hostile takeovers.
Research, Graduate School of Business Administration, Harvard University, Boston, J. Account. Econ. 16 (1–3), 167–198.
MA. Srinivasan, S., 2005. Consequences of financial reporting failure for outside directors:
Marcel, J.J., Cowen, A.P., Ballinger, G.A., 2017. Are disruptive CEO successions Evidence from accounting restatements and audit committee members. J. Account.
viewed as a governance lapse? Evidence from board turnover. J. Manage. 43 (5), Res. 43 (2), 291–334.
1313–1334. Stein, L.C., Zhao, H., 2019. Independent executive directors: How distraction affects
Masulis, R.W., Mobbs, S., 2014. Independent director incentives: Where do talented their advisory and monitoring roles. J. Corp. Finance 56, 199–223.
directors spend their limited time and energy? J. Financ. Econ. 111 (2), 406–429. Stuart, T.E., Yim, S., 2010. Board interlocks and the propensity to be targeted in private
Masulis, R.W., Zhang, E.J., 2019. How valuable are independent directors? Evidence equity transactions. J. Financ. Econ. 97 (1), 174–189.
from external distractions. J. Financ. Econ. 132 (3), 226–256. Sun, L., Abraham, S., 2021. Estimating dynamic treatment effects in event studies with
Matvos, G., Ostrovsky, M., 2010. Heterogeneity and peer effects in mutual fund proxy heterogeneous treatment effects. J. Econometrics 225 (2), 175–199.
voting. J. Financ. Econ. 98 (1), 90–112. Weisbach, M.S., 1988. Outside directors and CEO turnover. J. Financ. Econ. 20 (C),
Miller, D., 1991. Stale in the saddle: CEO tenure and the match between organization 431–460.
and environment. Manage. Sci. 37 (1), 34–52. Yermack, D., 2004. Remuneration, retention, and reputation incentives for outside
directors. J. Finance 59 (5), 2281–2308.
Zhang, S., 2021. Directors’s career concerns: Evidence from proxy contests and board
interlocks. J. Financ. Econ. 140 (3), 894–915.

22