TESIS - Mamun Et Al 2020 - Powerful CEO

Journal of Corporate Finance 62 (2020) 101582
Contents lists available at ScienceDirect
Journal of Corporate Finance

journal homepage: www.elsevier.com/locate/jcorpfin
Powerful CEOs and stock price crash risk

T
Md Al Mamuna, Balasingham Balachandrana, , Huu Nhan Duongb
⁎
a
Department of Economics and Finance, La Trobe University, Bundoora Vic 3086, Australia
b
Department of Banking and Finance, Monash University, Clayton, Vic 3800, Australia
ARTICLE INFO ABSTRACT
Keywords: We find that powerful chief executive officers (CEOs) are associated with higher crash risk. The
CEO power positive association between CEO power and crash risk holds when controlling for earnings
Delta management, tax avoidance, chief executive officer's option incentives, and CEO overconfidence.
CEO overconfidence Firms with powerful CEOs have higher probability of financial restatements, lower proportion of
CEO general skills
negative to positive earnings guidance, and lower ratio of negative to positive words in their
Stock price crash risk
Corporate governance
financial statements. The association between powerful CEOs and higher crash risk is mostly
evident among firms with higher sensitivity of CEO wealth to stock prices and when CEOs have
lower general skills. External monitoring mechanisms weaken but do not eliminate the asso-
ciation between powerful founder CEOs and higher crash risk.
1. Introduction
Managers have a strong incentive to withhold bad news from investors (e.g., Graham et al., 2005; Ball, 2009; Kothari et al.,
2009). Once the accumulated bad news reaches an overwhelming level, managers give up and release it all together, leading to a
stock price crash (Jin and Myers, 2006; Hutton et al., 2009). In line with this argument, Hutton et al. (2009) and Kim et al.
(2011a) show that managers use earnings management and tax avoidance to hoard bad news, which, in turn, leads to a stock
price crash. Graham et al. (2005) survey and interview the chief financial officers (CFOs) of US firms and show that managers are
willing to sacrifice economic value to manage financial reporting perceptions. Dichev et al. (2013) survey and interview CFOs of
US firms and find that firms manage earnings to influence stock prices because of outside and inside pressure to hit earnings
benchmarks and to avoid adverse compensation and career consequences for senior executives. Further, these CFOs (Dichev
et al., 2013, p. 30) believe that it is difficult for outside observers to unravel earnings management, especially when such
earnings are managed using subtle unobservable choices or real actions, and advocate paying close attention to the key managers
running the firm.
In the spirit of the findings of Graham et al. (2005) and Dichev et al. (2013), we expand the literature on stock price crashes by
paying close attention to the power dynamics of the key managers of firms, that is, the chief executive officers (CEOs). Prior studies
document that the success of managers in withholding bad news and manipulating performance hinges critically on their power to
influence decisions. For example, Friedman (2014) develops a model to demonstrate that a powerful CEO can pressure a CFO to
report biased performance measures. Feng et al. (2011) provide evidence that powerful CEOs with high equity incentives exert
significant pressure on CFOs to engage in accounting manipulation for firms subject to U.S. Securities and Exchange Commission
(SEC) enforcement actions. Further, Fracassi and Tate (2012) show that powerful CEOs are subject to weaker board monitoring, since
their firms are more likely to appoint directors with ties to the CEOs. Synthesizing this evidence, we argue that power in the hands of
⁎
Corresponding author.
E-mail address: b.balachandran@latrobe.edu.au (B. Balachandran).
https://doi.org/10.1016/j.jcorpfin.2020.101582
Received 9 May 2018; Received in revised form 28 October 2019; Accepted 23 January 2020
Available online 27 January 2020
0929-1199/ © 2020 Elsevier B.V. All rights reserved.
M. Al Mamun, et al. Journal of Corporate Finance 62 (2020) 101582
CEOs together with weaker internal monitoring gives CEOs the means and circumstances to withhold bad news from investors,
resulting in higher crash risk.
An alternate argument is also plausible. Prior studies emphasize that managers may accelerate the disclosure of bad news to
mitigate litigation risk (Kasznik and Lev, 1995; Skinner, 1994, 1997). Moreover, if CEO power emanates from his (her) superior
ability (e.g., Hermalin and Weisbach, 1998), there is little incentive for powerful CEOs to hoard negative news. Founder CEOs are
more likely to pursue an optimal shareholder-value maximizing strategy that results in more investment in research & development;
higher capital expenditure; more focused acquisitions; and higher stock market performance (Fahlenbrach, 2009). Powerful CEOs are
also less likely to be replaced based on their poor firm performance (Graham et al., 2017). Therefore, if hiding bad news is driven by
the agency motives such as personal compensation or job security, powerful CEOs care less about hiding bad news since they can
withstand any consequential pressure should they release any. Supporting these arguments, Jiraporn et al. (2014) show that firms
with more powerful CEOs experience greater corporate transparency, since these CEOs are well insulated and have fewer incentives
to conceal information. Therefore, it is possible that firms with powerful CEOs will have a lower stock price crash risk than firms with
non-powerful CEOs because powerful CEOs are not motivated to hoard bad news; or they deliver superior performance.
We test these competing empirical predictions using a large sample of - publicly listed US firms during 1993–2013. We use various
CEO power measures: CEO and founder (CEOFO); CEO founder and either the chair, the president, or both (CEOFEPCB); and CEO,
president, and chair (CEOPRCH) to examine the role of CEO power in hoarding bad news, which in turn, can lead to a stock price
crash. We use the negative conditional skewness of future firm-specific weekly returns (NCSKEW) and down-to-up volatility (DUVOL)
as our main crash risk measures. We find that CEO power is associated with higher stock price crash risk. The association between
CEO power and higher crash risk holds when controlling for other crash determinants, such as earnings management, tax avoidance,
CFO option incentives, and CEO overconfidence. Our result is also economically significant. We find the average crash risk associated
with firms with powerful CEOs is 11.4% higher than that associated with firms without powerful CEOs using CEOFO and NCSKEW as
CEO power and crash risk measures, respectively.
Our finding is not immune to endogeneity concerns. We therefore use several approaches to address this issue. First, we use an
instrumental variables approach in which we use dead founders and the number of founders who are alive as instrumental variables
to predict the presence of a powerful founder CEO (CEOFO and CEOFEPCB) that is arguably orthogonal to stock price crash risk. We
obtain consistent results on the positive association between powerful CEOs and crash risk. However, our approach to address
potential endogeneity concern using instrumental variable needs to be carefully interpreted as there may still be concern about the
validity of the exclusion restriction.
We also obtain consistent results when we use firm- and year- fixed effect specifications to control for firm-specific time-invariant
omitted variables. We further examine the changes in crash risk following exogenous CEO turnover events. We find that the changes
in industry-adjusted crash risk following exogenous CEO turnover events is significantly higher (lower) when the incoming CEO is
more (less) powerful than the outgoing CEOs compared to the events where there is no change in CEO power. We further use a
propensity score matching approach whereby in each fiscal year, we match each firm with powerful CEO with another firm with
comparable characteristics but does not have a powerful CEO. Our analysis shows that the stock price crash risk is higher for firms
with powerful CEOs than for comparable firms without powerful CEOs, irrespective of the CEO power measure is used. Our main
findings are also robust when we use alternative definitions for both CEO power and crash risk and when we control for various CEO
and firm characteristics.
We perform additional analyses to substantiate the argument that powerful CEOs facilitate bad news hoarding, which eventually
results in higher stock price crash risk. We show that firms with powerful CEOs experience higher probability of financial restate-
ments. These firms also issue less negative earnings guidance compared to positive earnings guidance and use fewer negative words
compared to positive words in their annual report filing. The use of fewer negative earnings guidance or negative words in annual
reports and the higher probability of financial restatements to correct accounting errors or misstatements for firms with powerful
CEOs indicates that these firms hoard bad news. Accordingly, our results imply that powerful CEOs have ability to hide bad news,
which, in turn, leading to positive association between CEO power and crash risk.
We conduct further tests to highlight the motivations for powerful CEOs to hide bad news, which, in turn, lead to positive
association between CEO power and crash risk. First, using the CEO delta, a measure of the sensitivity of CEO wealth to stock price
(Core and Guay, 2002; Coles et al., 2006), we find that the effect of powerful CEOs on crash risk is consistently positive and
significant for firms with a higher CEO delta, while insignificant in firms with a lower CEO delta. These findings indicate that
powerful CEOs affect future stock price crash risk by hiding bad news because inflating the stock price is personally beneficial to
them. Second, we find that the positive effect of CEO power on crash risk is consistently evident for CEOs with lower general skills.
Given that CEOs with higher general skills enjoy greater compensation when a firm is exposed to labor market conditions (Custódio
et al., 2013), and encourage firms to pursue risky innovation opportunities (Custódio et al., 2019), our findings of higher crash risk
for firms with powerful CEOs with lower general skills indicate that they hide bad news when they fail to deliver performance,
potentially due to their compensation or career concerns.
In the final set of analyses, we examine whether external governance mechanisms, such as the market for corporate control or
institutional investors, can curb the positive association between CEO power on crash risk. Employing the takeover index (Cain et al.,
2017) as a proxy for external corporate governance mechanism for takeover threats, we find that takeover threats mitigate crash risk.
We also find some evidence for dedicated institutional ownership (Chen et al., 2007; An and Zhang, 2013) in mitigating crash risk.
However, takeover threats and dedicated ownership do not mitigate the association between powerful founder CEOs and higher crash
risk in both high and low external monitoring environments.
Our paper contributes to the literature in several ways. First, we show that CEO power is associated with higher crash risk. The
2
association between CEO power and higher crash risk holds controlling for other incentives or mechanisms for hoarding bad news
such as financial opacity (Hutton et al., 2009), tax avoidance (Kim et al., 2011a), CFO option incentives (Kim et al., 2011b). We
further support our main story by showing that firms with powerful CEOs have higher probability of financial restatements, lower
proportion of negative to positive earnings guidance, and lower proportion of negative to positive words in their financial statements,
supporting their bad news hoarding behavior. Taken in their entirety, our findings suggest that, to understand determinants firm-
level stock price crash risk, it is important to consider the decision-making power structure of the firm.
Second, our study is related to the literature on the economic consequences of powerful CEOs. Our findings that powerful CEOs
are associated with negative stock price performance complement recent studies on the adverse consequences of CEO power, such as
rigging incentive contracts (Morse et al., 2011), a higher likelihood of committing fraud (Khanna et al., 2015), and lower firm
profitability and value (Bebchuk et al., 2011). We also extend this literature by showing that powerful CEOs with lower general skills
are associated with higher crash risk. This finding suggests that powerful CEOs engage in hiding bad news when they do not have the
necessary general skills, which can be transferable across firms or industries. Therefore, our result also contributes to the current
debate regarding the positive and negative consequences of powerful CEOs.
Finally, our study contributes to the literature on the effectiveness of external monitoring mechanisms in restraining powerful
CEOs. We provide the first empirical finding that the external market for corporate control, as reflected by the takeover index, plays
an important role in reducing crash risk. However, the presence of takeover threats does not mitigate the association between founder
CEOs and higher crash risk. Hence, our findings have significant policy implications on governance regulations to reduce the po-
tential negative consequences of CEO power, particularly for powerful founder CEOs.
Our paper is related to that of Adams et al. (2005), who show that powerful CEOs lead to variations in firm returns and per-
formance. However, our paper differs from theirs in many important respects. First, while Adams et al. (2005) concentrate on the
standard deviation of returns, our explicit concern is extreme negative stock returns. We find that CEO power has significant im-
plications on future extreme negative stock returns but not on extreme positive outcomes (price jumps). Second, they emphasize that
the effect of CEO power on return variability does not depend on the existence of an agency problem. We argue, instead, that the
concentration of decision-making power in the hands of CEOs support their intention and/or incentives to hoard bad news, and the
sudden release of accumulated bad news results in stock price crashes.
The remainder of the paper is structured as follows. Section 2 outlines the hypothesis development. Section 3 presents the
research design, variable measurements, sample selection, and descriptive statistics. Section 4 discusses the baseline empirical results
and results addressing endogeneity issues. Section 5 examines the link between CEO power and bad news hoarding. Section 6
explains the motivations for hiding bad news based on the sensitivity of CEO wealth to stock price and the level of CEO skills. Section
7 tests whether external governance mechanisms mitigate the effect of CEO power on crash risk. Section 8 presents additional
analyses and further robustness tests. Section 9 concludes the paper.
2. Related literature and empirical predictions
The literature on firm-specific determinants of stock price crash risk is built on the agency perspective of the hoarding of bad
news. Management, on average, delays the release of bad news to investors (Kothari et al., 2009) due to concerns on their career
(Baginski et al., 2018) or personal wealth (Andreou et al., 2017). However, when it is impossible for managers to hide bad news, the
sudden release of accumulated bad news leads to a significant decline in stock price or a stock price crash (Jin and Myers, 2006;
Hutton et al., 2009). Prior research shows that discretionary accrual-based earnings management (Hutton et al., 2009), tax avoidance
(Kim et al., 2011a), option incentives for CFOs (Kim et al., 2011b), short interest (Callen and Fang, 2015a), stock liquidity (Chang
et al., 2016), CEO age (Andreou et al., 2017), inefficient governance (Andreou et al., 2016), and CEO overconfidence (Kim et al.,
2016) can lead to a future stock price crash. In contrast, dedicated institutional ownership (An and Zhang, 2013), institutional
ownership by public pension funds (Callen and Fang, 2013), industry-specific auditors (Robin and Zhang, 2015), corporate social
responsibility (Kim et al., 2014), religiosity in the firm headquarters' county (Callen and Fang, 2015b), and accounting conservatism
(Kim and Zhang, 2016) mitigates future stock price crash risk.
Recent studies on CEO power suggest it has a negative impact on profitability and shareholder wealth. Daily and Johnson (1997)
argue that enhanced power provides CEOs with sufficient discretion to pursue objectives inconsistent with shareholder wealth
maximization. Morck et al. (1988) show that the presence of the founding family adversely affects Tobin's Q in older firms, where the
entrepreneurial talent of the founder might be less valuable. This finding supports a negative effect of powerful CEOs on performance
given that the power of a CEO can arise from his/her status as the founder of the firm (Adams et al., 2005). Grinstein and Hribar
(2004) find that CEOs with the power to influence board decisions receive significantly larger bonuses, engage in larger acquisition
deals, and experience more negative price reactions to their acquisition announcements. Morse et al. (2011) find that powerful CEOs
induce boards to shift the weights on performance measures toward better-performing measures, thereby rigging incentive pay.
Khanna et al. (2015) find appointment-based CEO connectedness is positively related to the likelihood of corporate fraud and
negatively related to the detection of fraud. Friedman (2014) demonstrates earnings management does not take place in isolation and
that firms with powerful CEOs can potentially pressure CFOs to develop biased performance measures to augment additional
compensation incentives. Feng et al. (2011) provide evidence that powerful CEOs with high equity incentives exert significant
pressure on CFOs to engage in accounting manipulation for firms subject to SEC enforcement actions, compared to matched firms of
similar size in the same industry not subject to SEC enforcement actions.
The case of WorldCom offers a stylized example of how powerful managers hide bad news from the investors. Records from the US
District Court, Southern District of New York [pp. 9–10, Indictment No. S2 O2 CR 1144 (BSJ)] reveal that.
3
In or about October 2000, rather than disclosing WorldCom's true financial condition and operating performance, BERNARD J.
EBBERS, the CEO, president, and director, and Scott D. Sullivan, the CFO, instructed subordinates, in substance and in part, to
falsely and fraudulently book certain entries in WorldCom's general ledger, which were designed to increase artificially
WorldCom's reported revenue and to decrease artificially WorldCom's reported expenses, resulting in, among other things, arti-
ficially-inflated figures for WorldCom's EPS, EBITDA, and revenue growth rate.
In a detailed press release from WorldCom issued June 25, 2002, the company confessed that its audit committee had uncovered
$3.8 billion in expenses that had been improperly booked as capital expenditures and were not made in accordance with generally
accepted accounting principles.
Synthesizing all of these evidences, we argue that power in the hands of CEOs gives them the means and justification to divert firm
resources for their personal gain and to withhold bad news from investors, which can result in a stock price crash.
A contrasting view is also plausible. That is, CEOs, in general, may disclose bad news as soon as it is available to them to mitigate
litigation risk (Kasznik and Lev, 1995; Skinner, 1994, 1997). CEO power can also emerge from various sources including their level of
ownership, the structural position in the organization, their prestige, and their expertise (Finkelstein, 1992; Daily and Johnson,
1997). If CEO power emerges in the form of ownership which is the case for the founder CEOs, or from the structural position, which
is the case for CEOs with multiple structural positions (i.e. CEO President and Chairman), then there is little incentive to hide bad
news that would translate into future crash risk. This is because, the powerful CEOs are generally more entrenched than nominal
CEOs. Being entrenched, powerful CEOs are more insulated against the risk of removal by the board compared with non-powerful
CEOs. Graham et al. (2017) also document that powerful CEOs are less likely to be replaced conditional on poor firm performance.
This result corroborates the fact that powerful CEOs should care less about hiding bad news, since they can withstand any con-
sequential pressure should they release any. When CEO power arises from unique influence or qualities such as self-concept (Norburn,
1989) or expertise (e.g., Hermalin and Weisbach, 1998) which translate into superior firm performance, there are little incentives to
hide bad news that would lead to a positive association between CEO power and future crash risk.
In addition, if CEO power emerges in the form of founder status, Fahlenbrach (2009) argues that the considerable equity stakes in
the firm and the lower likelihood of being removed encourage founder CEOs to pursue the optimal shareholder-value maximizing
strategy. As a result, firms run by founder CEOs have higher stock market performance; undertake more focused acquisitions; and
invest more in R&D and capital expenditure. These findings imply that crash risk or extreme negative stock market performance will
be less likely for firms with powerful CEOs.
Moreover, the “quiet life” hypothesis suggests that entrenched managers avoid activities involving costly efforts (Bertrand and
Mullainathan, 2003; Zhao and Chen, 2008) to cloud their firms' information environment. Jiraporn et al. (2014) document firms with
more powerful CEOs experience greater corporate transparency, since these CEOs are well insulated and have fewer incentives to
conceal information. This finding is also consistent with that of Armstrong et al. (2012), who document that financial statement
informativeness has increased since the passage of state antitakeover laws, suggesting that, insulated from the takeover market,
managers have less incentive to conceal bad information. Taken together, the incentives to withhold bad news may be less relevant to
powerful CEOs. Thus, CEO power will result in lower stock price crash risk.
Overall, the discussion in this section suggests that the effect of CEO power on crash risk ex-ante is unclear and ultimately is an
empirical matter. Accordingly, we formulate the following null hypothesis:
H: CEO power is not related to stock price crash risk.
3. Research design
In this section, we describe the measures of stock price crash risk and CEO power, describe the sample selection procedure, and
present descriptive statistics on measures of stock price crash risk, CEO power, and other control variables.
3.1. Measures of stock price crash risk
We follow Kim et al. (2011a, 2011b) to calculate firm-specific weekly returns (wj, t) by taking the natural log of one plus the
residual from the following extended market model:
r j, = j + 1j,rm, 2 + 2j,rm, 1 + 3j,rm, + 4j,rm, + 1 + 5j,rm, + 2 + j (1)
where, rj, τ is the return of stock j in week τ; rm, τ is the return on the CRSP value-weighted market index in week τ. We employ two
main measures of crash risk. Our first measure is negative conditional return skewness (NCSKEW). We define NCSKEW as the
negative of the third moments of the firm-specific weekly returns of each firm year divided by the standard deviation of firm-specific
weekly returns raised to the third power. Our second measure of stock price crash risk is down-to-up volatility (DUVOL). This measure
was developed by Chen et al. (2001) and followed by Hutton et al. (2009) and Kim et al. (2011a, 2011b). To calculate DUVOL, we
separate specific weekly returns into down and up weeks. Specifically, down (up) weeks refer to those weeks during which firm-
specific weekly returns are below (above) the annual average weekly return. We calculate DUVOL as the log of the ratio of the
standard deviation of firm-specific down weekly returns to the standard deviation of up weekly returns during the fiscal year. Similar
to Kim et al. (2011b) and Kim and Zhang (2016), we estimate our crash risk measures over a 12-month period starting three months
after the fiscal year-end.
4
3.2. Measures of CEO power
CEO power is multi-dimensional in nature. Finkelstein (1992) identifies four different dimensions of power in top management
teams: structural power, ownership power, prestige power and expert power. The structural power indicates the formal position(s) of
the CEO in the organization. Power, in the context of the ability of individual actors to exert their will, is mainly reflected in structural
power, since it directly provides top management the discretion over the decision making and reporting process. For example, the
dual role of CEO and chair implies CEOs' ability to direct board initiatives and impose their will on creating favorable board meeting
outcomes (Morse et al., 2011). If a CEO is not the chair of the board, the CEO will have less power, since the chair has a greater
influence on most strategic business decisions (Adams et al., 2005). Similarly, a CEO with president role can ensure that board has
limited choice in ensuring an in-training successor to tap if disagreement with the CEO ensues (Morse et al., 2011). Finkelstein (1992)
also acknowledges that structural power is perhaps the most commonly cited type of power. In terms of ownership power, Finkelstein
(1992) argues that founder executives gain power through their long-term interaction with the board and can use their unique
positions in the firm to establish implicit control over the board members. Since, crash risk is related to CEOs ability to hide bad news,
we primarily focus on power emanating from formal structural positions and ownership position of the CEOs. Adams et al. (2005)
also capture CEO power using structural power vis-à-vis the formal titles and positions as well as CEO founder status.
In line with the above arguments, we use CEO and founder (CEOFO) as our first CEO power measure, which takes a value of one if
the CEO is a founder and zero otherwise. Our second CEO power measure is a dummy variable taking a value of one if a firm has a
founder CEO who is either the president or chair or both (CEOFEPCB). This variable allows us to accumulate the most influential
positions in an organization's hierarchy of power, that is, the CEO, the founder, and either the president or chair or both. Our third
measure of CEO power is a dummy variable taking the value of one if the CEO is both the chair and president (CEOPRCH) and zero
otherwise.
3.3. Sample selection and descriptive statistics
We obtain a total of 147,480 firm–year observations for 1993–2013 from the Center for Research in Security Prices (CRSP)
database by applying the following data filtering conditions to construct our crash risk variables: a) a minimum of 26 weeks of stock
return data are available during the fiscal year, b) fiscal year-end prices are $1 or more, and c) firm–year observations have a positive
book value of total assets. To develop our CEO power measures, we use firm–year observations from ExecuComp for the period
1992–2012. We remove financial and regulated utility firms from the sample, as well as firms with less than two firm–year ob-
servations over the sample period. If more than one CEO is reported in a fiscal year, we check whether one of them served as CEO for
the previous or subsequent year. If neither did, we exclude that observation. These screening steps yield a sample of 25,375 firm–year
observations. Our final match of CEO power variables from ExecuComp, crash risk variables from the CRSP, and firm characteristic
variables from Compustat with non-missing values for various control variables generates a final sample of 24,300 firm–year ob-
servations. These correspond to our dependent variable at time t and independent and other control variables at time t - 1. We present
the definitions of the variables in the Appendix.
Table 1 presents the descriptive statistics of CEO power, crash risk measures, and control variables. The CEO power variables and
control variables driving the link between CEO power and future stock price crash risk represent the 1992–2012 period, while the
future stock price crash risk variables represent the 1993–2013 period. The mean values of NCSKEW and DUVOL are 0.1431
and − 0.0062, respectively.1 With regard to powerful CEO measures, 15.19% of our sample CEOs are founder CEOs, 12.21% of our
sample CEOs are the founder and either the president or chair or both, while 20.16% of our sample CEOs are both the chair and
president. With respect to our control variables, the mean DTURNt-1 is 0.0041. The average RETt-1 is −0.1953 and the average
SIGMAt-1 is 0.0534. These summary statistics are comparable to those reported in the literature on the determinants of stock price
crash risk.
3.4. Correlation matrix and univariate test
We present the correlation matrix for stock price crash risk, CEO power, and key control variables in Table A1 of the internet
appendix. As expected, we find that our main crash risk measures NCSKEWt and DUVOLt are positively correlated at the 1% level.
Similarly, our three measures of powerful CEOs (CEOFOt-1, CEOFEPCBt-1 and CEOPRCHt-1) are also highly positively correlated at the
1% level. More importantly, our powerful CEOs measures are positively and significantly related to our main crash risk measures
NCSKEWt and DUVOLt at least at the 5% level. This result provides an early sign of a potential positive link between powerful CEOs
1
The summary statistics for our crash risk measures differ from those of Kim et al. (2016). The mean (median) NCSKEW is 0.1431(0.0845) for our
sample and 0.068 (0.021) for Kim et al. (2016), whereas, the mean (median) DUVOL is −0.0062 (−0.0160) for our sample and 0.027 (0.012) for
Kim et al. (2016). Plausible reasons for the difference are (i) difference in lead and lag terms: Kim et al. (2016) use lead and lag terms for both
market and industry to calculate firm-specific weekly returns, while we use lead and lag terms on value-weighted market return following Kim et al.
(2011a & b); (ii) difference in sample period: Kim et al. (2016) study period is 1993–2010, while our study period is 1993–2013; (iii) difference in
sample selection: we restrict our sample to firms with a minimum $1.0 share price at the end of the fiscal year and firm–year observations have a
positive book value of total assets, Kim et al. (2016) do not discuss whether they impose these restrictions; and (iv) difference in winsorization: we
winsorize our crash risk measures at the 1% level, while no information on winsorizing the variables is available in Kim et al. (2016).
5
Table 1
Descriptive statistics.
Variable N Mean Min. Max SD Med. Skew.
CEO power variables at t - 1

CEOFO 24,300 0.1519 0.0000 1.0000 0.3590 0.0000 1.9393
CEOFEPCB 24,300 0.1221 0.0000 1.0000 0.3274 0.0000 2.3091
CEOPRCH 24,300 0.2016 0.0000 1.0000 0.4012 0.0000 1.4875
NCEOTITLE 24,300 3.4524 1.0000 8.0000 0.8285 3.0000 0.2188
Crash risk variables at t

DUVOL 24,300 −0.0062 −1.0795 0.9734 0.3909 −0.0160 0.1200
NCSKEW 24,300 0.1431 −2.8471 3.0383 1.0309 0.0845 0.2745
CRASHD 24,300 0.2062 0.0000 1.0000 0.4046 0.0000 1.4526
CRCOUNT 24,300 0.2148 0.0000 4.0000 0.4334 0.0000 1.8092
Other variables
ACQt-1 24,300 0.3127 0.0000 1.0000 0.4636 0.0000 0.8079
BIG4t-1 24,300 0.7255 0.0000 1.0000 0.4462 1.0000 −1.0106
BLOCKt-1 24,300 5.4190 0.0000 9.2304 3.0707 6.8864 −1.1321
BOIND t-1 16,334 0.6995 0.0000 1.0000 0.1675 0.7272 −0.9280
BOSIZEt-1 16,334 0.9851 0.0000 2.0413 0.1828 0.9542 1.1865
CEOAGEt-1 24,300 55.0004 29.0000 95.0000 7.6639 55.0000 0.2734
CEOGSK t-1 13,905 0.0048 −1.5043 6.5186 −0.1706 0.9802 1.0671
CEOOWNt-1 24,300 0.0110 0.0000 0.2137 0.0302 0.0012 4.9064
CEOOPINt-1 22,795 0.1349 0.0000 0.7267 0.1446 0.0870 1.8095
CFOOPINt-1 17,926 0.0820 0.0000 0.4601 0.0886 0.0536 1.9750
DELTAt-1 21,528 0.6373 0.0018 9.7543 0.1534 1.4382 4.3030
DFOt-2 24,300 0.3066 0.0000 1.0000 0.4611 0.0000 0.8388
DTURNt-1 24,300 0.0041 −0.2905 0.3089 0.0830 0.0027 0.1158
FAGE t-1 24,300 50.4117 3.0000 285.0000 41.0766 36.0000 1.2375
HOLD100t-1 21,706 0.4419 0.0000 1.0000 0.4966 0.0000 0.2337
INVESTt-1 23,996 0.32071 0.0312 1.9768 0.3108 0.2245 2.857
IODEDt-1 24,300 0.0837 0.0000 0.8129 0.0735 0.0681 1.5624
LATt-2 24,300 7.0702 3.8878 11.2316 1.5867 6.9146 0.3802
LEVt-1 24,300 0.1721 0.0000 0.5991 0.1541 0.1534 0.6740
LMVEt-1 24,300 7.1837 1.6796 11.5374 1.6048 7.0115 0.3680
LFOALIVEt-1 24,300 0.8908 0.0000 8.0000 0.8548 1.0000 1.5706
LFAGE t-1 24,300 3.5174 0.6931 5.6489 0.94303 3.5553 −0.4075
LATt-1 24,300 7.0730 1.2331 13.5896 1.6191 6.9147 0.4278
LCEOPAYt-1 24,300 6.6264 0.0000 11.2635 0.9392 6.6694 −2.2171
TENUREt-1 24,300 0.8112 0.3010 1.5563 0.3278 0.7782 0.1715
MTB t-1 24,300 3.3397 0.4903 20.9090 3.1882 2.3841 3.1055
NETOPOEG t 4420 0.2739 0.0000 5.0000 0.5861 0.0000 3.2948
NETPOW t 14,302 2.8243 0.8448 7.6428 1.2231 2.6198 1.2755
OPAQUEt-1 22,027 0.5026 0.0306 4.1183 0.7401 0.2451 3.2643
PRSHELTt-1 23,555 0.6185 0.0004 0.9993 0.3440 0.74958 −0.29453
R&DINTEN t-1 24,300 0.0333 0.0000 0.2762 0.0557 0.0026 2.1929
RETt-1 24,300 −0.1953 −3.9948 2.3515 0.85660 −0.15029 −0.4192
RESTt 12,147 0.0388 0.0000 1.0000 0.1932 0.0000 4.7723
ROA t 24,300 0.0545 −0.2644 0.2625 0.0947 0.0585 −0.8825
SIGMAt-1 24,300 0.0534 0.0162 0.1877 0.0265 0.0474 1.4744
TOINDt-1 22,965 0.1652 0.0279 0.4302 0.0976 0.1380 0.7465
Z-SCORE t-1 23,571 1.9773 −3.5876 5.3281 1.3735 2.0094 −0.7664
This table presents summary statistics for measures of stock price crash, CEO power, and other control variables. The crash variables cover the
period 1993–2013, while the control variables cover the period 1992–2012. All the variables, except the dummy variables, are winsorized at 1%. We
provide variable definitions in the Appendix.
and crash risk. We further conduct a univariate test to assess the initial indication of a possible link between crash risk and powerful
CEOs and present the results in the Table A2 of the internet appendix. The results presented in the Table A2 suggest that the average
values of crash risk measures are significantly higher for firms with powerful CEOs than those for firms without powerful CEOs.
Overall, our result in this section provides initial support of a positive relation between CEO power and stock price crash risk.
4. Empirical results: CEO power and stock price crash risk
4.1. Baseline results
To investigate the relation between CEO power and future stock price crash risk, we employ the following regression model:
CRit = 0 + 1 CEOPOWERi, t 1 + Control Variablesi, t 1 + i, t (2)
6
where, CRit is the stock price crash risk for firm i in year t and CEOPOWER is a dummy variable that takes the value of one for firms
with powerful CEOs and zero otherwise. We use NCSKEW and DUVOL as the main crash variables and CEOFO, CEOFEPCB, and
CEOPRCH as the main CEO power variables. Consistent with Kim et al. (2011a, 2011b), we use firm and year clustering to correct for
standard errors to alleviate concerns about potential time-scale and cross-sectional dependence in the panel. We also control for year
and industry (Fama and French's 49-industry classification) fixed effects in all models.
Following Chen et al. (2001), Hutton et al. (2009), and Kim et al. (2011a, 2011b), we include several control variables: DTURNt-1,
SIGMAt-1, RETt-1, LMVEt-1, MTBt-1, LEVt-1, and ROAt. DTURNt-1 is the detrended average monthly stock turnover in year t - 1, which
captures differences of opinion among investors; SIGMAt-1 is the standard deviation of weekly stock returns over the fiscal year t - 1;
RETt-1 is the average firm-specific weekly return over the fiscal year t - 1; LMVEt-1 is the log of the market value of equity; MTBt-1, a
proxy for growth, is measured as the market value of equity divided by the book value of equity; LEVt-1 is a ratio of long-term debt to
total assets; and ROAt is income before extraordinary items to total assets in year t.2
Since our focus is on the impact of powerful CEOs on stock price crash risk, we also use CEO tenure (measured as the log of tenure,
TENURE), CEO tenure squared (TENURESQ), CEO ownership (CEOOWN), and CEO ownership squared (CEOOWNSQ) as additional
control variables. Ali and Zhang (2015) show that earnings overstatement is more prevalent in the early years than in the later years
of a CEO's service. The authors also find that earnings overstatement is greater in the final year of a CEO's service after controlling for
earnings overstatement in the CEO's early years. Adams et al. (2005) suggest that CEOOWN is a proxy for managerial risk-taking
attitude, while Lilenfeld-Toal and Ruenzi (2014) show that firms with high CEO ownership have higher stock market returns than
firms with low ownership. Hence, we control for CEO ownership (CEOOWN) to better isolate the role of CEO power. Following Adams
et al. (2005), we also use ownership squared (CEOOWNSQ) to allow for a possible nonlinear relation.
Following Hutton et al. (2009), we use OPAQUE as a proxy for earnings management. Since the authors find a concave relation
between opacity and stock price crashes, we use OPAQUE and OPAQUESQ in our regressions. Following Kim et al. (2011a), we use
PRSHELT as a proxy for tax sheltering activities. The variable PRSHELT is the estimated probability of engaging in tax sheltering
activities based on Wilson's (2009). We use model 1 of Table 4 of Lisowsky (2010) to calculate PRSHELT. We follow Kim et al.
(2011b) to calculate a CFO's option incentives (CFOPOIN). We follow Campbell et al. (2011) to construct our CEO overconfidence
measure. We classify CEOs as overconfident if they hold options at least twice during the sample period that is more than 100% in the
money. Our measure of CEO overconfidence, HOLD100, takes the value of one if a CEO holds options that are more than 100% in the
money for at least two years and zero otherwise.
Table 2 present our main regression results using NCSKEW and DUVOL, respectively, as our crash variables. As can be seen in Table 2,
CEO power is significantly and positively related to stock price crash risk, irrespective of the CEO power and crash risk measures used. Our
finding is also economically significant. For example, the results in Model (2) in Table 2 show that the magnitude of the estimated
coefficient of CEOCFO on NCSKEWt is 0.114, which indicates that the average NCSKEWt associated firms with CEOCFOs will be 11.4%
higher than that associated with firms without CEOCFOs. This is equivalent to 79.66% (0.114/0.1431) of the mean value of NCSKEW.
We further document a concave relation between earnings management and stock price crash risk, as documented by Hutton et al.
(2009). Stock price crash risk is also positively related to tax avoidance (Kim et al., 2011a); CFO option incentives (Kim et al., 2011b);
and CEO overconfidence (Kim et al., 2016). More importantly, we show that the relation between CEO power and stock price crash
risk still holds after controlling for the effect of earnings management, tax avoidance, CFO option incentives, and CEO overconfidence
on crash risk. This finding implies that the association between powerful CEOs and higher stock price crash risk is independent and
incremental to the role of other determinants of stock price crash risk.
4.2. Addressing endogeneity concerns
In this section, we address potential endogeneity concerns regarding the relation between CEO power and stock price crash risk.
We undertake several approaches in this regard. Following Adams et al. (2005), we use the instrumental variable approach to isolate
the effect of CEO power on crash risk from other sources of variation in Section 4.2.1. The relation observed between CEO power and
stock price crashes could be driven by the presence of time-invariant firm-specific omitted variables. We address the omitted variable
bias by performing the firm and year fixed effects regression in Section 4.2.2. We further follow the difference-in-difference approach
of Bernile et al. (2017) to investigate the changes in crash risk following exogenous CEO turnover events in Section 4.2.3. Exogenous
CEO turnover events, such as the sudden death of a CEO, are arguably not related to the firm or macroeconomic conditions, and thus
help mitigate the potential issue that arises from endogenous CEO-firm matching. Finally, we perform additional tests utilizing
propensity score matching (PSM) analysis in Section 4.2.4.
4.2.1. Instrumental variable regression

Similar to Adams et al. (2005), we use two instruments for our power measures related to the founder: CEOFO and CEOFEPCB in the
instrumental variable regression analysis. We use the dead founder as our first instrument. In constructing the dead founder variable
(DFO), we manually collect the data from various sources.3 The dummy variable DFO takes the value of one if the founder died at year t-
2 to predict the CEO power at year t-1. For firms with multiple founders, we consider the death of the last founder. The death of a
2
Following Kim et al. (2011a) we use ROAt. We obtain similar results when we use ROAt-1 instead of ROAt.
3
We use google.com, ExecuComp, fundinguniverse.com, bloomberg.com, company websites, businessweek.com, and the SEC's Electronic Data
Gathering, Analysis, and Retrieval system (EDGAR) database to cross-check and construct the dead founder and the number of founders alive.
7
Table 2
CEO power and stock price crash risk: baseline results.
Power measure CEOFO CEOFEPCB CEOPRCH
NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
a a b a a a a a a c a
CEOPOWERt−1 0.099 0.114 0.022 0.029 0.142 0.126 0.035 0.034 0.070 0.031 0.028 0.018b
(4.08) (3.75) (2.45) (2.72) (4.92) (4.13) (3.56) (3.02) (3.58) (1.70) (3.92) (2.43)
NCSKEW t−1 0.055a 0.055a 0.020a 0.021a 0.055a 0.055a 0.020a 0.021a 0.054a 0.055a 0.020a 0.021a
(6.97) (7.01) (6.39) (6.64) (6.95) (6.99) (6.37) (6.63) (7.00) (7.04) (6.38) (6.67)
SIGMA t-1 -1.380b -1.412c −0.539a −0.532b −1.380b −1.413c −0.543a −0.533b −1.210c −1.260c −0.495b −0.488c
(−2.14) (−1.95) (−2.67) (−2.04) (−2.12) (−1.95) (−2.67) (−2.05) (−1.83) (−1.72) (−2.43) (−1.87)
RET t−1 0.138a 0.154a 0.052a 0.057a 0.138a 0.154a 0.052a 0.057a 0.137a 0.154a 0.052a 0.057a
(7.23) (6.95) (7.15) (7.17) (7.28) (6.95) (7.17) (7.18) (7.11) (6.83) (7.10) (7.10)
ROA t−1 -1.274a -1.361a −0.159a −0.206a −1.272a −1.359a −0.159a −0.205a −1.251a −1.340a −0.153a −0.199a
(−14.01) (−10.08) (−3.44) (−3.27) (−13.98) (−10.05) (−3.44) (−3.27) (−13.49) (−9.70) (−3.26) (−3.12)
DTURN t-1 0.227b 0.203c 0.107a 0.103a 0.230b 0.202c 0.108a 0.103a 0.216b 0.194c 0.104a 0.100a
(2.42) (1.79) (3.44) (2.67) (2.43) (1.78) (3.44) (2.67) (2.30) (1.72) (3.31) (2.59)
LEV t-1 −0.336a −0.359a −0.068a −0.077a −0.337a −0.361a −0.068a −0.077a −0.349a −0.370a −0.072a −0.080a
(−5.13) (−4.44) (−3.38) (−3.32) (−5.09) (−4.45) (−3.36) (−3.34) (−5.32) (−4.59) (−3.58) (−3.52)
LMVE t-1 0.056a 0.042a 0.017a 0.015a 0.056a 0.042a 0.017a 0.015a 0.055a 0.040a 0.017a 0.015a
(7.02) (4.92) (6.88) (5.02) (7.04) (4.99) (6.89) (5.06) (6.90) (4.75) (6.74) (4.94)
MTB t-1 0.014a 0.013a 0.001 0.002 0.014a 0.013a 0.001 0.002 0.014a 0.013a 0.001 0.002
(4.76) (3.33) (1.17) (1.11) (4.78) (3.34) (1.18) (1.11) (4.82) (3.33) (1.23) (1.13)
TENURE t-1 0.094 −0.076 0.056 −0.012 0.104 −0.081 0.060 −0.013 0.027 −0.145 0.037 −0.033
(0.69) (−0.47) (1.12) (−0.20) (0.77) (−0.51) (1.21) (−0.22) (0.19) (−0.91) (0.73) (−0.57)
TENURESQ t-1 −0.040 0.049 −0.028 0.009 −0.050 0.051 −0.031 0.009 0.010 0.105 −0.015 0.025
(−0.50) (0.51) (−0.95) (0.27) (−0.64) (0.55) (−1.09) (0.28) (0.13) (1.12) (−0.49) (0.75)
CEOOWN t-1 0.495 −0.846 0.098 −0.286 0.412 −0.875 0.072 −0.295 0.620 −0.671 0.116 −0.244
(0.61) (−1.39) (0.33) (−1.27) (0.51) (−1.45) (0.25) (−1.31) (0.77) (−1.08) (0.39) (−1.05)
CEOOWNSQ t-1 −2.517 4.022 −0.684 1.217 −2.263 4.154 −0.616 1.252 −2.592 3.951 −0.685 1.183
(−0.60) (1.24) (−0.44) (0.94) (−0.54) (1.28) (−0.39) (0.97) (−0.62) (1.19) (−0.44) (0.89)
OPAQUE t-1 0.142a 0.139a 0.034b 0.032c 0.142a 0.139a 0.034b 0.032c 0.144a 0.144a 0.034b 0.033c
(4.54) (3.33) (2.51) (1.84) (4.52) (3.32) (2.50) (1.84) (4.55) (3.39) (2.51) (1.88)
OPAQUESQ t-1 −0.033a −0.032a −0.009a −0.008c −0.033a −0.032a −0.009a −0.008c −0.034a −0.033a −0.009a −0.009b
(−4.24) (−3.19) (−2.59) (−1.95) (−4.22) (−3.19) (−2.58) (−1.95) (−4.27) (−3.27) (−2.60) (−1.99)
PRSHELT t-1 0.057a 0.016b 0.057a 0.016b 0.055a 0.016b
(2.92) (2.25) (2.91) (2.24) (2.83) (2.17)
CFOOPIN t-1 0.373a 0.083b 0.373a 0.082b 0.385a 0.084b
(2.61) (1.97) (2.60) (1.96) (2.69) (2.02)
HOLD100 t-1 0.059a 0.016a 0.060a 0.016a 0.061a 0.016a
(4.53) (2.69) (4.60) (2.73) (4.73) (2.78)
Adjusted R2 0.035 0.037 0.035 0.039 0.035 0.037 0.036 0.039 0.034 0.036 0.036 0.039
Constant Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Yr. & Ind. effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
N 22,027 16,393 22,027 16,393 22,027 16,393 22,027 16,393 22,027 16,393 22,027 16,393
This table presents the results on the impact of CEO power on stock price crashes. The dependent variable is the negative conditional skewness of
future firm-specific weekly returns (NCSKEW) and down-to-up volatility (DUVOL) in year t. The main independent variable is CEO power, CEO-
POWER, using various measures, such as CEOFO (CEO and founder), CEOFEPCB (CEO, founder, and either the president, the chair, or both), and
CEOPRCH (CEO, president, and chair) for year t - 1. We cluster standard errors by both firm and year. We present t-values in parentheses. Industry
and year fixed effects are included in all the regressions. The 1%, 5%, and 10% significance levels of the coefficients are denoted by a, b, and c,
respectively. All continuous variables are winsorized at the 1% level. We present the variable definitions in the Appendix.
founder CEO is a fairly exogenous event that should negatively affect the likelihood of a current CEO being the founder (CEOFO) or the
current CEO being the founder and either the chair, the president, or both (CEOFEPCB) and is arguably less likely to have an effect on
stock price crashes.4 We use the log of one plus the number of founders alive (LFOALIVE) as the second instrument. The number of
founders alive increases the probability that a CEO is a founder and is unlikely to have any impact on firm crash risk.5
4
We do acknowledge that the death of a founder CEO may affect future firm performance, which will be associated with stock market perfor-
mance and crash risk. To mitigate this potential effect, similar to Adams et al. (2005), we also consider the DFO as a dummy variable taking the
value of one if the death of the founder takes place before the firm enters into our sample; otherwise zero. We also use the death of a founder in year
t-5 as an instrument for CEOFO or CEOFEPCB in year t-1 to reduce the possibility that death of the founder may affect firm performance immediately
and thus in turn affect stock price crash risk. Using these alternative definitions of DFO, we find results consistent to those we report in this paper.
5
One may argue that dead founders and number of alive founders may still be correlated with stock price crash risk through firm age and
investment decision. Following the suggestion of Adams et al. (2009), we mitigate this concern by controlling for investment and firm age in our
robustness check (Panel C of Table 9) and in IV regression analysis (Table A3 in the internet appendix).
8
Given the binary nature of the endogenous variable (CEOFO or CEOFEPCB), we follow Adams et al. (2009) and use the following
the procedure to address the endogeneity concern: (i) we estimate a probit model of the determinants of CEO power status using the
instruments (DFO and LFOALIVE) and other relevant control variables; then, (ii) we estimate the crash risk model using the predicted
probability of CEOFO or CEOFEPCB obtained in step (i) as an instrument for powerful CEOs in the instrumental variable (IV) re-
gression. This approach does not require the probit model to be correctly specified (Adams et al., 2009). Moreover, the standard
errors from standard IV regression are still asymptotically valid (Wooldridge, 2010). This approach also overcomes the problem of the
binary nature of the endogenous variable that standard IV regression fails to address.
Following Fahlenbrach (2009), we use several control variables such as CEO age (CEOAGE), CEO ownership (CEOOWN), the log
of firm age (LFAGE), the log of total assets (LAT), dummy variables for a company listed on the American Stock Exchange (DAMEX)
and NASDAQ (DNASDAQ), a Delaware dummy (DDELAWARE), and a Standard & Poor's (S&P) 500 dummy (DS&P500) in our first-
stage regression in probit model of the determinants of CEO power status. We also control for CEO pay using the log of CEO pay
(LCEOPAY). We also use year and industry fixed effects. Panel A of Table 3 presents the results of the determinants for CEOFO and
CEOFEPCB. We use CEOFO and CEOFEPCB in year t - 1 as a dependent variable. The control variables are based on year t - 2.
The estimated coefficient of the dead founder variable is significantly negative for both CEO power measures: CEOFO and
CEOFEPCB. The coefficient of the number of founders alive is significantly positive for CEOFO and positive but insignificant for
CEOFEPCB. The coefficients of the other control variables are consistent with those of Fahlenbrach (2009). The Wald F-statistics are
significant at 1% level for both CEO power measures CEOFO and CEOFEPCB in models 1 and 2, respectively. Thus, we reject the null
hypothesis that all coefficients in the determinants of CEOFO and CEOFEPCB are zero.
Panel B of Table 3 presents the second-stage regression results. We observe that the association between CEO power and crash risk
is positive and significant across alternative measures of stock price crash risk. We also conduct several tests for checking the validity
and strength of the instrument. We test if CEO power variable is endogenous so that we really need an instrumental variable
approach. The augmented Durbin-Wu-Hausman test statistics range from 10.24 to 45.52 and are significant at 1% level for both crash
risk models, highlighting that CEOFO and CEOFEPCB have endogenous relation with crash risk, hence, supports the use of instru-
mental variable approach. The χ2 values for Anderson canonical correlations test are significant at the 1% level; thus, we reject the
null hypothesis that our models are under-identified. The Cragg-Donald Wald F statistic also shows that the instruments used in the
first stage are valid instruments, under the Stock and Yogo (2005) critical values.
4.2.2. Firm fixed effect regression

Another potential issue is the omitted variable bias whereby the relation observed between CEO power and stock price crashes
could be driven by the presence of time-invariant firm-specific omitted variables. We address this concern by performing the baseline
regression model using firm and year fixed effects specification. We show in Panel C of Table 3 that controlling for firm-specific
omitted variable bias, CEO power variables are still positively related to stock price crashes. Hence, we conclude that our main
findings on the positive relation between CEO power and crash risk are not driven by firm-specific omitted variable.
4.2.3. CEO power, exogenous CEO turnover events and stock price crash risk
We further examine the changes in crash risk following exogenous CEO turnovers. Similar to Bernile et al. (2017), we use the
exogenous CEO turnover events as classified by Eisfeldt and Kuhnen (2013).6 We use the data from Eisfeldt and Kuhnen (2013) for
the period 1992–2005. Since their dataset ends in 2005, we manually search in Factiva for CEO sudden death events to extend the
sample of exogenous CEO turnover events for the rest of our sample (for the period of 2006–2012). We follow Nguyen and Nielsen
(2010) to classify a sudden death as an unexpected and non-traumatic death that occurs instantaneously or within few hours of an
abrupt change in the person's previous clinical state. We include accidental and traumatic deaths that are unanticipated by investors
and unrelated to firm conditions.7
We identify 817 CEO turnover events as exogenous CEO turnover during our sample period. Out of 817 events, we could match
464 exogenous CEO turnover events with our sample. We then consider firms with single turnover event with at least a minimum 2
firm-year continuous observations prior to and after the events. These cleaning criteria result in a sample of 316 exogenous CEO
turnover events for our analysis. We document a total of 283, 284, and 221 exogenous turnover events that did not result in any
change in CEO power based on CEO power measures CEOFO, CEOFEPCB and CEOPRCH, respectively. On the other hand, we have 33,
32 and 95 exogenous CEO turnover events that result in a change in CEO power with at least two years of observation before or after
the events based on CEO power measures CEOFO, CEOFEPCB and CEOPRCH, respectively. We find 13, 12 and 56 exogenous CEO
turnover events that result in increase in CEO power, while, 20, 20 and 39 exogenous CEO turnover events that result in decrease in
CEO power based on CEO power measures CEOFO, CEOFEPCB and CEOPRCH, respectively.
We conduct a univariate test to examine the changes in crash risk following exogenous CEO turnover events. We compare the two
treatment groups (firms with increases or decreases in CEO power measured by CEOFO, CEOFEPCB, and CEOPRCH) against a control
group (firms with no change in CEO power).8 For example, when we use CEOFO as our CEO power measure, we compare firms that
6
We thank Camelia Kuhnen for providing the data on her website at http://public.kenan-flagler.unc.edu/faculty/kuhnenc/research/research.
html.
7
The detailed description of the steps to collect data on sudden death events is provided in Section 3.1 of Nguyen and Nielsen (2010).
8
We thank the anonymous referee's suggestion to construct two treatment groups of firms with increase and decrease in power against a control
group of firms with no change in power following turnover events. Moreover, following Bernile et al. (2017), we also conduct similar test to compare
9
Table 3
Endogeneity.
Panel A: The determinants of CEO power arising from founder status
Power measures CEOFOt-1 CEOFEPCBt-1
(1) (2)
a
DFOt-2 −5.135 −5.749a
(−4.96) (−5.60)
LFOALIVEt-2 0.387c 0.197
(1.86) (1.00)
LFAGE t-2 0.387a 0.261a
(5.26) (3.45)
LATt-2 −21.326 −35.104b
(−1.20) (−2.02)
CEOOWNt-2 −0.747a −0.739a
(−9.13) (−9.11)
CEOOWNSQ t-2 −0.146a −0.153a
(−2.96) (−3.18)
TENURE t-2 10.322a 12.506a
(2.85) (3.48)
TENURESQt-2 −0.738b −0.159
(−2.46) (−0.49)
CEOAGE t-2 0.005 0.009
(0.75) (1.21)
DDELWARE t-1 −0.066 −0.043
(−0.52) (−0.35)
DS&P500 t-1 0.083 0.708a
(0.96) (7.29)
DAMEX t-1 0.444 0.230
(1.33) (0.60)
DNASDAQ t-1 0.380a 0.303b
(2.74) (2.15)
LCEOPAY t-2 −0.012 0.080
(−0.17) (1.18)
2
Pseudo-R 0.284 0.271
Constant, Yr. & Ind. effects Yes YES
N 23,873 23,743
Wald test: all coef. =0 504.55a 672.28a
Panel B: CEO power and crash risk - second stage IV regression results
Power measures ECEOFO ECEOFEPCB
Variables NCSKEWt DUVOLt NCSKEWt DUVOLt
(1) (2) (3) (4) (5) (6) (7) (8)
a a a a a a a
ECEOPOWER t-1 0.419 0.475 0.094 0.107 0.557 0.544 0.124 0.120a
(6.59) (6.25) (3.92) (3.77) (7.83) (6.47) (4.62) (3.83)
OPAQUE t-1 0.139a 0.134a 0.035a 0.033b 0.140a 0.133a 0.035a 0.033b
(4.32) (3.62) (2.87) (2.41) (4.33) (3.58) (2.89) (2.35)
OPAQUESQ t-1 −0.032a −0.031a −0.009a −0.009b −0.032a −0.031a −0.009a −0.009b
(−3.81) (−3.24) (−2.89) (−2.45) (−3.86) (−3.22) (−2.93) (−2.40)
PRSHELT t-1 0.063b 0.019b 0.063b 0.018b
(2.53) (2.03) (2.52) (1.97)
CFOOPIN t-1 0.308a 0.067 0.306a 0.068
(2.62) (1.52) (2.60) (1.53)
HOLD100 t-1 0.051a 0.015b 0.053a 0.016b
(2.77) (2.13) (2.88) (2.27)
Adjusted R2 0.024 0.025 0.032 0.033 0.020 0.021 0.031 0.034
Const. & all controls Yes Yes Yes Yes Yes Yes Yes Yes
Yr. & Ind. effects Yes Yes Yes Yes Yes Yes Yes Yes
N 21,649 16,117 21,649 16,117 21,521 16,020 21,521 16,020
Instrument diagnostics tests:
Test of endogeneity:
Durbin-Wu-Hausman stats 28.91a 10.24a 45.52a 15.64a
Underidentification test (Anderson - LM statistic): 2431.64a 2431.64a 2176.97a 2176.97a
Weak identification test: (Cragg-Donald Wald F statistic) 2729.97 2729.97 2413.76 2413.76
(continued on next page)
10
Table 3 (continued)
Panel C: Impact of CEO power on stock price crash risk, controlling for firm fixed effects
CEO power CEOFO CEOFEPCB CEOPRCH
Variables NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt
(1) (2) (3) (4) (5) (6)
CEOPOWERt-1 0.139b 0.057b 0.151b 0.061a 0.101a 0.039a

(2.20) (2.42) (2.51) (2.76) (3.34) (3.43)
OPAQUE t-1 0.077 0.014 0.077 0.015 0.073 0.013
(1.58) (0.79) (1.58) (0.79) (1.50) (0.71)
OPAQUESQ t-1 −0.015 −0.003 −0.015 −0.003 −0.014 −0.003
(−1.20) (−0.70) (−1.21) (−0.70) (−1.15) (−0.64)
PRSHELT t-1 0.073c 0.024 0.074c 0.024 0.069c 0.022
(1.76) (1.53) (1.78) (1.55) (1.67) (1.43)
CFOOPIN t-1 0.113 0.058 0.112 0.058 0.093 0.051
(0.60) (0.85) (0.60) (0.85) (0.50) (0.74)
HOLD100 t-1 0.017 0.014 0.017 0.014 0.014 0.013
(0.51) (1.16) (0.51) (1.16) (0.41) (1.06)
Adjusted R2 0.079 0.069 0.079 0.069 0.079 0.069
Const. & all controls Yes Yes Yes Yes Yes Yes
Yr. & Firm effects Yes Yes Yes Yes Yes Yes
N 16,393 16,393 16,393 16,393 16,393 16,393
Panel D: Crash risk around CEO exogenous turnover events
(1) (2) (3) (4) (5) (6) (7)
Variables No change in power Increase in power Decrease in power Diff. (2–1) t-stats Diff. (3–1) t-stats
CEOFO Mean Mean
Δ Industry-Adjusted NCSKEWt 0.073 0.593 −0.674 0.520 2.140b −0.746 −3.967a

Δ Industry-Adjusted DUVOLt 0.033 0.189 −0.241 0.156 1.743 c −0.273 −3.881a
N 283 13 20
CEOFEPCB
Δ Industry-Adjusted NCSKEWt 0.067 0.593 −0.597 0.526 2.152b −0.664 −3.527a
Δ Industry-Adjusted DUVOLt 0.030 0.189 −0.200 0.159 1.769c −0.230 −3.256a
N 284 12 20
CEOPRCH
Δ Industry-Adjusted NCSKEWt 0.034 0.404 −0.417 0.370 3.070a −0.451 −3.182a
Δ Industry-Adjusted DUVOLt 0.007 0.161 −0.109 0.153 3.362a −0.116 −2.021b
N 221 56 39
Panel E: CEO power and crash risk around CEO exogenous turnover increasing CEO power
Power measures CEOFO CEOFEPCB CEOPRCH
Variables (1) (4) (5) (2) (3) (6)
CHANGEt-1 0.064 0.023 0.002 0.031 −0.002 0.003

(0.67) (0.66) (0.03) (0.86) (−0.03) (0.09)
INCREASEt-1 −0.613b −0.172 −0.442 −0.154 −0.052 −0.020
(−2.35) (−1.42) (−1.35) (−1.14) (−0.39) (−0.40)
CHANGE*INCREASE t-1 0.935a 0.278b 0.806b 0.293c 0.328c 0.158b
(2.82) (2.04) (2.08) (1.96) (1.80) (2.47)
OPAQUEt-1 0.347c 0.052 0.216 0.037 0.392b 0.096
(1.90) (0.78) (1.20) (0.55) (2.04) (1.32)
OPAQUESQ t-1 −0.077c −0.010 −0.046 −0.004 −0.068 −0.013
(−1.73) (−0.62) (−1.05) (−0.24) (−1.43) (−0.71)
PRSHELT t-1 0.033 −0.009 0.068 0.004 0.023 0.003
(0.30) (−0.22) (0.61) (0.11) (0.20) (0.08)
CFOOPIN t-1 1.194a 0.291c 1.001b 0.295c 0.749c 0.310c
(2.73) (1.70) (2.28) (1.70) (1.82) (1.86)
HOLD100 t-1 0.051 0.019 0.030 0.014 0.057 0.021
(0.70) (0.70) (0.40) (0.53) (0.73) (0.74)
Adjusted R2 0.044 0.043 0.032 0.048 0.044 0.061
(continued on next page)
11
Table 3 (continued)
Panel E: CEO power and crash risk around CEO exogenous turnover increasing CEO power
Variables (1) (4) (5) (2) (3) (6)
Const. & All controls Yes Yes Yes Yes Yes Yes
Year & Ind. effects Yes Yes Yes Yes Yes Yes
N 1049 1049 1047 1047 982 982
Panel F: CEO power and crash risk around CEO exogenous turnover decreasing CEO power
Variables (1) (4) (5) (2) (3) (6)
CHANGE t-1 0.031 0.030 0.044 0.034 0.065 0.018

(0.34) (0.87) (0.49) (0.96) (0.65) (0.48)
DECREASE t-1 0.273c 0.085c 0.301b 0.072c 0.204c 0.030
(1.96) (1.83) (2.20) (1.72) (1.71) (0.62)
CHANGE*DECREASEt-1 −0.647a −0.261a −0.530b −0.188b −0.501a −0.134b
(−3.16) (−3.56) (−2.48) (−2.32) (−2.85) (−2.01)
OPAQUEt-1 0.381b 0.107 0.394b 0.111 0.555a 0.181b
(2.15) (1.61) (2.19) (1.64) (2.64) (2.32)
OPAQUESQ t-1 −0.088c −0.023 −0.089c −0.023 −0.153a −0.046b
(−1.93) (−1.30) (−1.90) (−1.28) (−2.74) (−2.21)
PRSHELT t-1 0.019 −0.003 0.006 −0.014 0.013 −0.021
(0.19) (−0.09) (0.06) (−0.37) (0.12) (−0.50)
CFOOPIN t-1 0.876b 0.275c 0.870b 0.261 1.078b 0.356c
(2.13) (1.75) (2.09) (1.62) (2.34) (1.96)
HOLD100 t-1 0.041 0.021 0.037 0.020 −0.004 0.004
(0.60) (0.83) (0.54) (0.76) (−0.05) (0.15)
Adjusted R2 0.057 0.067 0.055 0.062 0.034 0.042
Const. & All controls Yes Yes Yes Yes Yes Yes
Year & Ind. effects Yes Yes Yes Yes Yes Yes
N 1099 1099 1101 1101 935 935
This table presents the results addressing potential endogeneity concerns in the relation between CEO power measure and stock price crashes. Panel
A presents the results of probit models for the determinants of CEO power, using CEOFEPCB and CEOFO as power measures, and two instruments:
dead founder (DFO) and number of founders alive (LNFOALIVE). Panel B presents the results of the second stage IV regression and diagnostic tests
for the validity of the instrument. CEO power measures (CEOFO and CEOFEPCB) are instrumented with the predicted probability that the CEO is
CEOFO or CEOFEPCB, using the probit model in Panel A. Panel C presents the results of firm fixed effect regression. Panel D reports the univariate
test on the effect of exogenous CEO turnovers on stock price crash risk. For each turnover event at t, the change in the firm's industry-adjusted crash
risk is the difference between average industry-adjusted value of the crash risk before the event and after the event. Panel E (panel F) reports the
impact of exogenous CEO turnovers on crash risk for firms with increases (decreases) in CEO power after the turnover events compared to firms with
no change in CEO power after the turnover events. For each turnover event occurring in year t, CHANGE is a dummy variable that equals one if the
firm-year observation is from the period after the exogenous CEO turnover event; it equals zero if the observation is from the period before the
exogenous CEO turnover event. INCREASE (DECREASE) is a dummy variable that equals one if the outgoing CEO is replaced by a more (less)
powerful incoming CEO and equals zero if the outgoing CEO is replaced by an equally powerful incoming CEO. We present t-values in parentheses.
Significance levels of 1%, 5%, and 10% are denoted a, b, and c, respectively. All continuous variables are winsorized at the 1% level. We present the
variable definitions in the Appendix.
experience an increase or decrease in CEOFO to firms that experience no change in CEOFO following exogenous turnover events. To
conduct the test, we restrict the window of a minimum of two firm-year and a maximum of four firm-year observations in both
periods before and after each exogenous CEO turnover event. We calculate the average Fama-French (FF)-49 industry-adjusted
NCSKEW and DUVOL for both periods before and after each turnover event. We then calculate the changes in the Fama-French (FF)-
49 industry-adjusted NCSKEW (DUVOL) as the difference between the pre- and post-period average industry-adjusted NCSKEW
(DUVOL).
We present our univariate test results in Panel D of Table 3. Specifically, Column (1) reports the mean change in industry-adjusted
NCSKEW and DUVOL around exogenous CEO turnover events where the outgoing CEO is replaced by an incoming CEO who have
(footnote continued)
the changes in crash risk following exogenous CEO turnover events for firms experiencing increases versus decreases in CEO power. Unreported
results are consistent with the overall findings of this section.
12
similar power status compared to outgoing CEOs (no change in power group). In column 2, we report the mean change in industry-
adjusted NCSKEW and DUVOL for the increase in CEO power group where outgoing CEOs are less powerful than incoming CEOs.
Similarly, in column 3, we report mean change in industry-adjusted NCSKEW and DUVOL for the decrease in CEO power group where
outgoing CEOs are more powerful than incoming CEOs. We then report the difference in the mean change in industry-adjusted
NCSKEW and DUVOL between columns (2) and (1) and associated t-statistics in columns 4 and 5, respectively. Similarly, we present
the difference in the mean change in industry-adjusted NCSKEW and DUVOL between columns (3) and (1) and associated t-statistics
in columns 6 and 7, respectively.
The univariate test results show that the mean difference in the changes in the industry-adjusted NCSKEW and DUVOL between
the increase in CEO power group and no change in CEO power group is positive and significant for all measures of CEO power. The
mean difference in the changes in industry-adjusted NCSKEW and DUVOL between the decrease in CEO power group and the no
change in CEO power group is negative and significant for all measures of CEO power.
We also formally test the effect of exogenous CEO turnovers on crash risk by comparing the effect of increase and decrease in CEO
power compared to no change in power following exogenous CEO turnover events. We use the exogenous change in CEO power
variables and corresponding control variables that are identical to our baseline regressions (Models 2, 4, 6, 8, 10, and 12 of Table 2).
In Panel E, we replace the CEO power variable with CHANGE, INCREASE, and the interaction variable, CHANGE × INCREASE.
CHANGE is a dummy variable that equals one if the firm-year observation is from the period after the exogenous CEO turnovers and
zero if the observation is from the period before the exogenous CEO turnovers. INCREASE is a dummy variable that equals one if the
incoming CEO has more power than the outgoing CEO and zero if the incoming CEO has similar power compared to the outgoing
CEO. We observe a positive and significant coefficients of the interaction variable CHANGE × INCREASE, suggesting that firms with a
more powerful incoming CEOs display significantly higher crash risk when compared to firms with no change in CEO power.
Similarly, in Panel F, we replace the CEO power variable with CHANGE, DECREASE, and the interaction variable, CHANGE ×
DECREASE. DECREASE is a dummy variable that equals one if the incoming CEO has less power than the outgoing CEO, and zero if
the incoming CEO has similar power compared to the outgoing CEO. The negative and significant coefficients for the interaction
variable CHANGE × DECREASE suggest that firms with a less powerful incoming CEOs display significantly lower crash risk when
compared to firms with incoming CEOs having similar power compared to outgoing CEOs after the exogenous CEO turnover.9
We do urge caution in interpreting a causal effect of CEO power on crash risk from our exogenous CEO turnover analysis. While
the timing of an outgoing CEO's exit is plausibly exogenous in our test, the board's choice of an incoming CEO may be related to crash
risk if the power granted to the incoming CEO is expected to foster the desired change in firm disclosure policy and eventually crash
risk. Without access to the data on the decision-making process of boards when appointing CEOs, we cannot rule out this possibility.
In addition, while we have controlled for CEO characteristics such as tenure, ownership and overconfidence in our analysis, other
characteristics of an incoming CEO may have a direct effect on crash risk. However, at the very least the results in Panels D to F
suggest a significant change in crash risk following exogenous CEO turnover events and the changes are more pronounced for
turnover events in which CEO power changes. The coefficients of CEO power are positive when turnover events result in increase in
CEO power and are negative when turnover events result in decrease in CEO power.
4.2.4. CEO power and stock price crash risk – propensity score matching
Finally, we perform additional tests utilizing propensity score matching (PSM) analysis. PSM does not require a clear source of
exogenous variation for identification (Roberts and Whited, 2013) and allows to reduce misspecification bias of the functional form of
the underlying relation between CEO power and crash risk (Armstrong et al., 2010). To implement propensity score matching (PSM),
we generate a robust control group of firms without powerful CEOs for firms with powerful CEOs. For each fiscal year, we match each
firm with powerful CEO to another firm without powerful CEO based on one to one matching of nearest neighbor with replacement
using several variables - SIGMA, RET, ROA, DTURN, LEV, LMVE, MTB, TENURE, and CEOOWN.
We present the results in Table A4 in the Internet Appendix. To ensure the quality of matching, we first compare the char-
acteristics of the treatment (Powerful CEOs) and control firms (Non-powerful CEOs) and find that the average values of these
variables are qualitatively similar across treatment and control firms. Panel B of Table A4 in the Internet Appendix presents the
regression result using the matched sample. The results show that powerful CEOs, within the restricted sample of firms, have a
significantly positive association with stock price crash risk, irrespective of the CEO power measure used.10
5. CEO power and bad news hoarding
The crash risk literature is built on the theory of hiding bad news (Jin and Myers, 2006). We argue that hiding bad news requires
both the incentive (Cheng and Warfield, 2005; Bergstresser and Philippon, 2006; Burns and Kedia, 2006) and the ability to hide. Our
result in the previous section that powerful CEOs lead to higher crash risk indirectly provides evidence on the potential role of
powerful CEOs' ability in hiding bad news. In this section, we further strengthen our main findings by examining the link between
9
Using strictly two-year window before and after the exogenous CEO turnover event, we find qualitative similar result. We also find similar result
when we do not exclude sample firms with multiple turnovers during our sample period.
10
We obtain consistent results when we use the residual short interest (a proxy for managers' private bad news (Bao et al., 2018)) as an additional
matching variable for PSM analysis. This result indicates that for the same level of bad news, more powerful CEOs withhold more bad news, leading
to positive association between CEO power and crash risk. We thank the anonymous referee for this insight.
13
powerful CEOs and proxies for hiding bad news.

Our first proxy for hiding bad news is financial restatements. Restatements represent a clear proxy for withholding bad news,
since they undo the reporting fraud and/or accounting errors committed by a firm. Archambeault et al. (2008) argue that a CEO, who
is holding a chairman position is likely to be related to higher levels of restatements. Efendi et al. (2007) document that the likelihood
of financial restatements is greater if the CEO is also the chair of the board and has sizable in-the-money stock options. Therefore, if
powerful CEOs engage in withholding bad news, the firms they run will more likely to restate their financial statements.
In a recent study, Edmans et al. (2018) show that CEOs strategically choose to release more discretionary news items in months in
which they are expected to sell equity, clearly highlighting the managerial agency motivation relating to news release. Earnings
guidance, a company's official prediction of its own profit or loss, represents a clear and direct measure for ‘news management’ that
can be tainted by top management to further their agency motivation. Hamm et al. (2018) document that on average, guidance,
particularly optimistic guidance, is associated with the higher future price crash risk. Hence, we also use a more direct measure of
hiding bad news which is the proportion of the total number of negative to positive earnings guidance (NETPOEG) issued by the firm.
We argue that if powerful CEOs engage in bad news hoarding, they will release less negative earnings guidance relative to positive
earnings guidance. As such, the NETPOEG should be lower in firms with powerful CEOs. We also use the tone of annual reports (i.e.
the total number of negative to positive words in the financial statement (NETPOW)) as an indirect but more pervasive measure of
hiding bad news. These measures reflect senior management's intention to portray a more positive image for the firm to the external
users of the news releases and financial statements.
We run a logit model with clustered errors to test the association between powerful CEOs at time t - 1 and the probability of
financial restatements (REST) at time t using restatements data from the US General Accounting Office (GAO). We closely follow
Archambeault et al. (2008) to include several variables that are likely to have an influence on financial restatements. In particular, we
use board independence (BOIND); board size (BOSIZE); block holding (BLOCK), that is, the proportion of shareholding by the top 5%
of institutional owners; firm age, that is, the number of years since the company's inception; return on assets (ROA); and an ac-
quisition activity dummy (ACQ). We include additional firm-related characteristics that account for firm-level heterogeneity which
can have a potential impact on financial reporting dynamics. We also include CEO tenure and CEO ownership as control variables.
The results in Models (1) to (3) in Table 4 shows that our measures of CEO power at time t - 1 has a positive and statistically
significant effect on REST at time t.
We further test whether powerful CEOs have any association with the proportion of negative to positive earnings guidance
(NETPOEG) and the ratio of negative to positive words (NETPOW) in 10-K filings (annual reports). To construct NETPOEG, we use the
RavenPack News Analytics that tracks and monitors all news relevant to corporations around the world from various sources in-
cluding Dow Jones Financial Wires, Wall Street Journal, Barron's, MarketWatch, regulatory and corporate press releases.11 Ra-
venPack provides further classification on earnings related news issues based on market reaction to the earnings guidance issued by
the firm. It categorizes an earnings guidance as either ‘earnings guidance up or above expectation’ or ‘earnings guidance down or below
expectation’. We count the total number of unique earnings-guidance news items categorized as ‘up or above expectation’ as well as
‘down or below expectation’ for each firm in each fiscal year. We then calculate the ratio of ‘total negative’ to ‘total positive’ (total
‘down or below expectation’ to total ‘up or above expectation’) earnings guidance (NETPOEG) for each firm-year. Since RavenPack data
are available from the year 2000 onward, our sample size for NETPOEG regression model is significantly lower compared to the
sample size of our baseline model. In the case of NETPOW, we collect the negative and positive word counts in 10-K filings from the
SEC Analytics Suite, available from Wharton Research Data Services (WRDS). The results in Models (4) to (9) in Table 4 show that all
three measures of our CEO power at time t - 1 are negatively related to both NETPOEG and NETPOW at time t.12 We interpret our
result as evidence that firms with powerful CEOs provides fewer negative than positive earnings guidance as well as employs fewer
negative than positive words in communicating their financial performance to market participants. Overall, our result on the ability
of powerful CEOs to withhold bad news through control of information released to the market further reinforces the relation between
CEO power and higher stock price crash risk.13
6. When do powerful CEOs hide bad news?
In this section, we examine the motivation and the circumstances that lead powerful CEOs to hide bad news, which, in turn, leads
to higher future crash risk. We postulate that powerful CEOs have a stronger motivation to hide bad news when their compensation is
more related to stock price and when they possess lower skills.
11
RavenPack News Analytics employs a variety of advanced textual analysis techniques to create news sentiment scores for business news stories
and provides assessment of the tone of the news (i.e., positive or negative) in a given article (Bushman et al., 2017).
12
Li (2008) finds that firms with lower earnings have less readable annual reports. In line with this argument, Lo et al. (2017) further show that
financial reports are more difficult to read for firms that are most suspicious of having managed earnings. In untabulated results, we also observe
that firms with powerful CEOs have less readable financial reports.
13
Crash risk is built on the idea of hiding bad news. We argue that managers require both the intention and the ability to hide bad news. Hence,
the act of hiding bad news can be accomplished far more easily by a powerful CEO. In unreported results, we find that the effects of earnings
management (Hutton et al., 2009) on crash are more pronounced for firms with powerful CEOs (particularly powerful founder CEOs) than non-
powerful CEOs. Similarly, the effect of tax avoidance (Kim et al., 2011a), and CFO option incentives (Kim et al., 2011b) on crash are more
pronounced for firms with powerful CEOs than non-powerful CEOs.
14
Table 4
CEO power and bad news hoarding.
Variables RESTt NETPOEGt NETPOWt
(1) (2) (3) (4) (5) (6) (8) (8) (9)
CEOFOt-1 0.706a −0.059b −0.116b

(3.65) (−2.50) (−2.36)
CEOFEPCB t-1 0.686a −0.056a −0.150a
(3.63) (−2.60) (−2.96)
CEOPRCH t-1 0.278b −0.054a −0.082b
(2.04) (−3.12) (−2.08)
BOSIZE t-1 0.423 0.430 0.321 −0.080c −0.080c −0.077c −0.092 −0.098 −0.083
(1.42) (1.45) (1.09) (−1.67) (−1.72) (−1.77) (−0.80) (−0.85) (−0.73)
BIG4 t-1 −0.214c −0.205 −0.223c −0.004 −0.004 −0.002 −0.002 −0.003 0.003
(−1.65) (−1.57) (−1.72) (−0.23) (−0.24) (−0.11) (−0.05) (−0.08) (0.07)
BOIND t-1 0.589 0.570 0.380 0.076 0.081 0.105 −0.005 −0.005 0.035
(1.57) (1.52) (0.98) (0.91) (0.99) (1.32) (−0.03) (−0.03) (0.19)
BLOCK t-1 0.016 0.015 0.017 −0.007b −0.007b −0.007b −0.000 −0.000 −0.000
(0.76) (0.72) (0.79) (−2.11) (−2.12) (−2.11) (−0.03) (−0.01) (−0.07)
LFAGE t-1 0.010 −0.008 −0.066 0.021c 0.022b 0.027b −0.083a −0.083a −0.070b
(0.12) (−0.09) (−0.89) (1.89) (2.10) (2.49) (−2.90) (−2.84) (−2.53)
ROA t−1 −1.487b −1.453b -1.381b −0.858a −0.857a −0.863a −1.343a −1.346a −1.360a
(−2.40) (−2.33) (−2.22) (−9.04) (−8.94) (−8.88) (−7.90) (−7.86) (−8.06)
ACQ t-1 0.203 0.207c 0.208c −0.048b −0.048b −0.049b −0.030 −0.030 −0.030
(1.63) (1.67) (1.67) (−2.29) (−2.27) (−2.34) (−1.07) (−1.07) (−1.08)
LEV t-1 0.831c 0.813c 0.774c 0.052 0.054 0.059 −0.195 −0.192 −0.183
(1.91) (1.84) (1.74) (0.67) (0.70) (0.79) (−1.54) (−1.52) (−1.44)
LMVE t-1 0.027 0.025 0.026 −0.006 −0.006 −0.005 0.017 0.017 0.018
(0.59) (0.53) (0.56) (−0.63) (−0.62) (−0.57) (1.14) (1.14) (1.21)
CEOOWN t−1 −1.573b -1.633b −2.049a 0.124 0.131 0.190 −0.634a −0.649a −0.541a
(−1.98) (−2.07) (−2.59) (0.92) (0.98) (1.40) (−3.22) (−3.33) (−2.85)
CEOOWNSQ t-1 22.774 22.343 15.335 −0.402 −0.387 −0.324 0.338 0.397 0.219
(0.88) (0.92) (0.69) (−0.36) (−0.35) (−0.29) (0.26) (0.30) (0.16)
TENURE t-1 0.805c 0.849c 1.209a −0.081 −0.087 −0.129c 0.340a 0.354a 0.271b
(1.69) (1.80) (2.58) (−1.09) (−1.16) (−1.69) (2.76) (2.90) (2.34)
TENURESQ t-1 −587.71 −541.10 −404.68 3.507 3.298 2.694 −2.147 −2.400 −2.017
(−0.66) (−0.68) (−0.59) (0.76) (0.70) (0.57) (−0.33) (−0.36) (−0.30)
Pseudo (adjusted) R2 0.098 0.097 0.093 0.027 0.027 0.027 0.072 0.072 0.071
Const. & all controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Yr. & Ind. effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
N 8320 8320 8320 4051 4051 4051 12,682 12,682 12,682
This table presents the results for the effect of CEO power on proxies for hiding bad news. We use three proxies for hiding bad news: REST, a dummy
variable that equals one to indicate financial reporting restatements and zero otherwise, NETPOEG, the ratio of negative to positive management
earnings guidance, and NETPOW, the ratio of negative to positive word counts in the annual report (10−K) report. All regressions include year and
Fama–French 49-industry fixed effects. We cluster standard errors by both firm and year. We present t-values in parentheses. The 1%, 5%, and 10%
significance levels of the coefficients are denoted a, b, and c, respectively. All continuous variables are winsorized at the 1% level. We present the
variable definitions in the Appendix.
6.1. CEO power, CEO delta, and crashes
Kothari et al. (2009) emphasize that career concerns, especially the effect of disclosure on management compensation, motivate
managers to withhold bad news. Building from this argument, we posit that, if the value of a CEO's personal wealth is at risk due to
depressed stock prices, a powerful CEO with the means and abilities will be motivated to conceal bad news, which eventually leads to
higher future crash risk. Since, the CEO delta measures the sensitivity of CEO wealth to stock prices (Core and Guay, 2002; Coles
et al., 2006), we argue that a higher CEO delta is likely to be one of the key motivations for powerful CEOs (with means and abilities)
to hide bad news and extract personal benefit from inflated stock prices. Thus, the positive association between CEO power and
higher stock price crash risk should be primarily driven by the subsample of firms with higher CEO delta. We examine the association
between CEO power and crash risk controlling for delta and present the results in Table A5 in the internet appendix. We find that CEO
delta has a positive association with future stock price crash risk, while the powerful CEOs has a positive association with future crash
risk except for CEO power measure CEOPRCH with crash measure NCSKEW.
Further, to test our main argument that higher sensitivity of CEO wealth to stock price is the key motivation for powerful CEOs to
hide bad news, we partition our sample into high and low CEO delta subsamples based on the sample median value of CEO delta. We
present the results in Table 5. We find that the coefficient of the CEO power variable on crash risk is positive and significant for the
high CEO delta (HDELTA) subsample, irrespective of CEO power and crash measures used, whereas the coefficient of the CEO power
15
Table 5
CEO power, sensitivity of CEO wealth to stock price, and crash risk.
Power measures NCSKEW DUVOL
CEOFO CEOFEPCB CEOPRCH CEOFO CEOFEPCB CEOPRCH
HDELTA LDELTA HDELTA LDELTA HDELTA LDELTA HDELTA LDELTA HDELTA LDELTA HDELTA LDELTA
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
a a a b a b
CEOPOWER t-1 0.106 0.059 0.125 0.037 0.096 −0.027 0.029 0.011 0.036 0.002 0.031 0.006
(2.91) (1.37) (3.39) (0.97) (2.71) (−1.09) (2.37) (0.61) (2.78) (0.10) (2.10) (0.74)
Diff in Coefficients 0.61 2.13 8.82 0.62 2.23 2.50
Chi-square test [0.43] [0.14] [0.00] [0.43] [0.13] [0.11]
OPAQUE t-1 0.119b 0.106b 0.118b 0.105b 0.128b 0.106b 0.031 0.019 0.030 0.018 0.033 0.018
(2.22) (2.30) (2.21) (2.29) (2.34) (2.30) (1.54) (0.90) (1.51) (0.89) (1.64) (0.88)
OPAQUESQ t-1 −0.024c −0.031b −0.024c −0.031b −0.026b −0.030b −0.007 −0.007 −0.007 −0.007 −0.007 −0.007
(−1.88) (−2.39) (−1.88) (−2.38) (−2.01) (−2.39) (−1.47) (−1.27) (−1.45) (−1.26) (−1.59) (−1.26)
PRSHELT t-1 0.082c 0.019 0.082c 0.019 0.077c 0.021 0.028c 0.002 0.028c 0.003 0.026c 0.002
(1.90) (0.64) (1.91) (0.65) (1.75) (0.69) (1.90) (0.19) (1.91) (0.20) (1.77) (0.18)
CFOOPIN t-1 0.352b 0.170 0.354b 0.169 0.349b 0.172 0.110c 0.008 0.111c 0.007 0.109c 0.006
(2.08) (0.80) (2.08) (0.80) (2.08) (0.82) (1.95) (0.11) (1.94) (0.11) (1.94) (0.09)
HOLD100 t-1 0.001 0.082a 0.003 0.083a −0.002 0.084a 0.004 0.021b 0.005 0.022b 0.003 0.022b
(0.05) (4.03) (0.11) (4.13) (−0.10) (4.16) (0.35) (2.24) (0.39) (2.30) (0.26) (2.27)
Adjusted R2 0.030 0.045 0.031 0.045 0.030 0.045 0.033 0.042 0.034 0.042 0.034 0.042
Const. & all controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
N 7327 7895 7327 7895 7327 7895 7327 7895 7327 7895 7327 7895
This table presents the results on how the impact of power varies between firms with higher and lower CEO delta. Both industry and year fixed
effects are included in all the regressions. We cluster standard errors by both firm and year. We present t-values in parentheses. The 1%, 5%, and
10% significance levels of the coefficient are denoted a, b, and c, respectively. All continuous variables are winsorized at the 1% level. We also
present the Chi-square value (with p-value in square brackets) to test for the significance of the difference in coefficients. We present the variable
definitions in the Appendix.
variable on crash risk is insignificant for the low CEO delta (LDELTA) subsample.14 Our findings suggest that the association between
powerful CEOs and future stock price crash risk is mostly evident in the subsample of firms where the CEOs' wealth is more tied to
stock prices.15 However, we do not find any significant difference on the estimated coefficients of CEO power between high CEO delta
and low CEO delta groups except using CEOPRCH as a power measure and NCSKEW as a crash measure.16
6.2. CEO power, CEO skills, and crashes
CEOs' decisions to hide bad news mainly represent compensation or career concerns due to the poor performance of the firm they
manage. However, CEOs with higher general skills or abilities tend to enjoy a very secure long-term career path. Therefore, they are
less sensitive to the risk of termination due to their more diverse career experiences and find it easy to move across industries to
pursue their career (Custódio et al., 2019). Moreover, CEOs with higher general skills enjoy much higher compensation packages than
others (Custódio et al., 2013) because of greater labor market demand for their general skills that potentially help firms to navigate
through the challenges of industry deregulation, changes in product market dynamics, increased foreign competition, and the
changing dynamics of technological and managerial practices (Garicano and Rossi-Hansberg, 2006; Cunat and Guadalupe, 2009a,
2009b). Thus, due to fewer concerns for their career or compensation, powerful CEOs with higher general skills are less likely to hide
bad news compared with powerful CEOs with lower general skills. Therefore, we argue that the positive effect of powerful CEOs on
future crash risk would be mostly evident among firms run by CEOs with lower general skills.
We use the index of generality of a CEO's human capital (General Ability Index) suggested by Custódio et al. (2013), which
measures a CEO's lifetime work experience in public limited firms prior to the CEO's current position. The CEO's general managerial
skills consider five different aspects of the CEO's professional life: the past number of positions, firms, and industries in which a CEO
worked; past position as a CEO at a different company, and work experience in a conglomerate (Custódio et al., 2013).17 We examine
14
We find similar results for cross-sectional grouping based on CEO option incentives, which provides further support for the complementary
effect of managerial power and managerial incentive to hide bad news leading to subsequent crash risk.
15
In an unreported result, we find that non-powerful CEOs do not have a positive and significant effect on crash risk in the high CEO delta
subsample. This reinforces our argument that while the material presence of the motive of hiding bad news is important, but it also requires the
power to design plans to hide bad news and secure those consequential motives.
16
Since we do not find any significant difference on the estimated coefficients of CEO power between HDELTA and LDELTA, we caution the
inferences drawn from these findings.
17
We thank Claudia Custódio, Miguel A. Ferreira, and Pedro Matos for making their CEO generality index, used in CEO turnover data, public for
others to use.
16
Table 6
CEO power, management skill, and crash risk.
HGSK LGSK HGSK LGSK HGSK LGSK HGSK LGSK HGSK LGSK HGSK LGSK
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
a a b a a
CEOPOWER t-1 0.102 0.136 0.111 0.147 0.037 0.080 0.010 0.052 0.007 0.058 0.015 0.037b
(1.39) (3.00) (1.44) (3.04) (1.13) (2.04) (0.38) (3.42) (0.23) (3.62) (1.49) (2.35)
Chi-square test [0.62] [0.61] [0.39] [0.12] [0.06] [0.26]
OPAQUE t-1 0.090 0.131 0.092 0.130 0.092 0.131 0.025 0.040 0.025 0.040 0.025 0.041
(1.42) (1.39) (1.45) (1.38) (1.46) (1.37) (1.14) (1.16) (1.15) (1.15) (1.15) (1.14)
OPAQUESQ t-1 −0.021 −0.043c −0.021 −0.043c −0.021 −0.043c −0.005 −0.014 −0.005 −0.014 −0.005 −0.014
(−1.31) (−1.86) (−1.34) (−1.86) (−1.34) (−1.84) (−0.97) (−1.46) (−0.98) (−1.46) (−0.98) (−1.45)
PRSHELT t-1 0.143a −0.027 0.143a −0.027 0.144a −0.034 0.046a −0.010 0.046a −0.010 0.045a −0.013
(5.07) (−0.77) (5.09) (−0.77) (5.12) (−0.95) (3.09) (−0.67) (3.11) (−0.66) (3.05) (−0.83)
CFOOPIN t-1 0.783a 0.123 0.784a 0.121 0.794a 0.167 0.220a −0.049 0.220a −0.050 0.219a −0.031
(3.26) (0.60) (3.27) (0.59) (3.22) (0.83) (2.59) (−1.03) (2.60) (−1.03) (2.58) (−0.66)
HOLD100 t-1 0.054c 0.069b 0.054b 0.069b 0.055b 0.068b 0.011 0.026b 0.011 0.026b 0.011 0.026b
(1.96) (2.33) (1.99) (2.32) (2.02) (2.32) (0.87) (2.24) (0.89) (2.22) (0.84) (2.23)
Adjusted R2 0.034 0.030 0.034 0.030 0.034 0.028 0.033 0.045 0.033 0.046 0.033 0.045
N 4960 4911 4960 4911 4960 4911 4960 4911 4960 4911 4960 4911
This table present the results on how the impact of CEO power on stock price crash risk varies among firms with higher and lower CEO general skills.
Industry and year fixed effects are included in all the regressions. We cluster standard errors by both firm and year. We present t-values in
parentheses. The 1%, 5%, and 10% significance levels of the coefficient are denoted a, b, and c, respectively. All continuous variables are winsorized
at the 1% level. We also present the Chi-square value (with p-value in square brackets) to test for the significance of the difference in coefficients. We
present the variable definitions in the Appendix.
the effect of CEO power on crash risk controlling for CEOs' general skills (CEOGSK) and present the results in Table A6 in the internet
appendix. We find some evidence for the negative relation between CEOGSK and stock price crash risk when we use NCSKEW as crash
measure. More importantly, the effect of powerful CEOs on future crash risk is still positive and significant controlling for CEOGSK.
We test our main argument related to CEOs' general skills by partitioning our full sample into two high and low CEO general skills
subsamples using the median value of CEOGSK. We present the results in Table 6. We find that the coefficients of the CEO power
variable on crash risk are positive and significant for the low CEOGSK subsample (LGSK), and insignificant for the high CEOGSK
subsample (HGSK), indicating that the significantly positive relation between CEO power and stock price crash risk mainly arises
among the firms with lower general skills CEOs.18
7. CEO power, external monitoring mechanisms, and stock price crashes
In this section, we examine whether external monitoring mechanisms attenuate the association between CEO power and stock
price crashes.19 We use two proxies for external monitoring mechanisms: the takeover index of Cain et al. (2017), and dedicated
institutional ownership, based on the work of Bushee (1998). Morck et al. (1989) find that corporate boards have not been the main
force behind removing unresponsive managers in poorly performing industries. They argue that takeovers come to play a role in
replacing managers when the board is unable or unwilling to discipline managers. Lel and Miller (2015) suggest that the threat of
takeover causes managerial discipline through the incentives that the market for corporate control provides to boards to monitor
managers. The takeover index (TOIND) of Cain et al. (2017) focuses on hostile takeovers, which are disciplinary in nature. Further,
the takeover index is constructed from takeover laws, aggregate capital liquidity, and firm age, all of which are outside the control of
the firm and are better measures of a firm's governance environment than the governance measures in prior studies, which have
focused almost exclusively on business combination law.20 Chen et al. (2007) show that dedicated institutional investors are effective
monitors of acquisitions. An and Zhang (2013) document that crash risk is negatively related to the ownership of dedicated
18
Since we do not find any significant difference on the estimated coefficients of CEO power between HGSK and LGSK, we caution the inferences
drawn from these findings.
19
We mainly use external governance, since powerful CEOs can withstand the pressure exerted by internal governance mechanisms. For example,
Fracassi and Tate (2012) show that powerful CEOs are exposed to weaker board monitoring, since their firms are more likely to appoint directors
with ties to the CEO.
20
In untabulated results, we find that governance indices based on firm-specific measures of takeover defense such as the Governance index
(Gompers et al., 2003) and the Entrenchment index (Bebchuk et al., 2009) do not reduce stock price crash risk.
17
institutions, who have a strong incentive to monitor. We calculate the yearly percentages of shares outstanding held by dedicated
institutional investors (IODED) by taking the average over the four quarters of the firm's fiscal year using data from the Thomson
Reuters Institutional Holdings (13F) database.
We examine the association between CEO power and crash risk controlling for both TOIND and IODED and present the results in
Table A7 in the internet appendix. We find that TOIND significantly curtails future stock price crash risk, irrespective of the CEO
power or crash measures used. However, the effect of IODED on crash risk is negative but not significant. More importantly, con-
trolling for TOIND and IODED, we still find a positive association between CEO power and crash risk.21 Further, we investigate the
impact of CEO power by splitting TOIND into high and low values based on the sample median and present the results in Panel A of
Table 7. We find that powerful founder CEOs have association with higher crash risk irrespective of either a high- or low-TOIND
environment. When we use CEOPRCH as a measure of CEO power, we find that powerful CEOs are associated with higher crash risk
only in a lower-TOIND environment.
Similarly, we further investigate the relation between CEO power and crash risk in high versus low IODED subsamples based on
the sample median and present the results in Panel B of Table 7. We find that in general powerful CEOs have association with higher
crash risk for firm with low IODED, except for using CEOPRCH as a CEO power measure and NCSKEW as crash risk measure. The
relation between CEO power and crash risk is inconsistent in the high IODED subsample. Interestingly, when we use CEOFEPCB as a
CEO power measure we find that CEO power has association with higher crash risk even for firm with high IODED. The tests for the
difference in the coefficients of CEO power measures across high- versus low-TOIND subsamples in Panel A and high- versus low-
IODED subsamples in Panel B are not significant. Overall, our findings indicate that TOIND reduces stock price crashes; however,
powerful CEOs, particularly powerful founder CEOs, presents significant monitoring challenges for current governance arrangements.
8. Further tests and robustness checks
8.1. CEO power and price jumps
In the preceding sections, we demonstrate that CEO power leads to stock price crash risk (downside risk). Since Adams et al.
(2005) show that powerful CEOs can lead to greater variability in firm performance, we further examine the effect of CEO power on
stock price jumps (upside risk) in this section. We use two alternative measures of jumps. First, JUMPD is a dummy variable taking
the value of one if at least one weekly firm-specific return in the entire fiscal year is 3.20 standard deviations above the mean weekly
return over the fiscal year. Second, JUMPCOU is the number of weeks during the entire fiscal year when the firm-specific weekly
return is 3.20 standard deviations above the mean weekly return.
We report the results of this test in Table 8. We do not find any evidence that firms with powerful CEOs have higher stock price
jumps than others. Consistent with Hutton et al. (2009), we find that MTB and LMVE are negatively related to JUMPD, irrespective of
the CEO power measures used in the regressions. Further, we find that RET is negatively related to JUMPD, while LEV and ROA are
positively related to JUMPD. In the case of models with JUMPCOU as the dependent variable, we find that RET and LMVE are
negatively related to JUMPCOU, while LEV is positively related to JUMPCOU. Overall, our findings in this section demonstrate that
CEO power has significant implications on future extreme negative stock returns but not on positive outcomes (price jumps), in-
dicating that the decision to hoard bad news, which, in turn, leads to extremely negative consequences, is more likely to be made
when CEOs are more powerful.
8.2. CFO option incentives
Kim et al. (2011b) examine whether CEO and CFO equity incentives are associated with firm-specific stock price crash risk and find
that the incentives from CFOs' option holdings are significantly and positively related to future crash risk, whereas, they find only weak
evidence that the incentives from CEOs' option holdings contribute to crash risk, and this weak effect disappears after the CFO option
incentives are included. In this section, we examine how the effect of CEO power on crash risk varies between firms with high and low
CFO's option incentives (CFOOPN). We categorize firm-year observations into high (low) CFOOPN subsample if the value of CFOOPN is
higher (lower) than the sample median. We present the results in Table A8 in the internet appendix. We find that the coefficients of all
three CEO power measures are positive and significant for firms in the high CFOOPN subsample. However, for firms in the low CFOOPN
subsample, we find that only the coefficients of CEOFO and CEOFEPCB are positive and significant with the crash measure NCSKEW.
These findings indicate that CEO power has association with higher crash risk mainly when CFOs have higher option incentive.
8.3. Further robustness checks
8.3.1. Alternative measures of stock price crash risk

In this section, we examine the relation between CEO power and stock price crashes using two alternative measures of stock price
crash risk. The first measure is the dummy variable CRASHD, which takes a value of one for a firm–year that experiences one or more
firm-specific weekly returns falling 3.2 standard deviations below the mean firm-specific weekly returns over the fiscal year and zero
21
We also find robust evidence of the effect of CEO power on crash risk, even after we control for other internal governance mechanisms such as
board independence and founder directors, together with the takeover index and dedicated institutional ownership.
18
Table 7
CEO power, external governance, and stock price crash risk.
Panel A: CEO power and crash risk – cross sectional test using TOIND
HTOIND LTOIND HTOIND LTOIND HTOIND LTOIND HTOIND LTOIND HTOIND LTOIND HTOIND LTOIND
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
CEOPOWER t-1 0.129a 0.111a 0.141a 0.123a 0.010 0.068b 0.035b 0.031b 0.036b 0.037b 0.011 0.029b
(2.92) (3.12) (3.48) (2.96) (0.43) (2.05) (2.07) (2.49) (2.26) (2.48) (1.36) (2.14)
Chi-square test [0.72] [0.74] [0.16] [0.87] [0.93] [0.22]
Adjusted R2 0.031 0.049 0.031 0.049 0.030 0.049 0.031 0.047 0.031 0.047 0.030 0.047
N 8245 7258 8245 7258 8245 7258 8245 7258 8245 7258 8245 7258
Panel B: CEO power and crash risk – cross sectional test using IODED
NCSKEW DUVOL
Power measures CEOFO CEOFEPCB CEOPRCH CEOFO CEOFEPCB CEOPRCH
HIODED LIODED HIODED LIODED HIODED LIODED HIODED LIODED HIODED LIODED HIODED LIODED
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
CEOPOWER t-1 0.092b 0.132a 0.114a 0.136a 0.030 0.033 0.022 0.036b 0.030b 0.037b 0.014 0.020b
(2.38) (3.45) (3.21) (3.36) (1.35) (1.37) (1.55) (2.49) (2.13) (2.46) (1.59) (2.32)
Chi-square test [0.44] [0.67] [0.93] [0.48] [0.73] [0.69]
Adjusted R2 0.033 0.045 0.034 0.045 0.033 0.044 0.036 0.043 0.036 0.043 0.036 0.043
N 8435 7958 8435 7958 8435 7958 8435 7958 8435 7958 8435 7958
This table presents the results on the relation between CEO power and stock price crash risk in subsamples based on high vs. low takeover index
(TOIND) and high vs. low dedicated institutional ownership (IODED), respectively. Industry and year fixed effects are included in all the regressions.
“all controls” refer to all independent variables in the baseline model, except the CEO power measure, as in Model 2 of Table 2. We cluster standard
errors by both firm and year. We present t-values in parentheses. The 1%, 5%, and 10% significance levels of the coefficient are denoted a, b, and c,
respectively. All continuous variables are winsorized at the 1% level. We also present the Chi-square value (with p-value in square brackets) to test
for the significance of the difference in coefficients. We present the variable definitions in the Appendix.
otherwise. The second measure is CRCOUNT, calculated as the number of firm-specific weekly returns exceeding 3.2 standard de-
viations below the mean firm-specific weekly return. We present the results of this robustness check in Panel A of Table 9. We still
find that CEO power is associated with higher stock price crash, irrespective of the alternative stock price crash measures.22 In terms
of economic significance, the coefficient estimates in Model (2) show that holding other variables constant, the magnitude of the
estimated coefficient of CEOCFO on CRCOUNT is 0.046, which indicates that the average CRCOUNT for firms with CEOCFOs will be
4.6% higher than that firms without CEOCFOs. This is equivalent to 21.4% of the mean value of CRCOUNT.
8.3.2. Alternative measures of CEO power

In this section, we examine the relation between CEO power and stock price crash risk using the number of titles a CEO holds
(NCEOTITLE), the CEO power index (CEOPIND), and CEO pay slice (CPS) as additional measures of CEO power.23 We calculate the
22
We also address the endogeneity concern with a method similar to that described in Section 4 and Table 3. We consistently find that the
predicted value of CEOFO (CEOFEPCB) is positively related to future stock price crash risk when we use CRASHD and CRCOUNT as proxies for crash
risk.
23
Given the multi-dimension nature of CEO dominance in the firm, and higher CPS reflects agency and governance problems (Bebchuk et al.,
2011), the use of CPS as an alternative proxy for CEO power can further capture the potential agency problems of hoarding bad news that leads to
crash risk. Bebchuk et al. (2011) argue that CPS captures the relative importance of the CEOs in terms of abilities, contribution, or power and it
provides a useful proxy for the relative centrality of the CEO in the top management team. Whisler et al. (1967) argue that creating pay differentials,
captured by CPS, compensation committee provide information on relative power of the executives. Moreover, controlling for CPS, we observe a
positive association between CEO power and crash risk, indicating the independence of various dimensions of CEO dominance.
19
Table 8
CEO power and stock price jumps.
Variables JUMPDt JUMPCOUt JUMPDt JUMPCOUt JUMPDt JUMPCOUt
(1) (2) (3) (4) (5) (6)
CEOPOWERt-1 −0.065 −0.008 −0.094 −0.013 0.004 −0.002

(−1.02) (−0.72) (−1.42) (−1.31) (0.08) (−0.41)
NCSKEW t-1 −0.010 −0.001 −0.010 −0.001 −0.010 −0.001
(−0.42) (−0.32) (−0.41) (−0.32) (−0.42) (−0.32)
SIGMA t-1 1.572 0.358 1.592 0.363 1.506 0.348
(1.32) (1.44) (1.34) (1.46) (1.27) (1.39)
RET t-1 −0.204a −0.030a −0.204a −0.030a −0.204a −0.030a
(−6.23) (−5.32) (−6.23) (−5.32) (−6.22) (−5.29)
ROA t-1 0.754a 0.104 0.757a 0.105 0.743a 0.103
(2.89) (1.53) (2.90) (1.54) (2.84) (1.50)
DTURN t-1 −0.394 −0.065 −0.395 −0.065 −0.390 −0.064
(−1.54) (−1.28) (−1.54) (−1.28) (−1.53) (−1.26)
LEV t-1 0.334b 0.052b 0.334b 0.051b 0.340b 0.052b
(2.17) (2.34) (2.17) (2.33) (2.21) (2.35)
LMVE t-1 −0.085a −0.011a −0.085a −0.011a −0.084a −0.011a
(−4.15) (−3.90) (−4.17) (−3.92) (−4.10) (−3.88)
MTB t-1 −0.015c −0.002 −0.015c −0.002 −0.015c −0.002
(−1.77) (−1.29) (−1.77) (−1.29) (−1.77) (−1.30)
TENURE t-1 0.213 0.040 0.205 0.039 0.247 0.045
(0.68) (0.92) (0.66) (0.92) (0.80) (1.11)
TENURESQ t-1 −0.129 −0.026 −0.121 −0.024 −0.160 −0.030
(−0.71) (−0.96) (−0.66) (−0.95) (−0.89) (−1.26)
CEOOWN t-1 −1.368 0.002 −1.315 0.011 −1.478 −0.010
(−0.60) (0.01) (−0.57) (0.03) (−0.64) (−0.03)
CEOOWNSQ t-1 5.505 −0.031 5.402 −0.045 5.508 −0.027
(0.48) (−0.02) (0.47) (−0.03) (0.48) (−0.02)
OPAQUE t-1 0.038 −0.006 0.039 −0.006 0.035 −0.007
(0.41) (−0.54) (0.42) (−0.53) (0.38) (−0.57)
OPAQUESQ t-1 −0.005 0.002 −0.005 0.002 −0.004 0.002
(−0.21) (0.73) (−0.21) (0.72) (−0.18) (0.76)
PRSHELT t-1 −0.034 −0.005 −0.034 −0.005 −0.033 −0.005
(−0.54) (−0.46) (−0.54) (−0.46) (−0.53) (−0.46)
CFOOPIN t-1 −0.167 −0.024 −0.165 −0.024 −0.174 −0.025
(−0.55) (−0.46) (−0.54) (−0.45) (−0.57) (−0.47)
HOLD100 t-1 −0.074 −0.012 −0.074 −0.012 −0.075 −0.012
(−1.56) (−1.46) (−1.57) (−1.46) (−1.59) (−1.47)
Pseudo / Adjusted R2 0.031 0.023 0.031 0.024 0.031 0.023
Yr. & Ind. effects Yes Yes Yes Yes Yes Yes
N 16,393 16,393 16,393 16,393 16,393 16,393
This table presents the results for the effect of CEO power on stock price jumps. We use two alternative jump measures: JUMPD and JUMPCOU,
where JUMPD is a dummy variable taking the value of one if at least one weekly firm-specific return in the entire fiscal year is 3.20 standard
deviations above the mean weekly return over the fiscal year and JUMPCOU is the number of weeks during the entire fiscal year when the firm
specific weekly return is 3.20 standard deviations above the mean weekly return. We cluster standard errors by both firm and year. We present t-
values in parentheses. The 1%, 5%, and 10% significance levels of the coefficients are denoted a, b, and c, respectively. All continuous variables are
winsorized at the 1% level. We present the variable definitions in the Appendix.
number of titles a CEO holds by assigning a value of one for each title held and summing the values. Specifically, a nominal CEO without
additional titles receives a value of one and receives an extra point for each additional title, namely, founder, chair, president, chief
operating officer (COO), CFO, vice president, vice chair, director, and a member on key committees. We measure the CEO power index
(CEOPIND) by running a principal component analysis of the various other titles a CEO holds, such as founder, COO, CFO, chairman,
president, chair and president, vice chair, vice-president, member of the board, member of the compensation committee, and member of
the audit committee. Following Bebchuk et al. (2011), we measure the CPS as the fraction of the total compensation the CEO receives to
the total compensation received by the top-five executives. Higher CPS reflects the importance of the CEO and the extent to which he/
she can extract rents. We present the results in Panel B of Table 9. The results indicate that the positive association between CEO power
and stock price crashes holds using NCEOTITLE, CEOPIND, and CPS as alternative measures of CEO power.
8.3.3. Robustness tests controlling for CEO and firm characteristics

We further perform a robustness test controlling for additional CEO and firm characteristics, such as tournament incentives,
CEO age, CEO turnover, company age, potential financial distress, and research and development (R&D) intensity and present the
20
Table 9
Robustness checks.
Panel A: Alternative proxies for stock price crash risk and CEO power
Variables CRASHD CRCOUNT CRASHD CRCOUNT CRASHD CRCOUNT
(1) (2) (3) (4) (5) (6)
a b a b a
CEOPOWERt-1 0.288 0.046 0.270 0.044 0.194 0.037a
(4.70) (2.54) (4.33) (2.42) (4.23) (4.72)
Adjusted R2 0.028 0.019 0.028 0.019 0.028 0.020
N 16,382 16,393 16,382 16,393 16,382 16,393
Panel B: Alternative proxies for CEO power and stock price crash risk
Power measures NCEOTITLE CEOPIND CPS
Variables NCSKEW DUVOL NCSKEW DUVOL NCSKEW DUVOL
(1) (2) (3) (4) (5) (6)
CEOPOWERt-1 0.027b 0.014a 0.021a 0.006a 0.146a 0.045b

(2.30) (3.12) (3.71) (2.68) (2.69) (2.14)
Adjusted R2 0.036 0.039 0.037 0.039 0.036 0.038
N 16,393 16,393 16,393 16,393 16,393 16,393
Panel C: CEO power and stock price crash risk, controlling for firm- and CEO-related variables
Variables NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt
(1) (2) (3) (4) (5) (6)
a b a b b
CEOPOWERt-1 0.081 0.025 0.087 0.028 0.032 0.017a
(2.69) (2.24) (2.88) (2.43) (2.05) (2.70)
TURINC t-1 −0.003 0.003 −0.003 0.003 −0.004 0.003
(−0.36) (1.01) (−0.38) (1.00) (−0.55) (0.84)
INVEST t-1 0.280a 0.082a 0.280a 0.082a 0.286a 0.084a
(5.25) (3.89) (5.26) (3.90) (5.35) (3.97)
FAGE t-1 −0.055a −0.008b −0.055a −0.008b −0.063a −0.010a
(−4.85) (−2.00) (−5.03) (−2.01) (−5.41) (−2.63)
TURNOVER t-1 −0.031 −0.009 −0.032 −0.009 −0.035 −0.010
(−0.82) (−0.61) (−0.82) (−0.62) (−0.92) (−0.69)
CEOAGE t-1 −0.000 −0.001 −0.001 −0.000 −0.001 −0.000
(−0.17) (−0.22) (−0.14) (−0.20) (−0.24) (−0.29)
ZSCORE t-1 0.064a 0.020a 0.063a 0.019a 0.064a 0.020a
(6.92) (5.91) (6.89) (5.88) (6.90) (5.91)
R&DINTEN t-1 −0.645a −0.216a −0.648a −0.216a −0.666a −0.219a
(−3.20) (−2.67) (−3.21) (−2.68) (−3.21) (−2.69)
Adjusted R2 0.047 0.043 0.047 0.043 0.046 0.043
N 15,853 15,853 15,853 15,853 15,853 15,853
Panels A and B present the results using alternative crash risk measures and alternative proxies of CEO power, respectively. Panel C presents the
results controlling for additional firm- and CEO-related variables. “all controls” refer to all independent variables in the baseline model, except the
CEO power measure, as in Model 2 of Table 2. We cluster standard errors by both firm and year. We present t-values in parentheses. The 1%, 5%,
and 10% significance levels of the coefficients are denoted a, b, and c, respectively. All continuous variables are winsorized at the 1% level. We
present the variable definitions in the Appendix.
21
results in Panel C of Table 9. Following Jia (2018), we use the pay gap between the CEO and other senior executives as a measure of
tournament incentives (TURINC). Jia (2018) argues that stronger TURINC can elicit more negative managerial behaviour by non-
CEO executives to increase their chance of promotion to CEO and documents that the TURINC is significantly and positively related
to the firm's future stock price crash risk. We control TURINC to address any concern that our measure of CEO power captures
stronger TURINC. We further control for R&D intensity as a proxy for excessive risk taking (Hoskisson et al., 1993); CEO age as a
proxy for CEO experience and CEO risk-taking attitude; firm investment; and firm age as a proxy for the level of maturity in dealing
with external and internal difficulties and opportunities that a firm faces in various stages of its life cycle. We use the modified Z-
score (Hasan et al., 2014) as a proxy for financial distress and CEO turnover to account for the possible effect of incoming CEOs
taking a big bath to reverse poor prior decisions (Weisbach, 1995).24 Results in Panel C of Table 9 show that, controlling for these
additional variables, CEO power still has a positive relation with crash risk.25 The coefficient estimate for CEOFO on NCSKEW is
0.081, which indicates that after controlling for additional firm and CEO characteristics, when a firm has a powerful founder CEO
(CEOFO), its NCSKEW will be 8.1% higher than firms without CEOFO. This is equivalent to 56.6% (= 0.081 / 0.1431) of the
average NCSKEW in our sample.
Our results after controlling for TURINC also indicate that the formal status such as founder or president and chairman, represent
a unique sources of CEO power that are not subsumed by the dimension of CEO power related to CEO pay.
8.3.4. CEO power – crash risk relation in crisis versus non-crisis periods
Our sample period includes two notable stock market crash periods: the dot-com bubble in the early 2000s (2001−2002) and the
global financial crisis (2007–2009). To ensure that our results are not driven by these events, we run our baseline regression model
for the crisis and non-crisis periods. Our findings in Table A10 of the internet appendix show that the positive relation between CEO
power and crash risk is evident in both crisis and non-crisis periods, except when we use CEOPRCH as the CEO power measure in
financial crisis period. Overall, we conclude that the association between CEO power and crash risk is robust in different sub-sample
periods and is not more pronounced during crisis time.
9. Conclusion
We investigate the role of managerial power in stock price crash risk. We provide evidence that CEO power is positively associated
with firm-specific stock price crash risk. Our results hold when controlling for previously documented crash determinants, such as
earnings management, tax avoidance, CFO option incentives, and CEO overconfidence. We address endogeneity concern by using
several approaches such as instrumental variable regression analysis; firm fixed effect regression; difference-in-difference analysis
using exogenous CEO turnover events; and propensity score matching approach. We still find a positive association between CEO
power and firm-specific stock price crash risk. Our analysis using instrumental variable regression needs to be carefully interpreted as
there may still be concern about the validity of the exclusion restriction.
Further, we find that powerful CEOs have positive relation with financial restatements; the proportion of negative earnings
guidance and use of negative words relative to positive words in financial reports. These findings suggest that the ability of powerful
CEOs to withhold bad news is related to the positive association between CEO power and crash risk.
Moreover, we examine the motivations of powerful CEOs to hoard bad news that result in subsequent stock price crashes. We
document that the positive association between powerful CEOs and crash risk is mostly evident in the subsample of firms where CEO
compensation is more sensitive to stock price or where the CEOs possess lower general skills. We do not find any evidence that firms
with powerful CEOs experience higher stock price jumps than others. Overall, our findings that CEO power is associated with future
extreme negative stock returns but not on positive outcomes (price jumps), suggesting that decisions with extreme negative con-
sequences are more likely to be made when the CEOs are more powerful.
Acknowledgments
We gratefully acknowledge helpful comments from three anonymous referees, Editor Bart Lambrecht, Renee Adams,
Robert Faff, Stephen Gray, Ferdinand Gul, Allan Hodgson, Janice How, Chang Mo Kang, Avanidhar Subrahmanyam, Michael
Theobald, Cameron Truong, Gary Twite, Peter Verhoeven, and participants at the 2015 Financial Research Network Conference,
the 2016 Financial Markets and Corporate Governance Conference, the 2017 Multinational Finance Conference, and the 2017
European Financial Management Conference, and seminar series of La Trobe University, the Queensland University of
Technology, the University of New England, the University of Queensland, the University of South Australia, and the University of
Southern Queensland.
24
To further account for the potential effect of CEO turnover, in an untabulated test, we restrict our sample to firms with CEO tenure of three years
or more and still observe a positive relation between CEO power and crash risk.
25
Tournament incentive (TURINC) is insignificant in Panel C of Table 9. However, when we control for TURINC in our baseline model without
controlling for CEO age, CEO turnover, company age, potential financial distress, and R&D intensity, we find that TURINC is positively related to
crash risk as documented in Jia (2018). We present this results in Table A9 of the internet appendix.
22
Appendix: Variable definitions
CEO power at t - 1
CEOFEPCB A binary variable equal to 1 if the CEO is the founder and either the president, chair, or both.
CEOFO A binary variable equal to 1 if the CEO is also the founder.
CEOPRCH A binary variable equal to 1 if the CEO is both the president and the chair.
ECEOFO Predicted value of CEO power using CEOFO as a power measure.
ECEOFEPCB Predicted value of CEO power using CEOFEPCB as a power measure.
NCEOTITLE The number of titles the CEO holds. A nominal CEO without an additional title receives one point and a powerful CEO receives one extra point for
each additional title, such as chair, president, COO, CFO, vice president, vice chair, director of the board, founder, and member of key
committees.
CEOPIND A CEO power measure constructed with principal component analysis using the various other titles a CEO holds, such as founder, COO, CFO,
chair, president, chair and president, vice chair, vice president, member of the board, member of the compensation committee, and member of the
audit committee.
CPS CPS is the fraction of the total compensation CEO receives to the total compensation received by the top-five executives.
Crash risk variables at t
CRCOUNT The number of firm-specific weekly returns exceeding 3.2 standard deviations below the mean firm-specific weekly return over the fiscal year.
CRASHD The crash dummy is an indicator variable taking the value of 1 for a firm–year that experiences one or more firm-specific weekly returns falling
3.2 standard deviations below the mean firm-specific weekly returns over the fiscal year. We follow Kim et al. (2011a, 2011b) to calculate this
variable.
DUVOL The log of the ratio of the standard deviation of firm-specific down weekly returns to the standard deviation of up weekly returns during the fiscal
year. We follow Kim et al. (2011a) to calculate DUVOL.
NCSKEW The negative skewness of firm-specific weekly returns over the fiscal year. We follow Kim et al. (2011a, 2011b) to calculate NCSKEW.
Other variables at t - 1 or t - 2
ACQ A binary variable equal to 1 if the firm reports acquisition activity in the fiscal year and zero otherwise.
BIG4 Indicator variable equal to 1 if the firm employs one of the four largest audit firms and zero otherwise.
BLOCK Natural log of the total shareholdings of the top 5% of owners.
BOIND Board independence, measured as the proportion of independent directors on the board.
BOSIZE Natural log of 1 plus board size.
CEOAGE The age of the firm's CEO.
CEOGSK CEO general skills, obtained through a principal component analysis on five proxies of general managerial ability: the number of past positions,
the number of past firms, the number of past industries worked, a CEO experience dummy, and a conglomerate experience dummy.
CEOOPIN The incentive ratio for CEO option holdings, defined similarly as for Kim et al. (2011a).
CEOOWN The percentage of shares the CEO owns.
CFOOPIN The incentive ratio for CFO option holdings, defined similarly as for Kim et al. (2011a).
DAMEX Dummy variable for firms listed on the American Stock Exchange (AMEX).
DDELAWARE Dummy variable for firms incorporated in Delaware.
DELTA The dollar changes in the value of the option or restricted stock grants, shareholdings, and any restricted stock and option holdings for a 1%
change in the stock price.
DFO A binary variable that takes the value of 1 if the founder died in year t-2. In the case of multiple founders, we consider the death of the last
founder.
DNASDAQ Dummy variable for firms listed on the NASDAQ.
DS&P500 Dummy variable for S&P 500 firms.
DTURN The average monthly share turnover over the fiscal year minus the average monthly share turnover over the last fiscal year. The monthly share
turnover is the ratio of the monthly trading volume of total shares outstanding during the month.
HDELTA A dummy variable takes a value of 1 if CEO delta is greater than the sample median value of CEO delta.
HGSK A dummy variable takes a value of 1 if CEO general skills is greater than the sample median value of CEOGSK.
HIODED A dummy variable takes a value of 1 if IODED is greater than the sample median value of IODED.
HOLD100 A measure of CEO overconfidence. We classify a CEO as overconfident if the CEO holds stock options that are more than 100% in the money for at
least two years during the sample period. We follow Campbell et al. (2011) to estimate average CEO stock option moneyness for each year. Then
we assign an indicator value of 1 for an overconfident CEO beginning with the second time the CEO exhibits the optimistic behavior.
HTOIND A dummy variable takes a value of 1 if TOIND is greater than the sample median value of TOIND.
INVEST Investment defined as (CAPX − SPPE)/ lag PPENT (Frank and Shen, 2016).
IODED IODED is yearly dedicated institutional ownership, following Bushee (1998), calculated as the average percentage of shares outstanding held by
dedicated institutional investors over the four quarters of the firm's fiscal year. The data are from the Thomson Reuters Institutional Holdings
(13F) database.
JUMPD A dummy variable taking the value of one if at least one weekly firm-specific return in the entire fiscal year is 3.20 standard deviations above the
mean weekly return over the fiscal year.
JUMPCOU The number of weeks during the entire fiscal year when the firm-specific weekly return is 3.20 standard deviations above the mean weekly return.
LAT Natural logarithm of 1 plus total asset.
LCEOPAY Natural logarithm of 1 plus the CEO's salary and bonus.
LEV Total long-term debt scaled by total assets.
LFAGE Natural logarithm of 1 plus firm age, in years. We manually collect the founding year of the firm or its ancestor to determine its age.
LFOALIVE Natural logarithm of 1 plus the number of founders alive.
LDELTA A dummy variable takes a value of 1 if CEO delta is lower than or equal to the sample median value of CEO delta.
LGSK A dummy variable takes a value of 1 if CEO general skills is lower than or equal to the sample median value of CEOGSK.
LIODED A dummy variable takes a value of 1 if IODED is lower than or equal to the sample median value of IODED.
LMVE Natural logarithm of 1 plus the market value of equity on the balance sheet date.
LTOIND A dummy variable takes a value of 1 if TOIND is lower than or equal to the sample median value of TOIND.
MTB The market value of equity divided by the book value of equity.
NETPOEG Ratio of the total negative earnings guidance to the total positive earnings guidance released by the firm as documented in the RavenPack News
Analytics.
23
NETPOW Ratio of the total negative word count to the total positive word count in the annual report.
OPAQUE The moving sum of the absolute value of discretionary accruals over the three-year period from t - 1 to t - 3, where discretionary accruals are
calculated based on the modified Jones model (Dechow et al., 1995).
OPAQUESQ Square of OPAQUE.
PRSHELT The estimated probability of engaging in tax-sheltering activities based on Wilson's (2009) model in Table 4 of Lisowsky (2010). We use model 1
of Table 4 of Lisowsky (2010, p. 1709) to calculate PRSHELT.
R&DINTEN R&D intensity is measured as the ratio of R&D expenditures to total assets, with missing values replaced by zero.
REST A dummy variable taking the value of 1 if the firm has a financial restatement in the fiscal year and zero otherwise.
RET Average firm-specific weekly return during the fiscal year times 100.
ROA Income before extraordinary items divided by lagged total assets.
SIGMA Standard deviation of firm-specific weekly returns over the entire fiscal year.
TENURE Natural logarithm of 1 plus the tenure of the CEO for a firm at the end of the fiscal year.
TENURESQ Square of TENURE.
TOIND The takeover index of Cain et al. (2017).
TURINC This is the pay gap between the CEO and other senior executives based on Jia (2018)
TURNOVER A dummy variable taking the value of 1 if there is a change in CEO.
ZSCORE Following Hasan et al. (2014), we use the modified Z-score = (1.2WCAP + 1.4RE + 3.3PI + 0.999SALE)/AT, where WCAP is the working
capital, RE is retained earnings, PI is pretax income, SALE is total sales, and AT is total assets. We exclude the market-to-book ratio, since it is a
control in our main regression.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcorpfin.2020.101582.
References
Adams, R.B., Almeida, H., Ferreira, D., 2005. Powerful CEOs and their impact on corporate performance. Rev. Financ. Stud. 18 (4), 1403–1432.
Adams, R., Almeida, H., Ferreira, D., 2009. Understanding the relationship between founder–CEOs and firm performance. J. Empir. Financ. 16 (1), 136–150.
Ali, A., Zhang, W., 2015. CEO tenure and earnings management. J. Account. Econ. 59 (1), 60–79.
An, H., Zhang, T., 2013. Stock price synchronicity, crash risk, and institutional investors. J. Corp. Finance. 21, 1–15.
Andreou, P.C., Antoniou, C., Horton, J., Louca, C., 2016. Corporate governance and firm-specific stock price crashes. Eur. Financ. Manag. 22 (5), 916–956.
Andreou, P.C., Louca, C., Petrou, A.P., 2017. CEO age and stock price crash risk. Rev Financ. 21 (3), 1287–1325.
Archambeault, D.S., DeZoort, F.T., Hermanson, D.R., 2008. Audit committee incentive compensation and accounting restatements. Contemp. Account. Res. 25 (4),
965–992.
Armstrong, C.S., Jagolinzer, A.D., Larcker, D.F., 2010. Chief executive officer equity incentives and accounting irregularities. J. Account. Res. 48 (2), 225–271.
Armstrong, C.S., Balakrishnan, K., Cohen, D., 2012. Corporate governance and the information environment: evidence from state antitakeover laws. J. Account. Econ.
53 (1), 185–204.
Baginski, S.P., Campbell, J.L., Hinson, L.A., Koo, D.S., 2018. Do career concerns affect the delay of bad news disclosure? Account. Rev. 93, 61–95.
Ball, R., 2009. Market and political/regulatory perspectives on the recent accounting scandals. J. Account. Res. 47, 277–323.
Bao, D., Kim, Y., Mian, G.M., Su, L., 2018. Do managers disclose or withhold bad news? Evidence from short interest. The Account. Rev. 94 (3), 1–26.
Bebchuk, L., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Rev. Financ. Stud. 22 (2), 783–827.
Bebchuk, L.A., Cremers, K., Peyer, U.C., 2011. The CEO pay slice. J. Financ. Econ. 102 (1), 199–221.
Bergstresser, D., Philippon, T., 2006. CEO incentives and earnings management. J. Financ. Econ. 80 (3), 511–529.
Bernile, G., Bhagwat, V., Rau, P.R., 2017. What doesn’t kill you will only make you more risk-loving: early-life disasters and CEO behavior. J. Financ. 72 (1), 167–206.
Bertrand, M., Mullainathan, S., 2003. Enjoying the quiet life? Corporate governance and managerial preferences. J. Political Econ. 111 (5), 1043–1075.
Burns, N., Kedia, S., 2006. The impact of performance-based compensation on misreporting. J. Financ. Econ. 79 (1), 35–67.
Bushee, B.J., 1998. The influence of institutional investors on myopic R&D investment behavior. Account. Rev. 73, 305–333.
Bushman, R.M., Williams, C.D., Wittenberg-moerman, R., 2017. The informational role of the media in private lending. J. Account. Res. 55 (1), 115–152.
Cain, M.D., McKeon, S.B., Solomon, S.D., 2017. Do takeover laws matter? Evidence from five decades of hostile takeovers. J. Financ. Econ. 124 (3), 464–485.
Callen, J.L., Fang, X., 2013. Institutional investor stability and crash risk: monitoring versus short-termism? J. Bank. Financ. 37 (8), 3047–3063.
Callen, J.L., Fang, X., 2015a. Short interest and stock price crash risk. J. Bank. Financ. 60, 181–194.
Callen, J.L., Fang, X., 2015b. Religion and stock price crash risk. J. Financ. Quant. Anal. 50 (1–2), 169–195.
Campbell, T.C., Gallmeyer, M., Johnson, S.A., Rutherford, J., Stanley, B.W., 2011. CEO optimism and forced turnover. J. Financ. Econ. 101, 695–712.
Chang, X., Chen, Y., Zolotoy, L., 2016. Stock liquidity and stock price crash risk. J. Financ. Quant. Anal. 52 (4), 1605–1637.
Chen, J., Hong, H., Stein, J.C., 2001. Forecasting crashes: trading volume, past returns, and conditional skewness in stock prices. J. Financ. Econ. 61 (3), 345–381.
Chen, X., Harford, J., Li, K., 2007. Monitoring: which institutions matter? J. Financ. Econ. 86 (2), 279–305.
Cheng, Q., Warfield, T.D., 2005. Equity incentives and earnings management. Account. Rev. 80 (2), 441–476.
Coles, J.L., Daniel, N.D., Naveen, L., 2006. Managerial incentives and risk-taking. J. Financ. Econ. 79 (2), 431–468.
Core, J., Guay, W., 2002. Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. J. Account. Res. 40 (3), 613–630.
Cunat, V., Guadalupe, M., 2009a. Executive compensation and competition in the banking and financial sectors. J. Bank. Financ. 33, 495–504.
Cunat, V., Guadalupe, M., 2009b. Globalization and the provision of incentives inside the firm: the effect of foreign competition. J. Labor Econ. 27, 179–212.
Custódio, C., Ferreira, M.A., Matos, P., 2013. Generalists versus specialists: lifetime work experience and chief executive officer pay. J. Financ. Econ. 108 (2), 471–492.
Custódio, C., Ferreira, M.A., Matos, P.P., 2019. Do general managerial skills spur innovation? Manag. Sci. 65 (2), 459–476.
Daily, C.M., Johnson, J.L., 1997. Sources of CEO power and firm financial performance: a longitudinal assessment. J. Manage. 23, 97–117.
Dechow, P.M., Sloan, R.G., Sweeney, A.P., 1995. Detecting earnings management. Account. Rev. 70 (2), 193–225.
Dichev, I.D., Graham, J.R., Harvey, C.R., Rajgopal, S., 2013. Earnings quality: evidence from the field. J. Account. Econ. 56 (2–3), 1–33.
Edmans, A., Goncalves-Pinto, L., Grown-Xu, M., Wang, Y., 2018. Strategic news releases in equity vesting months. Rev. Financ. Stud. 31 (11), 4099–4141.
Efendi, J., Srivastava, A., Swanson, E.P., 2007. Why do corporate managers misstate financial statements? The role of option compensation other factors. J. Financ.
Econ. 85 (3), 667–708.
Eisfeldt, A.L., Kuhnen, C.M., 2013. CEO turnover in a competitive assignment framework. J. Financ. Econ. 109 (2), 351–372.
Fahlenbrach, R., 2009. Founder–CEOs, investment decisions, and stock market performance. J. Financ. Quant. Anal. 44 (2), 439–466.
Feng, M., Ge, W., Luo, S., Shevlin, T., 2011. Why do CFOs become involved in material accounting manipulations? J. Account. Econ. 51 (1), 21–36.
Finkelstein, S., 1992. Power in top management teams: dimensions, measurement, and validation. Acad. Manag. J. 35 (3), 505–538.
24
Fracassi, C., Tate, G., 2012. External networking and internal firm governance. J. Finance. 67 (1), 153–194.
Frank, M.Z., Shen, T., 2016. Investment and the weighted average cost of capital. J. Financ. Econ. 119 (2), 300–315.
Friedman, H.L., 2014. Implications of power: when the CEO can pressure the CFO to bias reports. J. Account. Econ. 58 (1), 117–141.
Garicano, L., Rossi-Hansberg, E., 2006. Organization and inequality in a knowledge economy. Q. J. Econ. 121, 1383–1435.
Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. Q. J. Econ. 118 (1), 107–156.
Graham, J.R., Harvey, C.R., Rajgopal, S., 2005. The economic implications of corporate financial reporting. J. Account. Econ. 40, 3–73.
Graham, J.R., Kim, H., Leary, M.T., 2017. CEO Power and Board Dynamics. Available at SSRN:. https://doi.org/10.2139/ssrn.2938120.
Grinstein, Y., Hribar, P., 2004. CEO compensation and incentives: evidence from M&A bonuses. J. Financ. Econ. 73 (1), 119–143.
Hamm, S., Li, E.X., Ng, J., 2018. Management Forecasts and Bad News Hoarding: Evidence from Stock Price Crashes. Available at SSRN: https://ssrn.com/abstract=
2055008.
Hasan, I., Hoi, C.K.S., Wu, Q., Zhang, H., 2014. Beauty is in the eye of the beholder: the effect of corporate tax avoidance on the cost of bank loans. J. Financ. Econ. 113
(1), 109–130.
Hermalin, B.E., Weisbach, M.S., 1998. Endogenously chosen boards of directors and their monitoring of the CEO. Am. Econ. Rev. 88 (1), 96–118.
Hoskisson, R.E., Hitt, M.A., Hill, C.W., 1993. Managerial incentives and investment in R&D in large multiproduct firms. Organ. Sci. 4 (2), 325–341.
Hutton, A.P., Marcus, A.J., Tehranian, H., 2009. Opaque financial reports, R2, and crash risk. J. Financ. Econ. 94 (1), 67–86.
Jia, N., 2018. Tournament incentives and stock price crash risk. Account. Horiz. 32 (3), 101–121.
Jin, L., Myers, S.C., 2006. R2 around the world: new theory and new tests. J. Financ. Econ. 79 (2), 257–292.
Jiraporn, P., Liu, Y., Kim, Y.S., 2014. How do powerful CEOs affect analyst coverage? Eur. Financ. Manag. 20 (3), 652–676.
Kasznik, R., Lev, B., 1995. To warn or not to warn: management disclosures in the face of an earnings surprise. Account. Rev. 70 (1), 113–134.
Khanna, V., Kim, E., Lu, Y., 2015. CEO connectedness and corporate fraud. J. Financ. 70 (3), 1203–1252.
Kim, J.B., Zhang, L., 2016. Accounting conservatism and stock price crash risk: firm-level evidence. Contemp. Account. Res. 33 (1), 412–441.
Kim, J.B., Li, Y., Zhang, L., 2011a. Corporate tax avoidance and stock price crash risk: firm-level analysis. J. Financ. Econ. 100 (3), 639–662.
Kim, J.B., Li, Y., Zhang, L., 2011b. CFOs versus CEOs: equity incentives and crashes. J. Financ. Econ. 101 (3), 713–730.
Kim, Y., Li, H., Li, S., 2014. Corporate social responsibility and stock price crash risk. J. Bank. Financ. 43, 1–13.
Kim, J.B., Wang, Z., Zhang, L., 2016. CEO overconfidence and stock price crash risk. Contemp. Account. Res. 33 (4), 1720–1749.
Kothari, S.P., Shu, S., Wysocki, P.D., 2009. Do managers withhold bad news? J. Account. Res. 47 (1), 241–276.
Lel, U., Miller, D.P., 2015. Does takeover activity cause managerial discipline? Evidence from international M&a laws. Rev. Financ. Stud. 28 (6), 1588–1622.
Li, F., 2008. Annual report readability, current earnings, and earnings persistence. J. Account. Econ. 45 (2), 221–247.
Lilenfeld-Toal, U.V., Ruenzi, S., 2014. CEO ownership, stock market performance, and managerial discretion. J. Financ. 69 (3), 1013–1050.
Lisowsky, P., 2010. Seeking shelter: empirically modeling tax shelters using financial statement information. Account. Rev. 85 (5), 1693–1720.
Lo, K., Ramos, F., Rogo, R., 2017. Earnings management and annual report readability. J. Account. Econ. 63 (1), 1–25.
Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation: An empirical analysis. J. Financ. Econ. 20, 293–315.
Morck, R., Shleifer, A., Vishny, R.W., 1989. Alternative mechanisms for corporate control. Am. Econ. Rev. 79 (4), 842–852.
Morse, A., Nanda, V., Seru, A., 2011. Are incentive contracts rigged by powerful CEOs? J. Financ. 66 (5), 1779–1821.
Nguyen, B.D., Nielsen, K.M., 2010. The value of independent directors: evidence from sudden deaths. J. Financ. Econ. 98 (3), 550–567.
Norburn, D., 1989. The chief executive: a breed apart. Strat. Man. J. 10 (1), 1–15.
Roberts, M.R., Whited, T.M., 2013. Endogeneity in empirical corporate finance1. In: Handbook of the Economics of Finance. Vol. 2. Elsevier, Amsterdam, pp.
493–572.
Robin, A., Zhang, H., 2015. Do industry-specialist auditors influence stock price crash risk? Auditing - J. Pract. Th. 34 (3), 47–79.
Skinner, D., 1994. Why firms voluntarily disclose bad news. J. Account. Res. 32 (1), 38–61.
Skinner, D., 1997. Earnings disclosue and stockholders lawsuits. J. Account. Econ. 23 (3), 249–283.
Stock, J.H., Yogo, M., 2005. Testing for weak instruments in linear IV regression. In: Andrews, D.W., Stock, J.H. (Eds.), Identification and Inference for Econometric
Models: Essays in Honor of Thomas Rothenberg. Cambridge University Press, New York, pp. 80–108.
Weisbach, M.S., 1995. CEO turnover and the firm’s investment decisions. J. Financ. Econ. 37 (2), 159–188.
Whisler, T.L., Meyer, H., Baum, B.H., Sorensen, P.F., 1967. Centralization of organizational control: an empirical study of its meaning and measurement. J. Bus. 40 (1),
10–26.
Wilson, R.J., 2009. An examination of corporate tax shelter participants. Account. Rev. 84 (3), 969–999.
Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data, Edition 2. Vol. 1 MIT Press March.
Zhao, Y., Chen, K.H., 2008. Staggered boards and earnings management. Account. Rev. 83 (5), 1347–1381.
25

TESIS - Mamun Et Al 2020 - Powerful CEO

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TESIS - Mamun Et Al 2020 - Powerful CEO

Uploaded by

Copyright:

Available Formats

Journal of Corporate Finance 62 (2020) 101582

Contents lists available at ScienceDirect

Journal of Corporate Finance

Powerful CEOs and stock price crash risk

ARTICLE INFO ABSTRACT

2. Related literature and empirical predictions

3.1. Measures of stock price crash risk

r j, = j + 1j,rm, 2 + 2j,rm, 1 + 3j,rm, + 4j,rm, + 1 + 5j,rm, + 2 + j (1)

3.2. Measures of CEO power

3.3. Sample selection and descriptive statistics

3.4. Correlation matrix and univariate test

CEO power variables at t - 1

Crash risk variables at t

4. Empirical results: CEO power and stock price crash risk

4.1. Baseline results

CRit = 0 + 1 CEOPOWERi, t 1 + Control Variablesi, t 1 + i, t (2)

4.2. Addressing endogeneity concerns

4.2.1. Instrumental variable regression

NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt

4.2.2. Firm fixed effect regression

Power measures CEOFOt-1 CEOFEPCBt-1

Power measures ECEOFO ECEOFEPCB

Variables NCSKEWt DUVOLt NCSKEWt DUVOLt

(1) (2) (3) (4) (5) (6) (7) (8)

(continued on next page)

CEO power CEOFO CEOFEPCB CEOPRCH

Variables NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt

(1) (2) (3) (4) (5) (6)

CEOPOWERt-1 0.139b 0.057b 0.151b 0.061a 0.101a 0.039a

Panel D: Crash risk around CEO exogenous turnover events

(1) (2) (3) (4) (5) (6) (7)

CEOFO Mean Mean

Δ Industry-Adjusted NCSKEWt 0.073 0.593 −0.674 0.520 2.140b −0.746 −3.967a

Power measures CEOFO CEOFEPCB CEOPRCH

NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt

Variables (1) (4) (5) (2) (3) (6)

CHANGEt-1 0.064 0.023 0.002 0.031 −0.002 0.003

Power measures CEOFO CEOFEPCB CEOPRCH

NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt

Variables (1) (4) (5) (2) (3) (6)

Power measures CEOFO CEOFEPCB CEOPRCH

NCSKEWt DUVOLt NCSKEWt DUVOLt NCSKEWt DUVOLt

Variables (1) (4) (5) (2) (3) (6)

CHANGE t-1 0.031 0.030 0.044 0.034 0.065 0.018

5. CEO power and bad news hoarding

powerful CEOs and proxies for hiding bad news.

6. When do powerful CEOs hide bad news?

(1) (2) (3) (4) (5) (6) (8) (8) (9)

CEOFOt-1 0.706a −0.059b −0.116b

6.1. CEO power, CEO delta, and crashes

CEOFO CEOFEPCB CEOPRCH CEOFO CEOFEPCB CEOPRCH

6.2. CEO power, CEO skills, and crashes

CEOFO CEOFEPCB CEOPRCH CEOFO CEOFEPCB CEOPRCH

7. CEO power, external monitoring mechanisms, and stock price crashes

8. Further tests and robustness checks

8.1. CEO power and price jumps

8.2. CFO option incentives

8.3. Further robustness checks

8.3.1. Alternative measures of stock price crash risk

Power measures NCSKEW DUVOL

CEOFO CEOFEPCB CEOPRCH CEOFO CEOFEPCB CEOPRCH

Power measures CEOFO CEOFEPCB CEOPRCH CEOFO CEOFEPCB CEOPRCH

8.3.2. Alternative measures of CEO power