The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/2044-1398.htm

Managerial
Managerial overconfidence, overconfidence
firm transparency, and stock
price crash risk
271
Evidence from an emerging market
Quanxi Liang Received 16 January 2019
Revised 11 June 2019
School of Business, Guangxi University, Nanning, China 20 July 2019
Accepted 9 October 2019
Leng Ling
Bunting College of Business,
Georgia College and State University, Milledgeville, Georgia, USA
Jingjing Tang
School of Business, Guangxi University, Nanning, China
Haijian Zeng
Guangxi University, Nanning, China, and
Mingming Zhuang
Guangdong University of Foreign Studies, Guangzhou, China

Abstract
Purpose – The purpose of this paper is to empirically analyze whether and how managerial overconfidence
affects stock price crash risk.
Design/methodology/approach – Based on a large sample of Chinese non-state-owned firms from 2000 to
2012, this study employs methods including multiple linear regression model, Heckman two-stage treatment
effect procedure, firm fixed effects model and event study to clarify the causality relationship between
managerial overconfidence and crash risk.
Findings – The authors find that firms with overconfident managers (chief executive officer or board chairs)
are more likely to experience future stock price crashes than firms with non-overconfident managers. The
effect of overconfidence on crash risk is more pronounced for firms with low transparency, suggesting that
firm opacity facilitates overconfident managers’ bad news hoarding activities, which, in turn, increases stock
price crash risk. The authors also show evidence that overconfident managers tend to disclose good news in a
timely manner.
Originality/value – The authors add to the growing literature on stock price crash risk. Specifically, the
authors find that the cognitive bias of board chair plays an important role in the bad news hoarding activities,
thereby increasing the likelihood of stock price crash. This study also contributes to the literature that
addresses the effects of managerial overconfidence on corporate finance issues.
Keywords China, Managerial overconfidence, Stock price crash risk, Firm transparency
Paper type Research paper

1. Introduction
Stock price crash risk is a topic of significant interest in financial economics. Prior studies
have demonstrated that stock price crash risk is closely related to leverage effect (Christie,
1982), volatility feedback (Campbell and Hentschel, 1992; French et al., 1987), stochastic
bubbles (Blanchard and Watson, 1982) and differences among investors’ beliefs (Hong and
Stein, 2003; Chen et al., 2001). Considerably recent research provides new evidence that stock
China Finance Review
JEL Classification — M12, G14, G34 International
Vol. 10 No. 3, 2020
Liang would like to thank the financial support from the National Natural Science Foundation of pp. 271-296
© Emerald Publishing Limited
China (Grant Nos. 71362013 and 71762005). Tang would like to thank the financial support from the 2044-1398
National Natural Science Foundation of China (Grant No. 71764001). DOI 10.1108/CFRI-01-2019-0007
CFRI price crash could be caused partially by agency conflicts between firm managers and
10,3 outside investors. Ball and Shivakumar (2008), Graham et al. (2005) and Kothari et al. (2009)
argue that managers may strategically conceal negative information due to their concerns
regarding, for example, reputation, compensation and career development. When bad news
accumulates to a level that is considerably costly to hide, the release of clustered negative
information triggers a collapse of stock prices (Hutton et al., 2009; Jin and Myers, 2006; Kim
272 et al., 2011a, b). In the theoretical framework proposed by Jin and Myers (2006), poor investor
protection and firm opacity are considered as two key factors leading to crash risks, because
the opaque information environment provides the necessary condition for the entrenched
managers to hide bad news. Jin and Myers argue that opacity alone will not affect crash
risks if the manager is loyal to shareholders. They add that when a firm is completely
transparent, a lack of investor protection will not also cause a stock price crash.
This paper investigates the impact of managerial overconfidence on stock price crash
risk and how firm opacity may affect this relation in China. Unlike prior scholars who
analyzed the determinants of stock price crash risk from the agency conflict perspective, we
study the influence of managers’ personal traits, in this case, overconfidence, on their bad
news hoarding behavior. Although substantial evidence has proved the existence of
overconfidence in human judgment (Alicke, 1985; Baker and Wurgler, 2012), this cognitive
bias is considerably prominent among top managers (Cooper et al., 1988; Graham et al.,
2013). The finance literature documents that managerial overconfidence has a significant
influence on financial reports and information disclosure. Schrand and Zechman (2012)
document that overconfident managers are more likely to be involved in accounting frauds.
Further, Ahmed and Duellman (2013) find that overconfident managers tend to delay the
release of bad news but disclose good news in a timely manner. These findings imply that
managerial overconfidence could be one of the driving forces for managers’ hoarding bad
news, thereby resulting in an increasing stock price crash risk.
China provides a favorable environment for studying the impact of managerial
overconfidence and firm opacity on stock price cash risk. On the one hand, China’s
Confucian culture and its economic transition cultivate overconfidence among Chinese
managers. The psychology and sociology literature has provided extensive evidence that
respondents under the influence of Asian cultures (e.g. in China) exhibit markedly higher
degrees of overconfidence than do respondents affected by western cultures (e.g. in the
USA) (Yates et al., 1989, 1996, 1997). Asian cultures and the China’s Confucian culture in
particular are thought to make people behave in such a manner that has been characterized
in the literature as “Asian overconfidence” (Seok, 2007). The Confucian culture advocates an
organizational hierarchy that emphasizes the absolute authority of a leader. Therefore,
leaders of different kinds of organization in China often possess overwhelming power,
which leads to managerial overconfidence ( Jiang et al., 2009; Fast et al., 2012).
Meanwhile, numerous entrepreneurs have emerged from China’s economic reforms.
Many of these rich executives may overestimate their abilities and skills ( Jiang et al., 2009).
On the other hand, the information environment in China is considerably less transparent
than that of developed countries ( Jin and Myers, 2006; Piotroski and Wong, 2012),
providing a good setting to test the influence of firm opacity on the risk of stock
price crash.
For firms in developed countries such as the USA, chief executive officer (CEO) is usually
the person in charge of day-to-day business. For most of Chinese firms, however, the person
who is in charge is not the CEO. Oftentimes, it is chairman of the board of directors ( Jiang
and Kim, 2016). In China, the board chair is the legal representative of the firm and entitled a
lot of legal power. He/she is appointed by the largest shareholder of the firm (Kato and Long,
2006), and given that ownership structure is highly concentrated in China, this arrangement
suggests that board chair has a lot power in effect, not just legally ( Jiang and Kim, 2016).
In particular, in the study of Kato and Long (2006), when the CEO is also the board chair, Managerial
they designate that person as the CEO. However, when the CEO is not the board chair and if overconfidence
the board chair is (is not) on the payroll, the board chair (CEO) is designated as the CEO.
Therefore, in this study, we consider overconfidence of both CEO and board chair.
We explore a large sample of Chinese non-state-owned firms from 2000 to 2012 and find
that firms with overconfident managers (CEOs or board chairs) are associated with
substantially high stock price crash risk. Moreover, the predictive power of managerial 273
overconfidence with respect to crash risk is more pronounced for firms that have low
earnings quality; have been audited by non-international Big 4 auditors; have a large
dispersion in analysts’ earnings forecasts; and have a low rating on information disclosure.
These findings suggest that an opaque information environment facilitates overconfident
managers’ bad news hoarding activities, which contribute to a future stock price crash.
In addition, we determine a negative association between managerial overconfidence and
future positive jump risk in stock price. This finding indicates that overconfident managers
tend to disclose good news in a timely manner, and thus increase the crash risk of
stock prices.
Our findings contribute to the literature in several ways. First, we add to the growing
literature on stock price crash risk (e.g. Benmelech et al., 2010; Bleck and Liu, 2007; Callen
and Fang, 2013; Hutton et al., 2009; Jin and Myers, 2006; Kim et al., 2011a, b; Xu et al., 2014).
In particular, we find that managers’ cognitive bias (i.e. managerial overconfidence) is an
important driving force for their bad news hoarding behavior, thereby contributing to a
stock price crash. Our work is most closely related to Kim et al. (2016). Kim et al. (2016) also
find that CEO overconfidence is positively associated with future stock price crash risk in
US market. There are, however, significant differences that distinguish our work from Kim
et al. (2016). The most notable difference is that we consider not only CEO overconfidence,
but also that of board chair. An investigation on chairperson is important given that for
most of Chinese firms, board chair is actually the person who is the active controller in
charge of the day-to-day business ( Jiang and Kim, 2016). We find that the cognitive bias of
board chair also plays an important role in the bad news hoarding activities, thereby
increasing the likelihood of stock price crash.
Second, this study contributes to the literature that addresses the effects of managerial
overconfidence on corporate finance issues. Numerous studies using data from developed
countries have examined the overconfidence of CEOs and overlooked the influence of this
cognitive bias of the chairperson of the board. However, in certain countries such as China, it
is the board chairperson who actively controls and runs the firm ( Jiang and Kim, 2016).
Along this line, we provide evidence that CEO overconfidence and chairperson’s
overconfidence both have significant and positive effects on stock price crash risk.
Finally, this study also corresponds to Xu et al. (2013), who find that optimistic analysts
tend to release positive information to investors and ignore negative news, thus increasing
the stock price crash risk. Our research differs from theirs in that we consider overconfident
managers rather than financial analysts.
The rest of the paper is organized as follows. Section 2 presents the related literature and
develops the research hypotheses. Section 3 describes the sample data and explains the
measurement of the key variables used in the empirical analysis. Section 4 presents and
discusses the main results. Section 5 reports the results of the robustness checks and
additional analyses. Section 6 concludes the paper.

2. Related literature and hypothesis development

Economics and social psychology literature have provided extensive evidence that human
perception and decision making are subject to cognitive biases; a typical example is
overconfidence, which can lead to irrational judgment in social and economic activities
CFRI (Alicke, 1985; Baker and Wurgler, 2012; Camerer and Dan, 1999). In the field of psychology,
10,3 the effect of overconfidence is defined as a well-established bias, in which a person’s
subjective confidence in his or her judgments is reliably higher than the objective accuracy
of those judgments. Overconfident people tend to overestimate their probability of success
and underestimate the likelihood of failure (Weinstein, 1980). Furthermore, they are inclined
to attribute success to their abilities and failure to bad luck (Miller and Ross, 1975).
274 Several studies illustrate that overconfidence is considerably prominent among
corporate managers. Cooper et al. (1988) conducted a survey on managers of US firms and
found that these managers perceived the probability of success for their own business is
81 percent, compared with a 58 percent probability of success for other companies. However,
66 percent of the sample firms performed worse after the survey. Graham et al. (2013) find
that overconfidence is more common among CEOs than among other people. Various
studies show that managers’ overconfidence has an influential impact on their decision
making. In particular, managerial overconfidence is associated with low investment
efficiency ( Jiang et al., 2009; Malmendier and Tate, 2005), investments in high-risk projects
(Campbell et al., 2011; Galasso and Simcoe, 2011; Hirshleifer et al., 2012), value-destroying
mergers and acquisitions (Malmendier and Tate, 2005), aggressive debt financing policy
(Graham et al., 2013) and overoptimistic earnings forecasts (Libby and Rennekamp, 2012).
Schrand and Zechman (2012) analyze 49 firms subject to SEC Accounting and Auditing
Enforcement Releases and find that approximately 75 percent of the initial misstatements
reflect a managerial optimistic bias that is not necessarily intentional. Ahmed and Duellman
(2013) argue that overconfident managers tend to overestimate cash flows from firms’
investments, delay loss recognition and use less conservative accounting. They also
document robust evidence of a negative relation between CEO overconfidence and
accounting conservatism.
Based on the extant literature, we contend that managerial overconfidence may affect
stock price crash risk in two ways. First, overconfident managers tend to overestimate their
abilities: they may be overoptimistic about the prospect of the firm while underestimating
the risk of failure, and thus overestimate the free cash flows in the future (Heaton, 2002;
Malmendier and Tate, 2005). In other words, overconfident managers are more likely to
invest in negative NPV projects. Moreover, a rational manager should revise his or her
expectations on cash flows generated by a project and terminate the project if necessary. By
contrast, an overconfident manager tends to overemphasize positive signals and cast doubts
on negative signals (Taylor and Gollwitzer, 1995). This situation suggests that
overconfident managers could irrationally respond to a negative signal of information.
They may ignore the negative information and believe that the poor performance is
temporary. Thus, the value-destroying project will not be suspended or terminated until bad
news accumulates and eventually reaches a threshold that triggers stock price crash.
Although Bleck and Liu (2007) relate stock price crash to the agency conflict between
rational managers and outside investors, our proposed explanation for stock price crash
does not build on this agency conflict. Instead, we assume that managers act in the best
interest of firm owners. Managers invest in or delay the termination of negative NPV
projects because they suffer from the effects of overconfidence.
Second, even if overconfident managers are loyal to shareholders, they differ in the
manner by which they disclose good news vs bad news. On the one hand, overconfident
managers may underestimate the severity of bad news and refrain from disclosing it in a
timely manner. Schrand and Zechman (2012) argue that overconfident managers are
more likely to be in a position in which they are compelled to intentionally misstate
earnings, although the initial misstatement is not necessarily intentional. Consistently,
Myers et al. (2007) find that managers have incentives to window-dress performance
through earnings management in the hope that performance would improve and offset the
unintended misstatements. However, if a firm’s performance fails to improve as expected, Managerial
then the manager will be forced to undertake more aggressive earnings management to overconfidence
alleviate the pressure of reporting bad news.
On the other hand, overconfident managers may maintain their confidence in the project
(which is actually value destroying) even when they observe certain losses or problems. To
relieve the board of directors’ and shareholders concerns, and obviate their queries,
managers may opt to strategically delay the disclosure of bad news, providing the timely 275
release of good news to convince investors of a good prospect. Consequently, bad news is
stockpiled within the firm up to a certain point at which withholding the bad news becomes
impossible for managers. At this moment, all the hitherto unreleased news will come out
simultaneously, resulting in a stock price crash. Based on the preceding analysis, we
propose the first hypothesis as follows:
H1. All else being equal, firms with overconfident managers are more likely to experience
future stock price crashes than firms with non-overconfident managers.
If a firm is completely transparent, then any bad news will be released in a timely manner to
the public and subsequently incorporated into the stock prices. This scenario suggests that
an overconfident manager is unable to withhold bad news within the firm. Thus, he or she is
unlikely to accumulate negative information until the tipping point that leads to a large
adjustment in stock prices. However, an overconfident manager in an environment with
limited transparency would be considerably successful in hiding bad news. In this case,
stock prices do not have an opportunity to adjust to each instance of bad news, thereby
resulting in an increase in crash risk. Therefore, the extent of firm transparency plays an
important role in the effect of managerial overconfidence on stock price crash risk. We
predict that the impact of managerial overconfidence on crash risk is more pronounced for
firms with low transparency. This prediction leads to our second hypothesis as follows:
H2. All else being equal, the positive relation between managerial overconfidence and
crash risk is more pronounced for firms with more opaque information environment.

3. Sample and research design

3.1 Sample construction
Our sample consists of 1,243 non-state-owned firms listed in the Shanghai Stock Exchange
and Shenzhen Stock Exchanges from 2000 to 2012. We obtain these sample firms after the
following screening procedures: following Hutton et al. (2009) to exclude firms with stocks that
were traded less than 26 weeks in a given year; excluding firms in the financial and insurance
industries; and excluding firms that have missing values in the main variables used in the
analysis. The final sample includes 5,996 firm-year observations that are associated with 1,243
unique firms. The firm-level financial data are collected from the China Stock Market &
Accounting Research database (http://csmar.gtadata.com/). We hand-collect the information
regarding the CEO’s educational background. All continuous variables are winsorized at 1
and 99 percentiles to reduce the potential impact of outliers.

3.2 Measurement of stock price crash risk

We follow prior studies, such as Hutton et al. (2009) and Kim et al. (2011a), to compute the
firm-specific crash risk. Specifically, we first estimate the following extended market model:
RET it ¼ ai þbi1 M KRET it1 þbi2 I NDRET it1 þbi3 M KRET it þbi4 I NDRET it
þbi3 M KRET it þ 1 þbi6 I N DRET it þ 1 þeit ; (1)
where RET is the weekly stock return; MKRET the weekly value-weighted average return
CFRI of the market (excluding the return of the pertinent stock); INDRET the weekly
10,3 value-weighted average return of all stocks in the same industry (excluding the return of the
pertinent stock)[1]; ε the error term; and subscript i and τ refer to the firm and trading week,
respectively. The leading and lagged terms of the market and industry returns are employed
to ease the concern of potential nonsynchronous trading that may cause estimation bias
(French et al., 1987). Thereafter, we define firm-specific weekly return as w ¼ log ð1 þ e^Þ,
276 where e^ is the residual from the estimation of Equation (1). We construct two measures of
crash risk based on the firm-specific weekly return.
Following Jin and Myers (2006), our first measure of crash risk, COLLAR, accounts for
the frequency and severity of crashes. We use the firm-specific weekly returns as basis to
create an option portfolio with a long put and short call option. The strike prices are set to
the mean minus 3.2 standard deviations for the former and the mean plus 3.2 standard
deviations for the latter. Under a lognormal distribution, the initial investment for the
aforementioned strategy is zero. Thereafter, COLLAR is computed as the actual profits or
losses from this strategy as a percentage of the stock price. Accordingly, high values
indicate more frequency or more severity of crashes.
Following Chen et al. (2001) and Kim et al. (2011a, b), negative skewness of the
firm-specific weekly returns, NCSKEW, is also used to measure the crash risk. The larger
NCSKEW is, the greater probability that the crash will happen. For firm i in year t,
NCSKEW is defined as follows:
h X i  X 3=2 
3=2
NCSKEW it ¼  nðn1Þ wit = ðn1Þðn2Þ
3
w2it ; (2)

where n is the number of trading weeks over a year and τ indicates the week of the year.

3.3 Measurement of managerial overconfidence

Prior studies have developed various measures of managerial overconfidence, most of
which are constructed based on stock options exercised by managers (Campbell et al.,
2011; Malmendier and Tate, 2005, 2008), executives’ personal characteristics (Puri and
Robinson, 2005; Schrand and Zechman, 2012), media coverage (Hirshleifer et al., 2012),
biases between manager-forecasted earnings and actual earnings (Huang et al., 2011),
CEO’s relative salaries (Huang et al., 2011) and the firm’s investment and financing
activities (Campbell et al., 2011; Schrand and Zechman, 2012).
We aim to disentangle the effect of overconfidence on stock price crash from
that of agency conflict. The measurement of overconfidence based on stock options
exercise, management compensation, earnings forecast errors and investment/financing
activities may capture the information of agency conflicts. Thus, we construct our
overconfidence measure with the information provided by the manager’s personal
characteristics. In China, most listed firms are controlled by large shareholders. As
spokesperson for large shareholders, the chair of the board of directors is similar to the
CEO of firms in developed countries, and is actively involved in the company’s daily
operations ( Jiang et al., 2009). Therefore, we also measure the chair’s overconfidence apart
from that of the CEO.
For each CEO and chair, we first create several dummies based on the following five
personal characteristics:
(1) Gender: the psychology and finance literature show that males are more likely to be
overconfident than females, especially when they are making investment decisions
(Barber and Odean, 2001; Estes and Hosseini, 2001). We create a dummy variable
GenderDum, which equals 1 when the executive is male, and 0 otherwise.
 (2) Age: Taylor (1975) and Forbes (2005) explain that older managers collect more Managerial
information and spend more time before making decisions. They are more conservative overconfidence
when assessing the outcome of a decision. They are also less likely to overestimate
their abilities because they have more experience than younger managers.
Accordingly, younger managers are more likely subject to overconfidence than older
managers do. AgeDum is created as a dummy variable that takes the value of 1 if the
CEO’s (chairman’s) age is below the sample mean, and 0 otherwise. 277
(3) Expertise: the psychology literature contends that experts or professionals are more
confident than common people (Griffin and Tversky, 1992). While experts/professionals
intuitively have more field knowledge than common people to make better decisions,
they tend to believe that they can achieve “better than average” performance (Schrand
and Zechman, 2012), that is, being overconfident. Following Puri and Robinson (2005)
and Schrand and Zechman (2012), we create two dummy variables, EduDum and
MajorDum, to indicate managers’ education degrees and academic majors,
respectively. EduDum equals 1 if a CEO (chair) has obtained a master’s degree or
higher, and 0 otherwise. MajorDum has a value of 1 if the executive pursued a business
or economics major, and 0 otherwise.
(4) Tenure: although overconfident managers tend to overestimate their own
abilities, this cognitive bias decreases in their management experience as they
learn from past experience (Fraser and Greene, 2006). We define a dummy
variable, TenDum, which equals 1 if the manager’s tenure is below the sample
median, and 0 otherwise.
(5) CEO duality: prior studies have obtained evidence that CEOs with substantial
decision-making power tend to be overoptimistic (Adams et al., 2005). CEO duality
may reinforce a CEO’s self-esteem, thereby leading to a strong sense of overconfidence
in decision making. DualDum is defined as an indicator that equals 1 if the CEO also
holds the position of the chair of the board, and 0 otherwise.
Thereafter, we construct a composite measure of managerial overconfidence, that is, OC,
based on the aforementioned six dummy variables. For each firm year, OC_CEO (OC_CHM)
is defined as an indicator that equals 1 if the added value of the six dummy variables is
equal to or higher than 4, and 0 otherwise.

3.4 Measurement of firm transparency

Bushman et al. (2004) define firm transparency as the degree to which outside investors can
effectively obtain specific information about a listed company through annual reports,
announcements, analyst reports and voluntary information disclosure, among others.
Following Hutton et al. (2009), the first measure of firm opacity is earnings quality, which
is defined as the prior three-year moving sum of the absolute value of discretionary accruals.
First, we use the modified Jones model (Dechow et al., 1995) to estimate discretionary accruals.
For each industry and fiscal year, we run the following cross-sectional regression model:

TAit 1 DREV it PPE it

¼ b1 þb2 þb3 þx1it ; (3)
Ait1 Ait1 Ait1 Ait1

where TAit ¼ (ΔCAit−ΔCASHit)−(ΔCLit−ΔCLDit)−DEPit. ΔCAit is the change in current

assets; ΔCASHit the change in cash and cash equivalent; ΔCLit the change in current
liabilities; ΔCLDit the change in the current portion of long-term debt; ΔDEPit the annual
depreciation and amortization; Ait−1 the lagged total asset; ΔREVit the change in sales; and
PPEit the property, plant and equipment. After estimating Equation (3), we use the estimated
CFRI coefficients to compute the discretionary accruals as follow:
10,3  
TAit 1 DREV it DREC it PPE it
DiscAccit ¼ b^ 1 b^ 2  b^ 3 ; (4)
Ait1 Ait1 Ait1 Ait1 Ait1
where ΔRECit is the change in net account receivable and DD is defined as the moving sum of
the absolute value of discretionary accruals over the last three years (year t−1, t−2 and t−3):
278
DDit ¼ AbsVðDiscAccit1 ÞþAbsVðDiscAccit2 ÞþAbsVðDiscAccit3 Þ: (5)
For easy interpretation, we multiply DD with −1 so that a greater value of DD indicates a
higher level of earnings quality.
The second proxy for firm transparency is auditor quality (BIG4). Prior literature shows
that the quality of the auditing reports conducted by the international Big 4 auditors is
higher than that of reports conducted by other auditors. Lang and Maffett (2010) argue that
hiring the Big 4 can serve as a signal that a firm is willing to go through a strict auditing
procedure, which is associated with a high level of firm transparency. We define BIG4 as a
dummy variable that equals 1 if a firm is audited by one of the joint ventures of international
Big 4 audit firms and domestic audit firms, and 0 otherwise.
Following Jin and Myers (2006), the third measure of firm transparency, DIVSTY, is the
standard deviation of analysts’ forecasts of a firm’s earnings in the following year.
In particular, we provide the following definition:
^ m^
s=
DI V STY ¼ pffiffiffiffi; (6)
N
where m^ and s^ are the mean and standard deviation, respectively, of the earnings
forecast by financial analysts; and N the number of analysts covering the firm. Firms
followed by fewer than three analysts are excluded. We multiply DIVSTY by −1 so that a
higher value indicates a smaller variation in earnings forecasts, thereby suggesting higher
information transparency.
The fourth transparency proxy is built on the Information Disclosure Ratings released
annually by the Shenzhen Stock Exchange. Since 2001, the Shenzhen Stock Exchange has
assessed all listed firms every year with respect to the quality of the information disclosed
to the public, and ranked them into four categories: excellent, good, qualified and
unqualified. We collect this information from the exchange’s website and construct a
discrete variable, DSCORE, which has values of 4, 3, 2 and 1 for firms ranked into the
respective four categories.

3.5 Regression models

To test the relation between managerial overconfidence and stock price crash risk, we
estimate the following regression model:
X
CrashRiskt ¼ b0 þb1 OC t1 þ jCON TROLðt1
kÞ
þY earDum þI ndustryDum þxt ; (7)
k

where CrashRiskt is the stock price crash risk proxied by COLLARt or NCSKEWt; OCt−1
refers to the overconfidence variables, OC_CEOt−1 and OC_CHMt−1; CONTROL a set of
control variables; YearDum and IndustryDum represent the year and industry dummy
variables, respectively; and ξt the random error term. H1 predicts a positive coefficient
of OCt−1.
Following Chen et al. (2001), Hutton et al. (2009) and Kim et al. (2011a), we control a
set of variables that are deemed to be associated with crash risk, including lagged firm size
(SIZEt−1); return on net assets (ROEt); lagged market-to-book ratio (MtoBt−1); lagged Managerial
financial leverage (LEVt−1); detrended stock turnover ratio in year t−1 (DTURNt−1); overconfidence
lagged negative skewness (NCSKEWt−1); average firm-specific weekly returns in year t−1
(FSRETt−1); the volatility of firm-specific weekly returns in year t−1 (SIGMAt−1); and an
indicator for ST (PT) firms (STPTt)[2]. Year and industry fixed effects are included. The
Appendix presents all the variable definitions.
To test H2, we expand Equation (7) with firm transparency variables and their 279
interactions with managerial overconfidence variables as follows:

CrashRiskt ¼ b0 þb1 OC t1 þb2 OC t1  TRAN S t1 þb3 TRAN S t1
X
þ jk CON TROLðt1
kÞ
þY earDum þI ndustryDumþxt ; (8)
k

where TRANSt−1 refers to firm transparency, measured by DDt−1, BIG4t−1, DIVSTYt−1 and
DSCOREt−1, respectively. We expect that when a firm is more transparent, the effect of
managerial overconfidence on the crash risk is less pronounced, that is, β2 o0.

4. Empirical results
4.1 Descriptive statistics and univariate analysis
Table I reports the descriptive statistics of the key variables used in analysis. The mean
value of COLLAR is −0.081, standard deviation is 0.450 and maximum (minimum) value is
3.617 (−7.526). As per NCSKEW, its mean (median) is −0.270 (−0.251), the standard
deviation is 0.677 and maximum (minimum) value is 3.526 (−3.537). These results indicate
that our sample firms have a large variation in stock price crash risk. The mean values of
OC_CEO and OC_CHM are 0.201 and 0.167, respectively, suggesting that approximately
20 percent of the CEOs and 17 percent of the chairs in our sample firms exhibit
overconfidence. The unreported analysis shows that the correlation coefficient between
OC_CEO and OC_CHM is 0.369.

Variables n Mean SD Min. 25th Pctl. Median 75th Pctl. Max.

COLLARt 5,996 −0.081 0.450 −7.526 0.000 0.000 0.000 3.617

NCSKEWt 5,996 −0.270 0.677 −3.537 −0.628 −0.251 0.106 3.750
OC_CEOt−1 5,996 0.201 0.401 0.000 0.000 0.000 0.000 1.000
OC_CHMt−1 5,996 0.167 0.373 0.000 0.000 0.000 0.000 1.000
DDt−1 4,788 −0.219 0.155 −1.050 −0.282 −0.176 −0.109 −0.007
BIG4t−1 5,996 0.040 0.195 0.000 0.000 0.000 0.000 1.000
DIVSTYt−1 2,389 −0.092 0.094 −0.527 −0.112 −0.061 −0.033 0.000
DSCOREt−1 2,956 2.782 0.685 1.000 2.000 3.000 3.000 4.000
SIZEt−1 5,996 14.690 0.921 12.101 14.050 14.610 15.235 18.747
ROEt 5,996 0.047 0.230 −1.687 0.022 0.069 0.125 0.504
MtoBt−1 5,996 1.823 1.187 0.773 1.124 1.422 2.042 9.109
LEVt−1 5,996 0.471 0.190 0.095 0.331 0.480 0.610 0.990
DTURNt−1 5,996 −0.384 4.750 −11.640 −2.878 −0.139 2.086 11.560
NCSKEWt−1 5,996 −0.267 0.662 −3.863 −0.631 −0.251 0.108 3.750
FSRETt−1 5,996 −0.113 0.165 −9.664 −0.142 −0.086 −0.050 −0.001
SIGMAt−1 5,996 0.044 0.018 0.004 0.032 0.042 0.053 0.467
STPTt 5,996 0.117 0.321 0.000 0.000 0.000 0.000 1.000
Notes: This table shows the descriptive statistics of the variables measuring stock price crash risk,
managerial overconfidence, firm transparency and other control variables used in analysis. The sample Table I.
period is from 2000 to 2012. All variables are defined in the Appendix Descriptive statistics
CFRI Table II presents the results of the univariate test on the relation between managerial
10,3 overconfidence and crash risk. Panel A of Table II reports the difference in crash risk
between firms with overconfident and non-overconfident CEOs. Compared with firms with
non-overconfident CEOs, firms with overconfident CEOs have higher levels of crash risk,
regardless of using COLLAR or NCSKEW as the crash risk measure. The results of the
t- and Wilcoxon rank-sum tests show that the differences are statistically significant at least
280 at the 5 percent level. This is consistent with H1. In Panel B, we consider the difference
between firms with overconfident and non-overconfident chairs, and the results are
qualitatively similar to those reported in Panel A.

4.2 Multivariate test of H1

Table III presents the coefficient estimates for Equation (7). In Columns (1)–(3), COLLAR is the
dependent variable and replaced by NCSKEW in Columns (4)–(6). The t-statistics reported in
parentheses are based on robust standard errors clustered at the firm level. Columns (1) and
(4) show that the coefficients of OC_CEO are highly significant with expected positive sign
(0.047 with t ¼ 3.671 and 0.057 with t ¼ 2.899). Columns (2) and (5) present consistent results of
a positive effect of the overconfidence of a chair on crash risk, as indicated by the positive and
significant coefficients of OC_CHM (0.045 with t ¼ 3.210 and 0.071 with t ¼ 3.383). From
Columns (3) and (6), we observe that when OC_CEO and OC_CHM are included in the
regression model, the coefficients of these two variables remain positive and statistically
significant. This significantly positive association between managerial overconfidence and
future crash risk is consistent with H1, suggesting that overconfident managers are more
likely to conceal negative information or maintain negative NPV projects for extended periods,
which in turn leads to an increase in future crash risk.
The estimated coefficients of the control variables are generally consistent with the findings
in prior studies. For example, consistent with Hutton et al. (2009), the coefficients of SIZE are all
positive and significant in all columns and those of ROE are negative and significant when
NCSKEW is used as proxy for crash risk. The coefficients of MtoB are positive in all regressions
and those of LEV are positive in regressions where COLLAR is the dependent variable.

Panel A: CEO overconfidence and crash risk

COLLARt NCSKEWt
Overconfident Non-overconfident Overconfident Non-overconfident
(n ¼ 1,203) (n ¼ 4,793) (n ¼ 1,203) (n ¼ 4,793)
Mean −0.050 −0.089 −0.230 −0.280
Mean-Diff 0.039 0.050
t-test 2.705*** 2.310**
Wilcoxon test 2.578*** 2.174**
Panel B: chair overconfidence and crash risk
COLLARt NCSKEWt
Overconfident Non-overconfident Overconfident Non-overconfident
(n ¼ 999) (n ¼ 4,997) (n ¼ 999) (n ¼ 4,997)
Mean −0.055 −0.086 −0.231 −0.278
Mean-Diff 0.031 0.047
t-test 1.996** 2.001**
Wilcoxon test 2.938*** 1.759*
Notes: This table presents the mean values of crash risk for the sample of firms with overconfident
managers and firms with non-overconfident managers. The sample period is from 2000 to 2012. Panel A
reports the difference in crash risk between firms with overconfident and non-overconfident CEOs. Panel B
Table II. reports the difference in crash risk between firms with overconfident and non-overconfident chairs.
Univariate tests All variables are defined in the Appendix. *,**,***Significant at 10, 5 and 1 percent levels, respectively
 COLLARt NCSKEWt
Managerial
(1) (2) (3) (4) (5) (6) overconfidence
OC_CEOt−1 0.047*** 0.036** 0.057*** 0.037*
(3.67) (2.57) (2.90) (1.73)
OC_CHMt−1 0.048*** 0.034** 0.074*** 0.060***
(3.36) (2.22) (3.50) (2.61)
SIZEt−1 0.018* 0.018* 0.019* 0.037*** 0.037*** 0.037*** 281
(1.78) (1.80) (1.82) (2.87) (2.90) (2.93)
ROEt −0.009 −0.009 −0.008 −0.109*** −0.109*** −0.108***
(−0.19) (−0.20) (−0.18) (−2.68) (−2.66) (−2.65)
MtoBt−1 0.031*** 0.031*** 0.031*** 0.062*** 0.063*** 0.062***
(5.81) (5.87) (5.81) (6.72) (6.76) (6.73)
LEVt−1 0.053* 0.058** 0.055* 0.056 0.062 0.060
(1.84) (1.97) (1.90) (1.21) (1.33) (1.28)
DTURNt−1 −0.002 −0.002 −0.002 −0.002 −0.002 −0.002
(−0.97) (−1.00) (−1.00) (−0.60) (−0.63) (−0.64)
NCSKEWt−1 o0.001 o0.001 o 0.001 0.024* 0.025* 0.024*
(0.01) (0.05) (0.01) (1.69) (1.71) (1.68)
FSRETt−1 −0.057** −0.057* −0.057* −0.045 −0.044 −0.044
(−1.96) (−1.95) (−1.96) (−1.26) (−1.25) (−1.24)
SIGMAt−1 −6.683*** −6.678*** −6.699*** −5.850*** −5.857*** −5.878***
(−4.10) (−4.10) (−4.11) (−4.60) (−4.61) (−4.62)
STPTt 0.070*** 0.071*** 0.071*** 0.154*** 0.156*** 0.157***
(3.23) (3.27) (3.28) (5.29) (5.35) (5.37)
Intercept −0.216 −0.213 −0.222 −0.729*** −0.730*** −0.740***
(−1.11) (−1.09) (−1.14) (−3.08) (−3.07) (−3.12)
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
2
Adjusted R 0.071 0.071 0.072 0.089 0.089 0.089
No. of obs. 5,996 5,996 5,996 5,996 5,996 5,996
Notes: This table presents the regression results of the effect of managerial overconfidence on stock price Table III.
crash risk. The sample period is from 2000 to 2012. The dependent variables are the crash risk measures The effect of
COLLAR and NCSKEW. The t-statistics reported in parentheses are computed using the robust standard managerial
error clustered at the firm level. All variables are defined in the Appendix. *,**,***Significant at 10, 5 and overconfidence on
1 percent levels, respectively stock price crash risk

We further examine the impact of each component of managerial overconfidence on crash

risk and report the regression results in Table IV. Panels A and B document the results for
CEO overconfidence. As shown in Panel A, the estimated coefficients of all components of
CEO overconfidence are positively, but only the coefficients of Age and CEO duality are
statistically significant at the conventional levels. The results reported in Panel B are
similar, as all coefficients of overconfidence component are positive (but not statistically
significant). Panels C and D report the results of regressions of Chair’s overconfidence on
crash risk. We find that in both panels the majority coefficients of all components are
positive and statistically significant, except for that of MajorDum. Overall, our results
suggest that the results of main regressions demonstrated in Table III are not driven by only
one or two components of managerial overconfidence.

4.3 Test of H2
If opacity facilitates bad news hoarding activities for managers, then one can expect
the strength of the relation to be attenuated for firms with more transparency, as
proposed in H2. Table V shows the estimated results of Equation (8), in which the
managerial overconfidence proxies are interacted with the firm transparency proxies. In
CFRI (1) (2) (3) (4) (5) (6)
10,3
Panel A: COLLARt
GenderDum_CEOt−1 0.005 (0.48)
MajorDum_CEOt−1 0.010 (0.60)
EduDum_CEOt−1 0.014 (1.16)
TenDum_CEOt−1 0.018 (1.51)
282 AgeDum_CEOt−1 0.030** (2.50)
DualDum_CEOt−1 0.025* (1.67)
Other control variables Yes Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Adjusted R2 0.069 0.069 0.069 0.069 0.070 0.069
No. of obs. 5,996 5,996 5,996 5,996 5,996 5,996
Panel B: NCSKEWt
GenderDum_CEOt−1 −0.000 (−0.02)
MajorDum_CEOt−1 0.031 (1.26)
EduDum_CEOt−1 0.011 (0.58)
TenDum_CEOt−1 0.014 (0.71)
AgeDum_CEOt−1 0.025 (1.43)
DualDum_CEOt−1 0.038 (1.52)
Other control variables Yes Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Adjusted R2 0.087 0.087 0.087 0.087 0.087 0.087
No. of obs. 5,996 5,996 5,996 5,996 5,996 5,996
Panel A: COLLARt
GenderDum_CHMt−1 0.020 (1.54)
MajorDum_CHMt−1 −0.011 (−0.62)
EduDum_CHMt−1 0.022* (1.83)
TenDum_CHMt−1 0.040*** (3.34)
AgeDum_CHMt−1 0.047*** (3.89)
DualDum_CHMt−1 0.028**
(2.13)
Other control variables Yes Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Adjusted R2 0.089 0.070 0.070 0.070 0.071 0.072
No. of obs. 5,996 5,996 5,996 5,996 5,996 5,996
Panel B: NCSKEWt
GenderDum_CHMt−1 0.036* (1.81)
MajorDum_CHMt−1 −0.009 (−0.35)
EduDum_CHMt−1 0.041** (2.25)
TenDum_CHMt−1 0.044** (2.42)
AgeDum_CHMt−1 0.074*** (4.32)
DualDum_CHMt−1 0.039* (1.71)
Other control variables Yes Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Adjusted R2 0.088 0.087 0.088 0.088 0.090 0.088
Table IV. No. of obs. 5,996 5,996 5,996 5,996 5,996 5,996
The effect of Notes: This table presents the regression results of the effect of managerial overconfidence components on
managerial stock price crash risk. The sample period is from 2000 to 2012. The dependent variables are the crash risk
overconfidence measures COLLAR and NCSKEW. The t-statistics reported in parentheses are computed using the robust standard
components on stock error clustered at the firm level. All variables are defined in the Appendix. *,**,***Significant at 10, 5 and 1 percent
price crash risk levels, respectively
 COLLARt NCSKEWt
Managerial
(1) (2) (3) (4) overconfidence
Panel A: measuring firm transparency by DD
OC_CEOt−1 −0.006 (−0.17) 0.026 (0.35)
OC_CHMt−1 0.009 (0.37) −0.026 (−0.56)
OC_CEOt−1 ×DDt−1 −0.445** (−2.33) −0.311 (−0.67)
OC_CHMt−1 ×DDt−1 0.561* (2.15) 0.369 (0.57) 283
OC_CEOt−1 ×DDt−1 2
−0.312** (−2.78) −0.585** (−2.34)
OC_CHMt−1 ×DDt−1 2
0.377** (2.84) 0.347 (1.06)
DDt−1 0.093 (0.55) 0.139 (1.03) 0.254 (0.86) 0.214 (0.92)
DD2t−1 −0.018 (−0.08) −0.084 (−0.39) −0.175 (−0.44) −0.188 (−0.52)
Other control variables Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
Adjusted R2 0.071 0.070 0.100 0.101
No. of obs. 4,788 4,788 4,788 4,788
Panel B: measuring firm transparency by BIG4
OC_CEOt−1 0.049*** (4.68) 0.072*** (3.82)
OC_CHMt−1 0.050*** (4.10) 0.085*** (4.45)
OC_CEOt−1 ×BIG4t−1 −0.059*** (−3.48) −0.082* (−2.00)
OC_CHMt−1 ×BIG4t−1 −0.075** (−2.42) −0.208** (−2.35)
Other control variables Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
Adjusted R2 0.071 0.071 0.098 0.098
No. of obs. 5,996 5,996 5,996 5,996
Panel C: measuring firm transparency by DIVSTY
OC_CEOt−1 0.027 (1.52) −0.004 (−0.09)
OC_CHMt−1 0.017 (1.19) −0.009 (−0.19)
OC_CEOt−1 ×DIVSTYt−1 −0.241*** (−3.89) −0.386 (−1.10)
OC_CHMt−1 ×DIVSTYt−1 −0.380** (−2.33) −0.716** (−2.10)
DIVSTYt−1 −0.154* (−1.88) −0.142* (−1.89) 0.211 (1.27) 0.255 (1.50)
Other control variables Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
Adjusted R2 0.103 0.104 0.087 0.089
No. of obs. 2,389 2,389 2,389 2,389
Panel D: measuring firm transparency by DSCORE
OC_CEOt−1 0.137*** (4.32) 0.149*** (4.48)
OC_CHMt−1 0.152*** (3.83) 0.259*** (4.01)
OC_CEOt−1 ×DSCOREt−1 −0.024** (−2.43) −0.045*** (−3.65)
OC_CHMt−1 ×DSCOREt−1 −0.045** (−3.03) −0.077** (−2.96)
DSCOREt−1 −0.016 (−1.05) 0.004 (0.80) −0.008 (−0.82) −0.001 (−0.05)
Other control variables Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
2
Adjusted R 0.069 0.035 0.069 0.070
No. of obs. 2,956 2,956 2,956 2,956
Notes: This table presents the regression results of the impact of firm transparency on the relation between
managerial overconfidence and stock price crash risk. The sample period is from 2000 to 2012. The dependent
variables are the crash risk measures COLLAR and NCSKEW. In Panels A, B, C and D, DD, BIG4, DIVSTY
and DSCORE are used as proxy for firm transparency, respectively. The t-statistics reported in parentheses Table V.
are computed using the robust standard error clustered at the firm level. All variables are defined in the The impact of firm
Appendix. *,**,***Significant at 10, 5 and 1 percent levels, respectively transparency
CFRI Panel A, we use the measure of earnings quality DD as the proxy for firm transparency.
10,3 Following Hutton et al. (2009), we control for the quadratic term of DD in the regression. The
results are generally consistent with our expectations. Except for those under Column (3),
the coefficients of OC_CEO×DD and OC_CHM×DD are negative and statistically
significant at least at the 5 percent levels. The positive coefficients of OC_CEO×DD2 and
OC_CHM×DD2 are statistically significant in Columns (1) and (2), thus suggesting the
284 impact of transparency on the relation between managerial overconfidence and crash risk
diminishes as firm transparency increases.
Panel B presents the results when we use BIG4 to proxy for firm transparency. As seen
in Table IV, the positive coefficients of OC_CEO and OC_CHM remain statistically
significant in all four columns. Regardless of using COLLAR or NCSKEW as the dependent
variable, the coefficients of the interaction terms OC_CEO×BIG4 and OC_CHM×BIG4 are
negative and significant at the 5 percent level. This result implies that the effect of
managerial overconfidence on crash risk is less pronounced when the firm is audited by the
Big 4 auditors.
Panel C represents the results in which DIVSTY is used as a proxy for firm transparency.
The results illustrate that, except for Column (3), the coefficients of OC_CEO×DIVSTY and
OC_CHM×DIVSTY are negative and statistically significant at least at the 5 percent level,
indicating a considerably weak effect of overconfidence on crash risk in firms with a large
dispersion in analyst’s earnings forecasts.
When firm transparency is measured by DSCORE, we qualitatively obtain the same
results reported in Panel D. The coefficients of OC_CEO and OC_CHM are positive and
significant across all columns. Moreover, the coefficients of the two interaction terms
OC_CEO×DSCORE and OC_CHM×DSCORE are consistently negative and significant at
least at the 5 percent level.
In summary, firm transparency measurably affects the relation between managerial
overconfidence and crash risk. The marginal effect of overconfidence on the risk is
more pronounced for firms with low earnings quality, audited by non-Big 4 auditors, with
large dispersion in analyst earnings forecasts, and with low ratings on information
disclosure. These findings are consistent with H2.

5. Robustness checks and additional tests

5.1 Endogeneity Issues
We first consider the potential self-selection bias that might arise from the fact that firms
self-select their top managers. One cannot rule out the possibility that high-crash risk firms
are more likely to hire overconfident managers. In such a case, estimating the impact of
overconfident CEO (Chairman) vs non-overconfident CEO (Chairman) in a single equation
context may introduce a self-selection bias into the coefficients of CEO (Chairman)
overconfidence. We try to mitigate this issue using a two-stage regression approach.
In the first stage, we estimate a probit model in which the likelihood of CEO (Chairman)
overconfidence, denoted by Pr(OC_CEO/OC_CHM), is regressed on a set of firm-specific
characteristics that might influence the choice of CEO or board chair:
Pr OC_CEO=OC_CH M it
¼ j0 þj1 SI Z E it þj2 LEV it þj3 M toBit þ j4 ROE it

þj5 GROW TH it þj6 RETU RN it j7 RI SK it

þj8 CEOTU RN it =CH M TU RN it
þj9 TOPH OLDit þj10 BSI Z E it þ xit : (9)
The firm-specific variables include firm size (SIZE), financial leverage (LEV), market-to-book
ratio (MtoB), return on net assets (ROE), sales growth (GROWTH), yearly stock return
(RETURN), firm risk (RISK, measured as the volatility of weekly stock returns), CEO Managerial
(Chairman) turnover (CEOTURN/CHMTURN), the percentage of shares held by the largest overconfidence
shareholder (TOPHOLD) and board size (BSIZE). Year and industry dummies are included to
control for year and industry fixed effects.
In the second stage, we follow Gul et al. (2010) to estimate our main regression in two
different ways to address potential self-selection biases. First, we follow Heckman’s (1979)
two-stage treatment effect procedure. In particular, we calculate the inverse Mills ratio, 285
denoted as λ, from the first-stage probit regression of Equation (9), and then include λ in
the second-stage regression. Second, we estimate our main regression with the fitted
value of Pr(OC_CEO/OC_CHM), denoted by PredOC_CEO/PredOC_CHM, from the
first-stage probit regression as an instrument for the managerial overconfidence variables
in the second-stage regression.
Panel A of Table VI reports first-stage probit estimates. The results show that the
likelihood of a firm to choose overconfidence CEO (Chairman) is positively related to
leverage and CEO (Chairman) turnover and firm risk, and is insignificantly associated with
firm size, market-to-book ratio, sales growth and stock return, while it has a significantly
negative relation with return on net assets, firm age and ownership concentration.
Panel B of Table VI presents second-stage results with the inverse Mills ratio included in
Columns (1)–(4), while Columns (5)–(8) presents those with PredOC_CEO/PredOC_CHM lieu
of OC_CEO/OC_CHM. As shown in the table, the coefficients on OC_CEO/OC_CHM and
PredOC_CEO/PredOC_CHM are significantly positive, suggesting that corrections for
self-selection bias do not alter the main results presented in Table III.
We also employ an event study to clarify the causality relationship between managerial
overconfidence and crash risk using shocks associated with manager turnover. The sample
is selected based on the following procedures: identify the CEO (Chairman) turnovers that
cause switches between overconfident CEO (Chairman) and non-overconfident CEO
(Chairman) and require no CEO (Chairman) overconfidence status changes and CEO
(Chairman) turnovers either in the prior or the subsequent year. Thus, we obtain 69 cases of
managerial overconfidence status changes (OC–NonOC), among which 41 occur between
CEOs (OC_CEO–NonOC_CEO) and 28 between board Chairs (OC_CHM–NonOC_CHM).
Identify a matching firm for each event firm in the year of the CEO (Chairman) turnover.
The matching firm is required to have experienced CEO (Chairman) turnover, operate in the
same industry, and is closest in firm size but with no change in managerial overconfidence
status during the event window [t−1, t+1].
We then compute the changes in crash risk surrounding the event year (D_COLLAR and
D_NCSKEW) for event and matching firms, respectively. The changes in crash risk are
computed as year(t+1)−year(t−1) for switches from non-overconfidence CEO (Chairman) to
overconfidence CEO (Chairman) and as year(t−1)−year(t+1) for switches from overconfident
CEO (Chairman) to non-overconfident CEO (Chairman). We further compute the
differences in crash risk changes between event firms and matching firms. The
univariate results are shown in Table VII. For event firms, crash risk increases for both
groups of OC_CEO–NonOC_CEO and OC_CHM–NonOC_CHM. By contrast, matching
firms do not exhibit such a pattern in their crash risk measures. The t-test further suggests
that the differences in crash risk changes between event firms and matching firms are
statistically significant for the groups of OC–NonOC and OC_CEO–NonOC_CEO,
suggesting that in comparison with the matching sample, crash risk increases significantly
when a firm’s top manager changes from a non-overconfident one to an overconfident one,
and vice versa.
Managerial overconfidence and crash risk may be simultaneously driven by a few
unobserved, firm-level time-constant factors. Accordingly, the estimated coefficients from
the regression might be biased. In the robustness test shown in Table VIII, we control for
CFRI OC_CEOt OC_CHMt
10,3 (1) (2)
Panel A: first-stage regression
SIZEt −0.037 (−1.54) −0.022 (−0.81)
LEVt 0.304*** (2.71) 0.352*** (2.80)
MtoBt 0.030 (1.46) 0.005 (0.21)
286 ROEt −0.545* (−1.88) −0.822** (−2.54)
GROWTHt 0.004 (0.16) 0.029 (1.01)
RETURNt 0.026 (0.65) 0.010 (0.24)
RISKt 0.044 (0.89) 0.104* (1.83)
AGEt −0.028*** (−5.79) −0.010* (−1.80)
CEOTURNt 0.242*** (5.42)
CHMTURNt 0.195*** (3.35)
TOPHOLDt −0.290** (−2.09) −0.166 (−1.06)
BSIZEt −0.031*** (−2.99) −0.004 (−0.38)
Intercept 0.360 (0.69) −0.746 (−1.28)
Industry fixed Yes Yes
effects
Year fixed effects Yes Yes
Pseudo R2 0.024 0.020
No. of obs. 5,940 5,940
Panel B: second-stage regression
COLLARt NCSKEWt COLLARt NCSKEWt
(1) (2) (3) (4) (5) (6) (7) (8)
OC_CEOt−1 0.350*** 0.449***
(2.89) (2.79)
OC_CHMt−1 1.016*** 1.027***
(4.39) (3.39)
PredOC_CEOt−1 0.367*** 0.525***
(2.60) (3.05)
PredOC_CHMt−1 1.044** 0.840**
(2.24) (1.98)
λ −0.175** −0.517*** −0.241*** −0.534***
(−2.52) (−4.19) (−2.61) (-3.31)
Other control Yes Yes Yes Yes Yes Yes Yes Yes
variables
Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes
effects
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Adjusted R2 0.071 0.072 0.075 0.074 0.0704 0.071 0.074 0.073
No. of obs. 5,940 5,940 5,940 5,940 5,940 5,940 5,940 5,940
Table VI. Notes: This table presents two-stage regressions to examine self-selection bias associated with managerial
Results of two-stage overconfidence. Panel A reports the first-stage probit regression results and Panel B reports the second-stage
regressions to regression results. The sample period is from 2000 to 2012. The dependent variables in Panel A are the
examine self-selection managerial overconfidence measures OC_CEO and OC_CHM. The dependent variables in Panel B are
bias associated with the crash risk measures COLLAR and NCSKEW. The t-statistics reported in parentheses are computed
managerial using the robust standard error clustered at the firm level. All variables are defined in the Appendix.
overconfidence *,**,***Significant at 10, 5 and 1 percent levels, respectively

firm fixed effects to address the aforementioned endogeneity problem and conclude that the
results are consistent with our previous findings.

5.2 Alternative measures of managerial overconfidence

In the main analysis, the overconfidence measures are constructed based on the personal
characteristics of the CEO and chair. To assess the robustness of our main results, we redo
 Average crash risk
Managerial
Event sample (1) Matching sample (2) Difference (1)−(2) t-test of (1)−(2) n overconfidence
Panel A: D_COLLAR
OC–NonOC 0.196 −0.156 0.352 2.409** 28
OC_CEO–NonOC_CEO 0.267 −0.151 0.418 2.733*** 41
OC_CHM–NonOC_CHM 0.091 −0.164 0.255 0.896 28
287
Panel B: D_NCSKEW
OC–NonOC 0.213 −0.081 0.294 1.842* 69
OC_CEO–NonOC_CEO 0.222 −0.174 0.396 1.853* 41
OC_CHM–NonOC_CHM 0.198 0.054 0.418 0.600 28
Notes: This table presents results of the change of crash risk when a firm’s CEO or Chairman changes from an
overconfidence one to a non-overconfidence one, and vice versa. The event sample consists of firms that
experience managerial overconfidence switches between overconfidence CEO (Chairman) and non-overconfidence
CEO (Chairman) during the sample period. For each event firm, we find a matching firm also experiencing CEO
(Chairman) turnover, operating in the same industry, and closest in firm size but with no change in managerial
overconfidence status during the event window [t−1, t+1]. D_COLLAR and D_NCSKEW denote the changes in
crash risk surrounding the event window. They are computed as year(t+1)–year(t−1) for switches from
non-overconfidence CEO (Chairman) to overconfidence CEO (Chairman) and as year(t−1)–year(t+1) for switches Table VII.
from overconfidence CEO (Chairman) to non-overconfidence CEO (Chairman). The differences in the changes of Managerial
crash risk between the event sample and the matching sample, t-Statistics associated with these differences, overconfidence and
and the number of event are reported in the last three columns. *,**,***Significant at 10, 5 and 1 percent crash risk:
levels, respectively event study

COLLARt NCSKEWt
(1) (2) (3) (4)

OC_CEOt−1 0.062*** (3.02) 0.058** (1.98)

OC_CHMt−1 0.070*** (2.70) 0.073** (2.09)
Other control variables Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes
2
Adjusted R 0.088 0.088 0.111 0.111
No. of obs. 5,996 5,996 5,996 5,996
Notes: This table presents the regression results of crash risk on managerial overconfidence after controlling
for firm fixed effects. The sample period is from 2000 to 2012. The dependent variables are the crash risk
measures COLLAR and NCSKEW. The t-statistics reported in parentheses are computed using the robust Table VIII.
standard error clustered at the firm level. All variables are defined in the Appendix. *,**,***Significant at 10, Controlling for firm
5 and 1 percent levels, respectively fixed effects

the empirical analysis with three alternative measures of managerial overconfidence.

Following Schrand and Zechman (2012), we build the first additional overconfidence
variable OC_FIRM using four measures of firm-level investing and financing activities that
prior research has found to be related to managerial overconfidence. For a given firm in a
given year, OC_FIRM equals 1 if the firm meets the requirements of at least three of four
criteria following, and 0 otherwise. First, the excess investment is higher than the industry
median, where the excess investment is the residual from the regression of the total asset
growth on sales growth; second, the value of the net asset acquisition is higher than the
industry median, where net asset acquisition is calculated as the amount of asset acquisition
minus the amount of the stripped assets; third, the debt-to-equity ratio is higher than the
CFRI industry median, where the debt-to-equity ratio is defined as the long-term debt divided by
10,3 the total market capitalization of the equity; and fourth, the firm’s dividend payout ratio is 0.
Following Huang et al. (2011), we use the relative ratio of executive salary
(OC_RS) as another managerial overconfidence measure. For each firm in each year,
we compute the relative ratio of executive salary, which is the ratio of the sum of the
three highest-paid managers’ salaries to the sum of all managers’ salaries. Thereafter,
288 OC_RS is assigned a value of 1 if the relative ratio is higher than the industry median,
and 0 otherwise.
Following Lin et al. (2005) and Huang et al. (2011), the last overconfidence measure
OC_PB is based on the difference between manager’s forecast and actual earnings. If the
number of incidences in which the realized earnings are lower than the forecast earnings
exceeds the number of incidences in which the realized earnings are higher during the
sample period, then the CEO/chair of a firm is defined as being overconfident. Thus, OC_PB
takes a value of 1, and 0 otherwise.
Table IX presents new results with additional three measures of managerial
overconfidence. The coefficients of the three alternative overconfidence variables are
consistently positive and statistically significant at the 5 and 10 percent levels, except for
the insignificant coefficient of OC_CP in Column (2). These results are generally in line with
those presented in Table III.

5.3 Alternative measures of crash risk

In this section, we re-estimate Equation (7) with two alternative measures of crash risk.
Following Chen et al. (2001) and Kim et al. (2011a), the first risk measure DUVOL measures
stock return asymmetries. For each firm in each year, we identify the trading weeks with
firm-specific return below the annual mean (thereafter, down weeks) and the weeks with
firm-specific return above the annual mean (thereafter, up weeks). Consequently, we
compute the standard deviation of the returns for down and up weeks, respectively. In the
last step, we compute DUVOL using the following equation:
( )
X X
DU V OLit ¼ log ðnu 1Þ wit =ðnd 1Þ wit ;
2 2
(10)
down up

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

OC_FIRMt−1 0.023** (2.32) 0.035*** (3.37)

OC_RSt−1 0.020 (1.51) 0.048*** (2.66)
OC_PBt−1 0.046*** (4.40) 0.066*** (3.22)
Other control Yes Yes Yes Yes Yes Yes
variables
Industry fixed Yes Yes Yes Yes Yes Yes
effects
Year fixed Yes Yes Yes Yes Yes Yes
effects
Adjusted R2 0.076 0.081 0.084 0.087 0.090 0.099
No. of obs. 5,317 5,711 4,674 5,322 5,711 4,674
Notes: This table presents the regression results of crash risk on alternative measures of managerial
Table IX. overconfidence. The sample period is from 2000 to 2012. The dependent variables are the crash risk measures
Alternative measures COLLAR and NCSKEW. The t-statistics reported in parentheses are computed using the robust standard
of managerial error clustered at the firm level. All variables are defined in the Appendix. *,**,***Significant at 10, 5 and
overconfidence 1 percent levels, respectively
where w is the firm-specific weekly return and nd and nu are the number of down and up Managerial
weeks, respectively. The higher the value of DUVOL, the higher the crash risk. overconfidence
We also follow Jin and Myers (2006) to redefine crash risk presented by COUNT, which
is based on the frequency of firm-specific weekly returns that exceed 3.2 standard
deviations above and below the mean. In particular, COUNT is computed as the upside
frequencies subtracted by the downside frequencies, where a high value of COUNT
suggests a high frequency of crashes. 289
As indicated by the results reported in Table X, our main conclusions remain the same
despite using different crash risk measures.

5.4 Controlling for the influence of corporate governance

To disentangle the effects of overconfidence from those of agency conflicts, we build our main
measure of overconfidence on managers’ personal characteristics. However, one cannot rule
out the possibility that overconfident managers are self-selected by firms with a particular
governance structure. In addition, our composite measure of managerial overconfidence
includes a dummy variable for CEO duality, which may capture worse corporate governance
(Cornett et al., 2008)[3]. We perform two tests to mitigate the above-mentioned concerns.
First, we control for additional elements of corporate governance, including the percentage of
shares held by the largest shareholder (TOPHOLD) and executives (MSHAR) and the
proportion of independent directors (OUTR). The results presented in Columns (1)–(4) of
Table XI indicate that the sign and significance level of the estimated coefficients of OC_CHM
and OC_CHM do not change, and confirm the main findings presented in Table III. Second,
we exclude firms with CEO duality and re-estimate Equation (7). The results shown in
Columns (5)–(8) of Table XI are consistent with the previous main findings.

5.5 Additional analysis

Ahmed and Duellman (2013) find that overconfident managers are prone to disclose good
news more promptly than they would disclose bad news, which suggests that good news is
less likely to be concealed within a firm. Thus, a negative relationship should exist between
managerial overconfidence and the probability of a positive jump in stock price. Following
Hutton et al. (2009), we define JUMP as a dummy variable that equals 1 if, within a calendar
year, the firm-specific weekly return is over 3.2 standard deviations above the mean, and 0
otherwise. Table XII presents the results of the regression of the positive jump risk on
various measures of managerial overconfidence. Except for the insignificant coefficient of

DUVOLt COUNTt
(1) (2) (3) (4)

OC_CEOt−1 0.027*** (2.62) 0.034** (2.27)

OC_CHMt−1 0.034*** (3.14) 0.054* (1.82)
Other control variables Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes
Adjusted R2 0.147 0.148 0.357 0.357
No. of obs. 5,996 5,996 5,996 5,996
Notes: This table presents the regression results of alternative measures of crash risk on managerial
overconfidence. The sample period is from 2000 to 2012. The dependent variables are the crash risk measures
DUVOL and COUNT. The t-statistics reported in parentheses are computed using the robust standard error Table X.
clustered at the firm level. All variables are defined in the Appendix. *,**,***Significant at 10, 5 and 1 percent Alternative measures
levels, respectively of crash risk
CFRI Panel A: controlling for corporate governance Panel B: excluding firms with CEO
10,3 factors duality
COLLARt NCSKEWt COLLARt NCSKEWt
(1) (2) (3) (4) (5) (6) (7) (8)

OC_CEOt−1 0.046*** 0.058*** 0.042*** 0.042*

(3.51) (2.93) (2.67) (1.80)
290 OC_CHMt−1 0.049*** 0.077*** 0.043** 0.064**
(3.38) (3.60) (2.20) (2.26)
TOPHOLDt−1 −0.097** −0.100** −0.066 −0.071
(−2.27) (−2.33) (−1.10) (−1.18)
MSHARt−1 0.080** 0.084** 0.058 0.060
(1.97) (2.06) (0.87) (0.91)
OUTRt−1 −0.063 −0.056 −0.089 −0.084
(−0.59) (−0.52) (−0.52) (−0.49)
Other control Yes Yes Yes Yes Yes Yes Yes Yes
variables
Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes
effects
Year fixed Yes Yes Yes Yes Yes Yes Yes Yes
effects
2
Adjusted R 0.072 0.072 0.089 0.090 0.070 0.069 0.088 0.089
No. of obs. 5,940 5,940 5,940 5,940 4,772 4,772 4,772 4,772
Notes: This table presents the regression results of crash risk on managerial overconfidence with additional
control variables of corporate governance, including the percentage of shares held by the largest shareholder
(TOPHOLD), executive ownership (MSHAR) and the proportion of independent directors in the board
Table XI. (OUTR). The sample period is from 2000 to 2012. The dependent variables are the crash risk measures
Controlling for the COLLAR and NCSKEW. The t-statistics reported in parentheses are computed using the robust standard
influence of corporate error clustered at the firm level. All variables are defined in the Appendix. *,**,***Significant at 10, 5 and 1
governance percent levels, respectively

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

OC_CEOt−1 −0.131*** (−2.77)

OC_CHMt−1 −0.166** (−2.35)
OC_FIRMt−1 −0.109 (−1.58)
OC_RSt−1 −0.136** (−2.34)
OC_PBt−1 −0.188* (−1.78)
Other control Yes Yes Yes Yes Yes
variables
Industry fixed Yes Yes Yes Yes Yes
effects
Year fixed effects Yes Yes Yes Yes Yes
Pseudo R2 0.146 0.146 0.153 0.149 0.114
No. of obs. 5,996 5,996 5,322 5,711 4,674
Table XII. Notes: This table presents the logistic regression results of positive jump risk of stock price on the
The effects of managerial overconfidence. The sample period is from 2000 to 2012. The dependent variable is JUMP, which
managerial takes the value of one if a firm experiences one or more firm-specific weekly returns exceeding 3.2 standard
overconfidence on deviations above the mean within its fiscal year, and zero otherwise. The t-statistics reported in parentheses
positive jump risk are computed using the robust standard error clustered at the firm level. All variables are defined in the
of stock price Appendix. *,**,***Significant at 10, 5 and 1 percent levels, respectively
OC_FIRMt−1, the coefficients of the other four overconfidence measures are consistently Managerial
negative and statistically significant at least at the 10 percent level. overconfidence
Combined with the previous finding of a positive effect of managerial overconfidence on
crash risk, this result supports our prediction that overconfident managers have incentives
to hoard bad news within a firm, while they disclose good news in a timely manner.

6. Conclusion 291
This paper analyzes the impact of managerial overconfidence on stock price crash risk.
Using a large sample of Chinese non-state-owned firms from 2000 to 2012, we find that firms
with overconfident CEOs or board chairs are more likely to experience stock price crashes in
the future. In addition, the effects of managerial overconfidence on crash risk are more
pronounced for firms with lower transparency, indicated by lower earnings quality, being
audited by non-Big4 auditors, large dispersions in analysts’ earnings forecasts, and low
ratings on information disclosure. This finding implies that an opaque information
environment facilitates bad news hoarding activities by overconfident managers. We also
find that managerial overconfidence is negatively associated with future stock price jump
risk, suggesting that overconfident managers tend to release positive information in a
timely manner. The main results hold after a series of robustness tests. Our study
complements the research of Jin and Myers (2006), who emphasize the importance of
investor protection and firm transparency in reducing crash risk.

Notes
1. We re-estimate the model with value-weighted average return of the market and value-weighted
average industry return of the industry, and the results are qualitatively the same.
2. Liu and Lu (2007) state that “The stock exchanges will first label a firm in financial trouble as a
special treatment (ST) firm, then designate it a particular transfer (PT) firm if it fails to turn profitable
within one year” (p. 886).
3. We thank the reviewer for bringing this to our attention.

References
Adams, R.B., Almeida, H. and Ferreira, D. (2005), “Powerful CEOs and their impact on corporate
performance”, Review of Financial Studies, Vol. 18 No. 4, pp. 1403-1432.
Ahmed, A. and Duellman, S. (2013), “Managerial overconfidence and accounting conservatism”,
Journal of Accounting Research, Vol. 51 No. 1, pp. 1-30.
Alicke, M.D. (1985), “Global self-evaluation as determined by the desirability and controllability of trait
adjectives”, Journal of Personality and Social Psychology, Vol. 49 No. 6, pp. 1621-1630.
Baker, M. and Wurgler, J. (2012), “Behavioral corporate finance: a current survey”, in Constantinides, G.M.,
Harris, M. and Stulz, R.M. (Eds), Handbook of the Economics of Finance, Vol. 2, Elsevier/North
Holland, New York, NY, pp. 1-103.
Ball, R. and Shivakumar, L. (2008), “Earnings quality at initial public offerings”, Journal of Accounting
& Economics, Vol. 45 Nos 2-3, pp. 324-349.
Barber, B.M. and Odean, T. (2001), “Boys will be boys: gender, overconfidence, and common stock
investment”, Quarterly Journal of Economics, Vol. 116 No. 1, pp. 261-292.
Benmelech, E., Kandel, E. and Veronesi, P. (2010), “Stock-based compensation and CEO (dis)
incentives”, Quarterly Journal of Economics, Vol. 125 No. 4, pp. 1769-1820.
Blanchard, O.J. and Watson, M.W. (1982), “Bubbles, rational expectations, and financial markets”, in
Wachtel, P. (Ed.), Crises in Economic and Financial Structure, Lexington Books, Lexington, MA,
pp. 295-315.
CFRI Bleck, A. and Liu, X. (2007), “Market transparency and the accounting regime”, Journal of Accounting
10,3 Research, Vol. 45 No. 2, pp. 229-256.
Bushman, R., Piotroski, J. and Smith, A. (2004), “What determines corporate transparency?”, Journal of
Accounting Research, Vol. 42 No. 2, pp. 207-252.
Callen, J.L. and Fang, X. (2013), “Institutional investor stability and crash risk: monitoring versus short-
termism?”, Journal of Banking and Finance, Vol. 37 No. 8, pp. 3047-3063.
292 Camerer, C. and Dan, L. (1999), “Overconfidence and excess entry: an experimental approach”,
American Economic Review, Vol. 89 No. 1, pp. 306-318.
Campbell, J.Y. and Hentschel, L. (1992), “No news is good news: an asymmetric model of changing
volatility in stock returns”, Journal of Financial Economics, Vol. 31 No. 3, pp. 281-318.
Campbell, T., Gallmeyer, M., Johnson, S., Rutherford, J. and Stanley, B. (2011), “CEO optimism and
forced turnover”, Journal of Financial Economics, Vol. 101 No. 3, pp. 695-712.
Chen, J., Hong, H. and Stein, J. (2001), “Forecasting crashes: trading volume, past returns, and
conditional skewness in stock prices”, Journal of Financial Economics, Vol. 61 No. 3, pp. 345-381.
Christie, A.A. (1982), “The stochastic behavior of common stock variances: value, leverage and interest
rate effects”, Journal of Financial Economics, Vol. 10 No. 4, pp. 407-432.
Cooper, A.C., Woo, C.Y. and Dunkelberg, W.C. (1988), “Entrepreneurs perceived chances for success”,
Journal of Business Venturing, Vol. 3 No. 2, pp. 97-108.
Cornett, M.M., Marcus, A.J. and Tehranian, H. (2008), “Corporate governance and pay-for-performance:
the impact of earnings management”, Journal of Financial Economics, Vol. 87 No. 2, pp. 357-373.
Dechow, P., Sloan, R. and Sweeney, A. (1995), “Detecting earnings management”, The Accounting
Review, Vol. 70 No. 2, pp. 193-225.
Estes, R. and Hosseini, J. (2001), “The gender gap on Wall Street: an empirical analysis of confidence in
investment decision making”, Journal of Psychology, Vol. 122 No. 6, pp. 577-590.
Fast, N., Sivanathan, N., Mayer, N. and Galinsky, A. (2012), “Power and overconfident decision
making”, Organizational Behavior and Human Decision Processes, Vol. 117 No. 2, pp. 249-260.
Forbes, D. (2005), “Are some entrepreneurs more overconfident than others”, Journal of Business
Venturing, Vol. 20 No. 5, pp. 623-640.
Fraser, S. and Greene, F. (2006), “The effect of experience on entrepreneurial optimism and
uncertainty”, Economica, Vol. 73 No. 290, pp. 169-192.
French, K.R., Schwert, G.W. and Stambaugh, R.F. (1987), “Expected stock returns and volatility”,
Journal of Financial Economics, Vol. 19 No. 1, pp. 3-29.
Galasso, A. and Simcoe, A. (2011), “CEO overconfidence and innovation”, Managerial Science, Vol. 57
No. 8, pp. 1469-1484.
Graham, J., Harvey, C. and Puri, M. (2013), “Managerial attitudes and corporate actions”, Journal of
Financial Economics, Vol. 109 No. 1, pp. 103-121.
Graham, J., Harvey, C. and Rajgopal, S. (2005), “The economic implications of corporate financial
reporting”, Journal of Accounting and Economics, Vol. 40 No. 1, pp. 3-73.
Griffin, D. and Tversky, A. (1992), “The weighing of evidence and the determinants of confidence”,
Cognitive Psychology, Vol. 24 No. 3, pp. 411-435.
Gul, F.A., Kim, J.B. and Qiu, A.A. (2010), “Ownership concentration, foreign shareholding, audit quality,
and stock price synchronicity: evidence from china”, Journal of Financial Economics, Vol. 95
No. 3, pp. 425-442.
Heaton, J.B. (2002), “Managerial optimism and corporate finance”, Financial Management, Vol. 31 No. 2,
pp. 33-45.
Heckman, J.J. (1979), “Sample selection bias as a specification error”, Econometrica, Vol. 47 No. 1,
pp. 153-162.
Hirshleifer, D., Low, A. and Teoh, S. (2012), “Are overconfident CEOs better innovators?”, Journal of
Finance, Vol. 67 No. 4, pp. 1457-1498.
Hong, H. and Stein, J. (2003), “Differences of opinion, short-sales constraints, and market crashes”, Managerial
Review of Financial Studies, Vol. 16 No. 2, pp. 487-525. overconfidence
Huang, W., Jiang, F., Liu, Z. and Zhang, M. (2011), “Agency cost, top executives’ overconfidence, and
investment-cash flow sensitivity: evidence from listed companies in China”, Pacific-Basin
Finance Journal, Vol. 19 No. 3, pp. 261-277.
Hutton, A.P., Marcus, A.J. and Tehranian, H. (2009), “Opaque financial reports, R2, and crash risk”,
Journal of Financial Economics, Vol. 94 No. 1, pp. 67-86. 293
Jiang, F. and Kim, K.A. (2016), “Corporate governance in China: a modern perspective”, Journal of
Corporate Finance, Vol. 32, June, pp. 190-216.
Jiang, F., Yin, Z., Su, F. and Huang, L. (2009), “Managerial characteristics and firm over-investment”,
Management World, Vol. 1, pp. 130-139 (in Chinese).
Jin, L. and Myers, S.C. (2006), “R2 around the world: new theory and new tests”, Journal of Financial
Economics, Vol. 79 No. 2, pp. 257-292.
Kato, T. and Long, C. (2006), “CEO turnover, firm performance, and enterprise reform in china: evidence
from micro data”, Journal of Comparative Economics, Vol. 34 No. 4, p. 817.
Kim, J.-B., Li, Y. and Zhang, L. (2011a), “Corporate tax avoidance and stock price crash risk: firm-level
analysis”, Journal of Financial Economics, Vol. 100 No. 3, pp. 639-662.
Kim, J.-B., Li, Y. and Zhang, L. (2011b), “CFOs versus CEOs: equity incentives and crashes”, Journal of
Financial Economics, Vol. 101 No. 3, pp. 713-730.
Kim, J.-B., Wang, Z. and Zhang, L. (2016), “CEO overconfidence and stock price crash risk”,
Contemporary Accounting Research, Vol. 33 No. 4, pp. 1720-1749.
Kothari, S.P., Shu, S. and Wysocki, P.D. (2009), “Do managers withhold bad news?”, Journal of
Accounting Research, Vol. 47 No. 1, pp. 241-276.
Lang, M. and Maffett, M. (2010), “Economic effects of transparency in international equity markets: are
view and suggestions for future research”, Foundations and Trends in Accounting, Vol. 5 No. 3,
pp. 175-241.
Libby, R. and Rennekamp(, K. (2012), “Self-serving attribution bias, overconfidence, and the issuance of
management forecasts”, Journal of Accounting Research, Vol. 50 No. 1, pp. 197-231.
Lin, Y., Hu, S. and Chen, M. (2005), “Managerial optimism and corporate investment: some empirical
evidence from Taiwan”, Pacific-Basin Finance Journal, Vol. 13 No. 5, pp. 523-546.
Liu, S.Q. and Lu, Z. (2007), “Corporate governance and earnings management in Chinese listed
companies: a tunneling perspective”, Journal of Corporate Finance, Vol. 13 No. 5, pp. 881-906.
Malmendier, U. and Tate, G. (2005), “CEO overconfidence and corporate investment”, Journal of
Finance, Vol. 60 No. 6, pp. 2661-2700.
Malmendier, U. and Tate, G. (2008), “Who makes acquisitions? CEO overconfidence and the market’s
reaction”, Journal of Financial Economics, Vol. 89 No. 1, pp. 20-43.
Miller, D.T. and Ross, M. (1975), “Self-serving biases in the attribution of causality: fact or fiction?”,
Psychological Bulletin, Vol. 82 No. 2, pp. 213-225.
Myers, J.M., Myers, L.A. and Skinner, D.J. (2007), “Earnings momentum and earnings management”,
Journal of Accounting, Auditing, and Finance, Vol. 22 No. 2, pp. 249-284.
Piotroski, J.D. and Wong, T.J. (2012), “Institutions and information environment of Chinese listed
firms”, in Fan, J. and Morck, R. (Eds), Capitalizing China, c. 5, National Bureau of Economic
Research and University of Chicago Press, Chicago, IL, pp. 201-242.
Puri, M. and Robinson, D.T. (2005), “Optimism and economic choice”, Journal of Financial Economics,
Vol. 86 No. 1, pp. 71-99.
Schrand, C.M. and Zechman, S.L.C. (2012), “Executive overconfidence and the slippery slope to
financial misreporting”, Journal of Accounting and Economics, Vol. 53 Nos 1-2, pp. 311-329.
Seok, B. (2007), “Change, contradiction, and overconfidence: Chinese philosophy and cognitive
peculiarities of Asians”, Dao, Vol. 6 No. 3, pp. 221-237.
CFRI Taylor, R. (1975), “Age and experience as determinants of managerial information processing and
10,3 decision making performance”, Academy of Management Journal, Vol. 18 No. 1, pp. 74-81.
Taylor, S.E. and Gollwitzer, M.P (1995), “Effects of mindset on positive illusions”, Journal of Personality
and Social Psychology, Vol. 69 No. 2, pp. 213-226.
Weinstein, N. (1980), “Unrealistic optimism about future life events”, Journal of Personality and Social
Psychology, Vol. 39 No. 5, pp. 806-820.
294 Xu, N., Jiang, X., Kam, C.C. and Yi, Z. (2013), “Analyst coverage, optimism, and stock price crash risk:
evidence from China”, Pacific-Basin Finance Journal, Vol. 25, November, pp. 217-239.
Xu, N., Li, X., Yuan, Q. and Chan, K.C. (2014), “Excess perks and stock price crash risk: evidence from
China”, Journal of Corporate Finance, Vol. 25, April, pp. 419-434.
Yates, F.J., Lee, J. and Bush, J.G. (1997), “General knowledge overconfidence: cross-national variations,
response style, and ‘reality’ ”, Organizational Behavior and Human Decision Processes, Vol. 70
No. 2, pp. 87-94.
Yates, F.J., Lee, J. and Shinotsuka, H. (1996), “Beliefs about overconfidence, including its cross-national
variation”, Organizational Behavior and Human Decision Processes, Vol. 65 No. 2, pp. 138-147.
Yates, J.F., Zhu, Y., Ronis, D.L., Wang, D.F., Shinotsuka, H. and Toda, M. (1989), “Probability judgement
accuracy: China, Japan, and the United States”, Organizational Behavior and Human Decision
Processes, Vol. 43 No. 2, pp. 147-171.

Further reading
Bekaert, G. and Wu, G. (2000), “Asymmetric volatility and risk in equity markets”, Review of Financial
Studies, Vol. 13 No. 1, pp. 1-42.
Larwood, L. and William, W. (1977), “Managerial myopia: self-serving biases in organizational
planning”, Journal of Applied Psychology, Vol. 62 No. 2, pp. 194-198.
Taylor, S.E. and Brown, J.D. (1988), “Illusion and well-being: asocial psychological perspective on
mental health”, Psychological Bulletin, Vol. 103 No. 2, pp. 193-210.
Wright, G.N. and Phillips, L.D. (1980), “Cultural variation in probabilistic thinking: alternative ways of
dealing with uncertainty”, International Journal of Psychology, Vol. 15 Nos 1-4, pp. 239-257.
Yates, J.F., Lee, J.W., Levi, K.R. and Curley, S.P. (1990), “Measuring and analyzing probability judgment
accuracy in medicine”, Philippine Journal of Medicine, Vol. 28, pp. 21-32.
Xin, Q., Bao, A. and Hu, F. (2019), “West meets east: understanding managerial incentives in Chinese
SOEs”, China Journal of Accounting Research, Vol. 12 No. 2, pp. 177-189.

Appendix. Variable definitions

Dependent variables: crash (jump) risk measures

COLLAR is the actual profits or losses from an option portfolio of buying an out-of-money put option
and shorting a call option on the firm-specific weekly return, times 1,000. The strike price of the put is
set to the mean minus 3.2 standard deviations and the strike price of the call is set to the mean plus 3.2
standard deviations, with 3.2 chosen to generate frequencies of 0.1 percent in the normal distribution
during the fiscal year period.
NCSKEW is the negative skewness of firm-specific weekly returns over the fiscal year.
DUVOL is the natural logarithm of the ratio of standard deviations of downward week to upward
week firm-specific returns.
COUNT is the difference between the numbers of firm-specific weekly returns exceeding 3.2
standard deviations below and above the mean.
JUMP is a dummy variable that equals one if a firm experience one or more firm-specific weekly
returns exceeding 3.2 standard deviations above the mean within its fiscal year, and zero otherwise.
Managerial overconfidence variables Managerial
OC_CEO (OC_CHM) is a dummy variable that equals 1 if the CEO (chair) of a firm meets the overconfidence
requirements of at least four of six criteria following, and 0 otherwise. First, the CEO (chair) is male;
second, the CEO’s (chair’s) age is below the sample mean; third, the CEO (chairman) has obtained a
master’s degree or higher; fourth, the CEO (chair) studied a business or economics major; fifth, the
CEO’s (chair’s) tenure is below the sample median; and sixth, the CEO is also the chair of the board
of directors.
GenderDum is a dummy variable that equals 1 when the CEO (chair) is male, and 0 otherwise. 295
AgeDum is a dummy variable that equals 1 if the CEO (chair) is male, and 0 otherwise.
EduDum is a dummy variable that equals 1 if a CEO (chair) has obtained a master’s degree or
higher, and 0 otherwise.
MajorDum is a dummy variable that equals 1 if a CEO (chair) pursued a business or economics
major, and 0 otherwise.
TenDum is a dummy variable that equals 1 if a CEO’s (chair’s) tenure is below the sample median,
and 0 otherwise.
DualDum is defined as an indicator that equals 1 if the CEO also holds the position of the chair of
the board, and 0 otherwise.

Firm transparency variables

DD is the moving sum of absolute value of discretionary accruals over the last three years, multiplied
by −1, where discretionary accruals are estimated from the modified Jones model (Dechow et al., 1995).
BIG4 is a dummy variable that takes the value of 1 if a firm is audited by one of the joint ventures of
international Big 4 audit firms and domestic audit firms, and 0 otherwise.
DIVSTY is the standard deviation of analysts’ forecasts of the firm’s earnings in the next year,
normalized by the mean forecast, and divided by the square root of the number of analysts following
that firm, and then multiplied by −1.
DSCORE is the transparency measure based on the information disclosure rating released by the
Shenzhen Stock Exchange. DSCORE has values of 4, 3, 2 and 1 for firms ranked into the respective four
categories: excellent, good, qualified and unqualified.

Other control variables

SIZE is the size of a firm, which is calculated as the natural logarithm of year-end market value
of equity.
ROE is return on net assets, which is calculated as the net income divided by the book value
of equity.
MtoB is market-to-book ratio, which is measured as the ratio of market value of equity to book
value of equity.
LEV is leverage ratio, which represents the ratio of the book value of total liabilities to the book
value of total assets.
DTURN is the detrended stock trading volume, which is measured as the difference between the
average monthly share turnover over the current fiscal year period and that of the previous fiscal year,
where monthly share turnover is calculated as the monthly trading volume divided by the total number
of shares outstanding during the month.
FSRET represents the average firm-specific weekly return over a fiscal year, times 100.
SIGMA is the standard deviation of firm-specific weekly returns over a fiscal year.
STPT is a dummy variable that equals 1 for the ST (special treatment) firms that reported negative
earnings in the past two successive fiscal years, and for the PT (particular transfer) firms that reported
negative earnings in the past three successive fiscal years, and zero otherwise.
TOPHOLD is the percentage of shares held by the largest shareholder.
MSHAR is the percentage of shares held by the executives.
OUTR is the ratio of number of independent directors to board size.
GROWTH is the growth rate of sales.
RETURN is the yearly buy and hold stock return.
CFRI RISK is the standard deviation of weekly stock return.
10,3 CEOTURN is a dummy variable that equals 1 if CEO turnover occurs in a given year,
and 0 otherwise.
CHMTURN is a dummy variable that equals 1 if board chair turnover occurs in a given year,
and 0 otherwise.
BSIZE is the natural logarithm of the number of board directors.
296 Corresponding author
Jingjing Tang can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]