Abstract
CHINA JOURNAL OF ACCOUNTING STUDIES 2019, VOL. 7, NO. 1, 25–61 https://doi.org/10.1080/21697213.2019.1625577 ARTICLE Xixiong Xu, Wanli Li and Xichan Chen School of Economics and Business Administration, Chongqing University, Chongqing, China ABSTRACT KEYWORDS Confucianism; informal Using a sample of Chinese-listed firms during the period from institution; stock price crash 2008–2017, this paper investigates the impact of Confucian culture risk on stock price crash risk and its underlying mechanism. We find that Confucianism is significantly negatively associated with firm- specific crash risk. Further channel tests show that Confucian culture curbs crash risk mainly through mitigating agency conflict, improving financial information quality and restraining managerial overconfidence. Moreover, we also document that the negative association between Confucianism and crash risk is more promi- nent in firms with weaker corporate governance and lower analyst coverage. Our findings suggest that Confucian ethics, as an impli- cit norm and alternative mechanism of formal institutions plays a critical role in preventing stock price crash and promoting the healthy development of capital market. This study not only enriches the literature on stock price crash risk but also deepens theoretical cognition of the positive value of Confucian culture from firm-level. 1. Introduction Stock price crash refers to the phenomenon that stock price falls sharply in a short period of time without warning. It not only causes great losses to investors’ wealth but also destroys financial stability and even endangers the normal operation of real economy. Since the global financial crisis in 2008, the incidents of stock price plummet are emerging in endlessly around the world. China’s securities market is immature, where external investor protection and internal governance mechanism are still relatively weak and undeveloped (Allen, Qian, & Qian, 2005). Hence, the surge and plunge of stock price in China’s capital market is even more serious (Piotroski & Wong, 2010; Xu, Jiang, Yi, & Xu, 2012). For instance, in 2015, China’s stock market appeared a rare phenomenon of ‘1000 shares limit up’ and ‘1000 shares limit down’; in 2018, Changsheng Biotechnology had fallen for 32 consecutive trad- ing days due to the outbreak of the false vaccine incident, which refreshed the record of the longest limit down of individual stock in A-share market and led to 86.68% volatilisation of corporate market value. Therefore, stock price crash risk has recently become a hot research topic in the fields of economics and finance. CONTACT Wanli Li liwanlicqu@sina.com School of Economics and Business Administration, Chongqing University, China Paper accepted by Kangtao Ye. © 2019 Accounting Society of China 26 X. XU ET AL. Prior studies mostly rely on the bad news hoarding theory proposed by Jin and Myers (2006). This theory argues that managers tend to withhold bad news so as to pursue their private benefits. However, the amount of bad news that firms can absorb or accumulate has an upper limit. When the accumulation of bad news reaches the threshold, all of them will be released at once, leading to stock price falls sharply (Kim, Li, & Zhang, 2011a). Following this theory, some scholars have investigated the determinants of crash risk, including equity incentives (Kim et al., 2011a), tax avoid- ance (Kim, Li, & Zhang, 2011b), excess perks (Xu, Li, Yuan, & Chan, 2014), institutional investors (An & Zhang, 2013), financial reporting opacity (Hutton, Marcus, & Tehranian, 2009;Kim &Zhang, 2014), analyst coverage (Xu et al., 2012), accounting conservatism (Kim & Zhang, 2016), and CEO’s individual characteristics (Kim, Wang, & Zhang, 2016;Li &Liu, 2012). However, few literatures have explored the impact of cultural factors on crash risk (Callen & Fang, 2015). Upper echelons theory argues that the individual cognition and behavior preference of managers has been shaped by their growth environment and cultural soil, which will be reflected in corporate decision-making (Hambrick, 2007). The emerging literature on ‘culture and finance’ have also shown that, besides institutional and economic factors, informal institutions such as culture also play a pivotal role in corporate decision-making (Li, Griffin, Yue, & Zhao, 2013). Especially, compared with western-developed countries, China has a long history and cultural tradition, but its system construction and market mechanism are still not perfect. Therefore, culture may play a more important role in shaping corporate policies in the emerging market of China (Allen et al., 2005; Chen, Hu, Liang, & Xin, 2013). Confucius constructed the philosophy of Confucianism such as moral order, duty, ceremony, and the respect of family and authority during the Warring States Period (475–221 BCE) (Du, 2015). So far, Confucian culture has existed in China for thou- sands of years. As the core of Chinese traditional culture, Confucianism is the moral norm and action guide generally respected by individuals and organisations (Ip, 2009). Du (2003) believes that Confucian culture not only shapes the spirit of Chinese entrepreneur but also became a crucial spiritual pillar in the process of modernisation. Fu and Tsui (2003) further argue that Confucianism is pervasive in the values of Chinese entrepreneurs and is reflected in business decision-making. Recently, some empirical research also reveals that Confucian culture is beneficial to mitigate agency conflicts (Du, 2015;Gu, 2015) and improve the quality of internal control (Cheng, Pan, & Wang, 2016). Different from extant literature, this paper investigates the impact of Confucian culture on firm-specific stock price crash risk. Theoretically, Confucian culture may play a positive role in restraining firm-specific crash risk. On the one hand, Confucian philosophical emphasis on ‘loyalty’, ‘honesty’, ‘righteousness before profitableness’,and ‘self-discipline’,which helpstomitigate managerial agency conflicts and deter managerial bad news hoarding behavior. On the other hand, Confucian culture always advocates the doctrine of the Mean and stresses prudence in words and deeds, which may also reduce managerial over- confidence and improve accounting conservatism. Using a sample of Chinese-listed firms for the period 2008–2017, this paper system- atically examines the effect of Confucianism on firm-specific crash risk. We find that firms headquartered in regions with strong Confucianism atmosphere have lower crash risk CHINA JOURNAL OF ACCOUNTING STUDIES 27 than those headquartered in regions with weak Confucianism atmosphere, implying that Confucian culture can effectively reduce the likelihood of firm-specific future stock price crash. Furthermore, we examine the economic mechanisms through which Confucianism reduces crash risk. The evidences show that the negative effect is achieved mainly through mitigating agency conflict, improving financial information quality and reducing managerial overconfidence. In addition, we also document that the negative association between Confucianism and crash risk is more prominent in firms with weaker corporate governance and lower analyst coverage. The findings suggest that Confucian ethics, as an informal institution, helps to make up for the deficiency of formal institutions in the emerging capital markets and play an alternative governance func- tion. Our results are robust to a battery of sensitivity checks, including controlling for some possibly omitted variables, accounting for potential endogenous issues and adopting alternative measures of Confucianism and crash risk. Interestingly, after con- trolling for Confucianism and religion simultaneously, Confucianism is still significantly negatively correlated with crash risk while the significantly negative relationship between religion and crash risk suddenly disappear. This shows that Confucian ethics, in Chinese cultural context, rather than religious tradition is the more important cultural factor affecting firm-specific crash risk. We contribute to the extant literature in several ways. First, this study extends research in the stock price crash risk literature from the perspective of informal institu- tions. Previous studies primarily focus on firm-level characteristics and formal institu- tional factors to investigate the determinants of crash risk, including managers’ self- interest incentive (Kim et al., 2011a, 2011b; Xu et al., 2014), financial information quality (Hutton et al., 2009; Kim & Zhang, 2014, 2016), corporate governance (An & Zhang, 2013; Jiang, Cai, & Zhu, 2018; Wang, Cao, & Ye, 2015), and executive characteristics (Kim et al., 2016; Li & Liu, 2012). This paper examines the impact of Confucian culture and its ethical value on crash risk and its economic mechanism, and thus contributes to the literature on the determinants of crash risk. Second, Confucianism is the core component of Chinese traditional culture and plays a very critical role in the informal system of Chinese society. However, existing literature about Confucian ethical values mainly focuses on the fields of philosophy and sociology and adopts normative research methods. Far less attention has been paid to the role- played by Confucianism in corporate decision-making. Du (2012) argues that an impor- tant bottleneck in the development of Confucianism is ‘facing the challenge of scient- ism, exposing many defects’. Recently, a few scholars have empirically examined the influence of Confucian culture on corporate agency conflict (Du, 2015;Gu, 2015), internal control (Cheng et al., 2016), and risk-taking (Jin, Xu, & Ma, 2017). We provide new empirical evidence that Confucian ethics can deter managerial bad news hoarding behavior and thus reduce crash risk, which deepens the theoretical cognition for economic consequences of Confucianism from firm-level. Third, using data from American and Chinese listed firms, respectively, Callen and Fang (2015) and Li and Cai (2016) document that religion can also significantly reduce crash risk. However, different from western countries, the religious atmosphere in China are relatively weak. In contrast, Confucian tradition is the most influential cultural factor in Chinese society. Especially, the negative effect of Confucianism still remains while that of religion disappears suddenly after controlling for Confucianism and religiosity 28 X. XU ET AL. simultaneously, indicating that Confucianism rather than religion is the more important cultural factor in China. This is not only a beneficial supplement to the research of Li and Cai (2016), but also contributes the empirical evidence from the oriental culture to the emerging literature on ‘culture and finance’. Finally, as an interaction between society and technology, accounting is inevitability- restricted and influenced by culture (Pan, Li, Lin, & Su, 2012; Zhan, Xu, & Kang, 2017). The particularity and incompleteness of accounting standards and systems leave more discretionary space. Hence, empirically examining the influence of Chinese traditional culture on accounting information is an unavoidable crucial research issue. Our channel analyses suggest that Confucianism can curb earnings management, reduce the fre- quency of financial restatements, and improve accounting conservatism. In this sense, it also contributes to the literature that empirically investigates how Chinese traditional culture affects the generation of accounting information. The rest of this paper is organised as follows. Section 2 reviews related literature and develops hypotheses. Section 3 introduces the research design. Section 4 reports the empirical results. Section 5 performs further analyses. Section 6 presents robustness checks. Section 7 provides conclusions. 2. Literature review and hypotheses development 2.1. Literature review Stock price crash risk has become a hot research topic of economic and financial fields after the global financial crisis. The existing literature mainly attributed the formation mechanism of stock price crash risk to the fact that, managers tend to report good news and temporarily hide bad news due to career concerns, compensation contracts, and private benefits of control (Ball, 2009; Kim et al., 2011a; Kothari, Shu, & Wysocki, 2009). However, the amount of bad news that firms can absorb or accumulate has an upper limit. When the accumulation of bad news reaches the threshold, all of them will be released at once, leading to stock price crash (Kim et al., 2011a). Following this idea, subsequent scholars have investigated the potential determinants of crash risk. From the internal perspective of firms, the existing literature has examined the impact of information quality, corporate governance, and executive characteristics on stock price crash risk. First, many studies have found that financial reporting transparency is signifi- cantly negatively correlated with crash risk. For instance, Hutton et al. (2009) and Kim and Zhang (2014) find that financial reporting opacity increases firm’s future crash risk. Furthermore, tax avoidance activities facilitate managers to hide bad news and thus increase crash risk (Jiang, 2013;Kim et al., 2011b). In contrast, some studies show that accounting conservatism (Kim et al., 2016), internal Control (Ye, Cao, & Wang, 2015), and social responsibility disclosure (Song, Hu, & Li, 2017) can reduce crash risk. Furthermore, the self-interested motivation of managers is an important cause of stock price crash. Related literature suggests that stock option incentive (Kim et al., 2011a), excess perks consumption (Xu et al., 2014), post promotion (Piotroski, Wong, & Zhang, 2015), over- investment (Jiang & Xu, 2015), and management selling (Sun, Liang, Ruan, & Fu, 2017)are significantly positively associated with crash risk. However, the shares of the controlling shareholders (Wang et al., 2015), the independence of board (Liang & Zeng, 2016), and CHINA JOURNAL OF ACCOUNTING STUDIES 29 multiple large shareholders (Jiang et al., 2018) can mitigate crash risk. Recently, some scholars also explore the impact of executive characteristics on crash risk. For example, Li and Liu (2012) find that female CEOs tend to be more conservative and risk-averse than male CEOs, and hence firms with female CEOs have lower crash risks. Kim et al. (2016) argue that overconfident managers tend to overestimate the returns of investment projects, underestimate the risk of failure, ignore negative feedback, and release more optimistic forecast information, which will lead to higher crash risk. Jiang, Yao, and Chen (2018) also document that the higher the position of CFO in the top management team, the lower the risk of stock price crash. From the external perspective of firms, some studies find that external factors such as analyst coverage, institutional investors, media coverage and formal institution have also a critical influence on the firm-specific future crash risk. For instance, related research shows that analyst coverage and institutional investors, as an effective external governance mechanism, can reduce crash risk (An & Zhang, 2013; Pan, Dai, & Lin, 2011). However, analysts’ optimistic forecasts and the herd behavior of institutional investors increase crash risk (Xu et al., 2012;Xu, Yu,&Yi, 2013). Furthermore, Luo and Du (2014) suggest that media coverage helps to deter man- agerial bad news hoarding behavior, improve firms’ information transparency, and therefore curb crash risk. Zhao, Huang, and Liu (2018) also document that the opening of high-speed rail reduces crash risk by accelerating the flow of information. In addition, from the perspective of formal institutions, prior studies indicate that investor protection (Wang, Cao, Gao, & Li, 2014)and delistingregulation (Lin& Zheng, 2016) restrain crash risk, while the implementation of short-selling and margin-trading program worsen crash risk (Chu & Fang, 2016). From the perspective of informal institutions, recent research has also shown that religious traditions (Callen & Fang, 2015;Li&Cai, 2016) and social trust (Li, Wang, & Wang, 2017)can inhibit firms’ crash risk. Through the above literature review, we can conclude that extant studies pri- marily focus on firm-level characteristics and formal institutional factors to investi- gate the determinants of stock price crash risk, and have gained abundant achievements. However, few scholars pay attention to the impact of implicit normssuchasculture on firm-specific crash risk. At present, only Callen and Fang (2015)and LiandCai(2016) have examined the impact of religion on crash risk. Huntington and Harrison (2013) argue that culture is an important factor influen- cing social, political, and economic activities. It contains a set of persistent beliefs or values that often have a lasting and profound impact on individual thoughts, perceptions, preferences, and behaviors (North, 1990). Recently, the emerging literature on ‘culture and finance’ have also shown that, besides institutional and economic factors, informal institutional factors such as culture also plays a pivotal role in corporate decision-making (Li et al., 2013). However, different from western countries, religious atmosphere is relatively weak in Chinese society. In contrast, Confucianism, as the core of Chinese traditional culture, is actually the most influential culture in China. Therefore, this paper seeks to explore the impact of Confucianism on stock price crash risk from the perspective of informal institutions, so as to further enrich the research in this field. 30 X. XU ET AL. 2.2. Confucianism and crash risk: hypotheses development It should not be enough to understand China’s economic and management problems if we ignore informal institutions such as traditional culture, which have been slowly formed over the past thousands of years and have far-reaching influence (Allen et al., 2005; Chen et al., 2013). Confucianism is the most lasting and important cultural element in the Chinese philosophical system. Confucianism was founded by Confucius, a famous ideologist in the Spring and Autumn Period. After being inherited and developed by Mencius and Xunzi in the Warring States Period, Confucianism gradually formed an integrated ideological system (Li & Nie, 2011). Confucianism regards ‘Rén (benevolence), Yì (righteousness), Lǐ (ritual propriety), Zhì (wisdom), and Xìn (trustworthiness)’ as the core ideology. It advocates virtuous personal characteristics such as loyalty, honesty, obedience, righteousness, and self-discipline (Du, 2015). Since Emperor Wu of the Han Dynasty adopted the ‘spring and autumn annals’ unification outlook and the suggestion of ‘To abandon hundreds, Only Confucianism’ proposed by Dong Zhongshu, Confucianism had begun to become the mainstream culture and dominated Chinese Society for more than two thousand years (Zhang, 2013). For thousands of years, Confucianism has gradually become the moral norms and ethical principles respected by individuals and organisations who daily use without knowing (Ip, 2009). Although modern China has undergone tremendous historical changes such as Opium War, May Fourth Movement, and Cultural Revolution, the far-reaching influence of Confucianism on Chinese society has never been denied. Upper echelons theory argues that the individual cognition and behavior preference of managers has been shaped by their growth environment and cultural soil, which will be reflected in corporate decision-making (Hambrick, 2007). Hence, we predict that Confucianism, as the main ethical and philosophical system could have a crucial influ- ence on firm-specific stock price crash risk. Specifically, by the in-depth interpretation of relevant thoughts in Confucian classics, this paper argues that Confucianism may affect crash risk through the following three aspects. First, Confucian ethics help to mitigate agency conflicts. Extant studies show that managers tend to withhold bad news so as to pursue their private benefit such as option portfolio value (Kim et al., 2011a), excess perk consumption (Xu et al., 2014), post promotion (Piotroski et al., 2015), and private benefits of control (Jiang & Xu, 2015). The self-interested motivation of managers is an important cause of stock price crash. As an implicit moral norms and ethical principles, Confucian ethics can play an active role in restraining managers’ opportunistic behavior (Du, 2015; Gu, 2015). On the one hand, Confucianism advocates the value of ‘righteousness before profit’. The Master said, ‘Riches and honors acquired by unrighteousness, are to me as a floating cloud’ (Analects, 7:15) and ‘The mind of the superior man is conversant with righteousness; the mind of the mean man is conversant with gain’ (Analects, 4:16). Mencius explained the meaning of ‘righteousness’, and argued that ‘how to take what one has not a right In our study, the translations of Analects, Mencius, The Li Ki, and Book of Poetry are by James Legge. Specially, the detailed translation content, see The Chinese Classics, Volume I and II (Hong Kong: Hong Kong University Press, 1960, reprinted in 1970); and The Sacred Books of China (The Texts of Confucianism) in The Sacred Books of the East, Volumes III, XXVII, and XXVIII. Following conventional practice, the source is cited, for instance, as ‘Analects, 7:15,’ which shows Analects, chapter 7, section 15, and Mencius, 7A:33, which represents Mencius, chapter 7, part 1, and section 33. CHINA JOURNAL OF ACCOUNTING STUDIES 31 to is contrary to righteousness’ (Mencius, 7A:33). From the angle of agency relationship, the view of ‘Righteousness and Benefit’ will facilitates managers to respect the owner- ship of shareholders, and can’t ‘refuse to break through, or jump over, a wall’. On the other hand, Confucianism emphasises the ethical thought of ‘loyalty and trustworthi- ness’. The Master said, ‘The scholar does not consider gold and jade to be precious treasures, but loyalty and good faith’ (The Li Ki). Zengzi also argued that ‘whether, in transacting business for others, I may have been not faithful; whether, in intercourse with friends, I may have been not sincere’ (Analects 1:4). The ethical thoughts of ‘loyalty and trustworthiness’ require that managers should keep their contractual commitments, adhere to the principle of shareholders’ benefit maximisation, and work hard for share- holders. In addition, Confucianism also advocates that people should improve the levels of self-discipline or self-regulation through the self-cultivation and refinement of their character, and thus achieve the goal of ‘the superior man is watchful over himself, when he is alone’ (The Li Ki). This value helps managers regulate their own behavior to maintain an ethical standard of practice, even no one else around or lacking of effective supervision. Thus it can be seen, Confucian ethics, as an implicit moral norms and ethical principles, will form a strong internal moral constraint on managerial unethical behavior, and further could reduce firm-specific crash risk. Second, Confucianism could improve financial information quality. Managers tend to create an opaque information environment by disclosing fraud financial reporting or hiding bad news, which makes it difficult for investors to identify or know the actual operating conditions of companies in time, and eventually lead to stock price crash (Hutton et al., 2009; Kim & Zhang, 2014, 2016). We believe that Confucianism could have an important impact on the quality of corporate information disclosure, thus playing a positive role in curbing firms’ crash risk. For one thing, Confucianism regards ‘sincerity and faith’ as the most basic moral norms and ethical principles, and as the foundation of living in the world. The Master said, ‘If the people have no faith in their rulers, there is no standing for the state’ (Analects 12:7) and ‘I do not know how a man without truthful- ness is to get on’ (Analects 2:22). Mencius also argued that ‘Sincerity is the way of Heaven. The attainment of sincerity is the way of men’ (The Li Ki). Zhu Xi explained its meaning, and argued that ‘sincerity’ means ‘truth without delusion’, and ‘faith’ is the embodiment of ‘sincerity’. That is, ‘sincerity based on one’s heart is the same as faith’. The concept of sincerity and faith could deter managers from releasing fraud financial reporting or hiding bad news, and facilitate them to more timely and correctly reflect enterprise operating conditions. This will help to alleviate the information asymmetry between companies and investors, improve financial reporting transparency, and avoid stock price bubbles. For another thing, ‘caution’ was believed by Confucian sages as a kind of moral cultivation, which contains the meaning of circumspection, meticulous- ness, and rigorousness. The Book of Poetry vividly describes ‘caution’ as ‘We should be apprehensive and careful, as if we were on the brink of a deep gulf, as if we were treading on thin ice’. In dealing with people and things, Confucius advocates caution in speech and prudence in behavior. The Master said, ‘He is earnest in what he is doing, and careful in his speech’ (Analects 1:14) and ‘Hear much and put aside the points of which you stand in doubt, while you speak cautiously at the same time of the others: Then you will afford few occasions for blame. See much and put aside the things which seem perilous, while you are cautious at the same time in carrying the others into 32 X. XU ET AL. practice: Then you will have few occasions for repentance’ (Analects 2:18). The Li Ki also recorded that ‘The superior man is careful at the commencement; a mistake, then, of a hair’s breadth, will lead to an error of a thousand li’. The risk awareness of ‘caution’ will encourage managers to choose more conservative accounting policy, and reduce the likelihood of financial reporting errors, and thus result in a lower crash risk (Kim & Zhang, 2016). Third, Confucianism could also reduce managerial overconfidence. Based on beha- vioral finance theory, Kim et al. (2016) examine the impact CEO overconfidence on crash risk. They argue that overconfident managers tend to overestimate the returns of investment projects, underestimate the risk of failure, ignore or explain away privately observed negative feedback, and release more optimistic forecast information. This will lead to external investors can’t understand the actual operating conditions of compa- nies, and hence give rise to stock price bubbles. As we all know, Confucianism advocates the doctrine of the mean. Cheng Yi interpreted ‘the mean’ as ‘Zhong means bent neither one way or another, and Yong represents unchanging’. In the Commentary on Doctrine of the Mean, Zhu Xi also explained its meaning, and said that ‘The Zhong means not leaning or inclining and being without excess or deficiency’. Confucius himself regards the doctrine of the mean as a requirement of virtue. The Master said, ‘Perfect is the virtue which is according to the Constant Mean. Rare for a long time has been its practice among the people’ (Analects 6:27) and ‘The superior man embodies the course of the Mean; the mean man acts contrary to the course of the Mean’ (The Li Ki). Confucius opposed radical thoughts and actions, emphasising that people should grasp the appropriate degree in everything, which means ‘ impartiality, moderation and felicitous’. We predict that the doctrine of the mean could reduce the psychological bias of managerial overconfidence, which will help to avoid overestimating returns, underestimating risks, ignoring negative feedback, and releasing optimistic forecast information. In addition, recent research has suggested that people in individualistic cultures are more likely to form overconfident and overoptimistic personality traits, which will aggravate stock price crash risk (An, Chen, Li, & Lu, 2018). However, Confucianism advocates the thoughts of collectivism such as ‘When the Grand course was pursued, a public and common spirit ruled all under the sky’ (The Li Ki) and ‘In order for people to live, they cannot be without community’ (Xunzi). The thoughts of collecti- vism will further mitigate the psychological bias of managerial overconfidence, and thus reduce firm-specific crash risk. Through the above analysis, we argue that the ethical principles of Confucian ‘loyalty and trustworthiness’, ‘righteousness before profit’ and ‘self-discipline’ could curb man- agerial bad news hoarding behavior for their private benefit, and improve financial information quality. Meanwhile, the thoughts of Confucian ‘the Mean’, ‘collectivism’ and ‘caution’ may be also conducive to reduce managers’ overconfidence and improve accounting conservatism. Hence, we present the following hypothesis: Hypothesis 1: Firms headquartered in regions with strong Confucianism atmosphere have lower crash risk than those headquartered in regions with weak Confucianism atmosphere CHINA JOURNAL OF ACCOUNTING STUDIES 33 3. Research design 3.1. Sample selection Based on the background of global financial crisis, this paper selects Chinese A-share listed companies during the period from 2008 to 2017 as an initial sample. Following the existing studies (Hutton et al., 2009; Kim et al., 2011a), we exclude the samples as follows: (1) financial services firms; (2) firms with foreign capital holding; (3) firms with fewer than thirty trading weeks of stock returns in a fiscal year; (4) firms with missing values for all the variables. The final sample consists of 17,349 firm-year observations. The data on Confucianism are manually collected, and the specific process refers to the definition of variables. Other financial data and corporate governance data are obtained from the China Stock Market and Accounting Research (CSMAR) database, WIND data- base, and RESSET database. To rule out outlier effects, we winsorise the values of all continuous variables at the 1% and 99% levels. 3.2. Variables 3.2.1. Measuring confucianism The direct measure of culture or ideology will inevitably encounter some difficult problems, and its measure methods have always been controversial (North, 1990). With the rise of ‘culture and finance’ research, some scholars have recently begun to use historical information data to examine the impact of cultural and institutional factors on economic behaviors. For example, La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998) use the origin of law as the proxy of investor protection, and find that both Common Law and Continental Law were closely related to historical and political traditions. Guiso, Sapienza, and Zingales (2008) also show that the historical events 500 years can ago affect the level of trust among people, which explained the differ- ences in economic performance between the Southern and Northern Italy. Similarly, Du (2015), Jin et al. (2017), Gu (2015) and Cheng et al. (2016) use the number of Confucian temples or academies around a firm’s registered address as a proxy for Confucianism. Compared with other methods, using historical information data has the advantages of objectivity, repeatability and stability. Hence, this method has recently been widely used in the finance and accounting literature. The most prominent feature of Confucianism is that it guides people’s thoughts and behaviors through ethical education, and finally forms the humanistic environment of ‘Morality is woven into custom’. Since Emperor Wu of the Han Dynasty adopted the suggestion of ‘To abandon hundreds, Only Confucianism’ proposed by Dong Zhongshu, he ordered ‘all counties and states establish official schools’. Since then, school educa- tion has gradually become the main way to spread Confucianism. In the Tang Dynasty, Confucian schools mainly consisted of three types: Guozijian directly leaded by the central government, Confucian schools and academies built by local officials. In the Song Dynasty, Confucian schools were gradually divided into official schools and semi- official academies. In the Ming Dynasty, with the rose of Wang Yangming’s mind school, the academy gradually replaced the official school as the most important place for the spread of Confucianism. Therefore, following Gu (2015) and Cheng et al. (2016), we 34 X. XU ET AL. Figure 1. Map of Confucian academies in China. employ the number of Confucian academies around a firm’s registered address as the proxy for Confucianism, which can objectively reflect the influence of Confucianism. In particular, according to the records of ‘Synthesize of Chinese Local Records’ (Zhu, 1958) and ‘Chinese Academy Dictionary’ (Ji, 1996), we manually collect the names and addresses of Confucian academies in provincial administrative areas from Tang Dynasty to Qing Dynasty. Meanwhile, using Google-Earth and Baidu Map, we check the geo- graphic location of every academies site and obtain its longitude and latitude. Similarly, we locate the registered addresses of listed companies and recognise its longitude and latitude. Furthermore, we calculate the geodetic distance for each firm–academy pair. Finally, we use the number of Confucian academies within a radius of 100, 200 and 300 kilometers around a firm’s registered address (Confu_100, Confu_200, and Confu_300)to measure the influence of Confucianism. The more number of Confucian academies around a firm’s registered address suggests that the firm is more strongly influenced by Confucianism. Figure 1 depicts the geographic distribution of Confucian academies in different regions of China. 3.2.2. Measuring stock price crash risk Following prior studies (Hutton et al., 2009; Kim et al., 2011a), we use two proxies to measure stock price crash risk: the negative conditional return skewness of firm-specific weekly (NCSKEW) and the down-to-up volatility of crash likelihood (DUVOL). The calcu- lated process is as follows: CHINA JOURNAL OF ACCOUNTING STUDIES 35 First, using the weekly return data of firm i, we estimate firm-specific weekly returns (denoted by W ) by regressing the following expanded market model for each firm and i,t year: R ¼ α þ β R þ β R þ β R þ β R þ β R þ ε (1) i;t i m;t2 t1 m;t m;tþ1 m;tþ2 i;t 1;i 2;i 3;i 4;i 5;i where R is the return on stock i in week t, and R is the return on the value-weighted i,t m,t market index in week t. The lead and lag terms for the market index return are included to allow for nonsynchronous trading. ε is the residual in the model (1). The firm-specific i,t weekly return is calculated by W = Ln (1+ ε ). i,t i,t Second, based on W , we construct the following two proxies for stock price crash i,t risk: (1) The negative conditional return skewness of firm-specific weekly (NCSKEW)is calculated as follows: hi X X 3=2 3=2 3 2 NCSKEW ¼ nðn 1Þ W = ðn 1Þðn 2Þ W (2) i;t i;t i;t where n is the number of trading weeks on stock i in year t. The higher the value of NCSKEW, the greater left skewness in the distribution of firm-specific excess returns, implying that the firm has a higher crash risk. (2) The down-to-up volatility of crash likelihood (DUVOL) is calculated as follows: nhihio X X 2 2 DUVOL ¼ log ðn 1Þ W = ðn 1Þ W (3) i;t u down d up i;t i;t where n (n ) is the number of weeks that a firm’s W is higher (lower) than the mean of u d i,t W in year t. Firm with higher value of DUVOL are more likely to crash. i,t 3.2.3. Measuring financial reporting opacity Following Xu et al. (2014) and Kim and Zhang (2016), we employ the absolute value of abnormal accruals, calculated as the residuals from the modified Jones model, as the proxy for financial reporting opacity (Opaque). Specifically, the following regression equation is estimated for each industry and year: TA 1 ΔREV ΔREC PPE i;t i;t i;t i;t ¼ α þα þα3 þε (4) 1 2 i;t Asset Asset Asset Asset i;t1 i;t1 i;t1 i;t1 where TA is total accruals of firm i in year t, which is calculated as income minus cash flows from operation activities; Asset is total assets in year t-1; ΔREV is change in sales in t year; ΔREC is change in accounts receivable in t year; PPE is property, plant, and equipment in t year. The greater the absolute value of abnormal accruals, the higher financial reporting opacity. 3.3. Models Following Hutton et al. (2009) and Kim et al. (2011a), we construct the following model (5) to examine the impact of Confucianism on stock price crash risk: 36 X. XU ET AL. Table 1. Variable definitions. Variables Definition NCSKEW The negative conditional return skewness of firm-specific weekly in year t +1 t+1 DUVOL The down-to-up volatility of crash likelihood in year t +1 t+1 The number of Confucian academies within a radius of 100 kilometers around a firm’s registered Confu_100 address divided by 1000 in year t Confu_200 The number of Confucian academies within a radius of 200 kilometers around a firm’s registered address divided by 1000 in year t The number of Confucian academies within a radius of 300 kilometers around a firm’s registered Confu_300 address divided by 1000 in year t Turnover The average monthly share turnover in year t minus the average monthly share turnover in year t-1 Ret The mean of firm-specific weekly returns in year t The standard deviation of firm-specific weekly returns in year t Sigma NCSKEW The negative conditional return skewness of firm-specific weekly in year t Size The natural logarithm of total assets in year t Lev The ratio of total debt to total assets in year t Net profit divided by total assets in year t Roa MB The market value of shares plus the book value of total debt divided by the book value of total assets in year t Opaque The absolute value of abnormal accruals, calculated as the residuals from the modified Jones model in year t X X Crash Risk ¼ β þβ Confucian þβ Control þ YEARþ INDþε (5) i;tþ1 i;t i;t i;t 0 1 j where Crash_Risk is the dependent variable, which is measured by NCSKEW and i,t+1 DUVOL; Confucian reflects the influence of Confucianism, which is measured by i,t Confu_100, Confu_200, and Confu_300. Based on previous studies, we include the following control variables that have been shown to affect stock price crash risk: the detrended stock trading volume (Turnover ), the standard deviation of firm-specific i,t weekly returns (Sigma ), the mean of firm-specific weekly returns (Ret ), the lagged i,t i,t variable of crash risk (NCSKEW ), firm size (Size ), financial leverage (Lev), return on i,t i,t assets (Roa ), the market-to-book ratio (MB ), and financial reporting opacity (Opaque). i,t i,t In addition, we also include industry (IND) and year (YEAR)effects in our regression. It should be noted that all the dependent variables are measured in year t + 1, while all the independent and control variables are measured in year t. The detailed definitions of all variables are shown in Table 1. 4. Empirical results 4.1. Descriptive statistics Table 2 reports the summary statistics of the key variables in our analysis. The means of NCSKEW and DUVOL are −0.460 and −0.415, respectively, which are similar to the estimates by Wang et al. (2015) and Tian and Wang (2017). Meanwhile, the standard deviations of NCSKEW and DUVOL are 0.840 and 0.741, respectively, indicating that stock price crash risk varies greatly among firms. The maximum value of Confu_100 (Confu_200 and Confu_300) is 0.268 (0.783 and 1.250). However, the minimum value of Confu_100 (Confu_200 and Confu_300) is 0. This suggests that the strength of Confucianism atmo- sphere presents large variations among firms. The distribution of control variables is basically consistent with the existing research. CHINA JOURNAL OF ACCOUNTING STUDIES 37 Table 2. Descriptive statistics. Variables N Mean Std. Dev. Minimum Median Maximum NCSKEW 17,349 −0.460 0.840 −2.672 −0.453 1.500 t+1 DUVOL 17,349 −0.415 0.741 −2.255 −0.404 1.287 t+1 17,349 0.102 0.072 0.000 0.084 0.268 Confu_100 Confu_200 17,349 0.298 0.191 0.000 0.317 0.783 Confu_300 17,349 0.549 0.321 0.000 0.517 1.250 Turnover 17,349 0.118 0.787 −0.808 −0.138 2.930 17,349 0.055 0.018 0.026 0.052 0.112 Ret Sigma 17,349 0.000 0.008 −0.020 0.001 0.019 Size 17,349 21.982 1.281 19.317 21.816 25.830 Lev 17,349 0.462 0.215 0.051 0.464 0.970 17,349 0.038 0.055 −0.188 0.035 0.197 Roa MB 17,349 3.958 3.413 0.553 2.958 22.640 Opaque 17,349 0.076 0.077 0.001 0.054 0.442 According to the strength of Confucianism atmosphere, we divide the sample into two groups. A sample will be classified as firms with strong (weak) Confucianism atmo- sphere if the value of Confu_100 (Confu_200 and Confu_300) is higher (lower) than the median of all samples. Table 3 presents univariate comparisons of crash risk between firms with strong Confucianism atmosphere and those with weak Confucianism atmo- sphere. Panel A presents the results of the univariate tests grouped by Confu_100.We can find that the mean of NCSKEW (DUVOL)is −0.473 (−0.428) for firms with strong Confucianism atmosphere and −0.447 (−0.400), and the differences are both statistically significant at the 5% level. Meanwhile, we also document that the medians of NCSKEW and DUVOL are also significantly different between the two groups. We find similar results in Panel B and Panel C. The findings suggest that firms headquartered in regions of strong Confucianism atmosphere tend to have lower crash risk than those head- quartered in regions of weak Confucianism atmosphere, providing preliminary support for our hypothesis. The Pearson and Spearman correlation coefficients between variables are reported in Table 4. The results show that the crash risk measures, NCSKEW and DUVOL, are highly correlated. Furthermore, NCSKEW and DUVOL are significantly and negatively associated with Confu_100, Confu_200, and Confu_300, which are consistent with our expectation. However, NCSKEW and DUVOL are also significantly correlated with Turnover, Ret, Sigma, Table 3. Univariate comparisons. Firms with strong Confucianism Firms with weak Confucianism atmosphere atmosphere Differences Variables N Mean Median N Mean Median T value Z value Panel A：Grouping by Confu_100 NCSKEW 8480 −0.447 −0.438 8869 −0.473 −0.466 0.027** 3.024* DUVOL 8480 −0.400 −0.388 8869 −0.428 −0.419 0.028** 4.553** Panel B：Grouping by Confu_200 NCSKEW 8424 −0.446 −0.432 8925 −0.474 −0.470 0.028** 5.365** DUVOL 8424 −0.400 −0.384 8925 −0.428 −0.422 0.028** 8.025*** Panel C：Grouping by Confu_300 NCSKEW 8523 −0.442 −0.427 8826 −0.477 −0.476 0.035*** 9.363*** DUVOL 8523 −0.399 −0.383 8826 −0.430 −0.424 0.032*** 9.550*** *, ** The test for mean difference is the T-test. The test for median difference is Z test. and *** indicate the significance levels of 10%, 5%, and 1%, respectively. 38 X. XU ET AL. Table 4. Correlation matrix. Variables NCSKEW DUVOL Confu_100 Confu_200 Confu_300 Turnover Ret Sigma Size Lev Roa MB Opaque t+1 t+1 t t t t t t t t t t t NCSKEW 1.000 0.919*** −0.018** −0.021*** −0.022*** 0.087*** 0.079*** 0.093*** −0.113*** −0.036*** 0.050*** 0.155*** 0.031*** t+1 DUVOL 0.933*** 1.000 −0.024*** −0.028*** −0.028*** 0.110*** 0.095*** 0.087*** −0.114*** −0.033*** 0.035*** 0.149*** 0.024*** t+1 Confu_100 −0.022*** −0.027*** 1.000 0.795*** 0.693*** −0.015* 0.018** 0.038*** −0.080*** −0.117*** 0.085*** −0.010 −0.031*** Confu_200 −0.026*** −0.030*** 0.845*** 1.000 0.942*** −0.012 0.017** 0.033*** −0.062*** −0.075*** 0.076*** −0.009 −0.024*** Confu_300 −0.024*** −0.029*** 0.748*** 0.948*** 1.000 −0.009 0.009 0.028*** −0.058*** −0.066*** 0.061*** −0.015** −0.028*** Turnover 0.134*** 0.148*** −0.018** −0.017** −0.012 1.000 0.422*** −0.031*** 0.018** 0.058*** −0.061*** 0.116*** −0.001 Ret 0.101*** 0.107*** 0.018** 0.013* 0.004 0.405*** 1.000 −0.009 −0.259*** −0.019** −0.040*** 0.398*** 0.106*** Sigma 0.088*** 0.083*** 0.032*** 0.028*** 0.024*** 0.013* 0.000 1.000 −0.084*** −0.049*** 0.107*** 0.112*** 0.029*** Size −0.101*** −0.097*** −0.075*** −0.062*** −0.051*** 0.023*** −0.272*** −0.079*** 1.000 0.394*** 0.033*** −0.365*** −0.061*** Lev −0.035*** −0.034*** −0.112*** −0.081*** −0.060*** 0.061*** −0.027*** −0.046*** 0.419*** 1.000 −0.397*** 0.010 0.153*** Roa 0.061*** 0.048*** 0.101*** 0.082*** 0.065*** −0.071*** −0.009 0.109*** −0.021*** −0.426*** 1.000 0.063*** −0.075*** MB 0.199*** 0.193*** 0.021*** 0.014* −0.001 0.174*** 0.465*** 0.107*** −0.409*** −0.133*** 0.215*** 1.000 0.157*** Opaque 0.028*** 0.024*** −0.027*** −0.026*** −0.026*** 0.003 0.084*** 0.036*** −0.030*** 0.119*** −0.006 0.088*** 1.000 *, ** This table presents the Pearson (Spearman) correlation matrix in the upper (lower) triangle regions. and *** indicate the significance levels of 10%, 5%, and 1%, respectively. CHINA JOURNAL OF ACCOUNTING STUDIES 39 Size, Lev, Roa, MB, and Opaque. Therefore, it is important to examine the impact of Confucianism on crash risk within a multivariate framework. In addition, we compute the variance inflation factor (VIF) for independent variables, and find that all of them are less than 5. Thus, we conclude that there is no serious multicollinearity issue among variables. 4.2. Multivariate results Table 5 reports the baseline regression results of the main regression model. In columns (1)–(6), we use NCSKEW as the dependent variable. First three columns present the results without control variables. We find that NCSKEW is significantly and negatively correlated with each of three proxies for Confucianism. In columns (4)–(6), we introduce the full set of control variables. The coefficients of Confu_100, Confu_200, and Confu_300 are −0.216, −0.096, and −0.057, and all of them are statistically significant at the 1% level. Meanwhile, the negative relationship between Confucianism and crash risk is also economically significant. When Confu_100, Confu_200, and Confu_300 increase one- standard-deviation, crash risk will decrease 1.85%, 2.18%, and 2.18% of a standard deviation. The findings suggest that firms headquartered in regions of strong Confucianism atmosphere have lower crash risk than those headquartered in regions of weak Confucianism atmosphere. That is to say, Confucian culture can effectively reduce the likelihood of firm-specific future stock price crash, supporting our hypothesis. In columns (7)–(12), using DUVOL as the alternative measure of crash risk, the empirical results remain unchanged. The estimation results of the control variables are generally consistent with extant literature (Kim et al., 2011b; Xu et al., 2014). For instance, firms with higher Sigma, higher Ret, larger Size, lower Lev, higher MB, and higher Opaque are correlated with higher crash risk. 4.3. Economic mechanisms Hitherto, we provide strong evidence that Confucianism can significantly reduce firm- specific future crash risk. In this section, we seek to explore the economic mechanisms through which Confucianism influences crash risk. According to previous theoretical analysis, we argue that Confucianism may curb crash risk through the following three aspects: (1) mitigating agency conflict; (2) improving financial information quality; (3) reducing managerial overconfidence. To confirm our conjecture, we further examine the above three paths to reveal the transmission mechanism behind the negative effect of Confucianism on crash risk. 4.3.1. Confucianism and agency conflict Prior studies argue that managers tend to withhold bad news so as to pursue their private benefit, ultimately leading to stock price crash (Jiang & Xu, 2015; Kim et al., 2011a, 2011b; Xu et al., 2014). We argue that Confucian ethics, as an implicit moral norms and ethical principles will form a strong internal moral constraint on managerial unethical behavior. Therefore, Confucianism could reduce crash risk by alleviating the corporate agency conflict. We use the ratio of managerial expenses to total assets as the 40 X. XU ET AL. Table 5. Confucianism and stock price crash risk. NCSKEW DUVOL t+1 t+1 Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Confu_100 −0.161* −0.216*** −0.205*** −0.250*** (−1.94) (−2.64) (−2.85) (−3.53) Confu_200 −0.081*** −0.096*** −0.096*** −0.107*** (−2.63) (−3.17) (−3.59) (−4.09) Confu_300 −0.052*** −0.057*** −0.059*** −0.063*** (−2.82) (−3.16) (−3.71) (−4.02) Turnover −0.014 −0.014 −0.014 −0.007 −0.007 −0.007 (−1.40) (−1.41) (−1.40) (−0.82) (−0.84) (−0.82) Sigma 3.659*** 3.667*** 3.650*** 3.939*** 3.947*** 3.928*** (7.42) (7.44) (7.40) (9.26) (9.28) (9.24) Ret 14.851*** 14.826*** 14.836*** 12.277*** 12.249*** 12.259*** (12.80) (12.78) (12.79) (12.19) (12.16) (12.17) NCSKEW 0.089*** 0.089*** 0.089*** 0.074*** 0.074*** 0.074*** (9.54) (9.51) (9.50) (9.23) (9.19) (9.18) Size 0.018*** 0.018*** 0.018*** 0.017*** 0.017*** 0.017*** (2.81) (2.80) (2.79) (2.98) (2.97) (2.97) Lev −0.249*** −0.247*** −0.247*** −0.257*** −0.255*** −0.255*** (−6.79) (−6.74) (−6.74) (−8.10) (−8.04) (−8.03) Roa −0.048 −0.041 −0.046 −0.279*** −0.273** −0.279*** (−0.39) (−0.34) (−0.38) (−2.62) (−2.56) (−2.61) MB 0.024*** 0.024*** 0.024*** 0.021*** 0.021*** 0.021*** (11.12) (11.12) (11.12) (10.67) (10.68) (10.68) Opaque 0.170** 0.169** 0.168** 0.096 0.094 0.093 (2.21) (2.19) (2.18) (1.42) (1.40) (1.39) Constant 0.023 0.029 0.033 −0.581*** −0.575*** −0.568*** −0.014 −0.008 −0.003 −0.600*** −0.595*** −0.587*** (0.46) (0.56) (0.66) (−3.72) (−3.68) (−3.63) (−0.30) (−0.18) (−0.07) (−4.39) (−4.35) (−4.30) IND/YEAR YES YES YES YES YES YES YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 Adj. R 0.182 0.182 0.182 0.212 0.212 0.212 0.219 0.219 0.219 0.250 0.250 0.250 *, ** Values of heteroscedasticity robust t-statistics are in parentheses. and *** indicate the significance levels of 10%, 5%, and 1%, respectively. The same as in the following tables. CHINA JOURNAL OF ACCOUNTING STUDIES 41 proxy for the first kind agency cost (Agency_Cost). Meanwhile, Xu et al. (2014) and Jiang and Xu (2015) find that excess perks and overinvestment significantly exacerbate stock price crash risk. Hence, we further use excess perks (Excess_Perk) and overinvestment (Over_Invest) as the representative of managerial self-interest behavior. In addition, we also use other receivables scaled by total assets to measure the second kind agency cost (Tunneling). We construct the following models: X X Agency Cost ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (6) i;t 0 1 i;t j i;t i;t X X Over Invest ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (7) i;t i;t i;t i;t 0 1 j X X Excess Perk ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (8) i;t i;t i;t i;t 0 1 j X X Tunneling ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (9) i;t 0 1 i;t j i;t i;t Following prior studies, we control for the following variables: Size, Lev, Roa, CF (the net operating cash flow divided by the total assets), Growth (sales growth rate), Age (firm age), Dual (a dummy variable that equals 1 if a firm’s CEO and board chairman are the same person), Board (the number of board directors), Indep (the proportion of indepen- dent directors on the board), and Top1 (the proportion of shares held by the largest shareholder). Table 6 represents the regression results. We find that Confucianism is significantly negatively associated with Agency_Cost, Excess_Perk, Over_Invest, and Tunneling. This is consistent with our conjecture, suggesting that Confucianism does could reduce firms’ crash risk through mitigating agency conflict. 4.3.2. Confucianism and financial information quality Managers tend to create an opaque information environment by disclosing fraud financial reporting, hiding bad news, or early revealing good news. Prior studies show that financial opacity will increase crash risk, while accounting conservatism can induce crash risk (Hutton et al., 2009; Kim & Zhang, 2014, 2016). According to the previous analysis, we predict that Confucianism could reduce crash risk through improving financial information quality. This study uses three measures as the proxy for financial information quality: earnings manage- ment, financial restatement, and accounting conservatism. In particular, earnings manage- ment (EM) is measured by the absolute value of abnormal accruals, calculated as the residuals from the modified Jones model. Restate is a dummy variable that equals 1 if a firm announces a financial restatement and zero otherwise. We employ C-Score estimated by Khan and Watts (2009) as the proxy for accounting conservatism. A Logit regression is used in the estimation when the dependent variable is Restate. Control variables are the same as the model (6). We construct the following models: X X EM ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (10) i;t 0 1 i;t j i;t i;t X X Restate ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (11) i;t i;t i;t i;t 0 1 j The detailed definitions of excess perks and overinvestment refer to the research of Richardson (2006) and Quan, Wu, and Wen (2010), respectively. 42 X. XU ET AL. Table 6. Confucianism and agency conflict. Agency_Cost Over_Invest Excess_Perk Tunneling t t t t Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Confu_100 −0.118*** −0.017** −0.005* −0.069*** (−9.85) (−2.17) (−1.72) (−7.15) Confu_200 −0.058*** −0.011*** −0.003** −0.023*** (−12.57) (−3.60) (−2.56) (−5.54) Confu_300 −0.035*** −0.007*** −0.002*** −0.016*** (−12.81) (−3.82) (−2.94) (−6.34) Size −0.023*** −0.023*** −0.023*** −0.004*** −0.004*** −0.004*** −0.003*** −0.003*** −0.003*** −0.006*** −0.006*** −0.006*** (−21.42) (−21.59) (−21.57) (−7.19) (−7.21) (−7.22) (−9.92) (−9.93) (−9.95) (−6.46) (−6.43) (−6.45) Lev −0.091*** −0.089*** −0.089*** −0.005 −0.005 −0.005 0.005*** 0.005*** 0.005*** 0.033*** 0.034*** 0.034*** (−12.17) (−12.03) (−12.04) (−1.12) (−1.12) (−1.13) (3.47) (3.53) (3.52) (4.73) (4.85) (4.83) CF −0.000 0.001 0.001 0.049*** 0.050*** 0.050*** 0.035*** 0.035*** 0.035*** −0.099*** −0.099*** −0.099*** (−0.03) (0.07) (0.05) (5.90) (6.00) (5.99) (10.55) (10.58) (10.58) (−6.45) (−6.47) (−6.45) Roa −0.171*** −0.166*** −0.168*** −0.011 −0.011 −0.011 0.026*** 0.026*** 0.026*** −0.142*** −0.142*** −0.142*** (−5.65) (−5.52) (−5.58) (−0.75) (−0.72) (−0.75) (4.30) (4.34) (4.33) (−5.22) (−5.21) (−5.22) Growth −0.028*** −0.028*** −0.028*** 0.011*** 0.011*** 0.011*** −0.001* −0.001* −0.001* −0.009** −0.009** −0.009** (−7.10) (−7.22) (−7.25) (5.07) (5.03) (5.02) (−1.90) (−1.90) (−1.89) (−2.19) (−2.23) (−2.25) Age −0.000 −0.000 −0.000 −0.000*** −0.000*** −0.000*** 0.000 0.000 0.000 0.001*** 0.001*** 0.001*** (−0.20) (−0.42) (−0.41) (−3.65) (−3.71) (−3.70) (0.97) (0.92) (0.93) (3.24) (3.12) (3.11) Dual 0.012*** 0.011*** 0.011*** 0.003** 0.003** 0.003** 0.000 0.000 0.000 0.005*** 0.005** 0.005** (4.91) (4.90) (4.81) (2.04) (2.10) (2.10) (0.17) (0.15) (0.13) (2.59) (2.48) (2.46) Board 0.021*** 0.021*** 0.021*** 0.005 0.005 0.005 0.004*** 0.004*** 0.004*** −0.003 −0.003 −0.003 (4.15) (4.09) (4.18) (1.56) (1.51) (1.53) (3.09) (3.06) (3.09) (−0.60) (−0.58) (−0.56) Indep 0.105*** 0.098*** 0.100*** 0.020* 0.018 0.019 0.016*** 0.016*** 0.016*** 0.039** 0.037** 0.037** (5.74) (5.37) (5.44) (1.66) (1.52) (1.53) (3.35) (3.21) (3.22) (2.30) (2.17) (2.17) Top1 −0.044*** −0.043*** −0.045*** 0.005 0.005 0.005 0.003* 0.003* 0.003* −0.057*** −0.057*** −0.057*** (−7.14) (−7.05) (−7.26) (1.28) (1.39) (1.36) (1.65) (1.69) (1.67) (−10.31) (−10.36) (−10.42) Constant 0.627*** 0.636*** 0.638*** 0.123*** 0.125*** 0.126*** 0.048*** 0.048*** 0.049*** 0.220*** 0.221*** 0.223*** (27.58) (27.94) (27.99) (8.62) (8.75) (8.80) (8.01) (8.09) (8.15) (10.53) (10.50) (10.60) IND/YEAR YES YES YES YES YES YES YES YES YES YES YES YES N 17,144 17,144 17,144 6354 6354 6354 7619 7619 7619 17,248 17,248 17,248 Adj. R 0.329 0.332 0.332 0.095 0.096 0.096 0.139 0.139 0.139 0.130 0.129 0.130 CHINA JOURNAL OF ACCOUNTING STUDIES 43 X X CScore ¼ β þβ Confucian þβ Controls þ YEARþ INDþε (12) i;t i;t i;t i;t 0 1 j The empirical results are reported in Table 7.We documentthatConfucianism is signifi- cantly negatively correlated with EM and Restate but significantly positively associated with C-Score.The findings indicate that Confucianism can reduce firms’ earnings manage- ment and the likelihood of financial restatements and improve accounting conservatism. Therefore, we conclude that Confucianism does could curb crash risk by improving financial information quality, which is reflected by the decreased earnings management and frequency of financial restatements and increased accounting conservatism. 4.3.3. Confucianism and managerial overconfidence Kim et al. (2016) argue that overconfident managers tend to overestimate the returns of investment projects and ignore or explain away privately observed negative feedback. They find that firms with overconfident CEOs have higher crash risk than those with non-over- confident CEOs. However, we argue that the thoughts of ‘the Mean’ and ‘collectivism’ of Confucianism may be conducive to reduce managerial overconfidence and thus can reduce crash risk. Therefore, we further empirically investigate the potential economic mechanism. Following Jiang, Zhang, Lu, and Chen (2009) and Jiang and Xu (2015), we construct two proxies for managerial overconfidence. The first proxy, Overconfidence1,is basedon corporate earnings forecasts. In particular, we define the managers as overconfident (Overconfidence1 equals 1) if a firm’ forecast earnings are higher than its actual earnings and as non-overconfident otherwise (Overconfidence1 equals 0). Furthermore, we use the change of managers’ share holdings as the second proxy for managerial overconfidence. Overconfidence2 is a dummy variable that equals 1 if managers increase their stock holdings and zero otherwise. We construct the following models: X X Over Con ¼ β þ β Confucian þ β Controls þ YEAR þ IND þ ε (13) i;t i;t i;t i;t 0 1 j In the model (13), we control for the following variables: Size, Lev, MB, Age, Dual, Board, Indep, MAge (the average age of executives), MEducation (the education level of executives), and MGender (the proportion of male managers on top management team). We use Logit model to estimate the impact of Confucianism on managerial overconfidence. Table 8 reports the empirical results. We find a negative and significant relationship between Confucianism Overconfidence1 and Overconfidence2, implying that Confucianism can reduce managerial overconfidence. The results provide evidence that managerial overconfidence does could be a viable economic mechanism by which Confucianism curb crash risk. 5. Further analyses 5.1. The substitution effects of monitoring mechanisms North (1990) emphasises that the formal and informal institutions, as two important components of the institutional system, have an extremely pivotal influence on corpo- rate behaviors. The direction of their roles may be complementary or alternative. As China’s capital market is still in the primary stage of development, and the overall level The detailed calculation methods refer to the studies of Jiang et al. (2009) and Jiang and Xu (2015). 44 X. XU ET AL. Table 7. Confucianism and financial information quality. EM Restate C_Score t t t Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Confu_100 −0.014* −0.650** 0.017* (−1.80) (−2.12) (1.90) Confu_200 −0.006** −0.348*** 0.006* (−2.07) (−3.03) (1.82) Confu_300 −0.005*** −0.246*** 0.005** (−2.98) (−3.69) (2.27) Size −0.011*** −0.011*** −0.011*** −0.147*** −0.148*** −0.148*** 0.022*** 0.022*** 0.022*** (−16.25) (−16.27) (−16.29) (−6.75) (−6.79) (−6.81) (26.29) (26.29) (26.30) Lev 0.054*** 0.054*** 0.054*** 0.508*** 0.515*** 0.514*** 0.007 0.007 0.007 (13.45) (13.49) (13.48) (3.93) (3.98) (3.97) (1.38) (1.34) (1.34) CF −0.008 −0.007 −0.007 −1.978*** −1.948*** −1.946*** −0.094*** −0.094*** −0.094*** (−0.51) (−0.48) (−0.46) (−4.25) (−4.18) (−4.18) (−4.94) (−4.94) (−4.95) Roa 0.016*** 0.016*** 0.016*** 0.254*** 0.251*** 0.249*** −0.006*** −0.006*** −0.006*** (6.84) (6.82) (6.80) (3.93) (3.89) (3.87) (−2.86) (−2.83) (−2.82) Growth 0.001*** 0.001*** 0.001*** −0.003 −0.003 −0.003 −0.000*** −0.000*** −0.000*** (4.65) (4.61) (4.61) (−0.56) (−0.62) (−0.63) (−2.77) (−2.73) (−2.73) Age 0.005*** 0.005*** 0.005*** 0.041 0.040 0.040 −0.003 −0.003 −0.003 (3.82) (3.80) (3.81) (0.81) (0.80) (0.79) (−1.63) (−1.60) (−1.60) Dual −0.008** −0.008** −0.008** 0.001 −0.002 −0.000 −0.003 −0.003 −0.003 (−2.35) (−2.36) (−2.36) (0.01) (−0.02) (−0.00) (−0.76) (−0.77) (−0.77) Board 0.022* 0.021* 0.021* 0.208 0.165 0.161 −0.046*** −0.046*** −0.045*** (1.92) (1.85) (1.83) (0.45) (0.35) (0.35) (−3.19) (−3.14) (−3.14) Indep 0.012*** 0.012*** 0.012*** −0.224 −0.218 −0.222 −0.024*** −0.024*** −0.024*** (3.18) (3.18) (3.18) (−1.45) (−1.42) (−1.44) (−5.17) (−5.16) (−5.14) Top1 0.275*** 0.276*** 0.277*** 1.761*** 1.817*** 1.852*** −0.424*** −0.424*** −0.425*** (19.25) (19.26) (19.33) (3.38) (3.49) (3.56) (−23.88) (−23.87) (−23.88) IND/YEAR YES YES YES YES YES YES YES YES YES N 17,230 17,230 17,230 17,256 17,256 17,256 16,456 16,456 16,456 Adj./PseudoR 0.080 0.080 0.080 0.049 0.050 0.050 0.191 0.191 0.191 CHINA JOURNAL OF ACCOUNTING STUDIES 45 Table 8. Confucianism and managerial overconfidence. Overconfidence1 Overconfidence2 t t Variables (1) (2) (3) (4) (5) (6) Confu_100 −1.241*** −1.064*** (−3.34) (−2.79) Confu_200 −0.342** −0.274* (−2.45) (−1.88) Confu_300 −0.150* −0.148* (−1.75) (−1.67) Size −0.088*** −0.087*** −0.088*** 0.310*** 0.312*** 0.311*** (−2.69) (−2.65) (−2.69) (9.28) (9.35) (9.33) Lev 2.560*** 2.565*** 2.568*** 0.315* 0.322* 0.321* (14.92) (14.96) (14.99) (1.79) (1.83) (1.82) MB 0.010 0.011 0.011 −0.023** −0.023** −0.023** (1.09) (1.18) (1.15) (−2.05) (−2.02) (−2.05) Age 0.030*** 0.030*** 0.030*** 0.053*** 0.053*** 0.053*** (5.31) (5.27) (5.31) (9.10) (9.03) (9.05) Dual 0.180*** 0.184*** 0.186*** 0.059 0.064 0.064 (2.97) (3.03) (3.06) (0.92) (1.00) (1.00) Board −0.031 −0.023 −0.018 0.143 0.146 0.150 (−0.17) (−0.13) (−0.10) (0.80) (0.81) (0.84) Indep 0.445 0.435 0.467 0.423 0.416 0.440 (0.72) (0.70) (0.75) (0.69) (0.68) (0.72) MEducation 0.016 0.006 0.008 −0.165*** −0.171*** −0.170*** (0.29) (0.10) (0.15) (−2.86) (−2.95) (−2.93) MGender 0.121 0.138 0.152 0.365 0.378 0.390* (0.53) (0.61) (0.67) (1.55) (1.60) (1.65) MAge 0.014 0.015* 0.016* −0.017* −0.016* −0.016* (1.60) (1.71) (1.75) (−1.77) (−1.70) (−1.68) Constant 4.065*** 3.987*** 3.951*** −6.700*** −6.786*** −6.802*** (4.45) (4.37) (4.34) (−7.74) (−7.82) (−7.83) IND/YEAR YES YES YES YES YES YES N 8058 8058 8058 6807 6807 6807 Pseudo R 0.160 0.159 0.159 0.090 0.089 0.089 of corporate governance and information disclosure environment are relatively weak (Allen et al., 2005). At this stage, it is extremely crucial whether capital market can rely on informal institution to make up for the deficiency of formal institution. Therefore, we further examine whether Confucian ethics, as an informal institutional factor, can make up for the deficiency of formal institution in the emerging capital markets and play an alternative governance function. Specifically, we simultaneously consider internal and external monitoring mechanisms, and predict that the negative association between Confucianism and crash risk is more prominent in firms with weaker monitoring mechanisms. 5.1.1. Internal monitoring mechanisms: corporate governance Corporate governance as formal institutions can effectively alleviate corporate agency conflict. This paper uses the level of corporate governance as the proxy for the strength of internal monitoring. Following Tian and Wang (2017), we apply principal component analysis to compile a single composite index of corporate governance (CG). In particular, we construct CG by taking first principal component of the following ten variables: Dual, Board, Indep, Top1, Balance (the shares held by the second to fifth largest shareholders to the shares held by the largest shareholder), MH (the percentage of shares held by management), SOE (a dummy variable that equals zero if a firm’ ultimate controlling 46 X. XU ET AL. shareholder is government, and otherwise one), Sup (the size of supervisory board), Cdc (the number of board meetings), and Cjc (the number of supervisory board meetings). Furthermore, we divide the sample into two groups: the better corporate governance subsets with CG above the industry median in a given year (CG_good), and the poorer corporate governance subsets with CG below the industry median in a given year (CG_poor). Table 9 reports the regression results of the sub-sample analysis. We docu- ment that Confucianism is significantly negatively associated with crash risk in firms with poorer corporate governance, but statistically insignificant in firms with better corporate governance. The results suggest that the effect of Confucianism on crash risk is more pronounced in firms with poorer corporate governance. 5.1.2. External monitoring mechanisms: analyst coverage For a long time, as an important medium of information dissemination in the capital market, analysts’ professional background and unique information channels help to reduce information asymmetry between investors and companies and improve corporate information transparency. Prior studies show that financial analysts can play an external monitoring role (Kim et al., 2011b;Pan et al., 2011; Xu et al., 2012). Hence, we employ the level of analyst coverage as a proxy for the strength of external monitoring (AN). Anal is the log of one plus the number of analysts following. We divide the sample into two groups: the higher analyst coverage subsets with Anal above the industry median in a given year (AN_high), and the lower corporate governance subsets with Anal below the industry median in a given year (AN_low). The results are presented in Table 10.The estimated coefficients of Confucianism are negatively significant in firms with lower analyst coverage, but not significant in firms with higher analyst coverage. This indicates that the effect of Confucianism on crash risk is more prominent in firms with lower analyst coverage. Taken the above together, we document that the negative effect of Confucianism on crash risk is more prominent in firms with weaker monitoring mechanisms, such as poorer corporate governance and lower analyst coverage. The findings are consistent with our prediction, implying that Confucian ethics, as an informal institutional factor, helps to make up for the deficiency of formal institutions in the emerging capital markets and play an alternative governance function. Moreover, this also provides further evidence to support the previous conclusion. 5.2. The moderating effects of stock market quotation We further investigate whether the inhibitory effect of Confucianism on crash risk varies in different stock market quotation (bull or bear markets). In general, good and bad news of the same degree tend to make the influence of different degree on the stock market fluctuations. When the stock market is in a bear market, it means that the stock market is in a downturn, economic and industrial development is relatively depressed, and uncertainty risk is rising. In this case, investors react often more strongly to firms’ bad news due to the asymmetric reaction of the stock market (Li & Liu, 2012; Lu & Xu, 2004). Therefore, to obtain external financing and boost investor sentiment and con- fidence, firms are more likely to disclose false information and withhold bad news in a CHINA JOURNAL OF ACCOUNTING STUDIES 47 Table 9. The substitution effects of corporate governance. NCSKEW DUVOL t+1 t+1 CG_good CG_poor CG_good CG_poor CG_good CG_poor CG_good CG_poor CG_good CG_poor CG_good CG_poor Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Confu_100 0.035 −0.469** 0.029 −0.369** (0.21) (−2.45) (0.21) (−2.22) Confu_200 −0.023 −0.205*** −0.038 −0.170*** (−0.37) (−3.07) (−0.71) (−2.90) Confu_300 −0.014 −0.113*** −0.025 −0.088** (−0.35) (−2.89) (−0.76) (−2.58) Turnover −0.036* −0.006 −0.036* −0.006 −0.036* −0.006 −0.023 −0.022 −0.023 −0.023 −0.024 −0.023 (−1.69) (−0.25) (−1.69) (−0.26) (−1.69) (−0.26) (−1.30) (−1.19) (−1.31) (−1.21) (−1.32) (−1.21) Sigma 3.392*** 4.453*** 3.390*** 4.476*** 3.389*** 4.447*** 3.811*** 4.097*** 3.808*** 4.115*** 3.806*** 4.093*** (3.02) (3.90) (3.02) (3.92) (3.02) (3.89) (4.02) (4.13) (4.02) (4.15) (4.01) (4.12) Ret 10.801*** 11.123*** 10.801*** 11.062*** 10.801*** 11.128*** 9.570*** 10.510*** 9.567*** 10.454*** 9.568*** 10.515*** (4.72) (4.48) (4.72) (4.46) (4.72) (4.49) (4.91) (4.85) (4.91) (4.83) (4.91) (4.86) NCSKEW 0.076*** 0.056*** 0.076*** 0.056*** 0.076*** 0.056*** 0.064*** 0.053*** 0.063*** 0.053*** 0.063*** 0.053*** (3.57) (2.59) (3.56) (2.58) (3.56) (2.59) (3.56) (2.85) (3.54) (2.84) (3.54) (2.86) Size 0.017 0.055*** 0.016 0.055*** 0.016 0.054*** 0.014 0.046*** 0.014 0.046*** 0.014 0.046*** (1.22) (4.23) (1.19) (4.21) (1.19) (4.18) (1.16) (4.00) (1.11) (3.98) (1.11) (3.95) Lev −0.061 −0.379*** −0.061 −0.373*** −0.061 −0.373*** −0.129** −0.319*** −0.129** −0.315*** −0.129** −0.314*** (−0.86) (−4.55) (−0.86) (−4.48) (−0.87) (−4.48) (−2.11) (−4.39) (−2.10) (−4.34) (−2.11) (−4.33) Roa 0.191 0.065 0.201 0.070 0.200 0.057 0.004 −0.258 0.017 −0.251 0.016 −0.264 (0.87) (0.25) (0.91) (0.27) (0.91) (0.22) (0.02) (−1.18) (0.09) (−1.15) (0.08) (−1.21) MB 0.019*** 0.029*** 0.019*** 0.029*** 0.019*** 0.029*** 0.015*** 0.022*** 0.015*** 0.022*** 0.015*** 0.022*** (4.66) (5.94) (4.63) (5.97) (4.63) (5.95) (4.04) (4.99) (4.00) (5.02) (4.00) (4.99) Opaque 0.120 0.115 0.117 0.122 0.117 0.118 0.091 0.101 0.086 0.108 0.086 0.104 (0.81) (0.65) (0.79) (0.69) (0.79) (0.67) (0.70) (0.64) (0.67) (0.68) (0.67) (0.66) Constant −0.877*** −1.378*** −0.863*** −1.360*** −0.862** −1.337*** −0.791*** −1.141*** −0.772*** −1.126*** −0.769*** −1.110*** (−2.61) (−4.35) (−2.58) (−4.30) (−2.57) (−4.22) (−2.67) (−4.03) (−2.61) (−3.98) (−2.60) (−3.92) IND/YEAR YES YES YES YES YES YES YES YES YES YES YES YES Chow test P = 0.017 P = 0.024 P = 0.023 P = 0.030 P = 0.042 P = 0.091 N 3523 3528 3523 3528 3523 3528 3523 3528 3523 3528 3523 3528 Adj. R 0.199 0.251 0.199 0.252 0.199 0.252 0.244 0.288 0.244 0.289 0.244 0.288 48 X. XU ET AL. Table 10. The substitution effects of analyst coverage. NCSKEW DUVOL t+1 t+1 AN_high AN_low AN_high AN_low AN_high AN_low AN_high AN_low AN_high AN_low AN_high AN_low Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Confu_100 −0.073 −0.275** −0.120 −0.295*** (−0.64) (−2.34) (−1.19) (−2.99) Confu_200 −0.050 −0.113*** −0.068* −0.119*** (−1.18) (−2.63) (−1.77) (−3.30) Confu_300 −0.028 −0.071*** −0.037 −0.074*** (−1.09) (−2.79) (−1.60) (−3.50) Turnover −0.003 −0.026* −0.003 −0.026* −0.003 −0.026* −0.002 −0.015 −0.002 −0.015 −0.002 −0.015 (−0.20) (−1.83) (−0.20) (−1.84) (−0.20) (−1.84) (−0.18) (−1.27) (−0.18) (−1.28) (−0.18) (−1.27) Sigma 2.899*** 4.080*** 2.903*** 4.090*** 2.894*** 4.078*** 3.068*** 4.391*** 3.072*** 4.400*** 3.060*** 4.388*** (4.01) (5.93) (4.02) (5.95) (4.01) (5.93) (4.77) (7.68) (4.77) (7.70) (4.75) (7.68) Ret 14.230*** 13.664*** 14.214*** 13.658*** 14.220*** 13.678*** 13.046*** 10.849*** 13.024*** 10.842*** 13.032*** 10.863*** (8.88) (7.98) (8.87) (7.98) (8.87) (7.99) (9.11) (7.50) (9.09) (7.50) (9.10) (7.51) NCSKEW 0.080*** 0.070*** 0.080*** 0.070*** 0.080*** 0.070*** 0.086*** 0.050*** 0.086*** 0.050*** 0.086*** 0.050*** (5.94) (5.32) (5.93) (5.28) (5.93) (5.29) (7.08) (4.58) (7.07) (4.53) (7.06) (4.54) Size 0.049*** −0.024** 0.049*** −0.024** 0.049*** −0.024** 0.048*** −0.003 0.048*** −0.004 0.048*** −0.004 (5.20) (−2.12) (5.18) (−2.13) (5.18) (−2.14) (5.66) (−0.36) (5.63) (−0.38) (5.64) (−0.39) Lev −0.403*** −0.105** −0.402*** −0.102** −0.402*** −0.102** −0.368*** −0.157*** −0.367*** −0.154*** −0.367*** −0.154*** (−6.41) (−2.31) (−6.40) (−2.24) (−6.40) (−2.24) (−6.57) (−4.02) (−6.56) (−3.94) (−6.56) (−3.95) Roa 0.057 −0.183 0.067 −0.183 0.063 −0.187 −0.115 −0.174 −0.103 −0.174 −0.108 −0.178 (0.25) (−1.11) (0.30) (−1.11) (0.28) (−1.14) (−0.58) (−1.24) (−0.52) (−1.24) (−0.54) (−1.27) MB 0.037*** 0.013*** 0.037*** 0.013*** 0.037*** 0.013*** 0.033*** 0.012*** 0.033*** 0.012*** 0.033*** 0.012*** (8.86) (4.78) (8.85) (4.78) (8.84) (4.79) (8.95) (5.01) (8.95) (5.01) (8.94) (5.01) Opaque −0.005 0.209** −0.006 0.205* −0.006 0.204* −0.030 0.143 −0.031 0.140 −0.031 0.139 (−0.05) (1.97) (−0.06) (1.94) (−0.06) (1.93) (−0.29) (1.60) (−0.30) (1.57) (−0.30) (1.55) Constant −1.318*** 0.333 −1.307*** 0.338 −1.305*** 0.347 −1.320*** −0.155 −1.310*** −0.150 −1.308*** −0.141 (−5.86) (1.29) (−5.81) (1.31) (−5.80) (1.35) (−6.53) (−0.71) (−6.49) (−0.69) (−6.47) (−0.65) IND/YEAR YES YES YES YES YES YES YES YES YES YES YES YES Chow test P = 0.024 P = 0.058 P = 0.035 P = 0.042 P = 0.117 P = 0.078 N 8114 9235 8114 9235 8114 9235 8114 9235 8114 9235 8114 9235 Adj. R 0.264 0.188 0.264 0.189 0.264 0.189 0.317 0.214 0.317 0.214 0.317 0.214 CHINA JOURNAL OF ACCOUNTING STUDIES 49 bear market. We predict that the negative relationship between Confucianism and crash risk may obviously vary between a bear market and a bull market. Following Xu et al. (2012), using the judgment method of the average market return, we divide the sample into two groups: the bear market and the bull market. In particular, we define year 2008, 2010, 2011, 2012, 2013 and 2016 as the bear market and other remaining years as the bull market. Table 11 presents the empirical estima- tions of the sub-sample analysis. The results show that Confucianism is significantly negatively correlated with crash risk in the bear market, but statistically insignificant in the bull market. The findings suggest that the negative impact of Confucianism on crash risk is more pronounced in the bear market. 6. Robustness checks 6.1. Controlling for religious traditions Prior research has showed that religious tradition, as an informal institutional factor, can also mitigate agency conflicts, and improve corporate governance and financial informa- tion quality (Chen et al., 2013; Hilary & Hui, 2009). Recently, based on the data of American firms, Callen and Fang (2015) document a negative and significant association between religion and crash risk. Using the data of Chinese-listed companies, Li and Cai (2016) also find similar evidences. To eliminate the possible interference of religion, we add religion tradition into the empirical model. In particular, following Chen et al. (2013), we use the number of religious sites within a radius of 100, 200 and 300 kilometers around a firm’s registered address (Relig_100, Relig_200, and Relig_300) to measure the atmosphere of religion. The results are reported in Table 12. Columns (1)–(3) and (7)–(9) present the regres- sion results of religion on crash risk without Confucianism. We find that Relig_100 is not significantly associated with crash risk, while Relig_200 and Relig_300 are significantly negatively correlated with NCSKEW and DUVOL. The findings are basically consistent with the results of Callen and Fang (2015) and Li and Cai (2016), implying that religion can reduce crash risk. In Columns (4)–(6) and (10)–(12), we introduce Confucianism and religion in one regression simultaneously. Interestingly, Confucianism is still significantly negatively correlated with crash risk, while the significantly negative relationship between religion and crash risk suddenly disappear. This suggests that Confucian ethics, in the Chinese cultural context, rather than religious tradition is the more important informal institutional factor which affects firm-specific crash risk. 6.2. Controlling for regional factors Confucianism atmosphere may be closely related to the development level of regional economy, law, and marketisation. Hence, it is possible that the negative relationship between Confucianism and crash risk is driven by the related regional factors. In order to exclude this possible alternative explanation, we control for some regional factors, including the regional per capita GDP (Avrage_gdp), the regional economic growth The data of religious sites is obtained from Spatial Religious Analysis System, which is developed by Michigan State University, Purdue University, and Wuhan University. 50 X. XU ET AL. Table 11. The moderating effects of stock market quotation. NCSKEW DUVOL t+1 t+1 Bull Bull Bear Bull Bear Bull Bear Bull Bear Bull Bear Bear Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Confu_100 −0.007 −0.371*** −0.149 −0.317*** (−0.06) (−3.54) (−1.31) (−3.57) Confu_200 −0.018 −0.157*** −0.057 −0.145*** (−0.36) (−4.10) (−1.33) (−4.47) Confu_300 −0.014 −0.093*** −0.031 −0.089*** (−0.47) (−4.06) (−1.22) (−4.59) Turnover 0.021 −0.032*** 0.021 −0.032*** 0.021 −0.032*** 0.033** −0.027*** 0.033** −0.027*** 0.033** −0.027*** (1.19) (−2.61) (1.18) (−2.60) (1.18) (−2.59) (2.15) (−2.69) (2.14) (−2.68) (2.15) (−2.67) Sigma 3.861*** 3.383*** 3.867*** 3.403*** 3.863*** 3.390*** 3.990*** 3.544*** 3.988*** 3.563*** 3.971*** 3.552*** (5.01) (5.20) (5.02) (5.23) (5.01) (5.21) (5.86) (6.51) (5.86) (6.54) (5.83) (6.52) Ret 15.490*** 14.221*** 15.486*** 14.180*** 15.485*** 14.198*** 13.516*** 11.183*** 13.498*** 11.149*** 13.500*** 11.167*** (7.99) (9.74) (7.98) (9.72) (7.98) (9.73) (7.84) (9.03) (7.83) (9.01) (7.83) (9.02) NCSKEW 0.086*** 0.086*** 0.086*** 0.085*** 0.086*** 0.085*** 0.087*** 0.055*** 0.086*** 0.055*** 0.086*** 0.054*** (5.96) (6.95) (5.95) (6.91) (5.94) (6.90) (6.83) (5.34) (6.81) (5.29) (6.81) (5.28) Size −0.051*** 0.064*** −0.051*** 0.064*** −0.051*** 0.064*** −0.050*** 0.062*** −0.050*** 0.062*** −0.050*** 0.062*** (−4.58) (8.09) (−4.59) (8.10) (−4.60) (8.10) (−5.01) (9.05) (−5.01) (9.05) (−5.01) (9.05) Lev −0.153** −0.280*** −0.153** −0.276*** −0.153** −0.276*** −0.179*** −0.279*** −0.177*** −0.276*** −0.177*** −0.275*** (−2.51) (−6.06) (−2.51) (−5.98) (−2.51) (−5.97) (−3.34) (−7.00) (−3.32) (−6.93) (−3.32) (−6.93) Roa −0.028 −0.076 −0.023 −0.068 −0.023 −0.076 −0.293* −0.298** −0.291 −0.288** −0.295* −0.295** (−0.14) (−0.48) (−0.12) (−0.43) (−0.12) (−0.48) (−1.66) (−2.21) (−1.64) (−2.14) (−1.67) (−2.19) MB 0.020*** 0.025*** 0.020*** 0.025*** 0.020*** 0.025*** 0.019*** 0.021*** 0.019*** 0.021*** 0.019*** 0.021*** (4.85) (9.97) (4.85) (9.98) (4.84) (9.97) (5.18) (9.20) (5.19) (9.20) (5.19) (9.18) Opaque 0.033 0.256*** 0.032 0.254*** 0.031 0.254*** −0.011 0.155* −0.012 0.153* −0.012 0.153* (0.24) (2.75) (0.23) (2.73) (0.23) (2.72) (−0.09) (1.90) (−0.10) (1.87) (−0.10) (1.87) Constant 0.405 −1.712*** 0.409 −1.708*** 0.413 −1.700*** 0.367 −1.656*** 0.369* −1.649*** 0.373* −1.640*** (1.60) (−8.91) (1.62) (−8.90) (1.63) (−8.86) (1.64) (−9.96) (1.65) (−9.93) (1.67) (−9.87) IND/YEAR YES YES YES YES YES YES YES YES YES YES YES YES Chow test P = 0.018 P = 0.020 P = 0.010 P = 0.118 P = 0.054 P = 0.044 N 7366 9983 7366 9983 7366 9983 7366 9983 7366 9983 7366 9983 Adj. R 0.252 0.194 0.252 0.194 0.252 0.194 0.312 0.212 0.312 0.212 0.312 0.212 CHINA JOURNAL OF ACCOUNTING STUDIES 51 Table 12. Controlling for religious traditions. NCSKEW DUVOL t+1 t+1 Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Relig_100 0.012 0.025 −0.007 0.007 (0.65) (1.31) (−0.42) (0.42) Relig_200 −0.012* −0.001 −0.015*** −0.004 (−1.89) (−0.16) (−2.81) (−0.66) Relig_300 −0.008** −0.000 −0.009*** −0.002 (−2.23) (−0.05) (−3.12) (−0.46) Confu_100 −0.240*** −0.257*** (−2.85) (−3.53) Confu_200 −0.093*** −0.096*** (−2.58) (−3.10) Confu_300 −0.057** −0.056*** (−2.24) (−2.58) Turnover −0.014 −0.014 −0.014 −0.014 −0.014 −0.014 −0.007 −0.007 −0.007 −0.007 −0.007 −0.007 (−1.37) (−1.39) (−1.40) (−1.40) (−1.41) (−1.40) (−0.78) (−0.80) (−0.82) (−0.82) (−0.84) (−0.82) Sigma 3.643*** 3.637*** 3.647*** 3.669*** 3.666*** 3.650*** 3.914*** 3.912*** 3.925*** 3.942*** 3.942*** 3.929*** (7.38) (7.37) (7.39) (7.44) (7.43) (7.40) (9.20) (9.20) (9.23) (9.27) (9.27) (9.24) Ret 14.815*** 14.864*** 14.847*** 14.816*** 14.829*** 14.836*** 12.266*** 12.297*** 12.274*** 12.268*** 12.261*** 12.263*** (12.76) (12.81) (12.79) (12.77) (12.78) (12.79) (12.16) (12.20) (12.18) (12.17) (12.16) (12.17) NCSKEW 0.090*** 0.089*** 0.089*** 0.089*** 0.089*** 0.089*** 0.075*** 0.074*** 0.074*** 0.074*** 0.074*** 0.074*** (9.56) (9.54) (9.52) (9.54) (9.51) (9.50) (9.26) (9.22) (9.20) (9.23) (9.18) (9.17) Size 0.019*** 0.018*** 0.018*** 0.019*** 0.018*** 0.018*** 0.018*** 0.017*** 0.017*** 0.017*** 0.017*** 0.017*** (2.96) (2.78) (2.77) (2.90) (2.79) (2.79) (3.09) (2.92) (2.92) (3.01) (2.93) (2.94) Lev −0.247*** −0.245*** −0.246*** −0.250*** −0.247*** −0.247*** −0.254*** −0.253*** −0.254*** −0.258*** −0.255*** −0.255*** (−6.72) (−6.69) (−6.70) (−6.83) (−6.74) (−6.74) (−7.99) (−7.97) (−7.99) (−8.12) (−8.03) (−8.03) Roa −0.070 −0.056 −0.052 −0.051 −0.041 −0.046 −0.300*** −0.288*** −0.284*** −0.280*** −0.272** −0.278*** (−0.57) (−0.46) (−0.42) (−0.41) (−0.34) (−0.38) (−2.81) (−2.69) (−2.65) (−2.62) (−2.55) (−2.60) MB 0.024*** 0.024*** 0.024*** 0.024*** 0.024*** 0.024*** 0.021*** 0.021*** 0.021*** 0.021*** 0.021*** 0.021*** (11.22) (11.08) (11.07) (11.16) (11.10) (11.10) (10.73) (10.60) (10.60) (10.67) (10.62) (10.63) Opaque 0.174** 0.168** 0.168** 0.173** 0.169** 0.168** 0.098 0.093 0.093 0.097 0.093 0.093 (2.25) (2.18) (2.18) (2.24) (2.18) (2.17) (1.45) (1.37) (1.38) (1.43) (1.38) (1.38) Constant −0.614*** −0.580*** −0.578*** −0.597*** −0.574*** −0.568*** −0.622*** −0.596*** −0.595*** −0.604*** −0.589*** −0.585*** (−3.91) (−3.70) (−3.69) (−3.80) (−3.66) (−3.63) (−4.54) (−4.35) (−4.34) (−4.41) (−4.30) (−4.28) IND/YEAR YES YES YES YES YES YES YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 17,349 Adj. R 0.212 0.212 0.212 0.212 0.212 0.212 0.249 0.250 0.250 0.250 0.250 0.250 52 X. XU ET AL. Table 13. Controlling for related regional factors. NCSKEW DUVOL t+1 t+1 Variables (1) (2) (3) (4) (5) (6) Confu_100 −0.174* −0.174** (−1.78) (−2.05) Confu_200 −0.083** −0.083*** (−2.26) (−2.62) Confu_300 −0.054** −0.051*** (−2.53) (−2.79) Avrage_gdp −0.017* −0.015 −0.015 −0.008 −0.008 −0.008 (−1.69) (−1.48) (−1.48) (−0.92) (−0.92) (−0.91) gdp_growth 3.724*** 3.729*** 3.719*** 4.002*** 4.004*** 3.994*** (7.65) (7.55) (7.53) (9.39) (9.39) (9.37) Law 15.247*** 14.787*** 14.795*** 12.289*** 12.258*** 12.268*** (13.20) (12.74) (12.75) (12.18) (12.15) (12.16) Market 0.096*** 0.089*** 0.089*** 0.074*** 0.074*** 0.074*** (10.26) (9.47) (9.46) (9.24) (9.20) (9.18) Turnover 0.009 0.021*** 0.020*** 0.018*** 0.018*** 0.018*** (1.40) (3.12) (3.11) (3.12) (3.12) (3.12) Sigma −0.242*** −0.259*** −0.259*** −0.264*** −0.262*** −0.262*** (−6.83) (−7.00) (−7.01) (−8.25) (−8.19) (−8.20) Ret −0.037 −0.042 −0.045 −0.273** −0.270** −0.273** (−0.31) (−0.34) (−0.37) (−2.55) (−2.52) (−2.55) NCSKEW 0.023*** 0.024*** 0.024*** 0.021*** 0.021*** 0.021*** (10.85) (11.14) (11.12) (10.64) (10.65) (10.64) Size 0.171** 0.170** 0.169** 0.095 0.094 0.093 (2.24) (2.20) (2.18) (1.41) (1.39) (1.38) Lev −0.005 −0.004 −0.004 0.002 0.000 0.001 (−1.01) (−0.82) (−0.82) (0.36) (0.11) (0.17) Roa 0.558 0.673* 0.754* 0.100 0.136 0.214 (1.38) (1.66) (1.86) (0.28) (0.39) (0.61) MB −0.003 −0.003 −0.003 −0.006** −0.006* −0.006** (−0.78) (−0.83) (−0.97) (−2.21) (−1.92) (−2.11) Opaque 0.005 0.007 0.009 0.008 0.008 0.009 (0.57) (0.85) (1.00) (0.98) (1.04) (1.17) Constant −0.534*** −0.749*** −0.760*** −0.669*** −0.671*** −0.682*** (−3.22) (−4.28) (−4.34) (−4.36) (−4.38) (−4.45) IND/YEAR YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 Adj. R 0.209 0.212 0.212 0.250 0.250 0.250 rate (gdp_growth), the regional legal environment (Law), and the level of regional marketisation (Market). Table 13 reports the regression results. After controlling for related regional factors, we find that the significantly negative relationship between Confucianism and crash risk still holds. This further supports our conclusions. 6.3. Potential self-selection issue To a certain extent, the measurement method of Confucianism in this study depends on the firm’s registration address. Some studies have pointed out that the choice of firms’ registration address may be exogenous, and is mainly affected by labor costs, taxation, transportation and other factors (Du, 2015). However, there may also be the possibility that firms with better corporate governance prefer to choose regions with strong Confucianism atmosphere as registration sites, which leads to sample self-selection The data of Law and Market comes from Marketisation Index of China’s Provinces: NERI Report 2016. The value of missing years is supplemented by median insertion method. CHINA JOURNAL OF ACCOUNTING STUDIES 53 Table 14. Propensity score matching regression results. NCSKEW DUVOL t+1 t+1 Variables (1) (2) (3) (4) (5) (6) Confu_100 −0.267* −0.344*** (−1.93) (−2.89) Confu_200 −0.157*** −0.153*** (−2.82) (−3.22) Confu_300 −0.081*** −0.070*** (−2.59) (−2.62) Turnover −0.019 −0.013 −0.037* −0.014 −0.014 −0.038** (−1.09) (−0.64) (−1.80) (−0.93) (−0.82) (−2.23) Sigma 4.036*** 3.653*** 4.956*** 4.134*** 4.321*** 5.080*** (4.49) (3.67) (4.95) (5.46) (5.11) (6.04) Ret 13.444*** 11.330*** 10.731*** 10.647*** 10.361*** 10.553*** (6.47) (4.83) (4.54) (5.99) (5.15) (5.22) NCSKEW 0.095*** 0.089*** 0.078*** 0.071*** 0.075*** 0.068*** (5.64) (4.76) (4.13) (5.00) (4.71) (4.27) Size 0.008 0.008 0.005 0.011 0.016 0.012 (0.68) (0.67) (0.41) (1.10) (1.50) (1.15) Lev −0.239*** −0.326*** −0.362*** −0.311*** −0.331*** −0.347*** (−3.74) (−4.59) (−5.16) (−5.63) (−5.45) (−5.74) Roa 0.240 0.196 −0.139 0.065 −0.034 −0.296 (1.11) (0.85) (−0.61) (0.35) (−0.17) (−1.52) MB 0.026*** 0.024*** 0.022*** 0.022*** 0.021*** 0.017*** (6.45) (5.77) (5.53) (6.35) (5.86) (5.05) Opaque 0.244* 0.300** 0.290** 0.214* 0.231* 0.219* (1.80) (1.98) (1.99) (1.86) (1.74) (1.73) Constant −0.299 −0.402 −0.327 −0.419* −0.659** −0.538** (−1.08) (−1.33) (−1.10) (−1.79) (−2.55) (−2.11) IND/YEAR YES YES YES YES YES YES N 5466 4518 4362 5466 4518 4362 Adj. R 0.223 0.216 0.211 0.261 0.243 0.245 issue. Therefore, we utilise a propensity score matching approach (PSM) to alleviate the potential self-selection issue. First, according to the influence of Confucianism, we divided the samples into three groups. We regard the strongest group as the treatment group, and choose the weakest group as the control group. Second, for each treatment sample, we find a control sample with similar characteristics. Specifically, the character- istics include firm-level and regional factors, such as Size, Lev, MB, Dual, Board, Indep, and Top1, Avrage_gdp, gdp_growth, Law, Market, Industry and Year. The results (not tabu- lated) show that the above corporate governance and regional factors no longer have significant differences between the treatment group and the control group after PSM. Finally, we re-estimate the main regression model using the propensity score-matched sample. The results are presented in Table 14. We document that the previous findings are still robust. 6.4. Further test based on the birthplace of CEO and chairman The research in the fields of sociology, psychology and behavioral science shows that childhood is the formation stage of individual thinking mode and values, and individual early experience will directly affect their cognition and behavior preferences after Moreover, we divided the samples into two groups. The empirical results remain unchanged. To ensure a more accurate matching, we require the maximum difference between the propensity scores of the two groups does not exceed 0.01 in absolute value. 54 X. XU ET AL. Table 15. Based on the birthplace of CEO and Chairman. OLS DID NCSKEW DUVOL NCSKEW DUVOL t+1 t+1 t+1 t+1 Variables (1) (2) (3) (4) Confu_Dum −0.061** −0.044* (−2.17) (−1.83) Treat 0.081 0.061 (1.59) (1.45) Post 0.051*** 0.042** (2.64) (2.54) Post ×Treat −0.187** −0.122* t t (−2.39) (−1.81) Turnover −0.013 −0.014 −0.011 −0.011 (−0.92) (−1.23) (−0.70) (−0.89) Sigma 3.740*** 3.565*** 3.427*** 3.275*** (5.44) (6.01) (4.55) (5.06) Ret 15.249*** 13.576*** 14.989*** 13.388*** (9.88) (10.28) (8.98) (9.36) NCSKEW 0.086*** 0.067*** 0.082*** 0.064*** (6.78) (6.22) (5.95) (5.41) Size 0.041*** 0.044*** 0.046*** 0.048*** (4.90) (6.13) (5.12) (6.28) Lev −0.267*** −0.278*** −0.263*** −0.268*** (−5.39) (−6.50) (−4.87) (−5.80) Roa 0.086 −0.189 0.169 −0.141 (0.51) (−1.28) (0.90) (−0.86) MB 0.024*** 0.020*** 0.025*** 0.022*** (7.48) (7.17) (7.09) (7.19) Opaque 0.144 0.095 0.155 0.111 (1.39) (1.05) (1.40) (1.15) Constant −1.054*** −1.134*** −1.205*** −1.260*** (−5.24) (−6.54) (−5.63) (−6.83) IND/YEAR YES YES YES YES N 9692 9692 8355 8355 Adj. R 0.204 0.238 0.207 0.241 adulthood (Bernile, Bhagwat, & Rau, 2017). As the key decision-maker of a firm, CEO and chairman have a greater voice in the process of corporate decision-making, and thus their personal willingness and value orientation are more likely to be reflected in corporate decisions. Hence, we further use the Confucianism atmosphere in the birth- place of CEO and chairman as an alternative measure. In particular, we create a dummy variable, Confu_Dum, which equal one if a firm’s CEO or chairman were born in the regions with stronger Confucianism atmosphere and zero otherwise. Columns (1) and (2) in Table 15 present the results. The coefficients of Confu_Dum are significantly negatively correlated with crash risk. The findings are completely consistent with the previous conclusions. Furthermore, we also employ a difference-in-differences method to mitigate the potential endogenous issue by tracing the transitions of a firm’s CEO or chairman. Specifically, a firm will be treated as treatment sample when the CEO or chairman born in the region with weaker Confucianism atmosphere are replaced by CEO or chairman born in the region with stronger Confucianism atmosphere (Treat = 1). In We use the number of Confucian academies in a region to measure the influence of Confucianism. We divide the sample into two groups. A sample will be classified as regions with stronger (weaker) Confucianism atmosphere if the value of Confu_100 (Confu_200 and Confu_300) is higher (lower) than the median value of all regions. CHINA JOURNAL OF ACCOUNTING STUDIES 55 contrast, we consider a firm whose CEO or chairman has always been from the region with stronger Confucianism atmosphere as control sample (Treat = 0). Meanwhile, we also construct a dummy variable, Post, which equal one if a year is after CEO or chairman transition and zero otherwise. The estimation results of the difference-in-differences test are reported in columns (3) and (4) of Table 15.We find that the coefficients of Confu_Dum are significantly negatively correlated with crash risk, which further supports our conclusions. 6.5. Alternative measures of crash risk Following Hutton et al. (2009) and Kim et al. (2011a), we utilise more stringent measure- ment standard as proxy for crash risk. We create an indicator variable, CRASH_DUM,to measure the crash likelihood for each firm in each year. The calculation method is as follows: CRASH DUM ¼ 1 9t; W AverageðW Þ 3:09σ (14) i;t i;t i;t where W is the firm-specific weekly returns for firm i in week t; Average (W ) is the i,t i,t mean of firm-specific weekly returns; σ is the standard deviation of firm-specific weekly i,t returns. We call it crash week if the firm-specific weekly returns for firm i in week t satisfies Equation (12). CRASH_DUM equal one if a firm experiences one or more crash weeks during the fiscal-year period, and zero otherwise. Moreover, following Piotroski et al. (2015), we also use the frequency that the firm experiences material negative weekly stock returns as the alternative measure of crash risk (FREQUENCY). A large negative stock price drop is defined as a negative weekly excess return more than 2 standard deviations below the mean of firm-specific weekly returns. Firms with a higher level of FREQUENCY are interpreted as being more crash prone. A Logit and Tobit regression are used in the estimation when the dependent variables are CRASH_DUM and FREQUENCY, respectively. Table 16 reports the regression results, which is basically consistent with the previous findings. 6.6. Other robustness checks We also conduct the following robustness tests: (1) Following Kim et al. (2011a), the coefficient estimates are based on robust standard errors corrected for firm and year clustering; (2) Following Xu et al. (2014), we further control for corporate governance factors in the main model, including Dual, Board, Indep, and Top1; (3) Excluding the Observations of abnormal volatility years in stock markets such as 2008, 2015, and 2016. In these years, the sharp rise and fall of stock prices may be more likely to be caused by market factors than individual factors; (4) Eliminating the observations of companies that have experienced registration addresses changes. (5) The time span from Tang Dynasty to Qing Dynasty (618–1912 A.D.) is so long. In order to avoid the potential influence of historical and dynastic changes on culture, we recalculate the proxy variables of Confucianism only by using Confucian academies in Qing Dynasty. (6) Adopting the number of Confucian academies within a radius of 100, 200 and 300 kilometers around the office’s address of companies to measure the influence of Confucianism. (7) 56 X. XU ET AL. Table 16. Alternative measures of crash risk. CRASH_DUM FREQUENCY t+1 t+1 Variables (1) (2) (3) (4) (5) (6) Confu_100 −0.959** −0.290 (−2.53) (−1.20) Confu_200 −0.290** −0.184** (−2.05) (−2.06) Confu_300 −0.123 −0.142*** (−1.48) (−2.68) Turnover 0.015 0.016 0.016 −0.030 −0.030 −0.030 (0.35) (0.35) (0.37) (−1.03) (−1.04) (−1.04) Sigma −6.869*** −6.895*** −6.961*** 1.475 1.498 1.472 (−3.07) (−3.09) (−3.12) (0.99) (1.00) (0.99) Ret −0.080 −0.165 −0.126 14.599*** 14.559*** 14.583*** (−0.02) (−0.03) (−0.02) (4.19) (4.18) (4.19) NCSKEW −0.009 −0.010 −0.010 0.145*** 0.144*** 0.144*** (−0.21) (−0.23) (−0.22) (5.30) (5.27) (5.25) Size −0.101*** −0.101*** −0.100*** −0.086*** −0.086*** −0.087*** (−3.54) (−3.52) (−3.51) (−4.55) (−4.58) (−4.61) Lev −0.163 −0.152 −0.150 −0.190* −0.188* −0.188* (−1.06) (−0.99) (−0.97) (−1.76) (−1.74) (−1.75) Roa 0.346 0.340 0.310 −0.100 −0.076 −0.074 (0.65) (0.64) (0.58) (−0.28) (−0.21) (−0.21) MB 0.029*** 0.029*** 0.029*** 0.030*** 0.030*** 0.030*** (3.57) (3.62) (3.64) (4.64) (4.63) (4.62) Opaque 0.435 0.435 0.437 −0.016 −0.019 −0.023 (1.30) (1.30) (1.31) (−0.07) (−0.09) (−0.10) Constant 0.912 0.897 0.889 3.640*** 3.664*** 3.698*** (1.34) (1.32) (1.31) (8.03) (8.09) (8.16) IND/YEAR YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 Pseudo R 0.039 0.039 0.039 0.016 0.016 0.016 Employing the number of Confucian temples within a radius of 100, 200 and 300 kilometers around a firm’s registered address (Temple_100, Temple_200, and Temple_300) to measure the influence of Confucianism. Table 17 present the regression results of the above robustness tests. Overall, these results provide further evidence that Confucianism helps to reduce firms’ stock price crash risk. 7. Conclusions Stock price crash risk has become a hot research topic of economic and financial fields in the post-financial crisis era. Different from previous studies focused on firm-level char- acteristics and formal institutions, this study systematically examines the impact of Confucian culture on firm-specific crash risk and its underlying mechanism from the perspective of informal institutions. We document a negative and significant relationship between Confucianism and crash risk, suggesting that Confucian culture can effectively reduce firms’ future crash risk. Further channel analyses show that Confucianism curbs crash risk mainly through mitigating agency conflict, improving financial information quality and reducing managerial overconfidence. Moreover, we also find that the negative effect of Confucianism on crash risk is more prominent in firms with weaker monitoring mechanisms, such as poorer corporate governance and lower analyst cover- age. This indicates that Confucian ethics, as an informal institutional factor, helps to make up for the deficiency of formal institution in emerging capital markets and play an CHINA JOURNAL OF ACCOUNTING STUDIES 57 Table 17. Other robustness checks. NCSKEW DUVOL t+1 t+1 Variables (1) (2) (3) (4) (5) (6) Panel A: Based on robust standard errors corrected for firm and year clustering Confu_100 −0.216 −0.250* (−1.45) (−1.70) Confu_200 −0.096* −0.107** (−1.96) (−2.21) Confu_300 −0.057** −0.063*** (−2.43) (−2.67) Control variables YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 Adj. R 0.212 0.212 0.212 0.250 0.250 0.250 Panel B: Controlling for corporate governance factors Confu_100 −0.209** −0.243*** (−2.53) (−3.40) Confu_200 −0.091*** −0.103*** (−2.99) (−3.92) Confu_300 −0.054*** −0.060*** (−2.98) (−3.85) Control variables YES YES YES YES YES YES N 17,256 17,256 17,256 17,256 17,256 17,256 Adj. R 0.212 0.212 0.212 0.250 0.250 0.250 Panel C: Excluding the Observations of abnormal volatility years Confu_100 −0.146 −0.235** (−1.30) (−2.32) Confu_200 −0.076* −0.107*** (−1.87) (−2.91) Confu_300 −0.051** −0.067*** (−2.07) (−3.04) Control variables YES YES YES YES YES YES N 11,192 11,192 11,192 11,192 11,192 11,192 Adj. R 0.088 0.088 0.088 0.097 0.097 0.097 Panel D: Eliminating companies that have experienced registration addresses changes Confu_100 −0.239*** −0.266*** (−2.86) (−3.69) Confu_200 −0.105*** −0.114*** (−3.37) (−4.26) Confu_300 −0.063*** −0.068*** (−3.43) (−4.27) Control variables YES YES YES YES YES YES N 16,633 16,633 16,633 16,633 16,633 16,633 Adj. R 0.213 0.213 0.213 0.251 0.251 0.251 Panel E: Using Confucian academies in Qing Dynasty to measure the influence of Confucianism QConfu_100 −0.330** −0.350*** (−2.41) (−2.96) QConfu_200 −0.212*** −0.230*** (−3.39) (−4.24) QConfu_300 −0.150*** −0.158*** (−3.63) (−4.43) Control variables YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 Adj. R 0.212 0.212 0.212 0.250 0.250 0.250 Panel F: Adopting the office’s address of companies to recalculate the influence of Confucianism OConfu_100 −0.203** −0.239*** (−2.48) (−3.38) OConfu_200 −0.088*** −0.106*** (−2.89) (−4.03) OConfu_300 −0.051*** −0.061*** (−2.82) (−3.88) Control variables YES YES YES YES YES YES (Continued) 58 X. XU ET AL. Table 17. (Continued). NCSKEW DUVOL t+1 t+1 Variables (1) (2) (3) (4) (5) (6) N 17,348 17,348 17,348 17,348 17,348 17,348 Adj. R 0.212 0.212 0.212 0.250 0.250 0.250 Panel G: Employing the number of temples to measure the influence of Confucianism Temple_100 −4.108** −3.751*** (−2.47) (−2.59) Temple_200 −2.232*** −1.809** (−2.74) (−2.58) Temple_300 −0.866* −0.776* (−1.77) (−1.85) Control variables YES YES YES YES YES YES N 17,349 17,349 17,349 17,349 17,349 17,349 Adj. R 0.212 0.212 0.212 0.250 0.250 0.249 alternative governance function. Especially, after controlling for Confucianism and reli- gion simultaneously, Confucianism is still significantly negatively correlated with crash risk while the significantly negative relationship between religion and crash risk sud- denly disappear. This indicates that Confucian ethics, in the Chinese cultural context, rather than religious tradition is the more crucial informal institutional factor reducing firm-specific crash risk. Our study has important theoretical values and practical significances. In theory, this paper empirically investigates the impact of Confucianism on crash risk from the perspective of informal institutions. This not only enriches the literature on stock price crash risk but also deepens theoretical cognition of the positive value of Confucian culture from firm-level. In practice, our conclusions suggest that Confucian ethics, as an implicit restraint mechanism, can prevent the likelihood of stock price crash and pro- mote the healthy development of capital market, and thus help to make up for the imperfect formal system of emerging capital market. This provides theoretical basis and policy reference for giving full play to the era value and governance function of the excellent Chinese traditional culture in constructing modern commercial civilisation. Therefore, we should further enhance cultural self-confidence and attach great impor- tance to the inheritance and development of Chinese excellent traditional culture. Meanwhile, we also should continue to learn and obtain the nourishment from the excellent Chinese traditional culture, and seek more solutions to solve practical pro- blems from them. Acknowledgments We would like to thank Kangtao Ye (the editor), Hanwen Chen (the editor), and anonymous reviewers for their valuable comments and suggestions. This work was supported by the National Natural Science Foundation of China [Nos.71572019 and 71802169]; Graduate Research and Innovation Foundation of Chongqing [No. CYB17031]; Philosophy and Social Sciences of Ministry of Education of China [No. 18JHQ079]; Humanities and Social Sciences of Ministry of Education of China [No. 17YJA630017]; Fundamental Research Funds for the Central Universities [No. 2019 CDJSK 02 XK 11]. CHINA JOURNAL OF ACCOUNTING STUDIES 59 Disclosure statement No potential conflict of interest was reported by the authors. ORCID Wanli Li http://orcid.org/0000-0003-2081-0257 References Allen, F., Qian, J., & Qian, M. (2005). Law, finance, and economic growth in China. Journal of Financial Economics, 77(1), 57–116. An, H., & Zhang, T. (2013). Stock price synchronicity, crash risk, and institutional investors. Journal of Corporate Finance, 21(1), 1–15. An, Z., Chen, Z., Li, D., & Lu, X. (2018). Individualism and stock price crash risk. Journal of International Business Studies, 49(9), 1208–1236. Ball, R. (2009). Market and political/regulatory perspectives on the recent accounting scandals. Journal of Accounting Research, 47(2), 277–323. Bernile, G., Bhagwat, V., & Rau, P.R. (2017). What doesn’t kill you will only make you more risk- loving: Early-life disasters and CEO behavior. The Journal of Finance, 72(1), 167–206. Callen, J.L., & Fang, X. (2015). Religion and stock price crash risk. Journal of Financial & Quantitative Analysis, 50(1–2), 169–195. Chen, D.F., Hu, X.L., Liang, S.K., & Xin, F. (2013). Religious tradition and corporate governance. Economic Research Journal,71–84. (in Chinese). Cheng, B., Pan, F., & Wang, J.L. (2016). Confucian culture information environment and internal controls. Accounting Research,79–84. (in Chinese). Chu, J., & Fang, J.X. (2016). Margin-trading, short-selling and the deterioration of crash risk. Economic Research Journal, 143–158. (in Chinese). Du, W.M. (2003). Confucian ethics and East Asian enterprise spirit. Beijing: Zhong Hua Book Company. Du, W.M. (2012). Knowing confucianism. Zhejiang: Zhejiang University Press. Du, X. (2015). Does Confucianism reduce minority shareholder expropriation? Evidence from China. Journal of Business Ethics, 132(4), 661–716. Fu, P.P., & Tsui, A.S. (2003). Utilizing printed media to understand desired leadership attributes in the People’s Republic of China. Asia Pacific Journal of Management, 20(4), 423–446. Gu, Z.H. (2015). Confucian ethics and agency cost in the context of globalization. Management World, 113–123. (in Chinese). Guiso, L., Sapienza, P., & Zingales, L. (2008). Social capital as good culture. Journal of the European Economic Association, 6(2), 295–320. Hambrick, D.C. (2007). Upper echelons theory: An update. Academy of Management Review, 32(2), 334–343. Hilary, G., & Hui, K.W. (2009). Does religion matter in corporate decision making in America? Journal of Financial Economics, 93(3), 455–473. Huntington, S., & Harrison, L.E. (2013). The important role of culture: How values affect human progress. Beijing: Xinhua Publishing House. Hutton, A.P., Marcus, A.J., & Tehranian, H. (2009). Opaque financial reports, R , and crash risk. Journal of Financial Economics, 94(1), 67–86. Ip, P.K. (2009). Is Confucianism good for business ethics in China? Journal of Business Ethics, 88(3), 463–476. Ji, X.F. (1996). Chinese academy dictionary. Zhejiang: Zhejiang Education Publishing House. Jiang, D.Q., Yao, Z.Y., & Chen, D.H. (2018). CFO’s status and stock price crash risk. Management World, 153–166. (in Chinese). 60 X. XU ET AL. Jiang, F.X., Cai, X.N., & Zhu, B. (2018). Multiple blockholders and stock price crash risk. Accounting Research,68–74. (in Chinese). Jiang, F.X., Zhang, M., Lu, Z.F., & Chen, C.D. (2009). Managerial overconfidence, firm expansion and financial distress. Economic Research Journal, 131–143. (in Chinese). Jiang, X.Y. (2013). Tax enforcement, tax aggressiveness and stock price crash risk. Nankai Business Review, 152–160. (in Chinese). Jiang, X.Y., & Xu, N.H. (2015). Corporate over-investment and stock price crash risk. Journal of Financial Research, 141–158. (in Chinese). Jin, L., & Myers, S.C. (2006). R around the world: New theory and new tests. Journal of Financial Economics, 79(2), 257–292. Jin, Z., Xu, H., & Ma, Y.Q. (2017). Confucianism and corporate risk taking. The Journal of World Economy, 170–192. (in Chinese). Khan, M., & Watts, R.L. (2009). Estimation and empirical properties of a firm-year measure of accounting conservatism. Journal of Accounting & Economics, 48(2), 132–150. Kim, J.B., Li, Y., & Zhang, L. (2011a). CFOs versus CEOs: Equity incentives and crashes. Journal of Financial Economics, 101(3), 713–730. Kim, J.B., Li, Y., & Zhang, L. (2011b). Corporate tax avoidance and stock price crash risk: Firm-level analysis. Journal of Financial Economics, 100(3), 639–662. Kim, J.B., Wang, Z., & Zhang, L. (2016). CEO overconfidence and stock price crash risk. Contemporary Accounting Research, 33(4), 1720–1749. Kim, J.B., & Zhang, L. (2014). Financial reporting opacity and expected crash risk: Evidence from implied volatility smirks. Contemporary Accounting Research, 31(3), 851–875. Kim, J.B., & Zhang, L. (2016). Accounting conservatism and stock price crash risk: Firm-level evidence. Contemporary Accounting Research, 33(1), 412–441. Kothari, S.P., Shu, S., & Wysocki, P.D. (2009). Do managers withhold bad news? Journal of Accounting Research, 47(1), 241–276. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R.W. (1998). Law and finance. Journal of Political Economy, 106(6), 1113–1155. Li, J.B., & Nie, H.H. (2011). Confucian filial piety, economic growth and divergence among civiliza- tions. Social Sciences in China,41–55. (in Chinese). Li, K., Griffin, D., Yue, H., & Zhao, L. (2013). How does culture influence corporate risk-taking? Journal of Corporate Finance, 23(4), 1–22. Li, W.F., & Cai, G.L. (2016). Religion and stock price crash risk: Evidence from China. China Journal of Accounting Research, 9(3), 235–250. Li, X., Wang, S.S., & Wang, X. (2017). Trust and stock price crash risk: Evidence from China. Journal of Banking & Finance, 76,74–91. Li, X.R., & Liu, H. (2012). CEO vs CFO: Gender and stock price crash risk. The Journal of World Economy, 102–129. (in Chinese). Liang, Q.X., & Zeng, H.J. (2016). Reform of independent director system, independence of inde- pendent director and stock price crash risk. Management World, 144–159. (in Chinese). Lin, L., & Zheng, D.J. (2016). Delisting regulation and stock price crash risk. China Industrial Economics,58–74. (in Chinese). Lu, R., & Xu, L.B. (2004). The asymmetry information effect on bull and bear stock markets. Economic Research Journal,65–72. (in Chinese). Luo, J.H., & Du, X.Q. (2014). Media coverage, institutional environment and stock price crash risk. Accounting Research,53–59. (in Chinese). North, D.C. (1990). Institutions, institutional change and economic performance. Cambridge: Cambridge University Press. Pan, A.L., Li, B., Lin, Y.N., & Su, W.N. (2012). Culture’sinfluence on accounting: Literature review and future research prospects. Accounting Research,20–27. (in Chinese). Pan, Y., Dai, Y.Y., & Lin, C.Q. (2011). Information opacity, analyst coverage and individual stock Crash Risk. Journal of Financial Research, 138–151. (in Chinese). Piotroski, J., & Wong, T.J. (2010). Institutions and information environment of Chinese listed firms (Working Paper). Stanford University.. CHINA JOURNAL OF ACCOUNTING STUDIES 61 Piotroski, J.D., Wong, T.J., & Zhang, T. (2015). Political incentives to suppress negative information: Evidence from Chinese listed firms. Journal of Accounting Research, 53(2), 405–459. Quan, X.F., Wu, S.N., & Wen, F. (2010). Managerial power, private income and compensation rigging. Economic Research Journal,73–87. (in Chinese). Richardson, S. (2006). Over-investment of free cash flow. Review of Accounting Studies, 11(2), 159– Song, X.Z., Hu, J., & Li, S.H. (2017). Corporate social responsibility disclosure and stock price crash risk: Based on information effect and reputation insurance effect. Journal of Financial Research, 161–175. (in Chinese). Sun, S.W., Liang, S.K., Ruan, G.M., & Fu, Y.X. (2017). Management selling, suppression of information and stock price crash risk. Journal of Financial Research, 175–190. (in Chinese). Tian, L.H., & Wang, D.K. (2017). The “cover-up effect” of social responsibility disclosure and stock price crash risk of listed companies: A DID-PSM analysis from China’s stock market. Management World, 146–157. (in Chinese). Wang, H.C., Cao, F., Gao, S.H., & Li, Z.G. (2014). Investor protection and stock price crash risk. Finance & Trade Economics,73–82. (in Chinese). Wang, H.C., Cao, F., & Ye, K.T. (2015). Supervision or tunneling: The proportion of shares held by the major shareholders and stock price crash risk. Management World,45–57. (in Chinese). Xu, N.H., Jiang, X.Y., Yi, Z.H., & Xu, X.Z. (2012). Conflicts of interest, analyst optimism and stock price crash risk. Economic Research Journal, 127–140. (in Chinese). Xu, N.H., Li, X., Yuan, Q., & Chan, K.C. (2014). Excess perks and stock price crash risk: Evidence from China. Journal of Corporate Finance, 25(2), 419–434. Xu, N.H., Yu, S.R., & Yi, Z.H. (2013). The herding behavior of institutional investors and stock price crash risk. Management World,31–43. (in Chinese). Ye, K.T., Cao, F., & Wang, H.C. (2015). Can internal control information disclosure reduce stock price crash risk. Journal of Financial Research, 192–206. (in Chinese). Zhan, M.S., Xu, H., & Kang, J. (2017). The dilemma and breakthrough of research on how cultures influence accounting information. Accounting Research,25–30. (in Chinese). Zhang, W.Y. (2013). Norm entrepreneurs and Confucian social norms. Journal of Peking University (Philosophy and Social Sciences),16–35. (in Chinese). Zhao, J., Huang, J.C., & Liu, F. (2018). The opening of the high-speed rail and Stock Price Crash Risk. Management World, 157–168. (in Chinese). Zhu, S.J. (1958). Synthesize of Chinese local records. Beijing: Commercial Press.
Journal
China Journal of Accounting Studies
– Taylor & Francis
Published: Jan 2, 2019
Keywords: Confucianism; informal institution; stock price crash risk
