You are on page 1of 22

J Bus Ethics

DOI 10.1007/s10551-016-3025-x

Public Governance and Corporate Fraud: Evidence


from the Recent Anti-corruption Campaign in China
Jian Zhang1

Received: 1 September 2015 / Accepted: 12 January 2016


 Springer Science+Business Media Dordrecht 2016

Abstract Taking advantage of the China’s recent anti- country-level institutional factors.1 However, the cross-
corruption campaign, we attempt to examine the effect of country studies may have two problems: First, Fan et al.
public governance on a firm’s incentive to commit fraud. (2008) argue that cross-country studies may be subject to
Using enforcement actions data from the Chinese Securi- endogeneity problems, which may bias the results. Fan et al.
ties Regulatory Commission (CSRC) from 2004 to 2014, further contend that the potential endogeneity problems can
we find that, due to enhanced public governance, firms are be alleviated by studying the effect of country-level insti-
less likely to commit fraud in the post-campaign period tutional factors in a single country. Second, Allen et al.
than in the pre-campaign period. We further show that the (2005) argue that determining the assignment of weight for
effect of public governance is more evident in privately each country in the cross-country studies may create poten-
held listed firms, in firms with weak legal environment, and tial problems since large diverse countries and small
in firms in areas with poor local economies. In addition, we homogeneous countries are different in culture and politics.
find that older CEOs respond less actively to the public However, by examining the impact of public governance on
governance caused by anti-corruption regulations. This corporate fraud in China, we are able to effectively address
paper offers clear policy implications for business ethics by the concerns above by taking advantage of an exogenous
indicating that public governance provides external moni- shock to institutional factors, i.e., the recent anti-corruption
toring of corporate decisions. campaign initiated by President Xi Jinping.
China provides a unique setting for us to conduct a per-
Keywords Corporate fraud  Anti-corruption  Chinese suasive examination of the effect of public governance on
government  Public governance corporate fraud. First, the ongoing anti-corruption campaign
initiated by President Xi Jinping in China serves as an
JEL Classification G34 exogenous shock to public governance, thereby alleviating
potential endogeneity problems. The second challenge is the
insufficient variation in institutional factors resulting from
Introduction institutional stability within a single country. The anti-cor-
ruption campaign in China meets this challenge by providing
Several cross-country studies in the relevant literature show variations in public governance, which thus enables us to
that corporate decisions are significantly influenced by effectively investigate the effect of public governance within
one country. Third, unlike the U.S., the Chinese government
has discretion in implementing securities regulations and laws
(Li et al. 2014; Wu et al. 2014). This feature allows public
& Jian Zhang
1
zhangjian@swufe.edu.cn See Demirgüç-Kunt and Maksimovic (1998, 1999); Rajan and
Zingales (1995); Levine (1999); La Porta et al. (1997a, b, 1998,
1
School of Finance, Southwestern University of Finance and 1999b, 2000a, b); Booth et al. (2001); Beck and Levine (2002);
Economics, 555 Liutai Avenue, Wenjiang District, Giannetti (2003); Fan et al. (2012); La Porta et al. (1999a); Fan et al.
Chengdu 611130, China (2008).

123
J. Zhang

governance to significantly influence enforcement actions institutional background and develops hypotheses for the
against corporate misbehavior. Fourth, corporate fraud is analysis. ‘‘Data and Sample Selection’’ section describes
prevalent in China. Li et al. (2014) show that in 2012, the China the mode of construction of the data sample collected for
Securities Regulatory Commission (CSRC), whose functions the study. In ‘‘Research Design’’ section, the details of the
are similar to the U.S. Securities and Exchange Commission research design are illustrated. ‘‘Regression Results’’ sec-
(SEC), received 380 allegations and investigated 316 corporate tion shows the empirical results of the design method
fraud cases entailing violations of securities laws. At the cur- adopted, followed by the ‘‘Robustness Analysis’’ section
rent listing of 2494 firms in China, the CSRC fraud allegations that analyzes the robustness of the results. ‘‘Conclusion’’
involve 12.7 % of public firms in 2012. section finally provides conclusions based on the analysis
To conduct the empirical analysis, we collect the results of this research.
enforcement actions data against corporate fraud from the
CSRC. For the control group, we include all public firms
listed in China without fraud activities. In total, our sample Literature Review
from 2004 to 2014 contains 11,568 firm-year observations
with 993 fraud firm-year observations. The current literature on corporate fraud focuses on
We find that the ongoing anti-corruption campaign initi- explaining the likelihood of fraud with factors such as CEO
ated at the end of 2012 significantly lowers a firm’s fraud compensation and board characteristics. Beasley (1996)
likelihood due to the improvement of public governance. A examines the relation between board compositions and
firm’s fraud likelihood is reduced by 3.8 % in the post- financial statement fraud. He finds that a lower likelihood
campaign period compared with the pre-campaign period. of fraud is associated with smaller board size or higher
Given the unconditional mean of fraud likelihood of 8.6 %, board independence. Bergstresser and Philippon (2006)
this translates to an economically significant 46.51 % (3.8/ document that firms are more likely to manipulate earnings
8.6 %) increase in fraud likelihood from its average level. when the CEO’s total compensation consists of more stock
Moreover, we find that public governance has a stronger and option holdings. Similarly, Burns and Kedia (2006)
effect in privately held listed firms since they are lack of show that the propensity of misreporting increases with the
political connections to the government. In addition, we sensitivity of the CEO’s option portfolio value to stock
show that the location of the firm also matters. Firms are price. Efendi et al. (2007) find that financial misstatement
more active in response to the public governance in an is more pronounced at firms where the CEO holds more in-
unlawful region or in a poor area since anti-corruption reg- the-money stock options. Crutchley et al. (2007) argue that
ulations significantly increase the quality of public gover- fast-growing firms with fewer outsiders on the audit com-
nance in such areas. Finally, we find that older CEOs respond mittee and a larger number of busy outside directors are
less actively to the anti-corruption campaign since they are more likely to commit accounting fraud. Researchers have
reluctant to commit fraud in the pre-campaign period due to recently begun to link the role of executives’ social ties to
their risk-aversion. To further validate our results, we rule corporate fraud. Johnson et al. (2009) argue that manage-
out alternative explanations such as reverse causality, rial unrestricted stock holdings provide the strongest
macroeconomic factors, corporate culture, and partial incentive for firms to commit fraud. Dechow et al. (2011)
observability in fraud cases. find a red flag for earnings misstatement by developing a
This paper extends the research on a country’s institutions scaled probability (F-score), building on accrual quality,
on shareholder protections and corporate governance (La financial performance, nonfinancial measures, off-balance-
Porta et al. 1998, 2000a) in three ways. First, unlike previous sheet activities, and stock and debt market incentives.
cross-country studies, this paper attempts to examine the Chidambaran et al. (2011) show that professional con-
effect of institutional factors on corporate decisions in a signal nectedness is negatively related to financial fraud, while
country. Second, while extensive literature in law and finance nonprofessional connectedness is positively associated
compares the role of the origin of a country’s legal system in with fraud. Similarly, Khanna et al. (2015) find that a given
various countries, we emphasize the role of law enforcement CEO’s social connectedness with subordinate executives
within one country. Third, by employing the introduction of increases fraud incentives since the connected executives
an exogenous shock to institutional factors, the paper is help deter fraud detection. Connected executives reduce
immune from potential endogeneity problems. the probability of CEO dismissal after fraud discovery and
The rest of the paper is organized as follows: ‘‘Literature lower the coordination costs in committing fraud. Chen
Review’’ section provides a review of relevant past studies et al. (2006) find that the probability of restatement is lower
in the literature. In ‘‘Institutional Background and in companies whose boards or audit committees have an
Hypothesis Development’’ section, the author describes the independent director with financial expertise, and is higher

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

in companies where the CEO belongs to the founding in the leverage and debt maturity ratios of connected firms
family in China. after the arrest of the corrupt bureaucrat.
Another research stream focuses on external factors Our paper is also closely related to the literature on
affecting a firm’s fraud likelihood. Wang et al. (2010) show political connections. In this literature, various variables
that business conditions are related to fraud incentives. are used as proxy for political connections: family ties,
Fraud incentives increase with the level of investor beliefs personal relationship with officials, geographic proximity,
about industry prospects, yet decrease if beliefs are extre- and the appointment of the ex-government officials as
mely high when a firm goes public. Wang and Winton directors (Fan et al. 2007; Faccio et al. 2006; Faccio and
(2014) find that fraud incentives are on average higher in Parsley 2009). Agrawal and Knoeber (2001) find that the
competitive industries since competitive industries lack politically connected outside directors play an important
product market sensitivity to firm-level information, they role in dealing with the government in terms of sales to
lack firm-level information production, and they use more government and exports. Gupta and Swenson (2003) doc-
relative performance evaluation. Kedia and Rajgopal ument that a firm’s political contribution is associated with
(2011) find that the SEC is more likely to investigate firms its tax benefits. Faccio et al. (2006) analyze the likelihood
geographically closer to its offices as well as firms located of government bailouts by using the data from 35 countries
in an area with greater past SEC enforcement activity due during 1997–2002. They find that politically connected
to the resource-constraint. Parsons et al. (2014) argue that a firms are significantly more likely to be bailed out than
firm’s fraud incentives increase with the misconduct rates similar nonconnected firms. Faccio and Parsley (2009)
of nearby firms. This phenomenon is caused by peer effects argued that the sudden death of certain politicians lowers
rather than exogenous shock such as regional variation in the firm’s value if the firm’s headquarters are in the
enforcement. Li et al. (2014) find that provincial financial politician’s hometown. Using Indonesia data, Leuz and
development and block ownership decrease fraud likeli- Oberholzer-Gee (2006) show that political connections are
hood and increase detection likelihood in China. Wu et al. related to the long-term performance of the firm. Fan et al.
(2014) emphasize the role of institutional investors in (2007) point out that politically connected CEOs are
lowering the incidence of enforcement actions against associated with lower post-IPO stock returns and lower
corporate fraud in China. board professionalism since a significant number of con-
In addition, several authors have studied fraud detection. nected CEOs lack expertise in running the firm.
Karpoff et al. (1999) find that firms experience negative
abnormal returns when fraud is uncovered. Karpoff and
Lou (2010) show that short sellers detect firms misrepre- Institutional Background and Hypothesis
senting their financial statements months before the mis- Development
representation is publicly revealed. Dyck et al. (2010) find
that fraud detection is less a function of standard corporate Corruption in China
governance actors such as investors, the SEC, or auditors,
but instead on ‘‘environmental’’ factors such as employees, China is currently in a transition period from a planned
the media, and industry regulators. In particular, Dyck et al. economy to a free market economy. The Chinese govern-
argue that monetary incentives and career concerns moti- ment retains a strong influence on the economy, including
vate employee whistleblowing. resource allocation across firms. The cross-country studies
Our study contributes to the literature of corruption. A show that government intervention is significantly related
large number of papers in the literature of corruption to corruption since the state controls economic resources
examine whether corruption sand the wheel (Adit 2009; (La Porta et al. 1999a, b; Mauro 1995). Moreover, Shleifer
Ehrlich and Lui 1999; Mo 2001; Reinikka and Svensson and Vishny (1993) argue that the weakness of central
2004; Shleifer and Vishny 1993; Wei 2000) or grease the government allows multiple officials from various gov-
wheel (Acemoglu and Verdier 2000; Lui 1985) in emerg- ernment agencies to impose independent bribes on private
ing markets. Cai et al. (2011) find that entertainment and agents, which significantly increases the total bribe rate.
travel costs (expenditure on corruption) have significantly The political decentralization introduced in China’s post-
negative effects on firm productivity and substantial posi- 1979 economic reforms may aggravate the degree of cor-
tive returns to firms simultaneously in China. Fan et al. ruption. Fan et al. (2009) further show that uncoordinated
(2008) conduct a natural experiment by identifying 23 rent-seeking arises as government structures become more
corruption scandals involving high-level government complex in the process of political decentralization.
bureaucrats in China and a set of publicly traded companies The Transparency International Survey in 2004 states
connected to those corrupt officials through bribe or pre- that China’s Perception Index ranked 71 out of 145 in the
vious job affiliations. They further find a significant decline world, while China ranks 100 out 175 according to the

123
J. Zhang

survey in 2014 (The higher ranking implies more public Hypothesis Development
perceptions of corruption).2 Thus, according to the survey,
China is among the most corrupt countries. The anti-corruption campaign can significantly reduce a
firm’s fraud incentives through three mechanisms:
The Ongoing Anti-corruption Campaign Initiated enhancing corporate governance, raising fraud cost, and
in 2012 shaping corporate culture. First, CEOs are self-interested,
which results in agency problems between managers and
President Xi Jinping officially took office in the 18th shareholders.3 The high-profile corporate fraud cases pro-
National Congress of the Communist Party of China on the vide direct evidence supporting this view. Prior studies
14th November 2012. Shortly after his assumption of the show that corporate governance plays an important role in
presidency of the party, the party promulgated eight pro- reducing agency problems and limiting a firm’s fraud
visions and corrected ‘‘four winds’’ to regulate the party likelihood by monitoring managers (Beasley 1996; Agra-
members, which signals the beginning of the anti-corrup- wal and Chadha 2005). However, the level of corruption at
tion campaign. In China, the Central Commission for the country level lowers a firm’s corporate governance
Discipline Inspection (CCDI) was founded to enforce quality. Doidge et al. (2004) argue that, especially, in less-
internal rules and regulations in the party, which enforces developed countries, country characteristics, such as legal
the inner-party supervision system. Following Xi Jinping’s protections for minority investors and the levels of eco-
assumption of the party presidency in 2012, the CCDI nomic and financial developments, influence firms’ costs
experienced another round of reforms to strengthen its and benefits in implementing measures to improve their
power and independence in supervising the party members. own governance and transparency. Doidge et al. (2007)
According to the official record of CCDI, by the end of find that countrywide variation in intuitional quality sig-
2014, 41 provincial- and ministerial-level officials had nificantly affects corporate governance across business
been investigated and subsequently arrested during the firms. In particular, Shleifer and Vishny (1994) argue that
anti-corruption campaign. Before the anti-corruption cam- managers in more corrupt contexts have stronger incentives
paign, only 6–8 provincial- and ministerial-level officials to extract private rents by lowering corporate governance
had been investigated and arrested each year. During the standards. Both Shleifer and Vishny (1997) and et al.
current campaign, only five provinces have no provincial- (1998) show that corruption undermines legal enforcement,
level officials arrested, Further, nearly all of the high- which in turn weakens shareholder protection and aggra-
ranking officials in Shanxi province have been investi- vates agency problems between managers and sharehold-
gated. Therefore, the anti-corruption campaign launched by ers. In their survey paper, Claessens and Yurtoglu (2013)
President Xi can be viewed as the boldest and the most find that corruption is negatively associated with the
serious campaign against corruption in China. quality of corporate governance. Ng (2006) argues that
corruption weakens regulatory oversight and worsens cor-
China Securities and Regulatory Commission porate governance. Ng and Qian (2004) develop a theo-
(CSRC) retical model to illustrate that agency problems exist when
corruption undermines law enforcement. Consequently,
To stabilize the securities market, the CSRC has been corporate governance is negatively related to the level of
working as the major regulator of securities markets in corruption at the country level. Ng and Qian further pro-
China since 1998. The Securities Law also gives strong vide empirical evidence to support their model prediction
power to the CSRC to investigate and sanction firms and that the quality of corporate governance tends to be rela-
individuals for corporate fraud. In every province and in tively lower in a country with higher levels of corruption.
four cities, the CSRC has established a local office in Wu (2008) argues that the low transparency in the public
charge of investigating fraud. Investigations by the CSRC sector caused by corruption leads firms to adopt suspicious
are the primary way corporate fraud is detected in China. accounting practices, which increase fraud likelihood.
The decisions of the CSRC are strongly affected by the Kimbro (2002) shows that accounting standards are nega-
government since CSRC officials have career concerns tively related to corruption at the country level, which
about being replaced for disobeying orders from govern- increases fraud likelihood. The recent anti-corruption
ment officials. campaign initiated by President Xi Jinping is intended to
significantly lower the level of corruption in China.
Therefore, one may expect that the anti-corruption

2 3
For detailed information about the Transparency International See Jensen and Meckling (1976), Jensen (1986, 1993), Shleifer and
Survey, please visit https://www.transparency.org/research/cpi. Vishny (1989, 1997).

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

campaign enhances corporate governance, consequently penalties and the ex-ante probability of being caught for
lowering further a firm’s fraud incentives. committing fraud should increase significantly during the
Second, Xie and Lu (2003) build up theoretical models anti-corruption campaign. Since the firm’s fraud incentives
to analyze the collusion between fraud firms and corrupt depend on expected fraud benefits (Jayachandran 2006;
officials in China. They argue that the corrupt officials Aggarwal et al. 2012), the legal penalties the firm needs to
serving in regulatory authorities such as CSRC receive pay if incriminated (Karpoff et al. 2008a, b) and the CEOs’
bribes from the fraud firms to help conceal corporate perceptions of the ex-ante probability of being caught, we
misconducts.4 Xie and Lu’s models show that fraud firms expect that the anti-corruption campaign will lower firms’
tend to bribe the politicians to conceal their misconducts incentives to commit fraud.
when their expected fraud benefits are greater than the cost Third, culture can be viewed as shared values and
in bribery. They further show that politicians’ incentives to beliefs of people within a country (Donaldson and Lorsch
be corrupt depend on the bribery size, the expected com- 1983; Schein 1985; Kotter and Heskett 1992). Shared
pensation loss on being caught, and the expected value of values and beliefs can influence managerial decisions
punishment for corruption on being detected. They find that through norms of the individuals or of the groups within
fraud firms are less likely to bribe a politician to help cover the firm (Hackman 1992; Gintis 2003). Fisman and Miguel
their misdeeds when the discipline inspection for officials (2007) show that a corrupt culture in their country of origin
is stricter, when the compensation for officials is higher, may explain United Nations diplomats’ parking violations
and when the punishment for corruption is more severe. in New York City. Liu (2014) studies the effect of cor-
Finally, they provide empirical evidence to support their ruption culture on corporate misconduct. She adopts the
model predictions. Similarly, Lui (1986) presents a theo- country of origin of the executives and directors in a
retical model to illustrate that the incentives of officials to company to proxy for the corruption culture. If more
receive bribery does not depend on their preference and executives or directors are from countries with higher
political institutions, but rather rely on the expected value levels of corruption, a firm is regarded as having high
of punishment (Becker 1968; Rose-Ackerman 1978). Lui corruption culture. And she finds that firms with high
further argues that the higher expected value of punishment corruption culture, which tend to be more tolerant toward
determines that the anti-corruption campaign in 1952 is corrupt behavior, are more likely to engage in corporate
more successful than the anti-corruption campaign in 1982 misconducts. Clearly, China is one of high corruption
in China. countries according to the International Country Risk
During the campaign period, the vast majority of offi- Guide. The majority of officers and directors in a Chinese
cials at all government levels have been subjected to strict firm should be from China, which indicates that Chinese
supervision and monitoring by CCDI. Moreover, the CCDI firms have high corruption culture. The anti-corruption
encourages scrutiny from the public and social media. campaign lowers the corruption level in China, resulting in
Individuals can report suspicious officials to CCDI through lower corruption culture in Chinese firms. Thus, we expect
various ways such as regular mail, email, and phone calls. that the anti-corruption campaign leads to a reduced cor-
Therefore, the anti-corruption campaign should signifi- ruption culture in a firm which lowers the firm’s fraud
cantly enhance discipline inspection of officials, which is incentives. Based on the preceding arguments, we offer the
associated with a lower likelihood of being bribed. In following hypothesis:
addition, during the boldest and the most serious anti-
H1 Firms are less likely to commit fraud during the anti-
corruption campaign in the Communist Party of China,
corruption campaign.
corrupt officials face the most severe punishment including
the forfeiting of personal wealth, dismissal from the Party, In China, the government is the controlling shareholder
and imprisonment. Therefore, based on theoretical predic- for a significant number of listed firms due to the partial
tions, firms need to pay more to bribe officials due to privatization of SOEs. Managers in SOEs are appointed by
stricter discipline inspection and a higher expected severity the government, and many of them are high-ranking offi-
of punishment resulting from the anti-corruption campaign. cials (Chen et al. 2009). Wu et al. (2014) show that
Due to the significantly increased bribery costs and political connections are able to bring a firm certain priv-
unchanged benefits from committing fraud, firms are less ileges in the regulatory environment; these privileges imply
likely to bribe officials to help conceal corporate miscon- that enforcement actions in the form of fines and admin-
ducts. This implies that the CEOs’ perceptions of the legal istrative punishment may be eased or avoided. Although
the anti-corruption campaign raises the expected fraud cost
4 for firms, managers in SOEs may still have relatively lower
The paper is one part of the joint research project by the
Supervision Bureau and Research Bureau of The People’s Bank of expected fraud cost due to their political connections to the
China, the central bank of China. government. As a result, one may expect that SOEs will

123
J. Zhang

have a lower probability of being caught for committing extracting data from AAERs, Feng et al. (2011) find that
fraud and a higher probability of facing mild punishments. the involvement of the CFO in corporate fraud is due to
Therefore, the anti-corruption campaign should have less pressure from their CEO rather than motivated by their
effect on the fraud likelihood for the SOEs. Based on these personal financial benefits. The data from Accounting and
arguments, we propose the following hypothesis: Auditing Enforcement Releases (AAERs) suggest that the
CEO, rather than the CFO, is more likely to be treated by
H2 The impact of anti-corruption campaign on a firm’s
SEC as the person who orchestrated the accounting
fraud likelihood is less significant in SOEs.
manipulation and gained financial benefits from the
A large number of studies document that China has manipulation. Therefore, a CEO’s characteristics should be
strong regional disparity in terms of economy, market closely related to a firm’s fraud likelihood.
development, and legal protection (Chen et al. 2006, 2009; Younger CEOs are under more pressure regarding
Jian and Wong 2010). Consequently, the effect of anti- career development and are eager to signal their ability to
corruption may vary across different regions due to the market (Prendergast and Stole 1996; Serfling 2014).
regional disparity. First, legal enforcement is an important They may adopt more aggressive investment policies to
determinant of the level of corruption. The areas with poor signal their superior ability. Fraud commitment can be
legal enforcement tend to have a higher level of corruption viewed as a risky investment due to its high return and high
(Fisman and Miguel 2007). The anti-corruption campaign legal penalties if incriminated. Moreover, the younger
should have a stronger effect on limiting corruption in CEOs might be punished more for poor performance
those areas having higher levels of corruption originally. through reduced future career opportunity since they have
According to Hypothesis 1, the stronger the effect of anti- not been recognized as high quality managers by the labor-
corruption, the less likelihood that a firm will commit market. Consequently, younger CEOs may have more
fraud. Thus, firms in areas with poor legal enforcement incentives to inflate performance through committing
should have lower fraud likelihood in response to the anti- fraud. On the other hand, older CEOs are less aggressive in
corruption campaign. taking risks to demonstrate their abilities. If that is the case,
Second, prior literature suggests that the income level of older CEOs are more risk-averse and reluctant to commit
a country is positively related to the legal enforcement and fraud even in the pre-campaign period. Thus, the older
government efficiency, which lowers the level of corrup- CEOs should not respond to the anti-corruption campaign
tion in a country (Glaeser et al. 2004; Svensson 2003). significantly. Therefore, we offer the following hypothesis
Based on Lipset (1960), Glaeser and Saks (2006) find that regarding CEO characteristics:
the richer states are less corrupt since voters with more
H5 The impact of the anti-corruption campaign on a
income are more willing to and able to monitor public
firm’s fraud likelihood is less significant in firms with older
employees and to take action if they violate laws in US.
CEOs.
Higher income further accelerates the spread of education
and democratic institutions, which enables the public to
better identify corrupt officials (Dong and Torgler 2013).
Data and Sample Selection
Serra (2006) finds that the level of income is negatively
associated with the corrupt activities. Thus, the anti-cor-
We collect the CSRC enforcement actions data from 2004
ruption campaign should have a stronger effect in poor
to 2014 from the China Stock Market and Accounting
areas where is more corrupt previously. According to
Research (CSMAR) database. The CSRC does not publish
Hypothesis 1, we expect that firms in poor areas should
the cases that are not decided, therefore there is no false
have a lower fraud likelihood in response to the anti-cor-
detection case in our sample. To better describe the process
ruption campaign. Therefore, we offer the following
on an enforcement action, Fig. 1 reproduces a timeline
hypotheses:
highlighting major events.5 We also collect the firm’s
H3 The impact of the anti-corruption campaign on a ownership and financial data from the CSMAR database.
firm’s fraud likelihood is more significant in firms head- Following the literature, we exclude firms in the financial
quartered in areas with weak legal environments. industry (Wu et al. 2014).6 After merging all datasets, we
have 11,568 firm-year observations with 993 fraud firm-
H4 The impact of the anti-corruption campaign on a
firm’s fraud likelihood is more significant in firms head- 5
Source: China Securities Regulatory Commission Annual Report
quartered in poor areas.
2012. The figure is also used in Li et al. (2014).
6
In most of the corporate fraud cases, the CEO is the key The financial industry belongs to the regulated industry. Firms in
person who decides whether to commit fraud. By regulated industries are highly regulated by the government, which
distinguishes them from firms in typically unregulated industries.

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

Fig. 1 Timeline of CSRC Enforcement Procedure Source: China Securities Regulatory Commission Annual Report 2012 (Li et al. 2014)

year observations in the sample. The unconditional fraud (Panel D). Panel A indicates a statistically disproportionate
likelihood for a random firm is 8.6 % in the sample. All number of enforcement actions regarding the firms listed in
variables are winsorized at the 1 and 99 % level. Variable the Shenzhen stock exchange. One of the possible reasons
definitions are listed in Table 10 in Appendix. is that the Shenzhen stock exchange has more listed firms,
Table 1 presents the descriptive statistics for CSRC hence unconditionally has a higher probability to observe
enforcement during 2004–2014 by stock exchange (Panel fraud enforcement actions. Panel A also shows that the
A), province (Panel B), industry (Panel C), and fraud type number of fraud observations increases each year, peaks in
2011, then falls precipitously in later years. This pattern is
Footnote 6 continued consistent with the trend of strengthening the power of the
Therefore, following the accounting and finance literature, we
exclude firms in the financial industry from our sample.
CCDI to monitor government officials during the anti-

123
J. Zhang

corruption campaign. Facing stricter monitoring, the firms Table 1 Descriptive statistics for the CSRC enforcement during
cease their fraudulent behavior as the supervisory author- 2004–2014
ities are more likely to carry out their duties. On the other Panel A: By year and stock exchange
hand, the bribing firms may have fewer incentives to
Year Shanghai Shenzhen Total
commit fraud in order to shield corrupt politicians from
political scrutiny. 2004 5 4 9
There is no clear pattern in Panel B, suggesting that 2005 3 3 6
fraud incidence is not monotonically associated with 2006 6 15 21
regional economy. Although Guangdong province has the 2007 14 40 54
highest percentage of fraud cases in the fraud sample, firms 2008 16 52 68
charged with fraud make up only 13.97 % of firms head- 2009 40 75 115
quartered in Guangdong. Guangdong has more listed firms, 2010 40 90 130
hence unconditionally has a higher probability of observed 2011 62 171 233
fraud enforcement actions. Panel C of Table 1 shows the 2012 64 153 217
industry distribution of fraud firms.7 Percentage of frauds is 2013 19 84 103
calculated as the number of fraud cases in an industry 2014 11 26 37
divided by the total number of fraud cases in the fraud Total 280 713 993
sample. This number represents the industry distribution of
Panel B: By province
frauds in the fraud sample. The corporate fraud cases do
not appear to be concentrated in any specific industry. Province No. of fraud Percent (%) Ratio of fraud (%)
Ratio of frauds is calculated as the number of fraud cases Anhui 29 2.92 6.87
divided by the total number of firm-year observations in an Beijing 29 2.92 2.71
industry in the full sample. This number represents the
Fujian 49 4.93 12.53
industry distribution of fraud cases in the full sample. The
Gansu 6 0.60 5.83
large magnitude of the ratio of frauds implies that there are
Guangdong 227 22.96 13.97
more fraud cases or less listed firms8 in an industry in the
Guangxi 27 2.72 20.77
full sample.9
Guizhou 5 0.50 4.46
In Panel D, we realize that the most common violations
Hainan 7 0.70 9.46
are False Statements (12.99 %), Postponement/delay in
Hebei 5 0.50 2.14
Disclosure (13.59 %), Major Information Omission
Henan 64 6.45 18.34
(18.16 %), Mishandling the General Accounting
Heilongjiang 11 1.11 9.48
(12.01 %), and Others (31.03 %). We have 2340 violations
Hubei 8 0.81 2.16
Hunan 44 4.43 13.58
Jilin 30 3.02 17.34
Jiangsu 64 6.45 6.11
Jinagxi 27 2.72 14.21
Liaoning 18 1.81 5.68
7
We adopt the 2012 CSRC industry classification standard provided Neimenggu 11 1.11 8.09
by the CSRC. Since most of the firms belong to the manufacturing Ningxia 3 0.30 5.17
industry, the code for which begins with ‘‘C’’, we use first three codes Qinghai 4 0.40 9.76
to classify the industries within the manufacturing industry. We use
Shandong 38 3.83 5.20
the first one code to classify other industries (Wu et al. 2014).
8 Shanxi 26 2.62 12.50
The industry of Waste resources recycling has a relatively large
number of ratio of frauds. The industry code for the Waste resources Shan’xi 32 3.22 17.68
recycling industry is ‘‘C42’’. However, there is only one listed firm in Shanghai 42 4.23 5.07
the industry of Waste resources recycling. The firm went public in Sichuan 53 5.34 13.42
2010, and thus we only have 5 firm-year observations in our sample
for the industry of Waste resources recycling. The limited number of Tianjin 10 1.01 5.71
firm-year observations in the industry is the reason for why we end up Xinjiang 16 1.61 7.62
with a relatively large number of ratio of frauds for the industry. Xizang 5 0.50 15.15
9
For example, we have six fraud cases out of ten firm-year Yunnan 10 1.01 5.99
observations in a specific industry in the full sample. We also have
Zhejiang 69 6.95 5.88
993 fraud cases in our fraud sample. Thus, the percentage of frauds
will be around 0.6 % (=6/993), and the ratio of frauds will be 60 % Chongqing 24 2.42 15.59
(=6/10).

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

Table 1 continued Table 1 continued

Panel B: By province Panel C: By industry


Province No. of fraud Percent (%) Ratio of fraud (%) Industry No. of Percent Ratio of fraud
fraud (%)
Total 993 100.00 8.60
Culture, sports, and 4 0.40 3.15
Panel C: By industry entertainment
Industry No. of Percent Ratio of fraud Comprehension 4 0.40 6.35
fraud (%) Total 993 100 8.61

Agriculture 41 4.13 22.78 Panel D: By violation type


Mining 33 3.32 7.76 Violation type # of % of
Agriculture products 39 3.93 20.63 violations violations
processing
Food 13 1.31 10.24 Inflated profit 22 0.94
Wine and beverage 15 1.51 6.73 Fabrication of assets 3 0.13
Textile 14 1.41 10.45 False statements 304 12.99
Apparel 3 0.30 2.10 Postponement/delay in disclosure 318 13.59
Leather products 1 0.10 3.70 Major information omission 425 18.16
Timber 6 0.60 16.22 Major failure to disclose 53 2.26
information
Furniture 6 0.60 18.75
Fraudulent listing 3 0.13
Paper making 10 1.01 8.13
Unauthorized change in use of 41 1.75
Printing 1 0.10 3.45 funds
Sports goods 2 0.20 5.26 Firm’s asset occupation 60 2.56
Petroleum 2 0.20 2.35 Insider trading 1 0.04
Chemical materials 97 9.77 11.41 Illegal share buybacks 49 2.09
Medicine and biological 56 5.64 7.12 Stock price manipulation 1 0.04
products
Illegal loan guarantee 53 2.26
Chemistry 12 1.21 11.11
Mishandling the general accounting 281 12.01
Plastics 15 1.51 7.14
Others 726 31.03
Nonmetal mineral 32 3.22 8.96
Total 2340 100
Ferrous metals 13 1.31 4.87
Nonferrous metals 32 3.22 9.36 China has two stock exchanges: SHANGHAI stock exchange and
Metal 5 0.50 3.05 SHENZHEN stock exchange. In the CSMAR database. Ratio of
frauds is the number of fraud cases divided by the number of listed
Machinery, equipment 50 5.04 10.71
firms in a province/industry. The industry classification is based on
Special machinery 34 3.42 5.92 the 2012 CSRC industry classification standard. The types of fraud is
Automobile manufacturing 29 2.92 7.49 provided by CSRC
Railways 21 2.11 11.73
for 993 corporate fraud cases, which means that one fraud
Electrical manufacturing 76 7.65 10.50
case entails 2.36 violations on average.10 The types of
Electronics 92 9.26 9.24
fraud are provide by CSRC.
Instruments 8 0.81 10.13
Other manufacturing 5 0.50 9.09
industries
10
Waste resources recycling 3 0.30 60 A referee raises the point that the different types of fraud are
Power, gas, and water 35 3.52 7.64 important as not all frauds involve bribery or are connected to politicians.
We agree with the referee. However, after looking at the details of the
Construction 24 2.42 8.45 data, we are unable to make further analysis about the types of fraud for
Wholesale and retail 54 5.44 7.56 two reasons. First, most fraud cases involve multiple types of fraud.
Transportation 29 2.92 5.27 32.43 % of the fraud observations involve only one type of fraud.
25.88 % of the fraud observations involve two types of fraud. 21.55 % of
Lodging 9 0.91 17.31
the fraud observations involve three types of fraud. 14.7 % of the fraud
IT 41 4.13 6.86 observations involve four types of fraud. The rest of the observations
Leasing 9 0.91 6.57 involve more than four types of fraud. As the referee pointed out,
Tech services 1 0.10 1.92 different types of fraud might have different effects for the firm;
therefore, we cannot disentangle the effects of different types of fraud if a
Irrigation and environment 17 1.71 11.97 firm commits multiple types of fraud. Second, we also cannot study the

123
J. Zhang

Table 2 Comparison of fraud firms and non-fraud firms POST has a mean value of 0.36 in the fraud sample, indicating
Fraud Obs Non-fraud Obs T-statistics
that on average 36 % of observations in the fraud sample are
from the post-campaign period, and 64 % (1–36 % = 64 %)
POST 0.36 993 0.42 10,575 -3.52*** of observations in the fraud sample are from the pre-campaign
SIZE 21.77 993 22.01 10,575 -6.00*** period. This result indicates that firms are more likely to
LEV 0.45 993 0.42 10,575 3.56*** commit fraud in the pre-campaign period. When we compare
BM 0.65 993 0.65 10,575 0.87 the mean value of POST across the fraud sample and the non-
ROA 0.04 993 0.05 10,575 -7.78*** fraud sample (0.36 vs 0.42), we find that the non-fraud sample
TOBINQ 1.81 993 1.85 10,575 -1.32 tends to have a larger proportion of observations from the post-
BOARD 9.19 989 9.20 10,524 -0.17 campaign period. The univariate analysis of POST supports the
INDEP % 0.37 989 0.37 10,524 -0.04 hypothesis that firms are less likely to commit fraud in the post-
CEODUAL 0.28 969 0.22 10,134 4.50*** campaign period. Non-fraud firms tend to be larger than the
AUDIT 1.97 794 1.97 8089 -0.57 fraud firms, and the difference is statistically significant at the
IAUDIT 0.02 794 0.05 8089 -4.05*** 1 % significance level. One potential explanation is that larger
ANALYST 1.23 993 1.29 10,575 -2.77*** firms face heightened monitoring and public attention. This
INSTOWN 0.05 993 0.06 10,575 -2.50** raises the ex-ante probability of being caught for committing
GOV 0.09 993 0.13 10,575 -5.58 fraud and lowers a firm’s fraud likelihood. LEV is significantly
EXE 0.11 993 0.08 10,575 3.89*** higher in the fraud sample than in the non-fraud sample.
MKTRET 0.08 993 0.17 10,575 -5.95*** Leverage ratio is frequently regarded as a proxy for the
FINCRISIS 0.91 993 0.82 10,575 7.02*** closeness to covenant in accounting literature (Dechow et al.
1996). A large number of studies in accounting literature
POST is a dummy variable defined as 1 if the year is after 2012, 0
suggest that firms close to the debt covenant are more likely to
otherwise. SIZE is the log value of total asset. LEV is the ratio of total
debt over total asset. BM is the book-to-market value of the firm. ROA manage earnings (Healy and Wahlen 1999). The studies also
is the ROA of the firm. TOBINQ is the firm’s Tobin’s Q value. lead us to an intuitive argument that firms with higher risk of
BOARD represents the board size of the firm. INDP % is the per- being in default would be more inclined to commit fraud. ROA
centage of the independent director in the board. CEODUAL is a
suggests that financially healthy firms have fewer incentives to
dummy variable defined as 1 if the firm’s CEO is also the chairman of
the board. AUDIT is the number of auditors of the firm. IAUDIT is a commit fraud. CEODUAL is significantly higher in the fraud
dummy variable defined as 1 if the firm has an international auditor. sample. This is consistent with the literature about powerful
ANALYST is the number of analysts following the firm. INSTOWN is CEOs and low corporate governance quality (Beasley 1996;
total number of shares held by institution investors divided by the
Burns and Kedia 2006). The fraud sample tends to have a lower
total number outstanding shares of the firm. GOV is the government
ownership of the firm. EXE represents top executives’ ownership of value in IAUDIT, ANALYST, and INSTOWN, which implies
the firm. MKTRET is the annual return of Shanghai composite index. that governance quality is negatively related to the fraud
FINCRISIS is a dummy variable if the year is after 2008 financial incentives. Finally, we find that firms charged with fraud tend
crisis
to have lower ratios of executive ownership. In principle, the
*, ** and *** indicate statistical significance at the 10, 5, and 1 %
effect of insider ownership on the prevalence of fraud is
level, respectively
determined by two channels running in opposite directions:
higher insider ownership can reduce the likelihood of fraud as
Table 2 shows the comparison of fraud firms and non-fraud
stock ownership aligns the interests of managers and share-
firms. We assign the value of one to the variable POST if the
holders; meanwhile higher insider ownership can also
observation is taken from the post-campaign period. In con-
strengthen the incentive for fraud as the benefit of fraud
trast, we assign the value of zero to the variable POST if the
becomes larger.
observation is taken from the pre-campaign period. Conse-
quently, the mean value of POST indicates how much of the
observations are from the post-campaign period on average.11
Research Design
Footnote 10 continued
To capture the relationship between public governance and
effect of fraud type in the fraud observations with only one type of
fraud since a large proportion of them are included in the category of corporate fraud, we estimate the following probit
Others. regression:
11
For example, we have 100 observations in the sample, in which 36
observations are taken from the post-campaign period, and 64
observations are taken from the pre-campaign period. We assign the Footnote 11 continued
value of one to the variable POST if the observations are from the the number of observations (100). Therefore, the mean value of POST
post-campaign period, otherwise zero. The mean value of POST is 0.36, which indicates that we have 36 % (36/100) of observations
equals the total value of POST (36 9 1 ? 64 9 0 = 36) divided by from the post-campaign period.

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

PðFraudit Þ ¼ a þ b  Postit þ c  Controlsit significant effect on corporate fraud detection. Chen et al.
þ Industry dummies þ Year dummies þ eit ; (2006) find that international auditing companies have a
good reputation for auditing quality. We also believe that
where t indexes years, i indexes firms, and eit is the asso- the number of auditors should be positively related to the
ciated error term. We assign Postit, a dummy variable, with auditing quality. Consequently, we expect that the number
the value of 1 if the year is after 2012, otherwise 0. For of auditors (AUDIT) and the international auditors dummy
example, we assign the value of 1 to the variable POST for (IAUDIT) should lower a firm’s fraud incentives. Dyck
firm i at the year after 2012, and we assign the value of 0 to et al. (2010) document the active role analysts play in
the variable POST for firm i at the year before 2012. We uncovering fraud. Security analysts closely scrutinize a
follow the same rule for other firms. Controlsit is a vector firm’s financial disclosure and interact with the manage-
of firm-levelvariables controlling for firm size, capital ment more frequently; thus firms with more analysts
structure, firm performance, corporate governance, own- covering their stocks will have less incentives to commit
ership structure, and market conditions. Clustered standard fraud. Yu (2008) finds that analyst coverage leads to less
errors at the firm level are used to compute the t-statistics. earnings management. Shleifer and Vishny (1997) argue
Firm size (SIZE) might be negatively related to a firm’s that institutional investors have stronger incentives and
fraud incentives since larger firms face heightened moni- more resources to monitor management. Monitoring by
toring and public attention. This raises the ex-ante proba- institutional investors thus should lower a firm’s fraud
bility of being caught for committing fraud and lowers a incentives.
firm’s fraud likelihood (Cox and Thomas 2003; Dechow Furthermore, we attempt to control for the ownership
et al. 2011; Wang 2011; Yu and Yu 2012). We adopt structure of the firm. Due to Chinese partial privatization,
leverage (LEV) to measure the firm’s capital structure. a large number of listed firms are owned or partially
Leverage ratio is frequently regarded as a proxy for the owned by the government. Government ownership can be
closeness to covenant in accounting literature (Dechow viewed as a form of political connections. Enforcement in
et al. 1996). A large number of studies in accounting lit- the form of fines, public criticism, administrative pun-
erature suggest that firms close to the debt covenant are ishment, warnings, and even delisting may be eased or
more likely to manage earnings (Healy and Wahlen 1999). even avoided in politically connected firms (Wu et al.
We adopt three variables to proxy for firm performance 2014). Thus, firms with government ownership may have
including book-to-market (BM), ROA, and Tobin’s Q a higher likelihood of committing fraud due to the lower
(TOBINQ). On the one hand, Dechow et al. (2011) find that cost in punishment. In addition, executive equity incen-
firms manipulating earnings tend to show strong financial tives are regarded as one important factor related to the
performance prior to the manipulations. On the other hand, fraud commitment (Burns and Kedia 2006; Bergstresser
Khanna et al. (2015) argued that fraud is more likely when and Philippon 2006; Peng and Röell 2008). Executives
a firm is suffering operating troubles (Arlen and Carney commit fraud to boost firm performance to earn benefits
1992; Alexander and Cohen 1999). Therefore, there is no from their stockholdings. Thus, we expect that there is a
consensus about how firm performance affects a firm’s positive association between a firm’s fraud likelihood and
fraud incentives. executive equity incentives.
Moreover, we adopt several variables to proxy for Finally, we control for the market conditions such as
corporate governance including board size (BOARD), market return (MKRET) and financial crisis (FINCRISIS).
independent director fraction (INDEP %), CEO duality When the market is in the booming period, firms are
(CEODUAL), the number of auditors (AUDIT), interna- more likely to have better performance. Thus, market
tional auditors dummy (IAUDIT), analyst coverage return should be negatively related to a firm’s fraud
(ANALYST), and institutional ownership (INSTOWN). incentives. Managers have incentives to commit fraud to
Beasley (1996) finds that no-fraud firms have smaller boost firm performance for the purpose of avoiding
boards with significantly higher percentages of outside forced turnover and enhancing personal benefits due to
members than fraud firms. When the board acts as the shareholdings. During the financial crisis period, poor
corporate monitor, it should be a major deterrent of the performance is expected by the market and is regarded as
corporation fraud. CEO duality enhances the CEO’s not due to the managers’ skill or effort. As a result,
power in controlling the firm, which allows more man- managers are less likely to be replaced due to the poor
agerial discretion and impedes effective monitoring performance during the financial crisis period. Thus,
(Brickley et al. 1997; Cornett et al. 2008; Jensen 1993). managers have fewer incentives to commit fraud to boost
Firms with CEO duality are expected to have a higher firm performance for the purpose of avoiding a forced
probability of committing fraud. Auditors have a turnover. In addition, we admit that committing fraud

123
J. Zhang

may inflate the stock price. However, the price inflation Table 3 continued
might be very limited since investors are risk averse and (1)
reluctant to trade during the market turmoil. This means Probit
that managers may obtain limited personal benefits in
committing fraud to inflate stock price. Therefore, GOV -0.13
financial crisis may have a negative relation with fraud (-0.83)
incentives. EXE 0.09
In the simple probit model, we make a critical (0.51)
assumption that the ex-post fraud detection probability is MKTRET -0.84***
equal to one. Note that the ex-ante fraud detection proba- (-5.02)
bility, which is the CEOs’ perceived likelihood of being FINCRISIS -0.70***
caught, remains less than one. Otherwise, no fraud will (-3.29)
ever be committed due to the severe consequence of Constant 2.60***
detected fraud. (2.92)
Industry fixed effect Yes
Year fixed effect Yes
Regression Results Firm-year observations 8415
R-squared 0.09
Table 3 presents the results from estimating the simple The dependent variable is the dummy variable measuring whether the
probit regression. The coefficient of POST is negative and firm is committing fraud. POST is a dummy variable defined as 1 if
statistically significant at the 1 % significance level. A the year is after 2012, 0 otherwise. SIZE is the log value of total asset.
LEV is the ratio of total debt over total asset. BM is the book-to-
market value of the firm. ROA is the ROA of the firm. TOBINQ is the
Table 3 Anti-corruption and corporate fraud
firm’s Tobin’s Q value. BOARD represents the board size of the firm.
(1) INDP % is the percentage of the independent director in the board.
Probit CEODUAL is a dummy variable defined as 1 if the firm’s CEO is also
the chairman of the board. AUDIT is the number of auditors of the
POST -0.26*** firm. IAUDIT is a dummy variable defined as 1 if the firm has an
(-3.25) international auditor. ANALYST is the number of analysts following
the firm. INSTOWN is total number of shares held by institution
SIZE -0.17*** investors divided by the total number outstanding shares of the firm.
(-4.27) GOV is the government ownership of the firm. EXE represents top
LEV 0.80*** executives’ ownership of the firm. MKTRET is the annual return of
Shanghai composite index. FINCRISIS is a dummy variable if the
(4.28)
year is after 2008 financial crisis. We adopt the robust standard errors
BM 0.53** clustered in the firm level in the regression
(2.01) *, ** and *** indicate statistical significance at the 10, 5, and 1 %
ROA -1.88*** level, respectively
(-2.97)
TOBINQ 0.06 firm’s fraud likelihood is reduced by 3.8 % in the post-
(1.14)
campaign period compared with the pre-campaign period.
BOARD 0.04**
Given the unconditional mean of fraud likelihood of 8.6 %,
this translates to an economically significant 46.51 % (3.8/
(2.21)
8.6 %) increase in fraud likelihood from its average level.
INDEP % 0.46
This finding is consistent with our Hypothesis 1, that is, the
(0.81)
anti-corruption campaign improves public governance and
CEODUAL 0.16**
lowers a firm’s likelihood to commit fraud. SIZE is nega-
(2.54)
tively related to the fraud incentives since larger firms face
AUDITOR -0.00
more public scrutiny, which enhances the ex-ante detection
(-0.01)
probability (Dyck et al. 2010; Yu and Yu 2012). The
IAUDIT -0.20
coefficient of LEV is positive and significant at the 1 %
(-1.12)
significant level. These figures indicate that firms close to
ANALYST 0.02
the debt covenant are more likely to commit fraud (De-
(0.53)
chow et al. 1996; Healy and Wahlen 1999). The coefficient
INSTOWN 0.01 of BM is positive and significant at the 5 % significance
(0.02) level. The coefficient of ROA is significantly negative. This

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

suggests that better performing firms have less incentive to Table 4 continued
commit fraud. CEODUAL is positively associated with the (1) (2) (3)
fraud likelihood since powerful CEOs lower the corporate Central SOEs Local SOEs Privately
governance quality. MKTRET has a negative coefficient, held
which is significant at the 1 % significance level. That
Constant 3.30 2.78 1.17
indicates that when the overall market condition is good,
(1.60) (1.48) (0.96)
firms have less likelihood to commit fraud.
Industry fixed effect Yes Yes Yes
In order to test Hypothesis 2, we split the sample into
Year fixed effect Yes Yes Yes
central SOEs, local SOEs, and privately held listed firms.
Firm-year 1102 2081 4619
The central SOEs are the public firms controlled by the observations
central government or central government agency. The
R-squared 0.21 0.13 0.09
local SOEs are the public firms controlled by the local
government or local government agency. The privately The dependent variable is the dummy variable measuring whether the
firm is committing fraud. POST is a dummy variable defined as 1 if
Table 4 Anti-corruption and corporate fraud in SOEs the year is after 2012, 0 otherwise. SIZE is the log value of total asset.
LEV is the ratio of total debt over total asset. BM is the book-to-
(1) (2) (3) market value of the firm. ROA is the ROA of the firm. TOBINQ is the
Central SOEs Local SOEs Privately firm’s Tobin’s Q value. BOARD represents the board size of the firm.
held INDP % is the percentage of the independent director in the board.
CEODUAL is a dummy variable defined as 1 if the firm’s CEO is also
POST -0.30 -0.13 -0.33*** the chairman of the board. AUDIT is the number of auditors of the
(-1.09) (-0.65) (-3.42) firm. IAUDIT is a dummy variable defined as 1 if the firm has an
international auditor. ANALYST is the number of analysts following
SIZE -0.23** -0.15* -0.11**
the firm. INSTOWN is total number of shares held by institution
(-2.38) (-1.78) (-2.04) investors divided by the total number outstanding shares of the firm.
LEV 0.83* 1.77*** 0.52** GOV is the government ownership of the firm. EXE represents top
(1.76) (4.45) (2.28) executives’ ownership of the firm. MKTRET is the annual return of
Shanghai composite index. FINCRISIS is a dummy variable if the
BM 0.01 0.07 0.60* year is after 2008 financial crisis. The non-SOEs are the public firms,
(0.01) (0.13) (1.74) which are not controlled by the government or any government
ROA -1.95 0.83 -3.27*** agency. The local SOEs are the public firms controlled by the local
government or local government agency. The central SOEs are the
(-0.97) (0.64) (-4.31)
public firms controlled by the central government or central govern-
TOBINQ -0.16 0.03 0.09 ment agency. We adopt the robust standard errors clustered in the firm
(-0.82) (0.32) (1.40) level in the regression
BOARD 0.11** -0.01 0.05** *, ** and *** indicate statistical significance at the 10, 5, and 1 %
(2.33) (-0.44) (2.08) level, respectively
INDEP % 1.09 -0.92 1.11
held listed firms are the public firms, which are not con-
(0.71) (-0.78) (1.44)
trolled by the government or any government agency. We
CEODUAL 0.44 0.38** 0.07
reestimate our probit model in each subsample and present
(1.55) (2.23) (1.07)
the results in Table 4. We find that only privately held
AUDITOR 0.02 0.08 -0.01
listed firms respond to the anti-corruption campaign sig-
(0.12) (0.82) (-0.14)
nificantly. Our findings support Hypothesis 2, that is, the
IAUDIT 0.10 -0.03 -0.74**
impact of anti-corruption campaign is less significant in
(0.31) (-0.10) (-2.39)
SOEs with political connections to the government.
ANALYST 0.11 -0.09 0.04
We attempt to make subsample analysis based on local
(1.08) (-1.13) (1.00)
characteristics: the local legal environment and local econ-
INSTOWN -3.19** -0.38 0.42
omy at the provincial level. We adopt the local legal index
(-2.27) (-0.48) (0.74) compiled by Fan et al. (2011) to proxy for the local legal
GOV 0.49 0.18 -0.07 environment.12 We assign the firm to the subsample of strong
(1.35) (0.67) (-0.14) legal environment if the local legal index is above the sample
EXE 1.66 2.94** -0.06 median. Otherwise, we assign the firm to the subsample of
(0.40) (2.47) (-0.33) weak legal environment. Furthermore, we adopt log value of
MKTRET -0.39 -1.33** -0.84***
(-0.80) (-2.57) (-4.39) 12
Since the legal index is constructed from 1997 to 2009 and is
FINCRISIS -0.11 -1.09* -0.78***
stable across years, we use the average value of the index as proxy for
(-0.17) (-1.83) (-3.06) the local legal protection in each province.

123
J. Zhang

GDP per capita to measure the local economy. We assign the Table 5 continued
firm to the subsample of rich region if the log value of GDP Local legal Local economy
per capita is above the sample median. Otherwise, we assign environment
the firm in the subsample of poor area. We reestimate our
Strong Weak Rich Poor
regression model and present the results in Table 5. Our
finding supports Hypothesis 3 that firms respond actively to Year fixed effect Yes Yes Yes Yes
the public governance in areas with weak legal environment Firm-year 3437 4668 4734 3502
in terms of local legal index (North 1990; Faccio 2006). We observations
R-squared 0.13 0.09 0.11 0.09
Table 5 Anti-corruption and local characteristics The dependent variable is the dummy variable measuring whether the
firm is committing fraud. POST is a dummy variable defined as 1 if
Local legal Local economy the year is after 2012, 0 otherwise. SIZE is the log value of total asset.
environment LEV is the ratio of total debt over total asset. BM is the book-to-
Strong Weak Rich Poor market value of the firm. ROA is the ROA of the firm. TOBINQ is the
firm’s Tobin’s Q value. BOARD represents the board size of the firm.
POST -0.11 -0.36*** -0.33*** -0.79*** INDP % is the percentage of the independent director in the board.
CEODUAL is a dummy variable defined as 1 if the firm’s CEO is also
(-0.84) (-3.54) (-3.21) (-2.98)
the chairman of the board. AUDIT is the number of auditors of the
SIZE -0.31*** -0.06 -0.21*** -0.12* firm. IAUDIT is a dummy variable defined as 1 if the firm has an
(-4.85) (-1.14) (-4.10) (-1.94) international auditor. ANALYST is the number of analysts following
LEV 1.07*** 0.62** 0.73*** 0.72*** the firm. INSTOWN is total number of shares held by institution
investors divided by the total number outstanding shares of the firm.
(3.78) (2.50) (3.08) (2.65)
GOV is the government ownership of the firm. EXE represents top
BM 0.55 0.36 0.67* 0.59* executives’ ownership of the firm. MKTRET is the annual return of
(1.21) (1.12) (1.89) (1.66) Shanghai composite index. FINCRISIS is a dummy variable if the
ROA -3.54*** -1.13 -3.43*** -0.14 year is after 2008 financial crisis. We divided the sample according to
the sample media of legal index compiled by Fan et al. (2011) and log
(-3.41) (-1.37) (-3.86) (-0.16) value of GDP per capita. We adopt the robust standard errors clus-
TOBINQ 0.05 0.05 0.15** 0.00 tered in the firm level in the regression
(0.48) (0.82) (1.97) (0.05) *, ** and *** indicate statistical significance at the 10, 5, and 1 %
BOARD 0.04 0.03 0.04* 0.04* level, respectively
(1.44) (1.64) (1.77) (1.74)
INDEP % 1.30 -0.31 1.45** -1.05 also find evidence to support Hypothesis 4 that firms respond
(1.49) (-0.40) (2.10) (-1.27) actively to public governance in areas with a poor local
CEODUAL 0.18* 0.11 0.16** 0.19* economy measured by GDP per capita.
(1.86) (1.24) (2.06) (1.89)
We reestimate our regression model based on the CEO
AUDITOR 0.10 -0.06 -0.05 0.03
characteristics in Table 6. We find that older CEOs respond
(1.23) (-0.82) (-0.70) (0.42)
less actively to public governance. This finding supports
Hypothesis 5, suggesting that CEOs’ characteristics may
IAUDIT 0.04 -0.27 -0.10 -0.31
significantly influence their choice in committing fraud.
(0.15) (-0.97) (-0.43) (-1.18)
Due to the multicollinearity problem, IAUDIT and FIN-
ANALYST -0.02 0.05 -0.03 0.07
CRISIS are automatically removed from the regression by
(-0.35) (1.04) (-0.57) (1.38)
the statistical software for the younger CEO sample.
INSTOWN 0.30 -0.40 -0.03 -0.18
(0.41) (-0.79) (-0.04) (-0.34)
Robustness Analysis
GOV -0.34 -0.01 -0.53** -0.09
(-1.18) (-0.04) (-2.08) (-0.46)
Unobservable time-invariant firm characteristics may exist
EXE -0.20 0.28 0.04 0.10
that possibly generate a spurious correlation between
(-0.77) (1.19) (0.18) (0.33)
public governance and corporate fraud. To address this
MKTRET -1.32*** -0.65*** -0.71*** 3.29***
concern, we control for the firm-fixed effect in a logit
(-4.30) (-3.15) (-3.66) (2.91)
regression model.13 The results are shown in Table 7. In
FINCRISIS -1.37*** -0.39 -0.43 4.23***
Table 7, the coefficient of POST is still negative and highly
(-3.48) (-1.49) (-1.18) (3.16)
significant at the 1 % significance level.
Constant 4.85*** 0.43 3.00*** -2.07
(3.73) (0.38) (2.64) (-1.08) 13
There is no firm-fixed effect model for the probit regression.
Industry fixed Yes Yes Yes Yes Therefore, we can only adopt a logit model if we want to control for
effect the firm-fixed effect model.

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

Although the anti-corruption campaign is exogenous to Table 6 continued


an individual firm, reverse causality problems may still CEO age
arise: A large number of public firms may choose to
commit fraud since they can bribe corrupt officials to evade Old Young
investigation or minimize the penalty. The high-profile Year fixed effect Yes Yes
corporate misconduct may ruin the Chinese financial Firm-year observations 5951 2212
market and aggravate the problems of the Chinese econ- R-squared 0.12 0.07
omy. Thus, the government decides to launch the anti-
The dependent variable is the dummy variable measuring whether the
firm is committing fraud. POST is a dummy variable defined as 1 if
the year is after 2012, 0 otherwise. SIZE is the log value of total asset.
Table 6 Anti-corruption and corporate fraud with CEO LEV is the ratio of total debt over total asset. BM is the book-to-
characteristics market value of the firm. ROA is the ROA of the firm. TOBINQ is the
CEO age firm’s Tobin’s Q value. BOARD represents the board size of the firm.
INDP % is the percentage of the independent director in the board.
Old Young CEODUAL is a dummy variable defined as 1 if the firm’s CEO is also
the chairman of the board. AUDIT is the number of auditors of the
POST -0.31*** -0.67*** firm. IAUDIT is a dummy variable defined as 1 if the firm has an
(-3.06) (-2.68) international auditor. ANALYST is the number of analysts following
SIZE -0.12** -0.33*** the firm. INSTOWN is total number of shares held by institution
investors divided by the total number outstanding shares of the firm.
(-2.53) (-4.70) GOV is the government ownership of the firm. EXE represents top
LEV 0.81*** 0.91*** executives’ ownership of the firm. MKTRET is the annual return of
(3.57) (2.96) Shanghai composite index. FINCRISIS is a dummy variable if the
year is after 2008 financial crisis. We adopt the robust standard errors
BM 0.51* 0.81*
clustered in the firm level in the regression
(1.73) (1.84)
*, ** and *** indicate statistical significance at the 10, 5, and 1 %
ROA -2.04*** -1.82* level, respectively
(-2.60) (-1.82)
TOBINQ 0.07 0.06
(1.26) (0.66) corruption campaign to help reshape financial markets and
BOARD 0.04** 0.03 rebuild investors’ confidence in the Chinese economy.14
(2.34) (1.08) To alleviate the concern of reverse causality, we
INDEP % 0.09 1.17 examine whether the relationship between public gover-
(0.14) (1.23) nance and corporate fraud is stronger in large firms or
CEODUAL 0.18** 0.10 small firms. Considering that only corporate misconduct
(2.29) (0.93) in large firms is severe enough to influence the financial
AUDITOR -0.04 0.10 markets and the Chinese economy, one should expect that
(-0.63) (1.12) there is a strong association between public governance
IAUDIT -0.16 and corporate fraud in large firms if corporate fraud
(-0.86) induces the launch of the anti-corruption campaign. We
ANALYST -0.01 0.06 divide the sample between large firms and small firms
(-0.25) (0.99) based on the sample median of firm size. Then, we
INSTOWN -0.28 0.97 reestimate our probit model for each subsample. Table 8
(-0.57) (1.34)
GOV -0.11 -0.30 14
In reality, the situation described here is nearly impossible for
(-0.63) (-0.99) several reasons. First, as stated above, according to the statistics of
EXE 0.30 -0.24 CSRC, 12.7 % of public firms were involved in corporate misconduct
(1.32) (-1.00) in 2012. Although 12.7 % may not be a small proportion, it is still not
large enough to ruin financial markets and the Chinese economy.
MKTRET -0.82*** 2.58** Second, even if the fraud firms have a detrimental effect on financial
(-4.48) (2.42) markets and economy, the first choice for the government to restore
FINCRISIS -0.68*** the economy is to enact new security laws instead of disciplining
government officials. Third, the reason for President Xi Jinping to
(-2.92)
launch anti-corruption campaign may include various factors such as
Constant -0.82*** 2.58** economy, politics, culture, and so on. Corporate fraud is only one
(-4.48) (2.42) problem in the economy, and is not on its own sufficiently important
Industry fixed effect Yes Yes to influence President Xi to launch the boldest and the most serious
anti-corruption campaign.

123
J. Zhang

shows that the coefficient of POST is negative and highly Table 7 continued
significant in both large firms and small firms. Moreover, Variables (1)
the magnitude of the coefficient of POST in small firms is Model
even greater than that of large firms, which contradicts the
claim of reverse causality. This indicates that public Firm-year observations 1806
governance drives the change in fraud likelihood, and it is The dependent variable is the dummy variable measuring whether the
not the other way round. firm is committing fraud. POST is a dummy variable defined as 1 if
the year is after 2012, 0 otherwise. SIZE is the log value of total asset.
LEV is the ratio of total debt over total asset. BM is the book-to-
market value of the firm. ROA is the ROA of the firm. TOBINQ is the
Table 7 Anti-corruption and corporate fraud in logit model con- firm’s Tobin’s Q value. BOARD represents the board size of the firm.
trolling for firm-fixed effect INDP % is the percentage of the independent director in the board.
CEODUAL is a dummy variable defined as 1 if the firm’s CEO is also
Variables (1)
the chairman of the board. AUDIT is the number of auditors of the
Model
firm. IAUDIT is a dummy variable defined as 1 if the firm has an
POST -0.90*** international auditor. ANALYST is the number of analysts following
the firm. INSTOWN is total number of shares held by institution
(-3.84) investors divided by the total number outstanding shares of the firm.
SIZE 0.04 GOV is the government ownership of the firm. EXE represents top
(0.14) executives’ ownership of the firm. MKTRET is the annual return of
Shanghai composite index. FINCRISIS is a dummy variable if the
LEV 1.61**
year is after 2008 financial crisis
(2.08)
*, ** and *** indicate statistical significance at the 10, 5, and 1 %
BM 0.94 level, respectively
(1.17)
ROA -3.28*
(-1.80) Although we believe that the anti-corruption campaign
TOBINQ 0.09 is exogenous to an individual firm, it is possible that
(0.67) macroeconomic factors affect both the timing of the anti-
BOARD 0.11 corruption campaign and a firm’s choice to commit
(1.63) fraud. Economic problems may trigger the need to limit
INDEP % 0.31 corruption in China. In an economic-booming period,
(0.16) firms have less pressure to maintain firm performance
CEODUAL 0.12 and thus have less incentive to commit fraud. Therefore,
(0.50) the association between the anti-corruption campaign and
AUDITOR 0.11 a firm’s likelihood of committing corporate fraud may
(0.64) not be due to the change in public governance but due to
IAUDIT 0.28 the change in macroeconomic factors. To rule out this
(0.32) alternative explanation, we add log value of GDP per
ANALYST 0.05 capital at provincial level to control for the macroeco-
(0.47) nomic factors. The result is presented in column 1 of
INSTOWN 1.28 Table 9. The coefficient of POST is still positive and
(0.99) highly significant.
GOV -0.28 Another alternative explanation is that the firms may
(-0.55) become more ethical in China as Xi Jinping took office.
EXE -1.81** Shortly after Xi Jinping took office, he promoted his per-
(-2.00) sonal slogan, ‘‘Chinese Dream.’’ President Xi describes the
MKTRET -2.12*** dream as ‘‘national rejuvenation, improvement of people’s
(-4.22) livelihoods, prosperity, construction of a better society and
FINCRISIS -1.76** a strengthened military.’’ In the process of chasing the
(-2.52) ‘‘Chinese Dream,’’ corporations may change their culture
Firm-fixed effect Yes
to be more ethical and more socially responsible. The
Year fixed effect Yes
change in corporate culture may significantly alter a firm’s
fraud likelihood. To rule out this alternative explanation,

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

we add variables to proxy for corporate social responsi- Table 8 continued


bility to control for the corporate culture. Our proxy for Firm size
corporate social responsibility includes ten dimensions to
measure whether a firm is socially responsible. The ten Small Large
dimensions include shareholder protection, creditor pro- Firm-year observations 4319 3871
tection, staff protection, delivery protection, customer R-squared 0.10 0.14
The dependent variable is the dummy variable measuring whether the
firm is committing fraud. POST is a dummy variable defined as 1 if the
year is after 2012, 0 otherwise. SIZE is the log value of total asset. LEV is
Table 8 Reverse causality between anti-corruption and corporate
the ratio of total debt over total asset. BM is the book-to-market value of
fraud
the firm. ROA is the ROA of the firm. TOBINQ is the firm’s Tobin’s
Firm size Q value. BOARD represents the board size of the firm. INDP % is the
percentage of the independent director in the board. CEODUAL is a
Small Large dummy variable defined as 1 if the firm’s CEO is also the chairman of
the board. AUDIT is the number of auditors of the firm. IAUDIT is a
POST -0.30** -0.23* dummy variable defined as 1 if the firm has an international auditor.
(-2.57) (-1.94) ANALYST is the number of analysts following the firm. INSTOWN is
SIZE -0.22*** -0.17*** total number of shares held by institution investors divided by the total
number outstanding shares of the firm. GOV is the government own-
(-2.81) (-2.68)
ership of the firm. EXE represents top executives’ ownership of the firm.
LEV 0.70*** 1.02*** MKTRET is the annual return of Shanghai composite index. FINCRISIS
(3.10) (3.17) is a dummy variable if the year is after 2008 financial crisis. We adopt
BM 0.39 1.07*** the robust standard errors clustered in the firm level in the regression
(1.10) (2.73) *, ** and *** indicate statistical significance at the 10, 5, and 1 %
level, respectively
ROA -2.87*** -1.65
(-3.64) (-1.50) protection, environmental protection, public protection,
TOBINQ -0.01 0.28*** work safety, and firm deficiency. The data assigns the value
(-0.21) (3.31) of 1 to a firm if it does well in a dimension, 0 otherwise.
BOARD 0.05** 0.03 The total CSR score is the summation of the value of ten
(2.27) (1.22) dimensions. Therefore, the total value of a firm’s CSR
INDEP % 1.57** -0.74 score ranges from 0 to 10.15 The result is presented in
(2.06) (-0.92) column 2 of Table 9. After controlling for the corporate
CEODUAL 0.08 0.33*** culture, the coefficient of POST is still negative and highly
(0.98) (3.03) significant.
AUDITOR 0.05 -0.10 One caveat with an empirical study of corporate fraud is
(0.82) (-1.16) that we can only observe detected fraud: undetected fraud
IAUDIT -0.22 -0.15 cannot be accounted for empirically. To resolve this issue,
(-0.44) (-0.77) we follow Yu and Yu (2012) and focus on large fraud
ANALYST 0.07 -0.06 cases; we assume that the ex post fraud detection proba-
(1.47) (-1.08) bility for large cases is nearly one. Large firms tend to have
INSTOWN 0.01 0.01 more information disclosure, higher institutional holdings,
(0.01) (0.02) and wider analyst coverage. Presumably it is harder for
GOV -0.28 0.08 larger firms to ex post hide the fraud.16 We restrict our
(-1.20) (0.36)
15
EXE -0.16 1.08*** The CSR data is obtained from the China Stock Market and
(-0.83) (3.06) Accounting Research (CSMAR) database.
16
MKTRET -0.94*** -0.80*** The undetected fraud problem here is that we do not observe the
corporate misconducts if they are not caught by the CSRC. We may
(-3.85) (-3.22) treat those undetected fraud firms as no-fraud firms. If that is the case,
FINCRISIS -0.86*** -0.48 those fraud firms without being caught will be included in the non-
(-2.79) (-1.51) fraud sample. Thus, our regression results might be biased. Yu and Yu
(2012) argues that the fraudulent behaviors of a large firm are less
Constant 3.06* 2.66**
likely to be undetected. In another word, the issue of undetected fraud
(1.79) (2.01) is likely to be more pronounced among smaller firms. Therefore, if we
Industry fixed effect Yes Yes limit the sample in the large firms, we should have a lower probability
Year fixed effect Yes Yes to include the undetected fraud into the non-fraud sample, which
make our regression more reliable.

123
J. Zhang

sample to large firms (top 25 % in size in this sample) and Table 9 continued
adopt a simple probit model to reestimate the effect of Variables (1) (2) (2)
political connections on firms’ fraud incentives. The result Model Model Model

R-squared 0.1 0.1 0.14


Table 9 Anti-corruption and corporate fraud with additional controls
The dependent variable is the dummy variable measuring whether the
Variables (1) (2) (2) firm is committing fraud. POST is a dummy variable defined as 1 if
Model Model Model the year is after 2012, 0 otherwise. SIZE is the log value of total asset.
LEV is the ratio of total debt over total asset. BM is the book-to-
POST -0.22*** -0.26*** -0.29* market value of the firm. ROA is the ROA of the firm. TOBINQ is the
(-2.81) (-3.32) (-1.72) firm’s Tobin’s Q value. BOARD represents the board size of the firm.
INDP % is the percentage of the independent director in the board.
SIZE -0.15*** -0.19*** -0.19* CEODUAL is a dummy variable defined as 1 if the firm’s CEO is also
(-3.71) (-4.61) (-1.92) the chairman of the board. AUDIT is the number of auditors of the
LEV 0.72*** 0.82*** 1.34** firm. IAUDIT is a dummy variable defined as 1 if the firm has an
(3.79) (4.35) (2.43) international auditor. ANALYST is the number of analysts following
the firm. INSTOWN is total number of shares held by institution
BM 0.49* 0.55** 0.86 investors divided by the total number outstanding shares of the firm.
(1.83) (2.12) (1.41) GOV is the government ownership of the firm. EXE represents top
ROA -1.78*** -1.93*** -2.82 executives’ ownership of the firm. MKTRET is the annual return of
Shanghai composite index. FINCRISIS is a dummy variable if the
(-2.79) (-3.03) (-1.58)
year is after 2008 financial crisis. LNGDP is the log value of GDP per
TOBINQ 0.06 0.06 0.36*** capita. The corporate social responsibility (CSR) includes ten
(1.05) (1.15) (2.78) dimensions to measure whether a firm is social responsible. The ten
BOARD 0.03** 0.04** 0.00 dimensions include shareholder protection, creditor protection, staff
protection, delivery protection, customer protection, environment
(1.99) (2.19) (0.12) protection, public protection, work safety, firm deficiency. The data
INDEP % 0.38 0.45 -0.86 assign the value of 1 to a firm if it does well in a dimension, 0
(0.67) (0.79) (-0.70) otherwise. The CSR score is the summation of the value of ten
dimensions. Therefore, the total value of a firm’s corporate social
CEODUAL 0.18*** 0.17*** 0.32
responsibility ranges from 0 to 10. We adopt the robust standard
(2.73) (2.59) (1.46) errors clustered in the firm level in the regression
AUDITOR -0.02 0.00 0.24 *, ** and *** indicate statistical significance at the 10, 5, and 1 %
(-0.38) (0.00) (1.51) level, respectively
IAUDIT -0.16 -0.21 -0.05
(-0.88) (-1.15) (-0.21)
ANALYST 0.02 0.02 0.06 is presented in column 3 of Table 9. After controlling for
(0.57) (0.52) (0.70) undetected fraud, the coefficient of POST remains positive
INSTOWN -0.08 0.00 0.65 and highly significant.17
(-0.18) (0.01) (0.75)
GOV -0.18 -0.12 0.20
(-1.15) (-0.76) (0.67) Conclusion
EXE 0.12 0.09 1.77*
(0.69) (0.54) (1.96) Our study investigates the association between public
MKTRET -0.81*** -0.78*** -0.92***
governance and corporate fraud in China. In this paper, we
(-4.82) (-4.59) (-2.62)
treat the recent anti-corruption campaign initiated by
FINCRISIS -0.53** -0.66*** -0.33
President Xi Jinping as an exogenous shock to public
governance. We also collect the data of the enforcement
(-2.41) (-3.10) (-0.71)
actions against corporate fraud in China between 2004 and
LNGDP -0.27***
2014 from the Chinese Securities Regulatory Commission
(-4.17)
(CSRC). We find that firms are 3.8 % less likely to commit
CSR 0.01
fraud in the post campaign period than in the pre-campaign
(1.51)
period due to the enhanced public governance in China.
Constant 5.00*** 2.88*** 2.56
(4.80) (3.23) (1.21) 17
Following Wang (2013) and Poirier (1980), we also adopt a
Industry fixed effect Yes Yes Yes
bivariate probit model with partial observability to explicitly solve the
Year fixed effect Yes Yes Yes problem of partial observability. However, due to the data charac-
Firm-year observations 8394 8415 1657 teristics, the bivariate probit fails to converge and report any results.
Therefore, we are not able to report the results in the paper.

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

Given the unconditional mean of fraud likelihood of 8.6 %, orchestrated the accounting manipulation and gained
this translates to an economically significant 46.51 % (3.8/ financial benefits from the accounting manipulation. Thus,
8.6 %) increase in fraud likelihood from its average level. the CEO’s characteristics clearly affect a firm’s fraud
A principal characteristic of Chinese partial privatiza- likelihood. We find that older CEOs respond inactively to
tion of SOEs is that the state retains a significant ownership the public governance.
stake after listing. Moreover, a large number of non-SOEs This paper offers clear policy implications for business
maintain connections to politicians in China. Furthermore, ethics by suggesting that public governance can provide
jurisdictions may vary across provinces due to the dispar- external monitoring over corporate decisions. This finding
ities in the extent of market development and legal pro- implies that the government should continue disciplining
tection. Therefore, government ownership, political government officials, thereby reshaping the legal environ-
connections, and local institutional environment may all ment and enhancing law enforcement.
affect a firm’s choice to commit fraud. In this paper, we
show that the association between public governance and
corporate fraud is more evident in privately held listed Appendix
firms, in firms with weak legal environment, and in firms in
areas with poor local economy. See Table 10.
As the key person in the corporation, the CEO is more
likely to be treated by the SEC as the person who

Table 10 Variable definition


Variables Definition Data
Source

FRAUD 1 for firms with CSRC enforcement action, 0 otherwise CSMAR


POST 1 for the observations after 2012, otherwise 0 CSMAR
SIZE The log of total assets CSMAR
BOARD The board size of the firm CSMAR
INDP % The percentage of the independent director in the board CSMAR
CEODUAL A dummy variable defined as 1 if the firm’s CEO is also the chairman of the board CSMAR
AUDIT The number of auditors of the firm
IAUDIT A dummy variable defined as 1 if the firm has an international auditor CSMAR
ANALYST The number of analysts following the firm CSMAR
INSTOWN Total number of shares held by institution over the total number outstanding shares of the firm CSMAR
TOBINQ The firm’s Tobin’s Q value CSMAR
BM The book-to-market value of the firm CSMAR
LEV The ratio of total debt over total asset CSMAR
ROA The ROA of the firm CSMAR
GOV The government ownership of the firm CSMAR
EXE Top executives’ ownership of the firm CSMAR
MKTRET The annual return for Shanghai composite index CSMAR
FINCRISIS A dummy variable defined as 1 if the year is after 2008 financial crisis, otherwise 0 CSMAR
LNGDP Log value of GDP per capita at the provincial level CSMAR
CSR The corporate social responsibility (CSR) includes ten dimensions to measure whether a firm is social responsible. CSMAR
The ten dimensions include shareholder protection, creditor protection, staff protection, delivery protection,
customer protection, environment protection, public protection, work safety, firm deficiency. The data assign the
value of 1 to a firm if it does well in a dimension, 0 otherwise. The CSR score is the summation of the value of ten
dimensions

123
J. Zhang

References Crutchley, C. E., Jensen, M. R., & Marshall, B. B. (2007). Climate for
scandal: corporate environments that contribute to accounting
Acemoglu, D., & Verdier, T. (2000). The choice between market fraud. Financial Review, 42, 53–73.
failures and corruption. American Economic Review, 90, Dechow, P. M., Ge, W., Larson, C. R., & Sloan, R. G. (2011).
Predicting material accounting misstatements. Contemporary
194–211.
Ackerman, S. R. (1978). Corruption: A study in political econ- Accounting Research, 28, 17–82.
omy. New York: Academic Press. Dechow, P., Sloan, R. G., & Sweeney, A. (1996). Causes and
Aggarwal, R. K., Meschke, F., & Wang, T. Y. (2012). Corporate consequences of earnings manipulation: An analysis of firms
subject to enforcement actions by the SEC. Contemporary
political donations: Investment or agency? Business and Politics
14. Accounting Research, 13, 1–36.
Agrawal, A., & Chadha, S. (2005). Corporate governance and Demirgüç-Kunt, A., & Maksimovic, V. (1998). Law, finance, and
accounting scandals. Journal of Law and Economics, 48, firm growth. Journal of Finance, 53, 2107–2137.
371–406. Demirgüç-Kunt, A., & Maksimovic, V. (1999). Institutions, financial
Agrawal, A., & Knoeber, C. R. (2001). Do some outside directors markets, and firm debt maturity. Journal of Financial Eco-
play a political role? Journal of Law and Economics, 44, nomics, 54, 295–336.
179–198. Doidge, C., Andrew Karolyi, G., & Stulz, R. M. (2004). Why are
Aidt, T. S. (2009). Corruption, institutions, and economic develop- foreign firms listed in the US worth more? Journal of Financial
ment. Oxford Review of Economic Policy, 25, 271–291. Economics, 71, 205–238.
Alexander, C. R., & Cohen, M. A. (1999). Why do corporations Doidge, C., Andrew Karolyi, G., & Stulz, R. M. (2007). Why do
countries matter so much for corporate governance? Journal of
become criminals? Ownership, hidden actions, and crime as an
agency cost. Journal of Corporate Finance, 5, 1–34. Financial Economics, 86, 1–39.
Allen, F., Qian, J., & Qian, M. (2005). Law, finance, and economic Donaldson, G., & Lorsch, J. W. (1983). Decision making at the top:
growth in China. Journal of Financial Economics, 77, 57–116. The shaping of strategic direction. New York: Basic Books.
Arlen, J. H., & Carney, W. J. (1992). Vicarious liability for fraud on Dong, B., & Torgler, B. (2013). Causes of corruption: Evidence from
securities markets: Theory and evidence. U. Ill. L. Rev. China. China Economic Review, 26, 152–169.
Beasley, M. (1996). An empirical analysis of the relation between the Dyck, A., Morse, A., & Zingales, L. (2010). Who blows the whistle
board of director composition and financial statement fraud. The on corporate fraud? Journal of Finance, 65, 2213–2253.
Accounting Review, 71, 443–465. Efendi, J., Srivastava, A., & Swanson, E. P. (2007). Why do corporate
Beck, T., & Levine, R. (2002). Industry growth and capital allocation: managers misstate financial statements? The role of option compen-
Does having a market- or bank-based system matter? Journal of sation and other factors. Journal of Financial Economics, 85, 667–708.
Ehrlich, I., & Lui, F. T. (1999). Bureaucratic corruption and
Financial Economics, 64, 147–180.
Becker, G. S. (1968). Crime and punishment: An economic approach. endogenous economic growth. Journal of Political Economy,
The Journal of Political Economy, 76, 169–217. 107, S270–S293.
Bergstresser, D. B., & Philippon, T. (2006). CEO incentives and Faccio, M. (2006). Politically connected firms. The American
Economic Review, 96, 369–386.
earnings management. Journal of Financial Economics, 66,
511–529. Faccio, M., Masulis, R. W., & McConnell, J. J. (2006). Political
Booth, L., Aivazian, V., Demirguc-Kunt, A., & Maksimovic, V. connections and corporate bailouts. Journal of Finance, 61,
(2001). Capital structures in developing countries. Journal of 2597–2635.
Finance, 56, 87–130. Faccio, M., & Parsley, D. C. (2009). Sudden deaths: Taking stock of
Brickley, J. A., Coles, J. L., & Jarrell, G. (1997). Leadership structure: political connections. Journal of Financial and Quantitative
Separating the CEO and chairman of the board. Journal of Analysis, 44, 683–718.
Corporate Finance, 3, 189–220. Fan, C. S., Lin, C., & Treisman, D. (2009). Political decentralization
Burns, N., & Kedia, S. (2006). The impact of performance-based and corruption: Evidence from around the world. Journal of
compensation on misreporting. Journal of Financial Economics, Public Economics, 93, 14–34.
79, 35–67. Fan, J. P., Rui, O. M., & Zhao, M. (2008). Public governance and
corporate finance: Evidence from corruption cases. Journal of
Cai, H., Fang, H., & Xu, L. C. (2011). Eat, drink, firms, government: An
investigation of corruption from the entertainment and travel costs Comparative Economics, 36, 343–364.
of Chinese firms. Journal of Law and Economics, 54, 55–78. Fan, J. P., Titman, S., & Twite, G. (2012). An international
Chen, G., Firth, M., Gao, D., & Rui, Z. (2006). Ownership structure, comparison of capital structure and debt maturity choices.
corporate governance, and fraud: Evidence from China. Journal Journal of Financial and Quantitative Analysis, 47, 23–56.
of Corporate Finance, 12, 424–448. Fan, Joseph P. H., Wong, T. J., & Zhang, T. (2007). Politically-
Chen, G., Michael, F., & Lixin, X. (2009). Does the type of ownership connected CEOs, corporate governance and post-IPO perfor-
control matter? Evidence from China’s listed companies. mance of China’s partially privatized firms. Journal of Financial
Journal of Banking & Finance, 33, 171–181. Economics, 84, 330–357.
Chidambaran, N. K., Kedia, S., & Prabhala, N. R. (2011). CEO Fan, W., Wang, X., & Zhu, H. (2011). Neri index of marketization of
director connections and corporate fraud. Fordham University China Provinces. Beijing: National Economic Research Institute.
Feng, M., Ge, W., Luo, S., & Shevlin, T. (2011). Why do CFOs
Schools of Business Research Paper 1787500.
Claessens, S., & Burcin Yurtoglu, B. (2013). Corporate governance in become involved in material accounting manipulations? Journal
emerging markets: A survey. Emerging markets review, 15, of Accounting and Economics, 51, 21–36.
1–33. Fisman, R., & Miguel, E. (2007). Corruption, norms, and legal
enforcement: Evidence from diplomatic parking tickets. Journal
Cornett, M. M., Marcus, A. J., & Tehranian, H. (2008). Corporate
governance and pay-for-performance: The impact of earnings of Political Economy, 115, 1020–1048.
management. Journal of Financial Economics, 87, 357–373. Giannetti, M. (2003). Do better institutions mitigate agency prob-
Cox, J. D., Thomas, R. S., & Kiku, D. (2003). SEC enforcement lems? Evidence from corporate finance choices. Journal of
heuristics: An empirical inquiry. Duke Law Journal, 53, 737–779. Financial and Quantitative Analysis, 38, 185–212.

123
Public Governance and Corporate Fraud: Evidence from the Recent Anti-corruption Campaign in…

Gintis, H. (2003). The hitchhiker’s guide to altruism: Gene-culture La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R.
coevolution, and the internalization of norms. Journal of (1999b). The quality of government. Journal of Law Economics
Theoretical Biology, 220, 407–418. and Organization, 15, 222–279.
Glaeser, E. L., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (2000a).
(2004). Do institutions cause growth? Journal of Economic Investor protection and corporate governance. Journal of
Growth, 9, 271–303. Financial Economics, 58, 141–186.
Glaeser, E. L., & Saks, R. E. (2006). Corruption in America. Journal La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R.
of Public Economics, 90, 1053–1072. (2000b). Agency problems and dividend policy around the
Gupta, S., & Swenson, C. W. (2003). Rent-seeking by agents of the world. Journal of Finance, 55, 1–34.
firm. Journal of Law and Economics, 46, 253–268. Leuz, C., & Oberholzer-Gee, F. (2006). Political relationships, global
Hackman, R. J. (1992). Group influences on individuals in organi- financing, and corporate transparency: Evidence from Indonesia.
zations. Handbook of Industrial and Organizational Psychology Journal of Financial Economics, 81, 411–439.
(2nd ed., Vol. 3). Palo Alto, CA: Consulting Psychologists Press. Levine, R. (1999). Law, finance, and economic growth. Journal of
Healy, P. M., & Wahlen, J. M. (1999). A review of the earnings Financial Intermediation, 8, 8–35.
management literature and its implications for standard setting. Li, M., Makaew, T., & Winton, A. (2014). Cheating in China:
Accounting Horizons, 13, 365–383. Corporate fraud and the roles of financial markets. Available at
Jayachandran, S. (2006). The Jeffords effect. Journal of Law and SSRN 2521151.
Economics, 49, 397–425. Li, K., Wang, T., Cheung, Y.-L., & Jiang, P. (2011). Privatization and
Jensen, M. C. (1986). Agency costs of free cash flow, corporate risk sharing: Evidence from the split share structure reform in
finance, and takeovers. American Economic Review, 76, China. Review of Financial Studies, 24, 2499–2525.
323–329. Lipset, S. (1960). Political man: The social bases of politics. Garden
Jensen, M. C. (1993). The modern industrial revolution, exit, and the City, NY: Doubleday.
failure of internal control systems. The Journal of Finance, 48, Liu, X. (2014). Corruption Culture and Corporate Misconduct.
831–880. Available at http://www.cicfconf.org/sites/default/files/paper_
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: 965.pdf.
Managerial behavior, agency costs and ownership structure. Lui, F. T. (1985). An equilibrium queuing model of bribery. The
Journal of Financial Economics, 3, 305–360. Journal of Political Economy, 93, 760–781.
Jian, M., & Wong, T. J. (2010). Propping through related party Lui, F. T. (1986). A dynamic model of corruption deterrence. Journal
transactions. Review of Accounting Studies, 15, 70–105. of Public Economics, 31, 215–236.
Johnson, S. A., Ryan, H. E., & Tian, Y. S. (2009). Managerial Mauro, P. (1995). Corruption and growth. The Quarterly Journal of
incentives and corporate fraud: The sources of incentives matter. Economics, 110, 681–712.
Review of Finance, 13, 115–145. Mo, P. H. (2001). Corruption and economic growth. Journal of
Karpoff, J. M., & Lou, X. (2010). Short sellers and financial Comparative Economics, 29, 66–79.
misconduct. The Journal of Finance, 65, 1879–1913. Ng, D. (2006). The impact of corruption on financial markets.
Karpoff, J. M., Scott Lee, D., & Martin, G. S. (2008a). The cost to Managerial Finance, 32, 822–836.
firms of cooking the books. Journal of Financial and Quanti- Ng, D., & Qian, K. (2004). Corruption and corporate governance,
tative Analysis, 43, 581–611. mimeo. Ithaca, NY: Cornell University.
Karpoff, J. M., Scott Lee, D., & Martin, G. S. (2008b). The North, D. C. (1990). Institutions, institutional change and economic
consequences to managers for financial misrepresentation. performance. Cambridge: Cambridge University Press.
Journal of Financial Economics, 88, 193–215. Parsons, C. A., Sulaeman, J., & Titman, S. (2014). Peer Effects and
Karpoff, J., Scott Lee, D., & Vendrzyk, V. P. (1999). Defense corporate corruption. Working Paper.
procurement fraud, penalties, and contractor influence. Journal Peng, L., & Röell, A. (2008). Executive pay and shareholder
of Political Economy, 107, 809–842. litigation. Review of Finance, 12, 141–184.
Kedia, S., & Rajgopal, S. (2011). Do the SEC’s enforcement Poirier, D. J. (1980). Partial observability in bivariate probit models.
preferences affect corporate misconduct? Journal of Accounting Journal of Econometrics, 12, 209–217.
and Economics, 51, 259–278. Prendergast, C., & Stole, L. (1996). Impetuous youngsters and jaded
Khanna, V., Kim, E., & Lu, Y. (2015). CEO connectedness and old-timers: Acquiring a reputation for learning. Journal of
corporate fraud. The Journal of Finance, 70, 1203–1252. Political Economy, 104, 1105–1134.
Kimbro, M. B. (2002). A cross-country empirical investigation of Rajan, R. G., & Zingales, L. (1995). What do we know about capital
corruption and its relationship to economic, cultural, and structure? Some evidence from international data. The Journal of
monitoring institutions: An examination of the role of account- Finance, 50, 1421–1460.
ing and financial statements quality. Journal of Accounting, Reinikka, R., & Svensson, J. (2004). Local capture: evidence from a
Auditing & Finance, 17, 325–350. central government transfer program in Uganda. The Quarterly
Kotter, J. P., & Heskett, J. L. (1992). Corporate culture and Journal of Economics, 119, 679–705.
performance. New York: Free Press. Schein, E. H. (1985). Organizational culture and leadership. San
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1997a). Francisco: Jossey-Bass.
Legal determinants of external finance. Journal of Finance, 52, Serfling, M. A. (2014). CEO age and the riskiness of corporate
1131–1150. policies. Journal of Corporate Finance, 25, 251–273.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. Serra, D. (2006). Empirical determinants of corruption: A sensitivity
(1997b). Trust in large organizations. American Economic analysis. Public Choice, 126, 225–256.
Review, 87, 333–338. Shleifer, A., & Vishny, R. W. (1989). Management entrenchment:
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1998). The case of manager-specific investments. Journal of Financial
Law and finance. Journal of Political Economy, 106, 1113–1155. Economics, 25, 123–139.
La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (1999a). Corporate Shleifer, A., & Vishny, R. (1993). Corruption. Quarterly Journal of
ownership around the world. Journal of Finance, 54, 471–517. Economics, 108, 599–617.

123
J. Zhang

Shleifer, A., & Vishny, R. W. (1994). Politicians and firms. The Wei, S. (2000). How taxing is corruption on international investors?
Quarterly Journal of Economics, 109, 995–1025. Review of Economics and Statistics, 82, 1–11.
Shleifer, A., & Vishny, R. W. (1997). A survey of corporate Wu, X. (2008). Public sector transparency and corporate accounting
governance. The Journal of Finance, 52, 737–783. practices in Asia. Available at SSRN: http://ssrn.com/abstract=
Svensson, J. (2003). Who must pay bribes and how much? Evidence 1404016.
from a cross section of firms. The Quarterly Journal of Wu, Wenfeng, Johan, Sofia A., & Rui, Oliver M. (2014). Institutional
Economics, 118, 207–230. investors, political connections, and the incidence of regulatory
Wang, T. Y. (2011). Corporate securities fraud: Insights from a new enforcement against corporate fraud. Journal of Business Ethics,.
empirical framework. Journal of Law Economics and Organi- doi:10.1007/s10551-014-2392-4.
zation, 31, 1–34. Xie, P., & Lu, L. (2003). Unwilling bribery and collusion. Journal of
Wang, Tracy Yue. (2013). Corporate securities fraud: Insights from a Financial Research, 277, 1–15. (in Chinese).
new empirical framework. Journal of Law, Economics, and Yu, F. F. (2008). Analyst coverage and earnings management.
Organization, 29(3), 535–568. Journal of Financial Economics, 88, 245–271.
Wang, T. Y., & Winton, A. (2014). Product market interactions and Yu, F., & Xiaoyun, Yu. (2012). Corporate lobbying and fraud
corporate fraud. Available at SSRN 2398035. detection. Journal of Financial and Quantitative Analysis, 46,
Wang, T. Y., Winton, A., & Xiaoyun, Yu. (2010). Corporate fraud 1865–1891.
and business conditions: Evidence from IPOs. Journal of
Finance, 65, 2255–2292.

123

You might also like