Do Women Top Managers Always Mitigate Earnings Management? Evidence from China

Abstract

This study investigates the non-linear association between women top managers (WTM) and earnings management and further examines the moderating effect of monitoring intensity. Drawing on leadership style theory and using a sample of Chinese listed firms over the period of 2001–2011, our findings reveal the inverted-U relationship between WTM and earnings management, suggesting that different proportions of WTM can shape distinct leadership styles and processes of decision-making, and thus bring out asymmetric earnings qualities. Moreover, the monitoring intensity attenuates the inverted-U relationship between WTM and earnings management. These findings are robust to a variety of sensitivity tests and our conclusions are still valid after controlling for endogeneity. As additional tests, we find that the inverted-U relationship between WTM and earnings management is valid for both female directors and female managers, but only stands for the subsample with low business complexity, low information transparence, and state-owned enterprises.

Keywords:

• earnings management
• inverted-U relationship
• monitoring intensity
• women top managers (WTM)

1. Introduction

In recent years, the role of women in affecting earnings quality has attracted an increasing interest. One branch of prior literature argues that women are more ethical, risk averse, and conservative than men, and thus women top managers (WTM) are associated with a lower extent of earnings management (Srinidhi, Gul, & Tsui, 2011; Zahra, Priem, & Rasheed, 2005). In contrast, another stream of previous literature finds no significant influence of WTM on earnings management, suggesting that women’s effect on earnings management is similar to that of men (Sun, Liu, & Lan, 2011; Ye, Zhang, & Rezaee, 2010). Moreover, some scholars argue that WTM can lead to important changes in an organisation only if the proportion of WTM is relatively higher. Borrowing support from the concept of “a critical mass” in nuclear physics, Dahlerup (1988, p. 276) emphasises that “a large minority can make a difference”. Extending this line of literature, recent studies examine whether a critical mass of WTM can have an impact on corporate behaviour (Chesterman, Ross-Smith, & Peters, 2005; Jia & Zhang, 2013). Overall, previous studies imply that WTM does change corporate behaviour significantly only if the number (proportion) of WTM is more than a critical value (usually three or 30%). Findings in extant studies are mixed at best, and thus extant studies provide little evidence on whether the effects of different proportions of WTM on corporate culture, behaviour, and decision-making are similar or asymmetric. This gap in prior literature inspires our study to explore the non-linear relationship between WTM and earnings management, which may help understand the mixed results in extant studies about women’s effects on earnings management.

Specifically, leadership style or management style in an organisation varies with gender diversity in the top management team. It is well-known that women are associated with participative and democratic leadership style, but men are related with authoritative and autocratic leadership style (Eagly & Carli, 2003; Eagly & Johnson, 1990; Eagly, Karau, & Johnson, 1992; Hull & Umansky, 1997; Kushell & Newton, 1986; Mukhtar, 2002; Rosener, 1995; Rozier & Hersh-Cochran, 1996). As a result, different proportions of WTM can signal different leadership styles and processes of decision-making, which bring out different earnings qualities. First, when the proportion of WTM is less than the inflexion point, male top managers dominate the firm and manifest relatively autocratic leadership, so men’s aggressive and immoral opinions on earnings management are inclined to be adopted. In this case, female top managers can not assert their opinions but follow men’s decisions (Chesterman et al., 2005; Rose, 2007). Even worse, more women top managers may strengthen the tendency to follow and thus induce a higher degree of earnings management (e.g. Srinidhi et al., 2011; Zahra et al., 2005). Second, when the proportion of WTM exceeds the inflexion point, according to the “critical mass” theory (Dahlerup, 1988), women top managers significantly influence leadership style and corporate decisions to an increasing extent although women are still a minority group. As a result, firms with higher proportions of WTM are supposed to practise democratic leadership and women’s opinions are respected and well responded. Since women top managers are inclined to be more ethical, risk-averse, and conservative, we can expect that a sufficient proportion of WTM is negatively associated with earnings management (Adams & Ferreira, 2009; Carter, D’Souza, Simkins, & Simpson, 2010; Carter, Simkins, & Simpson, 2003; Krishnan & Parsons, 2008; Peni & Vähämaa, 2010). To sum up, the foregoing arguments, taken together, motivate us to predict an inverted-U relationship between WTM and earnings management.

Most extant studies examine the role of women in corporate decisions using the context of developed countries like U.S., but the evidence from emerging markets is relatively rare (Jin, Song, & Yang, 2014). China is the largest developing country and the second largest economy in the world, and thus is appropriate for researchers to investigate whether WTM can affect earnings management. In addition, the social status of women in China may be quite different from those in developed countries. Due to paternalism and Confucian doctrines, women in ancient China possessed low social status (Rowlinson, Ho, & Po-Hung, 1993). However, the situation has prominently changed since 1949. Although gender discrimination in workplaces is still inevitable nowadays, women play an increasingly important role in contemporary China (Jia & Zhang, 2011, 2013; Tao, Zheng, & Mow, 2004; Xiang, He, & Cheng, 2014). In this regard, a stream of literature suggests that women managers (directors) can exert significant influence on corporate behaviour in Chinese enterprises.

Using a sample of 10,337 firm-year observations from the Chinese stock market over the period 2001–2011, our findings reveal the following aspects: First, WTM is significantly positively associated with earnings management when the proportion of WTM is lower than the inflexion point, while WTM is significantly negatively associated with earnings management when the proportion of WTM is higher than the inflexion point. Second, the monitoring intensity attenuates the inverted-U relation between WTM and earnings management. Third, our findings are robust to a variety of sensitivity tests using alternative measures of WTM and earnings management, and further our conclusions are still valid after controlling for the endogeneity issue. Fourth, the inverted-U relationship between WTM and earnings management is valid both for female members on the board/supervisory board and for female senior managers, but the moderating effect of monitoring intensity on the inverted-U relationship only stands for female members on the board/supervisory board. Finally, the inverted-U relationship only stands for the subsample with low business complexity, the subsample with low information transparency, and the subsample of state-owned enterprises (SOEs).

Our study contributes to the existing literature in several ways: First, most prior studies focus on the contexts of developed countries to investigate the impact of WTM on corporate decisions (Adams & Ferreira, 2009; Barua, Davidson, Rama, & Thiruvadi, 2010; Mukhtar, 2002; Srinidhi et al., 2011), but provide insufficient evidence on whether WTM affects corporate behaviour in developing economies. According to the global gender gap index of 2012 published by the World Economic Forum (WEF), there is a huge gap in the status of women between developing countries and developed countries (Hausmann, 2012). The five countries with the highest scores are developed countries (Iceland, Finland, Norway, Sweden and Ireland), but the five countries with the lowest scores are developing countries (Saudi Arabia, Syria, Chad, Pakistan and Yemen). As a result, findings based on developed countries may not fit well with those based on developing countries, and thus it is necessary for researchers to separately examine women’s role in corporate decisions using the contexts of developing countries.

Moreover, there is very sparse evidence on the relation between WTM and earnings management in emerging markets like China. Chinese enterprises are suffering from weak policies enforcement, poor legal protection, and incomplete corporate governance mechanisms (Allen, Qian, & Qian, 2005; Du, 2013; Fan, Wong, & Zhang, 2007). Therefore, earnings management in China may be more severe. Poor earnings quality is harmful to the interest of investors. As a result, our findings based on the Chinese context can add to the existing literature.

Second, to our knowledge and literature in hand, our study is one of few studies, if not the first, to explore the inverted-U relationship between WTM and earnings management. We argue that different proportions of WTM imply democratic or autocratic style, and thus affect earnings management differently. Our study validates the inverted-U relationship between WTM and earnings management and thus differentiates itself from previous literature, which either proposes that WTM is negatively associated with earnings management or finds that WTM has insignificant impact on earnings quality. In this regard, the inverted-U relation between WTM and earnings management can inspire future studies to examine whether the impacts of WTM on other corporate decisions are also non-linear.

Third, we explore the moderating effect of the monitoring intensity on the inverted-U relationship between WTM and earnings management. Extant studies recognise that WTM improves the monitoring efficiency of corporate boards (Adams & Ferreira, 2009; Srinidhi et al., 2011). China’s Securities Regulatory Commission (CSRC), Shanghai stock exchange (SSE), and Shenzhen stock exchange (SZSE) can serve as monitoring mechanisms to mitigate earnings manipulation (Du, Jian, Lai, Du, & Pei, 2015). Our results indicate the moderating effect of the monitoring intensity (an external monitoring mechanism) on the inverted-U relation between WTM (an internal governance mechanism) and earnings management. In this regard, our study contributes to the existing literature on the substitutive effect between internal and external governance mechanisms on corporate decisions.

Finally, we specify WTM as female directors on corporate boards, female members of supervisory board, and female senior managers, which enable us to comprehensively examine women’s role in earnings management. Previous literature verifies that women directors and managers play an important role in corporate decisions and behaviour (Adams & Ferreira, 2009; Campbell & Mínguez-Vera, 2008; Peni & Vähämaa, 2010; Srinidhi et al., 2011). With regard to earnings management, senior managers are deeply involved in the process of financial reporting while directors play an important monitoring role. Therefore, with regard to women’s influence on earnings management, neither women directors nor women managers should be ignored. If this happens researchers will face the risk of underestimating the impact of women on earnings management. Furthermore, after distinguishing female members on the board of directors/supervisory board from female senior managers in additional tests, we find that the inverted-U relationship between WTM and earnings management is valid for both female members on the board of directors (supervisory board) and female senior managers, but the moderating effect of the monitoring intensity on the inverted-U relationship only stands for female members on the board/supervisory board.

The remainder of the paper is organised as follows. In Section 2, we introduce the institutional background, review related literature, and develop research hypotheses. In Section 3, we specify empirical models and key variables. In Section 4, we report the sample, data sources, descriptive statistics, and Pearson correlation analysis. In Section 5, we report and discuss empirical results. In Section 6, we conduct robustness checks and discuss endogeneity between WTM and earnings management. Finally, we summarise our conclusions.

2. Institutional background, literature review, and hypotheses development

2.1. The current status of women in China

The social status of women in China has experienced remarkable change during the past decades. In ancient China, women acted the subordinate roles in families and the society. First, paternalism prevails in ancient China. The male head of the family made all decisions via his undisputed authority (Rowlinson et al., 1993), and thus women had to obey men’s decisions. Second, Confucian doctrines emphasise that women should follow “the three obediences and the four virtues”. Specifically, according to “the three obediences”, women are compelled to obey: (1) their fathers before their marriage; (2) their husbands when they get married; and (3) their sons if the husbands pass away (Confucianism, Ceremonial Etiquette Vol. 11, see Jin, 2003). As a result, men and women have their respective and distinct obligations, i.e., “the man goes out to work while the woman looks after the house” (a Chinese saying). Traditionally, women are supposed to take more responsibilities in housekeeping, raising children, and assisting husband. Therefore, women used to be adjuncts of men rather than independent individuals (Du, 2016; Johnson, 2009; Li, 2000; Mitchell, 1966; Teng, 1996) .

However, the situation changed after the establishment of the People’s Republic of China (Liu, 2007; Ono, 1989). First, the Chinese Communist Party adopts an egalitarian ideology, and thus the central and local governments and All-China Women’s Federation (ACWF) carry out a variety of policies to upgrade women’s social status (Leung, 2003). For example, there were twelve female members of the national committee of Chinese People’s Political Consultative in the first central government in 1949, four of which were members of the standing committee. Second, women constitute a high proportion of the labour force in contemporary China and women engage in more social work rather than only devote themselves to their families. In 1958, Mao Zedong put forward the slogan that “women can hold up half of the sky”, which encouraged women to participate in productive labour and work (Tao et al., 2004). According to the China’s National Bureau of Statistics and China’s Ministry of Human Resources and Social Security’s (2012), about 36.27% of employees are women. Third, Chinese women play an increasingly important role in various social activities. It is common for women to take part in politics, business, education, and philanthropy in recent years, and thus the degree of women’s participation is increasingly significant (e.g. Wu, 2009; Yuan, 2005). Finally, laws, regulations and rules in China ensure women’s social status to some extent. For example, the National People’s Congress published “Law of the People’s Republic of China on the Protection of Rights and Interests of Women” in 1992 to prevent women from infringement.¹ In response, recent studies (Jia & Zhang, 2011, 2013; Xiang et al., 2014) find that women directors and managers affect corporate behaviour in Chinese context.

To sum up, the current status of women rests on legal requirements because of the egalitarian ideology in gender issues adopted by the Chinese Communist Party (Bowen, Wu, Hwang, & Scherer, 2007; Lee Cooke, 2003; Ye et al., 2010). Considering that China is the second largest economy in the world and the largest developing country with its population of more than 1.3 billion, China can be a typical context to investigate the influence of women on corporate decisions. Therefore, in this study, we focus on the Chinese context to address whether and how WTM influences earnings management.

2.2. Literature review

Recent studies validate the important role of women in corporate boards, leadership, and organisation effectiveness. Eagly, Karau, & Makhijani (1995, p.135) suggest that women leaders are more effective than men in some fields. Harel, Tzafrir, and Baruch (2003) indicate that an organisation with high-quality human resource management and organisational effectiveness is more likely to give women a fair process of promotion. Peterson and Philpot (2007) find that female directors sit on public affairs committees more but on executive committees less. Adams and Ferreira (2009, p. 291) find that women directors attend meetings more frequently and are more likely to serve as members of monitoring committees, indicating that “gender-diverse boards allocate more effort to monitoring”. Huse, Nielsen, and Hagen (2009) and Nielsen and Huse (2010) find that women directors have significantly positive influences on strategic types of control and CSR.

Another stream of literature focuses on differences in ethical perceptions between women and men. Betz, O’Connell, and Shepard (1989) and Ruegger and King (1992) find that women are more ethical in their responses. Eynon, Hills, and Stevens (1997) validate that female accountants display significantly higher moral reasoning abilities than their male counterparts. With regard to agent-related duties, Simga-Mugan, Daly, Onkal, and Kavut (2005) argue that women managers manifest higher ethical sensitivity than men managers. Bernardi, Bosco, and Columb (2009) reveal that women on the board increase the likelihood of a firm’s appearance in the lists of the “most ethical companies”. Some other studies address the association between WTM and specific firm behaviour. Williams (2003) shows that appointing more women in the boardroom significantly increases corporate charitable giving. Carter et al. (2003) and Campbell and Mínguez-Vera (2008) validate that women directors are positively associated with firm value. Francoeur, Labelle, and Sinclair-Desgagne (2008) examine the impact of women directors (managers) on corporate performance and find that firms with a high proportion of women officers have significantly higher abnormal returns. Carter et al. (2010) indicate that the number of women directors significantly increases return on assets.

Relevant to the influence of WTM on earnings management, findings in the existing literature are mixed at best. A stream of previous literature finds that women do not behave differently in earnings management and thus the influence of women on earnings quality is indistinguishable from men. Giacomino, Bellovary, and Akers (2006) find that the difference in ethical perceptions between women and men is insignificant for 16 actions. Habbash (2010) examines the effectiveness of women directors on reducing earnings management in UK companies and finds that the negative relationship between women directors and earnings management is not supported. Ye et al. (2010) find that discretionary accruals are irrelevant to female top executives in Chinese listed firms. Similarly, Sun et al. (2011) do not find a significantly negative relationship between female directors on independent audit committee and earnings management in S&P 500 firms.

Another stream of prior literature validates the constraining effect of women on earnings management. Zahra et al. (2005) argue that gender factor can serve as a moderator for managerial fraud. Krishnan and Parsons (2008) find that earnings management for high gender diversity group is lower on average compared to low gender diversity group. Peni and Vähämaa (2010) and Barua et al. (2010) document that female CFOs decrease earnings management. Srinidhi et al. (2011) and Thiruvadi and Huang (2011) link female directors to earnings quality, and find that women directors improve the effectiveness and independence of board, and thus constrain companies’ earnings management.

Overall, the inconclusive findings in extant studies motivate us to explore a non-linear relationship between WTM and earnings management in this study.

2.3. Women top managers and earnings management

Borrowing from “critical mass” theory, this study aims at exploring the non-linear relationship between WTM and earnings management. Next, drawing on leadership style theory, we propose the inverted-U relationship between WTM and earnings management.

Previous literature suggests that women behave differently from men (Deaux & Lewis, 1984; Eagly & Steffen, 1984), and thus women and men have different leadership styles. According to Deaux and Lewis (1984, p. 994), men are supposed to manifest masculine traits including “self- confident, stands up well under pressure, and feels superior”, while women are supposed to show feminine traits such as “gentle, kind, aware of the feelings of others, helpful to others”. Eagly and Steffen (1984) indicate that women are more selfless, more concerned with others, less self-assertive and controlling. Moreover, a stream of literature on gender difference in leadership argues that women tend to practise a democratic or participative leadership (managerial) style, but men are inclined to adopt an autocratic or directive leadership style (Eagly & Carli, 2003; Eagly & Johnson, 1990; Eagly, Karau et al., 1992; Hull & Umansky, 1997; Kushell & Newton, 1986; Mukhtar, 2002; Rosener, 1995; Rozier & Hersh-Cochran, 1996). Lewin and Lippitt (1938, p. 294) define the democratic leadership as a type of atmosphere in which the leaders “attempt to be a group member in spirit but not in the actual work” and subordinates are encouraged to provide advice in decision-making process. First, democratic leadership is more consistent with female characteristics. Women usually consider others’ feelings more, pay greater attention to understanding and possess higher interpersonal skills (Bowes-Sperry, Veiga, & Yanouzas, 1997; Deaux & Lewis, 1984; Eagly & Steffen, 1984; Strebler, 1997), which ensure women to fit in well with the democratic leadership style. Second, women can suffer from being devalued if they choose the autocratic leadership (Eagly, Makhijani, & Klonsky, 1992; Hull & Umansky, 1997). Therefore, some scholars argue that women should practise a more collaborative mode, i.e. democratic leadership style, to remove barriers of effectiveness and obtain the acceptance by colleagues and subordinates (Eagly & Johnson, 1990).

The impact of WTM on corporate behaviour varies with the proportion of women (Chesterman et al., 2005; Dahlerup, 1988; Ely, 1995; Kanter, 1977). When women are relatively few, men dominate the leadership style and determine accounting policies and reporting earnings. In the gender stereotypes literature, men are commonly inclined to be self-assertive and controlling, and thus tend to adopt the autocratic leadership style. In other words, a relatively low proportion of WTM signals that a firm’s leadership style is autocratic. Therefore, for firms with a relatively low proportion of WTM, women’s opinions are easily ignored but men’s decisions are usually adopted. Moreover, women tend to follow or be in compliance with men’s decisions. Extant studies find that men are more likely to be associated with aggressive earnings manipulation (Barua et al., 2010; Krishnan & Parsons, 2008; Peni & Vähämaa, 2010; Srinidhi et al., 2011; Thiruvadi & Huang, 2011; Zahra et al., 2005). Further, the increase in the proportion of WTM reinforces aggressive earnings manipulation because women are inclined to follow men’s decisions (Chesterman et al., 2005; Rose, 2007). Based on the aforementioned discussions, we predict that the proportion of WTM is positively associated with earnings management when the proportion of WTM is lower than the inflexion point.

Next, we turn to discuss the influence of WTM on earnings management when the proportion of WTM exceeds the inflexion point. According to the “critical mass” theory, once the proportion of the minority becomes large enough, this minority group can result in substantial influence on decisions (Chesterman et al., 2005; Dahlerup, 1988; Ely, 1995; Kanter, 1977). Similarly, a critical mass of women significantly influences corporate behaviour including managerial culture and decision-making process, corporate governance, financial performance, innovation level, and corporate philanthropy. (Chesterman et al., 2005; Jia & Zhang, 2013; McKinsey & Company, 2007; Torchia, Calabro, & Huse, 2011). As a result, when the proportion of WTM is higher than the inflexion point, women can exert a significant impact on the leadership style, and a higher ratio of WTM implies the democratic leadership style. In addition, firms with large proportion of WTM are believed to be more open and liberal (e.g. Eagly & Carli, 2003; Hull & Umansky, 1997; Mukhtar, 2002; Rosener, 1995). As such, women’s opinions are respected and considered seriously during the process of decision making, which enables women to constrain earnings management because women behave more ethically than men (Bernardi et al., 2009; Betz et al., 1989; Eynon et al., 1997; Krishnan & Parsons, 2008; Peni & Vähämaa, 2010; Ruegger & King, 1992; Simga-Mugan et al., 2005; Zahra et al., 2005) and thus increase the effectiveness of monitoring (Adams & Ferreira, 2009; Carter et al., 2003, 2010; Nielsen & Huse, 2010; Srinidhi et al., 2011). Based on the foregoing arguments, we predict that when the proportion of WTM is higher than the inflexion point, the proportion of WTM is negatively associated with earnings management.

Overall, different proportions of WTM imply autocratic or democratic leadership style, and thus affect earnings quality differently. Therefore, we formulate hypothesis H1 in an alternative form:

H1: Ceteris paribus, there is an inverted-U relationship between the proportion of WTM and earnings management.

2.4. The moderating role of monitoring intensity

Extant studies (DeFond, Francis, & Hu, 2011; Du et al., 2015; Kedia & Rajgopal, 2011) find that the geographic proximity between a firm and regulatory centres significantly influences the quality of financial information. Kedia and Rajgopal (2011) argue that SEC’s regulation is more effective for firms closer to SEC offices, and thus geography distance adds extra costs to SEC enforcement. That is, local firms have better information about SEC enforcement and thus reduce financial misreporting behaviour. Furthermore, Kedia and Rajgopal (2011) validate the geographic proximity between a firm and the SEC is negatively associated with the likelihood of financial reporting restatement. DeFond et al. (2011) find that firms closer to SEC offices are less likely to receive a going concern opinion. Du et al. (2015) find that firms closer to three regulatory centres have lower extent of earnings management, echoing findings in DeFond et al. (2011) and Kedia and Rajgopal (2011).

China’s Securities Regulatory Commission (CSRC), Shanghai stock exchange (SSE) and Shenzhen stock exchange (SZSE) are three principal supervision bodies in the Chinese stock market, located in Beijing, Shanghai, and Shenzhen, respectively. Because of limited resources for supervision, regulators pay more attention to the irregular behaviour of nearby firms and conduct more efficient regulatory oversight of them (DeFond et al., 2011; Kedia & Rajgopal, 2011). Therefore, the reciprocal value of the average distance between a firm and these three regulatory centres can serve as a positive proxy for the monitoring intensity. As a result, we can rationally infer that a firm closer to regulatory centres faces more intensive monitoring, and thus less engages in earnings management.

Next, we further discuss the moderating effect of monitoring intensity on the inverted-U relation between WTM and earnings management. A stream of prior literature (Felix & Gramling, 2001; Starks, Hartzell, & Gillan, 2004; Weir, Laing, & McKnight, 2002; Young, Peng, Ahlstrom, Bruton, & Jiang, 2008) indicates the substitutive effect between internal governance mechanisms and external monitoring mechanisms. For example, Felix and Gramling (2001) find a negative relationship between internal audit contribution and external audit fees. Weir et al. (2002) verify the substitutive effects between internal and external governance mechanisms. Starks et al. (2004) suggest that board independence is negatively associated with the extent of firms’ exposure to takeovers. Young et al. (2008, p. 208) find that “concentrated ownership substitutes for poor external governance mechanisms in emerging economies to reduce traditional principal–agent monitoring conflicts”. Analogously, the geographic proximity between a firm and regulatory centres, as an external monitoring mechanism, plays a moderating role in the inverted-U relationship between WTM and earnings management. Specifically, for firms closer to regulatory centres, stronger monitoring intensity reduces the motivation and extent of earnings management, and thus: (1) when the ratio of WTM is lower than the inflexion point, the reinforced effect of WTM on earnings management is alleviated; when the ratio of WTM is higher than the inflexion point, WTM’s mitigating effect on earnings management is weakened. Based on above discussions, we formulate hypothesis H2 in an alternative form:

H2: Ceteris paribus, monitoring intensity attenuates the inverted-U relationship between WTM and earnings management.

3. Measurement and empirical models

3.1. The measurement of earnings management

Discretionary accruals are partially attributed to managers’ self-serving decisions. Jones (1991) specifies nondiscretionary accruals as a function of change in revenue and gross property, plant and equipment (PPE). Extending Dechow, Sloan, and Sweeney (1995), Jones (1991) develop the modified Jones model. In this study, we employ the modified Jones model of Dechow et al. (1995) to estimate the extent of earnings management. In addition, we will use the augmented Jones model of Ball and Shivakumar (2006) to calculate two additional variables to conduct robustness checks. Specifically, following Dechow et al. (1995), we calculate discretionary accruals (DA) using equations (1) and (2):(1) $$\frac{AC{C}_{jt}}{T{A}_{jt-1}}={\mathit{\alpha }}_1\frac{1}{T{A}_{jt-1}}+{\mathit{\alpha }}_2\frac{\mathrm{\Delta }RE{V}_{jt}}{T{A}_{jt-1}}+{\mathit{\alpha }}_3\frac{PP{E}_{jt}}{T{A}_{jt-1}}+{\mathit{\epsilon }}_{jt}$$(1)

In equation (1), for firm j and year t, ACC equals to net income minus cash flows from operations. TA denotes the total assets of last period. ΔREV represents change in sales revenue and PPE represents gross property, plant and equipment. We estimateα₁, α₂, α₃ for each industry and year using the original Jones model. Following Dechow et al. (1995), we exclude changes in accounts receivable from total revenue (ΔREC) and then calculate nondiscretionary accruals using equation (2). Our measure of DA equals to actual total accruals less the predicted value of nondiscretionary accruals.(2) $$\frac{AC{C}_{jt}}{T{A}_{jt-1}}={\mathit{\alpha }}_1\frac{1}{T{A}_{jt-1}}+{\mathit{\alpha }}_2\frac{(\mathrm{\Delta }RE{V}_{jt}-\mathrm{\Delta }RE{C}_{jt})}{T{A}_{jt-1}}+{\mathit{\alpha }}_3\frac{PP{E}_{jt}}{T{A}_{jt-1}}+{\mathit{\epsilon }}_{jt}$$(2)

3.2. Empirical model specification for hypothesis H1

To test H1, we construct equation (3) to link earnings management with WTM, firm-specific control variables, industry dummies, and year dummies: (3) $$DA={\mathit{\beta }}_0+{\mathit{\beta }}_{1}WMR+{\mathit{\beta }}_{2}WM{R}^2+Controls+IndustryDummies+YearDummies+\mathit{\mu }$$(3)

In equation (3), the dependent variable is earnings management, labeled as DA. WMR, the label for the proportion of women top managers, is the main independent variable. In Eq. (3), if the coefficient on WMR (β₁) is significantly positive and the coefficient on WMR² (β₂) is significantly negative simultaneously, H1 is supported by empirical evidence.

Following previous studies, we include a set of firm-specific control variables in Eq. (3). First, Becker, DeFond, Jiambalvo, and Subramanyam (1998) suggest that auditor size is negatively associated with earnings management, and thus we include BIG4 in equation (3). Myers, Myers, and Omer (2003) find that longer auditor tenure results in lower discretionary accruals, and thereof we include TENURE.

Second, previous studies argue that external governance mechanisms can alleviate earnings management. Yu (2008) finds that analyst coverage is significantly negatively associated with earnings manipulations. Chung, Firth, and Kim (2002) validate that institutional shareholdings inhibit managers’ opportunistic earnings management. Leuz, Nanda, and Wysocki (2003) suggest that outsider rights reduce earnings management. Thus, we include three variables, i.e., ANALYST, INSTSHR and MKT, in equation (3) to control for the effects of analyst coverage, institutional ownership, and institutional development of each province in China on earnings management, respectively.

Third, following Choi, Kim, Qiu, and Zang (2012) and Wang, Li, and Chen (2015), LNBGS, CHGSALES and BTM are included in equation (3) to control for firm segments and firm growth.

Fourth, prior studies find that firms in financial distress or intending to issue equities have stronger incentives to manipulate earnings (Barton, 2001; Teoh, Welch, & Wong, 1998). Ashbaugh, LaFond, and Mayhew (2003) employ the lag total accruals to control for the reversal of accruals over time. Thus, we include a modified financial distress score MZ (Altman, 1968; Omer, Sharp, & Wang, 2010), a dummy variable of equity issuing (ISSUE), and one-year lagged total accruals (LAGACCR) in equation (3).

Fifth, internal governance mechanisms are significantly negatively associated with earnings management (Davidson, Goodwin-Stewart, & Kent, 2005; Klein, 2002; Xie, Davidson, & DaDalt, 2003). As a response, we include five variables, FIRST, DUAL, INDR, MANSHR and BOARD in equation (3).

Sixth, firm-specific variables, OCF, ROA, SIZE, LEV and STATE, are included in equation (3) to control for the impact of operating cash flows, returns on total assets, firm size, financial leverage and the nature of ultimate owners (Jiang, Lee, & Yue, 2010), respectively, on earnings management.

Finally, to control for the fixed effect of calendar years and industries, YEAR dummies and INDUSTRY dummies are both included in the model. All the variables are defined in Table 1.

Table 1. Variable definitions.

Variable	Definition	Data source
DA	Discretionary (abnormal) accruals using modified Jones model of Dechow et al. (1995)	A’ C based on CSMAR
WMR	The ratio of women top managers, equaling to the number of women top managers scaled by the total number of top managers;	A’H based on CSMAR
REGLIST	The reciprocal value of the average distance between a listed firm and regulatory centres (i.e., Beijing, Shanghai, Shenzhen), multiply 1,000;	A’H based on Google- earth map
BIG4	A dummy variable, equaling to 1 when the auditor is a Big 4 accounting firm (including affiliated firms) and 0 otherwise (Fan & Wong, 2005);	www.cicpa.org.cn
TENURE	The number of years since a client hires an auditor to audit its annual reports;	CSMAR
ANALYST	Analyst coverage, measured as the natural log of (1+the number of analyst coverage);	CSMAR
INSTSHR	The percentage of shares owned by institutional investors;	CSMAR
MKT	Fan et al.’s Marketization Index which measures the whole institution development of each province in China and the investor protection of each region;	Fan et al. (2010)
LNBGS	The natural log of the sum of business, product, and geographic segments minus 2 (Choi et al., 2012);	A’ C based on WIND
CHGSALES	Change in sales deflated by lagged total assets (Choi et al., 2012);	CSMAR
BTM	The ratio of book-to-market;	CSMAR
ISSUE	A dummy variable, equaling 1 if debt and equity issued during the past three years is more than 5% of the total assets, and 0 otherwise (Choi et al., 2012);	A’H based on CSMAR
MZ	According to Altman (1968)’s Z-score, equaling 2 if Z-score is less than 1.80, 1 if Z-score is between 1.80 and 3.00, and 0 otherwise (Omer et al., 2010);	A’ C based on CSMAR
LAGACCR	One-year lagged total accruals, and accruals are calculated as net income before extraordinary items minus operating cash flows deflated by the lagged total assets;	A’ C based on CSMAR
FIRST	The percentage of shares owned by controlling shareholders;	CSMAR
DUAL	An indicator variable, equaling 1 if the CEO and the chairman are the same person, and 0 otherwise;	CSMAR
INDR	The number of independent directors to the total number of the directors in the boardroom;	CSMAR
MANSHR	The percentage of shares owned by a firm’s managers;	CSMAR
BOARD	The natural logarithm of the number of directors in the boardroom;	CSMAR
OCF	Operating cash flows deflated by lagged total assets;	CSMAR
ROA	Return on total assets, measured as earnings before interest and tax deflated by total assets;	CSMAR
SIZE	Firm size, measured by natural log of total assets (Jiang et al., 2010);	CSMAR
LEV	The ratio of total liabilities to total assets (Jiang et al., 2010);	CSMAR
STATE	A dummy variable, equaling to 1 when the ultimate controlling shareholder of a listed firm is a (central or local) government agency or government controlled enterprises and 0 otherwise (Jiang et al., 2010).	CSMAR
DA_ΔCF	Discretionary accruals based on augmented Jones model of Ball and Shivakumar (2006) considering the change of operating cash flow;	A’ C based on CSMAR
DA_RET	Discretionary accruals based on augmented Jones model of Ball and Shivakumar (2006) considering abnormal returns;	A’ C based on CSMAR
DA_PMADJ	Discretionary accruals based on Kothari et al. (2005)’s model;	A’ C based on CSMAR
RM	The sum of the inverse value of abnormal cash flows from operations and abnormal discretionary expenses (Cohen & Zarowin, 2010; Zang, 2011);	A’ C based on CSMAR
LISTAGE	The number of years since firms’ initial public offerings on Chinese A-Share market (Srinidhi et al., 2011);	CSMAR
NONEXEDR	The percentage of non-executive directors, equaling to the number of non-executive directors scaled by the total number of directors on board (Srinidhi et al., 2011);	A’H based on CSMAR
NUMIND	The number of industries from which firms’ sales revenues are generated (Srinidhi et al., 2011);	A’H based on CSMAR
SD	Standard deviation of firms’ daily return over a fiscal year, standardize to a mean of 0 and a standard deviation of 1 (Srinidhi et al., 2011);	A’C based on CSMAR
RET	The annual return of each firm (Srinidhi et al., 2011);	CSMAR
MARRET	The value weighted market value over the corresponding period (Srinidhi et al., 2011);	CSMAR
FEMALE	The ratio of female employees to total employees at the industry level (Srinidhi et al., 2011);	China Statistic Yearbook
GDP_PC	The natural logarithm of GDP per capita at the province level.	China Statistic Yearbook

Notes: A’C: Author’s calculation; A’H: Author’s Hand-collection; CSMAR: China Stock Market and Accounting Research.

3.3. Empirical model specification for hypothesis H2

To test H2, we construct equation (4) by adding monitoring intensity (REGLIST) and the interactive item between WTM and monitoring intensity (WMR × REGLIST and WMR² × REGLIST):(4) $$DA={\mathit{\lambda }}_0+{\mathit{\lambda }}_{1}WMR+{\mathit{\lambda }}_{2}WM{R}^2+{\mathit{\lambda }}_{3}REGLIST+{\mathit{\lambda }}_{4}WMR\times REGLIST+{\mathit{\lambda }}_{5}WM{R}^2\times REGLIST+Controls+IndustryDummies+YearDummies+\mathit{\xi }$$(4)

In equation (4), REGLIST denotes monitoring intensity. A stream of previous literature validates the economic implications of geographic proximity (Audretsch & Feldman, 1996; Coval & Moskowitz, 2001; Krugman, 1991). For firms closer to CSRC, SSE and SZSE, their irregular behaviour is more likely to be detected and then investigated by the supervisory bodies. H2 predicts that monitoring intensity attenuates the inverted-U relation between WTM and earnings management. Thus, if the coefficient on WMR × REGLIST (λ₄) is significantly negative and the coefficient on WMR² × REGLIST (λ₅) is significantly positive, H2 is supported by empirical evidence. Control variables in equation (4) are the same as those in equation (3).

4. Sample, data and descriptive statistics

4.1. Sample construction and data sources

Our initial sample consists of all the Chinese A-share listed firms over the period of 2001–2011. Panel A of Table 2 displays the sample selection process. Beginning with 17,091 firm-year observations, we select the research sample according to the following principles: (1) Eliminate firms pertaining to the banking, insurance, and other financial industries because of their unique financial structures. (2) Delete firms that have been listed in the market for less than one year. (3) Exclude firm-year observations marked as ST (special treatment), *ST (suspension of trading), and PT (particular transfer). (4) Delete observations whose stocks are listed in two or more stock markets. (5) Eliminate observations whose data on firm-specific control variables are unavailable. Finally, we obtain a sample of 10,337 firm-year observations which consists of 1,437 unique firms. Panel B of Table 2 presents the year and industry distribution of observations. Moreover, all the continuous variables used in our study are winsorized at 1% and 99% to mitigate the influence of outliers.

Table 2. Sample selection and distribution.

Panel A: Firm-years selection process
Initial observations														17,091
Eliminate observations pertaining to the banking, insurance, and other financial industries														(186)
Eliminate observations that do not meet the criteria that listed for at least 1 year														(1,501)
Eliminate observations whose transaction statuses are ST,*ST or PT														(1,486)
Eliminate observations whose stocks are cross-listed														(1,222)
Eliminate observations whose data required to measure firm-specific control variables are unavailable														(3,169)
Available firm-year observations														10,337
Unique firms														1,437
Panel B: Sample distribution by year and industry
Industry	Code	Year											Total by Industry	%
		2001	2002	2003	2004	2005	2006	2007	2008	2009	2010	2011
Agriculture, forestry, husbandry and fishery	A	12	14	21	22	22	21	24	24	25	28	30	243	2.35
Mining	B	7	9	12	14	14	16	19	16	22	26	27	182	1.76
Food and beverage	C0	26	31	34	37	43	43	43	46	49	53	56	461	4.46
Textile, garment manufacturing and products of leather and fur	C1	24	28	35	37	36	39	45	47	49	54	60	454	4.39
Wood and furniture	C2	1	1	2	1	1	1	2	3	3	3	5	23	0.22
Papermaking and printing	C3	13	13	13	16	16	15	19	19	24	26	30	204	1.97
Petroleum, chemical, plastics, and rubber products	C4	77	88	99	109	107	112	116	112	125	125	131	1,201	11.62
Electronics	C5	15	15	18	21	25	31	30	33	38	51	53	330	3.19
Metal and non-metal	C6	56	62	77	80	93	92	89	89	98	107	113	956	9.25
Machinery, equipment, and instrument manufacturing	C7	92	100	107	122	135	138	154	165	170	188	203	1,574	15.23
Medicine and biological products manufacturing	C8	33	37	44	52	60	64	67	70	74	74	77	652	6.31
Other manufacturing	C9	3	8	11	12	11	10	12	12	15	18	19	131	1.27
Production and supply of electricity, steam and tap water	D	28	29	36	39	41	45	48	50	48	48	51	463	4.48
Construction	E	9	12	13	15	16	21	24	23	31	30	31	225	2.18
Transportation and warehousing	F	18	22	28	31	33	35	37	39	42	41	44	370	3.58
Information technology	G	38	46	50	51	53	53	53	54	64	71	78	611	5.91
Wholesale and retail	H	65	66	71	76	79	76	78	75	77	79	86	828	8.01
Real estate	J	24	25	28	31	33	34	35	34	36	39	39	358	3.46
Social services	K	17	23	30	30	30	32	33	33	36	40	41	345	3.34
Communication and culture	L	6	7	6	7	7	6	6	6	6	6	8	71	0.69
Conglomerates	M	59	61	61	61	65	59	56	58	56	59	60	655	6.34
Total by year		623	697	796	864	920	943	990	1,008	1,088	1,166	1,242	10,337	100
%		6.03	6.74	7.70	8.36	8.90	9.12	9.58	9.75	10.53	11.28	12.02

Data sources in this study are reported as below: (1) Referring to Dechow et al. (1995) and based on China Stock Market and Accounting Research (CSMAR), we calculate discretional accruals (DA). (2) We compute the ratio of WTM base on CSMAR database. (3) We hand-collect the longitudes and latitudes of regulatory centres (i.e. Beijing, Shanghai and Shenzhen) and each firm via the well-known Geographic Information System (GIS), and then calculate the reciprocal value of the average distance between a firm and three regulatory centres (regulatory distance) to define REGLIST. (4) We obtain BIG4 from the accounting firms’ ranking of Chinese Institute of Certified Public Accountants (CICPA). (5) MKT, the province-level Marketization index, is obtained from Fan, Wang, and Zhu (2010). (6) We hand-collect data on corporate segments (LNBGS) based on WIND database. Except for above variables, the data on other control variables are collected or calculated based on the CSMAR database.

4.2. Descriptive statistics and Pearson correlation analysis

Table 3 presents the descriptive statistics. The mean (median) value of DA is 0.01 (0.00), suggesting the basic characteristics of discretionary accruals in Chinese listed firms. The mean (median) value of WMR is 0.14 (0.12), indicating that only 14% of top managers in Chinese listed firms are women. REGLIST has a mean (median) value of 0.98 (0.99), revealing the average distance between a firm and three regulatory centres is about 1,002 kilometres.

Table 3. Descriptive statistics.

Variable	N	Mean	Std. Dev.	Min	Q1	Median	Q3	Max
DA	10,337	0.01	0.10	−0.31	−0.04	0.00	0.05	0.39
WMR	10,337	0.14	0.10	0.00	0.06	0.12	0.20	0.67
REGLIST	10,337	0.98	0.25	0.33	0.78	0.99	1.20	1.32
BIG4	10,337	0.04	0.19	0.00	0.00	0.00	0.00	1.00
TENURE	10,337	6.82	4.09	1.00	4.00	6.00	9.00	20.00
ANALYST	10,337	0.98	1.13	0.00	0.00	0.69	1.79	3.66
INSTSHR	10,337	0.14	0.15	0.00	0.02	0.08	0.22	0.54
MKT	10,337	7.94	2.22	0.33	6.27	7.66	9.55	11.71
LNBGS	10,337	1.72	0.65	0.00	1.39	1.79	2.19	2.99
CHGSALES	10,337	0.12	0.26	−0.47	0.00	0.07	0.19	1.54
BTM	10,337	0.55	0.26	0.06	0.34	0.54	0.76	1.18
ISSUE	10,337	0.85	0.35	0.00	1.00	1.00	1.00	1.00
MZ	10,337	0.83	0.81	0.00	0.00	1.00	2.00	2.00
LAGACCR	10,337	−0.02	0.10	−0.30	−0.07	−0.02	0.02	0.36
FIRST	10,337	0.38	0.16	0.09	0.26	0.36	0.51	0.76
DUAL	10,337	0.13	0.33	0.00	0.00	0.00	0.00	1.00
INDR	10,337	0.33	0.10	0.00	0.33	0.33	0.36	0.71
MANSHR	10,337	0.01	0.06	0.00	0.00	0.00	0.00	0.42
BOARD	10,337	2.22	0.21	1.10	2.20	2.20	2.40	2.94
OCF	10,337	0.04	0.09	−0.23	0.00	0.03	0.09	0.33
ROA	10,337	0.05	0.06	−0.22	0.03	0.05	0.08	0.23
SIZE	10,337	21.52	1.02	19.19	20.81	21.43	22.12	25.16
LEV	10,337	0.50	0.18	0.07	0.37	0.51	0.63	0.91
STATE	10,337	0.69	0.46	0.00	0.00	1.00	1.00	1.00

Note: All the variables are defined in Table 1.

As for control variables, on average and approximately, 4% of Chinese listed firms are audited by big four auditors (BIG4), auditor tenure (TENURE) is 6.82 years, the number of analyst coverage (ANALYST) is about 1.67 (e^0.98−1), the percentage of shares held by institutional investors (INSTSHR) is 14%, the Marketization index (MKT) is 7.94, the number of corporate segments (LNBGS) is about 7.61 (e^1.72+2), the change of sales for Chinese listed firms (CHGSALES) is 12% of lagged total assets, the book-to-market ratio (BTM) is 0.55, 85% of firms have substantial debt or/and equity issuing (ISSUE) during the past three years, more than 25% of firms are likely to face financial distress (MZ), The lagged total accruals (LAGACCR) is −0.02, the controlling shareholders own 38% of shares (FIRST), CEO and the chairman of the board are the same person in 13% of firms (DUAL), the ratio of independent directors (INDR) in Chinese listed firms is around the criteria of CSRC (i.e., 1/3), the shareholding of top managers (MANSHR) is only 1%, the number of directors in the boardroom (BOARD) is nine (e^2.22), operating cash flows is 4% of lagged total assets (OCF), return on assets (ROA) is 5%, firm size (SIZE) is 2.23 billion Chinese Yuan (e^21.52), the ratio of total liabilities to total assets (LEV) is around 50%, 69% of sample firms are state-owned enterprises (STATE), respectively.

Table 4 reports the Pearson correlation analysis of variables used in our study. There is no significant relation between WMR and DA, consistent with Ye et al. (2010). This result preliminarily suggests a non-linear relation between WTM and earnings management.

Table 4. Pearson Correlation Matrix.

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)	(17)	(18)	(19)	(20)	(21)	(22)	(23)	(24)
(1)	DA	1.000
(2)	WMR	−0.005	1.000
(3)	REGLIST	0.012	−0.012	1.000
(4)	BIG4	0.003	−0.032^***	0.088^***	1.000
(5)	TENURE	0.003	0.077^***	0.094^***	−0.052^***	1.000
(6)	ANALYST	0.091^***	−0.043^***	0.047^***	0.098^***	0.089^***	1.000
(7)	INSTSHR	0.009	0.051^***	−0.024^**	0.018*	0.015	0.213^***	1.000
(8)	MKT	0.037^***	0.081^***	0.605^***	0.081^***	0.178^***	0.261^***	0.069^***	1.000
(9)	LNBGS	0.052^***	−0.036^***	0.004	−0.022^**	0.082^***	0.170^***	0.004	0.177^***	1.000
(10)	CHGSALES	0.039^***	−0.029^***	0.051^***	0.038^***	−0.005	0.191^***	0.077^***	0.034^***	0.054^***	1.000
(11)	BTM	0.035^***	−0.091^***	0.016^*	0.066^***	−0.045^***	−0.209^***	−0.197^***	−0.109^***	−0.050^***	−0.073^***	1.000
(12)	ISSUE	0.076^***	−0.038^***	0.033^***	−0.048^***	0.021^**	−0.024^**	−0.057^***	−0.005	0.088^***	0.018*	0.066^***	1.000
(13)	MZ	−0.064^***	−0.046^***	−0.058^***	−0.024^**	−0.017*	−0.180^***	−0.107^***	−0.101^***	0.005	−0.154^***	0.346^***	0.240^***	1.000
(14)	LAGACCR	0.118^***	0.014	0.014	−0.009	−0.039^***	0.027^***	−0.013	0.039^***	0.026^***	−0.005	0.009	0.049^***	−0.028^***	1.000
(15)	FIRST	0.019^*	−0.143^***	0.025^**	0.110^***	−0.135^***	0.036^***	−0.109^***	−0.091^***	−0.082^***	0.087^***	0.217^***	−0.040^***	0.000	0.001	1.000
(16)	DUAL	−0.000	0.078^***	0.004	−0.003	0.001	0.012	0.031^***	0.061^***	0.028^***	−0.008	−0.074^***	−0.020^**	−0.046^***	0.034^***	−0.072^***	1.000
(17)	INDR	0.022^**	0.049^***	0.035^***	0.021^**	0.142^***	0.265^***	0.056^***	0.326^***	0.232^***	0.057^***	−0.003	0.003	0.006	−0.015	−0.101^***	0.038^***	1.000
(18)	MANSHR	0.053^***	0.047^***	0.090^***	−0.020^**	−0.111^***	0.156^***	−0.004	0.206^***	0.067^***	0.038^***	−0.113^***	0.016*	−0.106^***	0.074^***	−0.136^***	0.120^***	0.094^***	1.000
(19)	BOARD	−0.004	−0.113^***	−0.027^***	0.053^***	−0.043^***	0.087^***	0.035^***	−0.072^***	−0.012	0.0373^***	0.108^***	0.022^**	0.066^***	−0.056^***	0.029^***	−0.104^***	−0.163^***	−0.079^***	1.000
(20)	OCF	−0.280^***	−0.026^***	−0.022^**	0.020^**	−0.016	0.099^***	0.022^**	0.007	−0.000	0.122^***	−0.022^**	−0.061^***	−0.093^***	−0.062^***	0.079^***	−0.019^*	0.024^**	−0.010	0.032^***	1.000
(21)	ROA	0.253^***	0.001	0.082^***	0.070^***	0.047^***	0.405^***	0.147^***	0.141^***	0.023^**	0.300^***	−0.132^***	−0.089^***	−0.403^***	0.017*	0.099^***	−0.014	0.083^***	0.083^***	0.039^***	0.187^***	1.000
(22)	SIZE	0.103^***	−0.145^***	0.072^***	0.181^***	0.115^***	0.536^***	0.029^***	0.187^***	0.177^***	0.178^***	0.230^***	0.114^***	0.204^***	0.034^***	0.203^***	−0.063^***	0.171^***	−0.053^***	0.188^***	0.043^***	0.206^***	1.000
(23)	LEV	−0.052^***	−0.017^*	−0.014	−0.029^***	0.042^***	0.016	−0.005	0.017^*	0.090^***	0.116^***	−0.030^***	0.343^***	0.684^***	−0.016	−0.045^***	−0.029^***	0.071^***	−0.076^***	0.054^***	−0.079^***	−0.269^***	0.319^***	1.000
(24)	STATE	−0.037^***	−0.186^***	−0.028^***	0.040^***	−0.003	−0.036^***	−0.021^**	−0.151^***	−0.057^***	0.012	0.155^***	−0.021^**	0.063^***	−0.081^***	0.273^***	−0.116^***	−0.134^***	−0.286^***	0.188^***	0.022^**	−0.052^***	0.155^***	0.024^**	1.000

Notes: ***: Significant at the 1% level; **: significant at the 5% level; *: significant at the 10% level. All the variables are defined in Table 1.

As for control variables, the results show that: (1) ANALYST, MKT, LNBGS, CHGSALES, BTM, ISSUE, LAGACCR, FIRST, INDR, MANSHR, ROA and SIZE are significantly positively correlated with DA, while MZ, OCF, LEV and STATE are significantly negatively correlated with DA. The foregoing results, taken together, suggest the need to include these variables in our regressions. In addition, the coefficients on pair-wise correlation among control variables are generally low (less than 0.5), suggesting that our regression tests are less likely to suffer from serious multicollinearity problem.

5. Empirical results

5.1. Multivariate test of hypothesis H1

H1 predicts the inverted-U relationship between WTM and earnings management. Table 5 reports the OLS regression results and all t statistics are computed and reported based on standard errors adjusted for clustering at the firm and the year levels (Petersen, 2009).

Table 5. Regression results of discretionary accruals on WMR and other determinants.

Variable	The dependent variable: DA
	Coefficient	t-value
WMR	0.056***	2.85
WMR^2	−0.165***	−2.94
BIG4	−0.010*	−1.82
TENURE	−0.000	−1.11
ANALYST	−0.003**	−2.21
INSTSHR	−0.005	−0.67
MKT	−0.001**	−2.29
LNBGS	0.005***	2.90
CHGSALES	−0.009	−1.40
BTM	0.026***	3.06
ISSUE	0.022***	4.49
MZ	0.004	1.60
LAGACCR	0.085***	4.74
FIRST	0.006	0.57
DUAL	0.000	0.02
INDR	−0.002	−0.12
MANSHR	0.025	1.48
BOARD	−0.003	−0.77
OCF	−0.181**	−2.03
ROA	0.540***	11.69
SIZE	0.006**	2.17
LEV	−0.028**	−2.05
STATE	−0.003	−1.41
INTERCEPT	−0.134*	−1.93
INDUSTRY	Control
YEAR	Control
adj_ R^2	20.47%
Observations	10,337
F- (p-value)	51.18***

Notes: ***,** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level and the year level (Petersen, 2009). All the variables are defined in Table 1.

As shown in Table 5, the coefficient on WMR is positive and significant at the 1% level (0.056 with t = 2.85). Moreover, the coefficient on WMR² is negative and significant at the 1% level (−0.165 with t = −2.94). These results, taken together, suggest the inverted-U relationship between WTM and DA, lending strong support to H1. Furthermore, based on the coefficients on WMR and WMR², the inflexion point of WTM is 17% (−0.056/(2 × −0.165)), meaning that the number of women top managers on the inflexion point is three (17% × 17.84; 17.84 denotes the average total number of top managers). As a result, when the number of women top managers is more than three, WTM is significantly negatively associated with earnings management; Otherwise, WTM is significantly positively associated with earnings management when the number of women top managers is less than three.

As for control variables, we note the following aspects: (1) The coefficients on BIG4, ANALYST and MKT are all significantly negative, suggesting that external auditor, analyst coverage and Marketization can serve as external mechanisms to reduce earnings management. (2) LNBGS has a significant and positive coefficient, indicating that corporate diversification is positively associated with earnings management. (3) The coefficient on BTM is significantly positive, revealing that high-growth companies exhibit less discretionary accruals. (4) ISSUE is significantly positively associated with DA, providing support to the argument that seasoned-equity-offering firms have stronger incentives to manipulate earnings. (5) LAGACCR has a significantly positive coefficient, implying that the lagged total accruals are positively associated with discretionary accruals. (6) The coefficient on OCF is significantly negative, echoing to the findings in prior literature (see Aharony, Yuan, & Wang, 2005; Choi et al., 2012; Kim, Chung, & Firth, 2003). (7) ROA has a significantly positive coefficient, indicating that profitable companies have greater extent of earnings management. (8) SIZE has a significantly positive coefficient and LEV has a significantly negative coefficient, reflecting that large companies and companies with low financial leverage exhibit more substantial earnings management.

5.2. Multivariate test of hypothesis H2

H2 predicts that monitoring intensity attenuates the inverted-U relationship between WTM and earnings management. Table 6 reports the regression results.

Table 6. Regression results of discretionary accruals on WMR, monitoring intensity, and other determinants.

Variable	The dependent variable: DA
	Coefficient	t-value
WMR	0.044**	2.07
WMR^2	−0.134**	−2.13
REGLIST	−0.012***	−2.71
WMR × REGLIST	−0.247***	−3.00
WMR^2 × REGLIST	0.588***	3.85
BIG4	−0.010*	−1.75
TENURE	−0.001	−1.12
ANALYST	−0.003**	−2.17
INSTSHR	−0.006	−0.77
MKT	−0.000	−0.29
LNBGS	0.004***	2.84
CHGSALES	−0.009	−1.30
BTM	0.027***	3.20
ISSUE	0.022***	4.65
MZ	0.004	1.62
LAGACCR	0.084***	4.69
FIRST	0.006	0.60
DUAL	−0.000	−0.10
INDR	−0.003	−0.22
MANSHR	0.026	1.54
BOARD	−0.003	−0.81
OCF	−0.181**	−2.03
ROA	0.541***	11.72
SIZE	0.006**	2.16
LEV	−0.029**	−2.08
STATE	−0.003	−1.21
INTERCEPT	−0.130*	−1.87
INDUSTRY	Control
YEAR	Control
adj_ R^2	20.54%
Observations	10,337
F- (p-value)	48.71***
Joint F-test:
WMR+ WMR × REGLIST
F- (p-value)	5.39** (0.0203)
WMR^2+ WMR^2 × REGLIST
F- (p-value)	5.93** (0.0149)

Notes: ***,** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level and the year level (Petersen, 2009). All the variables are defined in Table 1.

As presented in Table 6, the coefficient on WMR is positive and significant at the 5% level (0.044 with t = 2.07) while the coefficient on WMR² is negative and significant at the 5% level (−0.134 with t = −2.13), which are again consistent with H1. REGLIST is significantly negatively associated with DA at the 1% level (−0.012 with t = −2.71), suggesting that a firm closer to regulatory centres tends to engage less in earnings management. More importantly, the coefficient on WMR × REGLIST is negative and significant at the 1% level (−0.247 with t = −3.00) and the coefficient on WMR² × REGLIST is positive and significant at the 1% level (0.588 with t = 3.85). These results, taken together, provide support to H2 by indicating that monitoring intensity attenuates the association between WTM and earnings management. Furthermore, as shown in the bottom of Table 6, the F value of WMR × REGLIST and WMR² × REGLIST are both significant at the 5% level (F-value = 5.39 and F-value = 5.93, respectively). The results of joint-F tests imply that, the sum of the coefficients on WMR (WMR²) and WMR × REGLIST (WMR² × REGLIST) is statistically different from zero.

5.3. Robustness checks using alternative measures of earnings management

Ball and Shivakumar (2006) suggest that researchers should consider the asymmetric recognition of gains and losses and then construct a non-linear model of accruals. Kothari, Leone, and Wasley (2005) indicate that corporate performance should be taken into account. Dechow, Kothari, and Watts (1998) and Roychowdhury (2006) find that managers manipulate real operating activities to deviate from normal operational practices to achieve certain financial goals (real earnings management). Next, we use alternative measures of earnings management (DA_ΔCF, DA_RET, DA_PMADJ and RM) for robustness checks. Specifically, DA_ΔCF (DA_RET) denotes discretionary accruals based on the augmented Jones model of Ball and Shivakumar (2006), considering the change of operating cash flows (abnormal returns). DA_PMADJ denotes discretionary accruals using Kothari et al. (2005)’s method. RM is the sum of the inverse value of abnormal cash flows from operations and abnormal discretionary expenses (Cohen & Zarowin, 2010; Zang, 2011). Table 7 presents the regression results.

Table 7. Robustness checks using alternative measures of discretional accruals.

Panel A: Robustness checks of H1 using alternative measures of discretionary accruals
Variable	(1)		(2)		(3)		(4)
	Dependent variable: DA_ΔCF		Dependent variable: DA_RET		Dependent variable: DA_PMADJ		Dependent variable: RM
	Coefficient	t-value	Coefficient	t-value	Coefficient	t-value	Coefficient	t-value
WMR	0.035**	2.44	0.059***	3.27	0.095***	2.65	0.347*	1.75
WMR^2	−0.098***	−2.91	−0.192***	−3.58	−0.258**	−2.51	−0.932*	−1.77
Other control variables	Control		Control		Control		Control
INTERCEPT	−0.010	−0.29	−0.086***	−2.69	−0.166	−1.57	−0.282	−0.58
INDUSTRY/ YEAR	Control		Control		Control		Control
adj_ R^2	39.74%		20.41%		7.53%		27.15%
Observations	10,337		10,337		10,337		10,209
F-(p-value)	129.61***		49.77***		16.88***		72.77***
Panel B: Robustness checks of H2 using alternative measures of discretionary accruals
Variable	(1)		(2)		(3)		(4)
	Dependent variable: DA_ΔCF		Dependent variable: DA_RET		Dependent variable: DA_PMADJ		Dependent variable: RM
	Coefficient	t-value	Coefficient	t-value	Coefficient	t-value	Coefficient	t-value
WMR	0.030**	2.03	0.052***	2.73	0.076**	1.98	0.322*	1.70
WMR^2	−0.088**	−2.48	−0.175***	−3.01	−0.209*	−1.85	−0.837*	−1.68
REGLIST	−0.011***	−3.87	−0.009***	−3.66	−0.011	−1.17	−0.108*	−1.78
WMR × REGLIST	−0.105*	−1.95	−0.187**	−2.20	−0.453***	−2.97	−1.825**	−2.30
WMR^2 × REGLIST	0.208**	2.36	0.387***	3.32	0.991***	4.78	4.741**	2.17
Other control variables	Control		Control		Control		Control
INTERCEPT	−0.007	−0.19	−0.080**	−2.55	−0.165	−1.58	−0.199	−0.44
INDUSTRY/ YEAR	Control		Control		Control		Control
adj_ R^2	39.83%		20.48%		7.57%		27.18%
Observations	10,337		10,337		10,337		10,209
F-(p-value)	123.16***		47.29***		16.11***		69.02***

Notes: ***,** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level and the year level (Petersen, 2009). All the variables are defined in Table 1.

In columns (1)–(4) of Panel A, the coefficients on WMR are significantly positive, and WMR² are significantly negatively associated with discretionary accruals. These findings support H1. In columns (1)–(4) of Panel B, the coefficients on WMR × REGLIST are significantly negative, but the coefficients on WMR² × REGLIST are positive and significant. These results provide additional support to H2, suggesting that the geographic proximity between a firm and regulatory centres attenuates inverted-U relationship between WTM and earnings management.

Following prior studies (Cohen, Dey, & Lys, 2008; Frankel, Johnson, & Nelson, 2002; Klein, 2002), we further use the absolute value of discretionary accruals (|DA|) as the proxy of earnings management to re-examine Hypotheses 1 and 2. Untabulated results show that WMR (WMR²) is significantly positively (negatively) associated with |DA|, supporting H1. Moreover, using |DA| as the dependent variable, the coefficient on WMR × REGLIST (WMR² × REGLIST) is significantly negative (positive), consistent with H2.

We also use two alternative measures of monitoring intensity to re-test H2. The first one is the distance between a firm and regulatory centres (REGLIST_S). We redefine regulatory centres as Beijing and Shanghai for firms listed on Shanghai stock exchange and Beijing and Shenzhen for those listed on Shenzhen stock exchange, and then calculate the reciprocal value of the average distance between a listed firm and the corresponding regulatory centres. The second one is the enforcement number by CSRC and stock exchanges at the provincial level (PENALTY). Untabulated results show that the coefficient on WMR × REGLIST_S (WMR × PENALTY) is significantly negative while the coefficient on WMR² × REGLIST_S (WMR² × PENALTY) is significantly positive, consistent with H2.

5.4. Robustness checks considering the endogeneity between WTM and earnings management

To address the endogeneity between WTM and earnings management, we employ two-stage least squares regression procedure (2SLS) to re-examine equation (3) and equation (4). Following extant studies, we construct the predicted model of WTM (Adams & Ferreira, 2009; Boone, Field, Karpoff, & Raheja, 2007; Campbell & Mínguez-Vera, 2008; Cheng, 2008; Srinidhi et al., 2011) in the first stage.

In Section A of Table 8, we employ eight exogenous variables of LISTAGE, SD, NONEXEDR, NUMIND, RET, MARRET, FEMALE and GDP to control for the endogeneity between WTM and earnings management. Specifically, Srinidhi et al. (2011) find that firm age is positively associated with the likelihood of women on the board, and thus we predict that the coefficient on LISTAGE is positive. Cheng (2008) argues that stock volatility influences the demand for monitoring and female are believed to increase monitoring efficiency. As a result, we expect a negative coefficient on SD. Srinidhi et al. (2011) indicate that a firm with larger demand for additional network tends to employ female directors. As a response, we predict that NONEXEDR is positively associated with WTM. Boone et al. (2007) suggest that firms with diversified product market have additional need for board independence, so we expect NUMIND has a positive coefficient. Moreover, Campbell and Mínguez-Vera (2008) and Adams and Ferreira (2009) show that financial and market performance are related to the presence of women on the board, so we expect positive signs for RET and MARRET. Hillman, Cannella, and Harris (2002) and Srinidhi et al. (2011) use the percentage of female employees in each industry as an indicator for women presentation on corporate boards, so we adopt the ratio of industry-level women employees (FEMALE) as another instrumental variable and predict a positive coefficient on FEMALE. Finally, we include provincial GDP per capital (GDP_PC) to control for economic environment. We predict that firms located in more developed regions employ more WTM. In addition, all control variables in equations (3) and (4) are included as exogenous variables to predict the ratio of WTM.

Table 8. Robustness checks using the 2SLS regression to control for potential endogeneity.

Variable		Section A: Eight instrumental variables									Section B: Two instrumental variables
		The first stage			The second stage						The first stage			The second stage
		Dependent variable: WMR			Dependent variable: DA						Dependent variable: WMR			Dependent variable: DA
		(1)			(2) H1			(3) H2			(4)			(5) H1			(6) H2
		Coefficient	t-value		Coefficient	t-value		Coefficient	t-value		Coefficient	t-value		Coefficient	t-value		Coefficient	t-value
LISTAGE		0.002***	3.16
SD		0.000	0.34
NONEXEDR		−0.001	−0.04
NUMIND		0.002	1.40
RET		0.000	0.11
MARRET		0.001	0.16
FEMALE		−0.007	−1.25								−0.007	−1.44
GDP_PC		0.026***	4.02								0.027***	4.25
WMR					0.521*	1.79		1.264**	1.99					0.467**	2.26		0.770**	2.23
WMR^2					−2.425**	−2.45		−2.272**	−2.50					−2.049***	−3.16		−3.005***	−2.74
REGLIST								−0.079**	−2.52								−0.055***	−2.76
WMR × REGLIST								−0.804*	−1.92								−0.493*	−1.91
WMR^2 × REGLIST								3.042**	2.14								2.133**	2.30
Other control variable		Control			Control			Control			Control			Control			Control
INTERCEPT		0.408***	3.59		−0.103	−1.14		−0.149	−1.29		0.406***	3.65		−0.118	−1.38		−0.108	−1.19
INDUSTRY/YEAR		Control			Control			Control			Control			Control			Control
adj_ R^2		16.79%			19.75%			19.87%			16.27%			20.58%			20.73%
Observations		10,156			10,156			10,156			10,156			10,156			10,156
F-(p-value)		42.83***			48.15***			45.98***			46.66***			51.53***			49.25***

Notes: ***,** and * represent the 1%, 5% and 10% levels of significance, respectively, for a two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level and the year level (Petersen, 2009). All the variables are defined in Table 1.

As shown in column (1), firms with longer duration after initial public offerings on A-share market and located in more developed provinces are inclined to appoint higher ratio of WTM, echoing findings in Srinidhi et al. (2011). Columns (2)–(3) report the results of the second stage regression, and results are qualitatively the same with those in Tables 5 and 6. Therefore, our conclusions are still valid when using 2SLS procedure to control for endogeneity.

In addition, we only keep two provincial variables of FEMALE and GDP in the first stage. Columns (4)–(6) report results using two instrumental variables, which remain qualitatively similar to those in columns (1)–(3), respectively.

In sum, H1 and H2 are supported after we control for the endogeneity using the 2SLS procedure.

6. Additional tests

6.1. Additional tests considering different female members

It is well-known about a unique two-tier board structure in Chinese listed firms, and thus both the board of directors and the supervisory board play a monitoring role in the firms (Firth, Fung, & Rui, 2007; Jia & Zhang, 2011; Xiao, Dahya, & Lin, 2004). As a response, we employ two additional variables, i.e., the ratio of female members of the board/supervisory board and the ratio of female senior managers, to re-examine Hypotheses 1 and 2. Table 9 presents regression results.

Table 9. Additional tests considering different female members.

Variable		Section A: Female members of the board/supervisory board						Section B: Female senior managers
		(1)			(2)			(3)			(4)
		H1			H2			H1			H2
		Coefficient	t-value		Coefficient	t-value		Coefficient	t-value		Coefficient	t-value
WMR		0.069**	2.39		0.064**	2.16		0.011**	2.11		0.011**	2.10
WMR^2		−0.374***	−2.81		−0.352***	−2.58		−0.022**	−2.50		−0.022***	−2.59
REGLIST					−0.015***	−3.05					−0.010*	−1.95
WMR × REGLIST					−0.274***	−2.59					0.004	0.13
WMR^2 × REGLIST					1.038**	2.35					−0.033	−1.03
BIG4		−0.010*	−1.83		−0.009*	−1.67		−0.010*	−1.82		−0.010*	−1.76
TENURE		−0.001	−1.07		−0.001	−1.09		−0.0001	−1.08		−0.001	−1.07
ANALYST		−0.003**	−2.26		−0.003**	−2.23		−0.003**	−2.32		−0.003**	−2.26
INSTSHR		−0.005	−0.60		−0.005	−0.67		−0.005	−0.59		−0.005	−0.65
MKT		−0.001**	−2.17		−0.000	−0.26		−0.001**	−2.18		−0.000	−0.20
LNBGS		0.005***	2.82		0.004***	2.70		0.005***	2.99		0.005***	2.89
CHGSALES		−0.009	−1.38		−0.009	−1.29		−0.009	−1.38		−0.009	−1.29
BTM		0.026***	3.06		0.027***	3.24		0.025***	2.99		0.026***	3.13
ISSUE		0.022***	4.50		0.022***	4.62		0.022***	4.48		0.022***	4.60
MZ		0.003	1.55		0.003	1.51		0.004	1.61		0.004*	1.66
LAGACCR		0.085***	4.77		0.084***	4.76		0.085***	4.73		0.085***	4.72
FIRST		0.006	0.61		0.006	0.66		0.006	0.63		0.006	0.61
DUAL		0.000	0.09		−0.000	−0.01		0.000	0.06		0.000	0.03
INDR		−0.002	−0.13		−0.003	−0.18		−0.002	−0.15		−0.003	−0.19
MANSHR		0.025	1.46		0.026	1.50		0.027	1.58		0.027	1.61
BOARD		−0.004	−0.99		−0.005	−1.12		−0.003	−0.61		−0.003	−0.70
OCF		−0.181**	−2.03		−0.181**	−2.03		−0.181**	−2.03		−0.181**	−2.03
ROA		0.539***	11.73		0.540***	11.73		0.539***	11.65		0.539***	11.60
SIZE		0.006**	2.19		0.006**	2.20		0.006**	2.20		0.006**	2.18
LEV		−0.028**	−2.02		−0.027**	−1.97		−0.029**	−2.14		−0.030**	−2.16
STATE		−0.003	−1.33		−0.003	−1.27		−0.003	−1.44		−0.003	−1.34
INTERCEPT		−0.131*	−1.91		−0.124*	−1.83		−0.136*	−1.92		−0.135*	−1.87
INDUSTRY/ YEAR		Control			Control			Control			Control
adj_ R^2		20.48%			20.55%			20.44%			20.48%
Observations		10,337			10,337			10,337			10,337
F-(p-value)		51.21***			48.73***			51.10***			48.53***

Notes: ***,** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level and the year level (Petersen, 2009). All the variables are defined in Table 1.

Columns (1) and (2) show the impact of WTM on monitoring positions on earnings management. The coefficient of WMR is significantly positive and the coefficient of WMR² is significantly negative in column (1), consistent with H1. In column (2), the interactive term WMR × REGLIST (WMR² × REGLIST) is significantly negatively (positively) associated with DA, implying that monitoring intensity attenuates the relation between WTM and earnings management. As for WTM on the monitored positions, column (3) shows that the coefficient on WMR is significantly positive and the coefficient on WMR² is significantly negative, verifying the inverted-U relationship between WTM and earnings management and thus providing support to H1. However, the interactive term between WMR (WMR²) and REGLIST is insignificantly related with DA.

In brief, Hypotheses H1 and H2 are supported using the percentage of female members of the board/ supervisory board, but only H1 stands using the percentage of female senior managers.

6.2. Additional tests considering business complexity

Extant studies argue that firm-specific characteristics impact earnings management or earnings quality (Choi et al., 2012; Hunton, Libby, & Mazza, 2006; Jo & Kim, 2007; Lee, Lev, & Yeo, 2007). As a response, we re-examine Hypotheses 1 and 2 considering corporate business complexity.

We partition our sample into two subsamples: the subsample with high business complexity and the subsample with low business complexity. For the subsample with high (low) business complexity, they have more (less) business, product, and geographic segments than the mean value of the sample. For the subsample with high business complexity, in column (1) of Panel A in Table 10, both WMR and WMR² are insignificant. Also, in column (2), we find insignificant coefficients on both WMR × REGLIST and WMR² × REGLIST. However, column (3) shows that, for the subsample with low business complexity, the coefficient on WMR is significantly positive and the coefficient on WMR² is significantly negative, providing additional support to H1. In column (4), the interactive term of WMR × REGLIST (WMR² × REGLIST) is significantly negatively (positively) associated with DA, and thus H2 is supported. H1 and H2 only stand for the subsample with low business complexity.

Table 10. Additional tests.

Panel A: Additional tests considering business complexity

Variable

Section A: High business complexity

Section B: Low business complexity

(1) H1

(2) H2

(3) H1

(4) H2

Coefficient

t-value

Coefficient

t-value

Coefficient

t-value

Coefficient

t-value

WMR

0.038

1.21

0.024

0.72

0.083***

3.51

0.073***

2.71

WMR^2

−0.150

−1.42

−0.114

−0.99

−0.211***

−2.80

−0.180**

−2.04

REGLIST

−0.014**

−2.34

−0.013*

−1.79

WMR × REGLIST

−0.181

−1.10

−0.298**

−2.49

WMR^2 × REGLIST

0.529

1.38

0.624***

2.67

Other control variables

Control

Control

Control

Control

INTERCEPT

−0.132

−1.56

−0.122

−1.44

−0.134*

−1.81

−0.134*

−1.77

INDUSTRY/ YEAR

Control

Control

Control

Control

adj_ R^2

17.92%

17.94%

23.68%

23.77%

Observations

5,510

5,510

4,827

4,827

F-(p-value)

23.69***

22.50***

29.26***

27.88***

Panel B: Additional tests considering information transparency

Variable

Section A: High information transparency

Section B: Low information transparency

(1) H1

(2) H2

(3) H1

(4) H2

Coefficient

t-value

Coefficient

t-value

Coefficient

t-value

Coefficient

t-value

WMR

−0.055

−0.41

−0.083

−0.63

0.187**

2.32

0.132*

1.79

WMR^2

0.474

1.14

0.554

1.54

−0.436**

−2.02

−0.320*

−1.65

REGLIST

−0.022

−0.72

−0.047***

−3.38

WMR × REGLIST

−1.044

−1.26

−0.760***

−3.74

WMR^2 × REGLIST

3.635

1.52

1.124**

2.16

Other control variables

Control

Control

Control

Control

INTERCEPT

−0.296*

−1.90

−0.266*

−1.68

−0.222*

−1.72

−0.194

−1.64

INDUSTRY/ YEAR

Control

Control

Control

Control

adj_ R^2

46.85%

46.99%

18.31%

18.88%

Observations

399

399

3,015

3,015

F-(p-value)

8.31***

7.92***

14.51***

14.23***

Panel C: Additional tests considering the nature of ultimate owners

Variable

Section A: SOEs

Section B: Non-SOEs

(1) H1

(2) H2

(3) H1

(4) H2

Coefficient

t-value

Coefficient

t-value

Coefficient

t-value

Coefficient

t-value

WMR

0.086**

2.52

0.077**

2.26

−0.043

−0.96

−0.031

−0.65

WMR^2

−0.246***

−2.74

−0.213**

−2.35

0.083

0.86

0.060

0.63

REGLIST

−0.031***

−3.62

−0.015

−1.04

WMR × REGLIST

−0.226*

−1.91

−0.146

−1.00

WMR^2 × REGLIST

0.448*

1.95

0.414

1.34

Other control variables

Control

Control

Control

Control

INTERCEPT

−0.098*

−1.74

−0.080

−1.52

−0.268*

−1.93

−0.246*

−1.68

INDUSTRY/ YEAR

Control

Control

Control

Control

adj_ R^2

25.14%

25.31%

17.30%

17.27%

Observations

6,971

6,971

3,366

3,366

F-(p-value)

46.89***

44.73

14.53***

13.77***

Notes: ***,** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level and the year level (Petersen, 2009). All the variables are defined in Table 1.

6.3. Additional tests considering information transparency

Next, we employ the rating of information transparency by the Shenzhen stock exchange to distinguish high and low transparency. Firms with A disclosure rating are defined as high transparency, but firms with B/C/D disclosure rating are defined as low transparency. In doing so, the number of observations is reduced to 3,414 because we can only obtain data on transparency rating for firms listed on the Shenzhen stock exchange. For the subsample with high information transparency, column (1) of Panel B shows that, the influence of WTM is insignificant. Also, in column (2), we can not find the moderating effect of regulatory intensity on the association between WTM and earnings management. For the subsample with low information transparency, in column (3) of Panel B, WMR is significantly positively associated with DA and WMR² is significantly negatively associated with DA, verifying H1 again. In column (4), the coefficient on WMR × REGLIST (WMR² × REGLIST) is significantly negative (positive). These results, token together, indicate that H1 and H2 are validated only for the subsample with low information transparency.

6.3. Additional tests considering the nature of ultimate owners

Following extant studies, (Chen, Chen, Lobo, & Wang, 2011; Du et al., 2015; Wang & Yung, 2011), we further examine whether the influence of WTM on earnings management in state-owned enterprises (SOEs) is different from that in non-SOEs. The results are presented in Panel C of Table 10.

For State-owned enterprises (SOEs), the coefficient on WMR is significantly positive and the coefficient on WMR² is significantly negative. More importantly, the interactive term WMR × REGLIST is significantly negatively associated with DA but WMR² × REGLIST is significantly positively associated with DA. Overall, H1 and H2 are valid for SOEs. However, for non-SOEs, in columns (3)–(4), WMR, WMR², WMR × REGLIST and WMR² × REGLIST are all insignificant. These results, taken together, suggest that H1 and H2 are not supported for non-SOEs.

7. Summary and conclusions

Prior studies find mixed results at best on the relation between WTM and earnings management. In this study, using a sample of Chinese listed firms, we find a non-linear relationship between WTM and earnings management, and further monitoring intensity attenuates above inverted-U relationship.

Our study has several managerial implications. First, we explore the asymmetric impacts of different proportions of WTM on earnings management. WTM has a constraining effect on earnings management only if the proportion of WTM is higher than the inflexion point. Alternatively, WTM has an opposite effect and is significantly positively associated with earnings management. Therefore, a firm aiming at improving financial reporting quality will benefit from more WTM.

Second, our findings lend important support to regulators that encourage the employment of women top managers. At present the opportunities of being promoted to serve as top managers are unequal between male and female in China’s contemporary corporations. Considering the role of women top managers in business operation and corporate governance, it is necessary for regulators to inspire Chinese firms to promote more women top managers because WTM behave more ethically in the practice of earnings management when the proportion of women is relatively higher.

Third, our study suggests that the geographic proximity between a firm and regulatory centres can serve as an external monitoring mechanism to mitigate earnings management. This finding suggests that firms far away from regulatory centres are likely to engage more in earnings manipulations since it is easier for them to evade exposure to CSRC, SSE and SZSE. Therefore, regulators should pay more attention to firms whose locations are far away from regulatory centres and further closely monitor their financial misconducts. More importantly, our findings about the moderating effect of monitoring intensity on the inverted-U relationship between WTM and earnings management imply that firms should balance between internal governance mechanisms and external monitoring mechanisms.

Finally, our findings attach the importance to distinguish firms hiring WTM only to cover the negative images from those with more WTM to improve corporate governance. In other words, WTM do not necessarily always play an efficient governance role in corporate decisions, especially for the case of lower proportion of WTM. Therefore, for WTM, it is important for investor to differentiate tokenism from an effective governance role.

Our study, of course, has its limitations. First, our study only focuses on the influence of WTM on earnings management, but we do not examine the influence of WTM on other dimensions of earnings quality such as earnings persistence and the informativeness of earnings. Second, our study is conducted based on the Chinese context, and thus our findings may not fit in well with other developed markets. Finally, we do not identify other characteristics of WTM such as age, education, and work experience, which may attenuate or reinforce the influence of WTM on earnings management. As a response to such limitations, future research can focus on the international setting to examine whether the inverted-U relation between WTM and earnings management exists. In addition, it is necessary for researchers to examine whether WTM non-linearly impact other corporate decisions.

Funding

This work was supported by The National Natural Science Foundation of China [grant number NSFC-71572162].

Notes

^* Paper accepted by Kangtao Ye.

¹ Gender discrimination such as employment discrimination, salary inequity, and the ceiling effects is still inevitable in labor market and workplaces. To some extent, rapid economy growth induces gender discrimination (Cooke, 2010, 2011; Loi & Ngo, 2009; Woodhams, Lupton, & Xian, 2009).