CEO and CFO Gender on Earnings Management:
Evidence from the Netherlands
Master’s thesis Accountancy and Controlling
University of Groningen, Faculty of Economics and Business
Author: Wieger Schoonen w.schoonen@student.rug.nl s2361795 Supervisor: Shuo Wang shuo.wang@rug.nl
Second assessor: dr. Teye Marra t.a.marra@rug.nl
Date: January 18, 2018 Words (excluding references): 10.697
1. ABSTRACT
Earnings management research is still based on male behavior, since males dominate executive top positions. Even though more and more countries have policies in place to stimulate female participation in management top functions, little research has been done about the gender influence on earnings management. The Netherlands is one of these countries that stimulate female participation in executive functions. In this thesis a gender comparison was made between CEOs and CFOs and the use of earnings management. This thesis focuses on both accrual and real earnings management, where previous research mainly focused on accrual-based earnings management. This is an addition to the field. Archival data was used and all companies in the dataset were listed between 2010 and 2016 on the Dutch AEX, AMX, or ASCX exchange. The earnings management proxy for accrual-based earnings management was calculated according to the modified Jones model (Dechow, Sloan, & Sweeney, 1995). Roychowdhury (2006) was followed for the calculation of real earnings management proxy. In the Netherlands, female CEOs tend to be more aggressive in the use of accrual-based earnings management. CFOs are conservative in accrual use, and more inclined to use real earnings management, but both not to a significant matter. These results can be influenced by the small dataset and the limited women on CEO and CFO positions in the Netherlands. This thesis contributes to the discussion on female executives in top management functions. Following the reason of Lakhal, Aguir, Lakhal, & Malek (2015), there is still a glass ceiling in the Netherlands for top management functions. This thesis can have insights for policy makers in how gender influences accounting decisions.
Keywords: CEO and CFO gender; earnings management; discretionary accruals; real earnings management
2
2. INTRODUCTION
Previous earnings management research is still based on male behaviour and little research has been done about the gender influence on earnings management. Men still dominate top executive positions, despite the fact that since the seventies, the employment rate of women rose from around 35 percent to 65 percent in 20161. It is known that women differ from men, and could employ different earnings management practices. Women are for example more risk-averse (Byrnes, Miller, & Schafer, 1999; Powell & Ansic, 1997; Croson & Gneezy, 2009) and conduct less unethical behaviour (Bersoff, 1999). Due to the growing female workforce, one can expect that more and more women will take seat on top executive positions. To stimulate this, the European Union published the Strategic Engagement for Gender Equality 2016 – 2019, to promote ‘equality between women and men in decision-making’ among other things. In addition, in 2014 the Dutch minister of Education, Culture and Science, Mrs. Bussemaker and Hans de Boer from VNO-NCW (the largest employers’ organization in the Netherlands2), initiated a database in which the names of women are collected who are seen fit for top positions in companies3. Their website, topvrouwen.nl, is useful for firms who are looking for high qualified women but struggle to find them in their own network. The ‘bigger purpose’ is to get rid of the stigma that there are not enough women fit for top positions in the Netherlands such as CEO and CFO.
It is known that the positions of chief executive officer (CEO) and chief financial officer (CFO) have considerable influence on earnings management within organizations (Bergstresser & Philippon, 2006; Jiang, Petroni, & Wang, 2010; Arun, Almahrog, & Aribi, 2015). Earnings management is a tool for managers to influence the reported earnings by choosing for certain accounting options or choosing to do more/less of certain economic activities, such as cutting research and development, and increase production (Walker, 2013). Sarbanes and Oxley legislation (SOx) makes the individuals on these positions personally responsible for the financial statements and internal control of the firm4. In light of current developments and initiatives of the EU and the Netherlands, the aim is that more and more women will take seat in these top positions in corporate organizations in the Netherlands. But, earnings management research to date is not based on female behaviour, as up until now men dominated the positions
1 https://www.cbs.nl/nl-nl/nieuws/2017/37/groei-arbeidsdeelname-afgevlakt accessed October 4, 2017 2 https://www.vno-ncw.nl/over-vno-ncw/english accessed May 30, 2017
3 https://www.nrc.nl/nieuws/2014/12/10/waar-blijven-die-vrouwen-aan-de-top-a1498626 accessed May 17, 2017 4 Section 302
3
of CEO and CFO (Ho, Li, Tam, & Zhang, 2015). Focusing more on women in this line of research would therefore be relevant. Taking this into consideration, it may be the case that women participate differently in earnings management than men.
According to previous research, women are more risk-averse and this risk aversion will be reflected in the use of accrual and real earnings management practices (Byrnes et al., 1999; Powell & Ansic, 1997; Croson & Gneezy, 2009). Also, women tend to be less involved in unethical behaviour (Bersoff, 1999). Next to that, previous research shows that females influences earnings management which results in a lower earnings management level (Peni & Vähämaa, 2010; Arun et al., 2015; Lakhal et al., 2015). In this thesis a comparison is made between gender and earnings management use. Based on previous research my expectation is that female executives are less involved in earnings management, both accrual-based earnings management as well as real earnings management. This is done by using archival data from the period 2010-2016.
The main findings are as follows. Overall, Dutch CEOs and CFOs are conservative in their earnings management practices. Female CEOs use accruals on a significantly different manner compared to male CEOs. For CFO gender, there was no significant difference found for the earnings management proxies. Analysis of the merged CEO and CFO data resulted in a significantly less cash flow from operations for women compared to men. Taking these results in mind, gender has influence on earnings management practices. Especially CEO gender has an influence, since no earnings management proxy was significant for CFOs. The results were checked with five additional and robustness checks after the general analysis. First, the data is winsorized to prevent skewed results as a result of outliers. Second, this thesis focuses on the absolute value of accruals for the main analysis. Therefore an extra analysis was done on positive and negative accruals. Third, the Altman’s z-score to indicate bankruptcy is calculated for additional analysis. A low z-score is highly associated with the use of both types of earnings management. Fourth, an additional analysis was performed on hypothesis 3 to check for robustness. Finally, an extra analysis was conducted for the firms with the highest sales growth, but are not associated with accrual or real earnings management.
This research will contribute to existing literature in several ways. First, it will provide more evidence on the relationship between women and earnings management. As mentioned earlier, research on earnings management was mainly focused on top management functions which are men dominated. Second, a comparison can be made between insider and outsider countries as depicted by Leuz, Nanda, & Wysocki (2003). Third, it will provide more evidence
4
of earnings management and female CEOs and CFOs from an insider country. A comparison can be made with the research of Lakhal et al. (2015) which provided evidence from France. Fourth, this thesis contributes to the discussion on gender diversity in corporate boards. This can be useful for legislators and managers, since this thesis provides insights on how gender influences accounting decisions. Fifth, this thesis follows Walker’s (2013) definition of earnings management and therefore focuses on accruals and real earnings management. Therefore, this thesis expands previous research of Peni and Vähämaa (2010), Arun et al. (2015), and Lakhal et al. (2015) by focusing as well on real earnings management next to accrual-based earnings management. Previous research only focused on the latter. Sixth, few papers make a comparison between CEOs and CFOs and earnings management. There is plenty of research about roles of these two functions and earnings management, but a comparison is seldom a topic within one paper.
The structure of this paper is as follows. In the next chapter relevant literature will be discussed and the hypothesis are presented. The relevant literature is divided in the chapters gender, earnings management, and CEOs and CFOs. In chapter 4 the research design will be elaborated on and chapter 5 will show the main results and the results from the additional and robustness checks. In chapter 6 the results are discussed and will be compared to the hypothesis from chapter 4 and previous literature. Finally, chapter 7 contains the references used.
3. LITERATURE REVIEW AND HYPOTHESIS
3.1 Gender
According to research of Khazanchi (1995), unethical actions and behaviour are better recognized by women than by men (Bersoff, 1999), and women are less likely to conduct unethical behaviour (Bersoff, 1999). In terms of risk taking, men take more risk than women (Byrnes et al., 1999; Powell & Ansic, 1997; Croson & Gneezy, 2009). Differences in corporate risk taking has been recorded by Faccio, Marchica, and Mura (2015). They found that a female CEO takes less risk in corporate decision making than a male CEO. These firms ran by female CEOs and CFOs ‘have lower leverage, less volatile earnings, and a higher chance of survival
than otherwise similar firms run by male CEOs’ (p. 193). Broverman, Vogel, Broverman,
Clarkson, and Rosenkrantz (1972) found several stereotypical perceived differences between men and women, one of which is that men ‘almost always act as a leader’ and women mostly do not. Social preferences of women are subject to the situation, which means that their choices
5
are more dependent on the circumstances in which the decision is made (Croson & Gneezy, 2009). On top of that they find that women are less inclined to competition.
Earnings management is less when the audit committee has a higher degree of gender diversity (Thiruvadi & Huang, 2011). More diverse boards hold their CEOs more accountable when stock prices are performing badly (Adams & Ferreira, 2009). Adams and Ferreira (2009) also find evidence that women take more time for monitoring, which could reduce earnings management. Moreover, Srinidhi, Gul, and Tsui (2011) found evidence that more women in the board can lead to a higher earnings quality. Discretionary accruals, which reflects voluntary management choices in accruals, are a sign of reduced earnings quality (Chen, Lin, & Lin, 2008). Krishnan and Parsons (2008) found the same result and that the higher net earnings are not a result of earnings management or a decrease in quality of the earnings. Next to that, earnings quality of female CFOs is higher when compared to men, because female CFOs frequently have lower accruals than men (Barua, Davidson, Rama, & Thiruvadi, 2010). More risk is taken by boards that only consist of men, which endorses that women take less risk (Baixauli-Soler, Belda-Ruiz, & Sanchez-Marin, 2015). When accrual quality is compared between sexes, then women deliver higher accrual quality (Barua et al., 2010), since the height of accruals of women are lower than men. Peni and Vähämaa (2010) found no evidence for a difference in earnings management magnitude and the gender of the CEO, but Ho et al. (2015) did find a significant relation between female CEOs and conservative accounting. It can be concluded that women have a positive influence on the quality of financial statements, since they differ in personal characteristics compared to men. Consequently, differences between men and women are expected to be found in this thesis.
Previous research of Peni and Vähämaa (2010), Arun et al. (2015), and Lakhal et al. (2015) researched the influence of women in top positions on earnings management practices across various jurisdictions, namely in the United States, the United Kingdom, and France respectively. This research seeks to extend these papers to the Netherlands, which is a different country compared the US and UK (outsider economies), but is similar to France (insider economy) according to Leuz et al. (2003). Leuz et al. (2003) argues that investor protection within a country has influence on the level of earnings management based on an international comparison of countries. Countries are divided in two types: insider (for example the Netherlands, Germany, France & Sweden) and outsider economies (e.g. the UK, US & Australia). Insider economies have smaller stock markets, more ownership concentration, less disclosure and a legal system with less force than outsider economies (Leuz et al., 2003). Insider
6
economies are split in two groups based on the strength of the legal system. Leuz et al. (2003) conclude that in outsider economies earnings management occurs less than in insider economies. The research of Peni and Vähämaa (2010), Gavious et al. (2012), Arun et al. (2015), and Lakhal et al. (2015) already show a difference between insider and outsider countries. Peni and Vähämaa (2010) conducted a research about female executives and earnings management in Standard & Poor’s (S&P) 500 firms. This research in the US found that female CFOs follow more conservative strategies of earnings management, however they did not find a relationship between CEO gender and earnings management. This is an indication that CFOs have more influence on the financial reporting process than CEOs. However, Gavious, Segev, & Yosef (2012) did find evidence for reduced earnings management and CEOs and CFOs in high-technology firms. Arun et al. (2015) focus primarily on CFOs because of their accountability for the financial statements. They conducted a similar research as Peni and Vähämaa (2010), but with data from the UK. Arun et al. (2015) found that organizations that have more (independent) female directors, have a higher chance of following a more conservative earnings management policy. In these situations, earnings management is used more often to decrease earnings instead to increase earnings. Evidence from France states that female CEOs and CFOs do not have much influence on earnings management when she is the only female in the board (Lakhal et al., 2015). However, earnings management reduces when with three or more women in the board, which is in line with the critical mass theory (Lakhal et al., 2015). The critical mass theory applies also for board, because when women are with three or more in a board, then they do not feel the need to find a place within the group (Kramer et al., 2006). This is confirmed by a quote from a female executive (Kramer et al., 2006, p. 21): ‘Three is like three
legs on a stool. Strong. It is clear you are not there because of gender but because of your talent.’ From these papers, one can conclude that female directors are more dominant regarding
earnings management in outsider countries compared to insider countries. However there is only evidence from one insider country on which the comparison is made. Francis, Hasan, Park, and Wu (2015) found that female CFOs are related to more conservative accounting policies and the researchers directly relate this to risk aversion. Their results show that the relation between conservative accounting of females CFOs is stronger when their firms face higher risks of for example litigation, default, systematic, and management turnover. Ho et al. (2015) found a similar result that accounting tends to be more conservative for female CEOs when risks increases. This effect is stronger for companies with more rigid corporate governance and which are smaller in size.
7 3.2 Earnings management
Earnings management is a way for managers to influence reported earnings of a company. This is mostly used as a tool to achieve expectations from stakeholders or achieve targets for personal gain. Walker (2013, p. 446) defines earnings management as: ‘the use of
managerial discretion over (within GAAP) accounting choices, earnings reporting choices, and real economic decisions to influence how underlying economic events are reflected in one or more measures of earnings.’ Two types of earnings management techniques can be identified
in the definition of Walker (2013), namely accrual-based and real earnings management (AEM and REM) (Cohen & Zarowin, 2010). Both are ways to manage earnings, but significantly differ on one point: cash flow. Accrual-based earnings management has no influence on the cash flow of an organization, whereas real earnings management does (Cohen & Zarowin, 2010). Roychowdhury (2006) uses the next definition of real earnings management: ‘Management
actions that deviate from normal business practices, undertaken with the primary objective of meeting certain earnings thresholds’ (p. 336). Examples are making extra provisions and
change the timing of write-offs. According to the same author, managers do not solely rely on accrual management, despite of the costs involved. Reasons to do so are that regulators and/or the auditor may subject these practices to extra investigation. Second, it is not possible to manipulate real earnings management at year end when accrual management is deficient (Roychowdhury, 2006). For these reasons, this thesis focuses on real earnings management as well.
There are two reasons for earnings management: there is not always a clear benchmark to which the earnings can be compared to, and managers have certain motives to show certain results (Dunmore, 2008). First, regarding the benchmark argument, Dunmore gives the example of the depreciation of buildings. He states that a manager is unable to objectively determine for how long a building will be used. This can be ten, twenty or fifty years. Possibly even longer. The decision to depreciate over X years is a decision on judgment in the domain of earnings management: usually, the longer the period, the lower the annual depreciation and thus the higher the profit. This also works vice versa.
The second reason for earnings management is discussed next. There are several motives to conduct earnings management. Walker (2013) summarizes three of them: contractual, capital market, and third-party motives. First, contractual motives to conduct earnings management are motives to serve the self-interest of managers. For example, rewards and compensation can be based on financial performance of a firm (Xie, Davidson, & DaDalt,
8
2003). When this is the case, a manager can influence the presentation of the financial performance of an organization by earnings management, leading to rewards and compensation for that individual. The influence of CEOs and CFOs has been covered by research in the past. Research of Bergstresser and Philippon (2006) shows that CEOs are more likely to conduct earnings management when their compensation is more closely linked to the share price, e.g. share options and stocks CEOs already own. Jiang et al. (2010) researched this similar topic and found that CFOs are more likely to conduct earnings management when their compensation is more closely linked to the share price. However, Feng, Ge, Luo, & Shevlin (2011) find that earnings management by CFOs is mostly not initiated by the prospect of personal financial gain, but by coerce of the CEO. This implies that the CEO is the highest function, although the CFO is mainly responsible for the financial statements. Complying with debt covenants is another contracting motive (Walker, 2013). Failing to comply with debt covenants generally results in negative outcomes, such as revising the loan into a more strict contract (Nini, Smith, & Sufi, 2012), a going-concern opinion, and a rise in audit fee (Bhaskar, Krishnan, & Yu, 2017). The results of Bhaskar et al. (2017) are in line with risk estimation of auditors. Covenant violations increase business risks and auditors will respond to mitigate these risks, which can result in a higher audit fee as a result of more audit work performed that needs to be performed. The second type of motive named by Walker (2013) is capital motives, which contains managing share prices in certain instances, e.g. initial public offerings (IPO), seasoned equity offering (SEO), and takeover bids. Ball and Shivakumar (2008) find evidence that companies are more conservative in their reporting when planning an IPO. Venture capitalists that invest prior the IPO and have a good reputation, reduce the use of earnings management (Lee and Masulis, 2011). Shivakumar (2000) argues that in the period before a SEO, accruals are used to manage earnings upwards, because shareholders naturally make a discount in earnings when a SEO is announced. Shareholders do this because they assume that earnings are overstated prior to a SEO and cannot estimate correctly what the volume of earnings management actually was. Next to contractual motive, debt covenants are a capital motive as well (Beneish, Press, & Vargus, 2012). When there is a credible risk to a technical default, accruals are used to mitigate that risk. Especially when this technical default may have legal consequences. Finally, third party motives. These are related about the opinions of stakeholders (e.g. suppliers, employees, and the government) about the financial health of the company (Walker, 2013). Walker (2013) gives the example of companies managing earnings downwards or smooth earnings to make sure third parties do not view earnings as ‘monopoly rents’. Firms are followed by analysts. The more analysts that follow a firm, the less earnings are managed by managers (Yu, 2008). CSR
9
minded firms use less earnings management practices (Chih, Shen, & Kang, 2008). The same research also finds that earnings smoothing and earnings management practices are less used to mitigate losses and declines in earnings. But when these practices are used, they tend to be relatively aggressive. However, countries that have a strong legal enforcement system – outsider economies – can counteract this. The Netherlands is an insider economy as described by Leuz et al. (2003), so is theoretically less able to withstand aggressive earnings management. In the definition of Walker mentioned previously, earnings management is a result of managerial decision making and is mostly viewed as an opportunistic tool for managers, but it can also be used beneficially for the firm. It may reduce agency costs and can be used to improve the reflection fundamental value in the earnings (Jiraporn, Miller, Yoon, &Kim, 2008; Huang, Zhang, Deis, & Moffitt, 2009). Even firm value (measured by Tobin’s Q) can be enhanced by the use of earnings management (Jiraporn et al., 2008). Huang et al. (2009) find evidence that Tobin’s Q increases when financial derivatives are used for earnings smoothing. On the other hand is SOx legislation to reduce earnings management, which implies the negative view on this topic by US Congress (Jiraporn et al., 2008). As a result, real earnings management became a more frequently used post-SOx and accrual based earnings management use declined (Cohen, Dey, & Lys, 2008). Jiang et al. (2010) even found that post-SOx the positive relation between CEO and CFO incentives and earnings management based on accruals is no longer there. Which confirms the effectiveness of SOx in reducing discretionary accruals. Burgstahler and Dichev (1997) found evidence for the use of earnings management to smooth earnings to prevent earning declines and going into the red. In this way, earnings management is used as a tool to prevent negative variations in the earnings. Hsieh, Bedard, & Johnstone (2014) relate earnings management to the overconfidence of CEOs. They found evidence that after the implementation of SOx, overconfident CEOs did not feel more restricted by the legislation than pre-SOx. However, they did find that earnings management was reduced in that period, probably because of the change of behavior of investors post-SOx.
The article of Walker (2013), as discussed in the previous paragraph, focuses on the downsides of earnings management, since motives to prevent negative outcomes are discussed. Bergstresser and Philippon (2006) and Cohen and Zarowin (2010) even use the words ‘earnings
manipulation’ in their papers. Following the main body of previous research, earnings
management is argued to have more negative effects than positive. This is – for this reason – also the viewpoint of this thesis.
10 3.3 CEOs and CFOs
CEOs and CFOs are argued to be the most influential positions regarding earnings management. These are the two top functions in the firm, but the literature is not unanimous either on who - CEO or CFO - is the biggest driver in this. Bergstresser and Philippon (2006) found that when CEO pay is more closely linked to the share prices, earnings management increases, but Jiang et al. (2010) found an even stronger effect for CFOs. Further, CFOs have the responsibility over the firm’s financial reporting and therefore have the ability to manage earnings (Feng et al., 2011). Geiger and North (2006) found that the appointment of a new CFO resulted in a significant decline discretionary accruals, especially when the new CFO had not worked for the firm before. This indicates that a CFO has considerable influence on the height of the accruals and operates a stricter accrual use policy. The opposite is true for a newly appointed CEO from outside the firm. They tend to use accruals more regularly to increase earnings to survive in the beginning years of their tenure (Kuang, Qin, & Wielhouwer, 2014). The findings of Hazarika, Karpoff, and Nahata (2012) and Liu, Wei, and Xie (2016) support this result. They find that earnings management is used more often when the likelihood of getting fired as CEO increases. After all, the chance of getting fired decreases with time, because the CEO can build backing in the firm during his/her tenure (Goyal and Park, 2002).
The degree of earnings management not only depends on the individual, also corporate governance has its influence. Xie et al. (2003) find that earnings management will be reduced when directors in boards have more experience in the corporate world and are more independent. Klein’s (2002) research concludes with a similar result: independent boards and audit committees reduce the amounts of abnormal accruals. Earnings management tends to reduce when higher standards for corporate governance are employed (Bekiris & Doukakis, 2011). Earnings management is better monitored when the audit committees and board are larger, e.g. have more members (Ghosh, Marra, & Moon, 2010). This is possibly the result of the combined knowledge of these executives. Monitoring of earnings management is also improved when the board is more independent (García-Meca & Sánchez-Ballesta, 2009).
Based on the previous paragraphs, the expectation is that females are less involved in earnings management, both in accrual-based earnings management as well as in real earnings management. Women tend to be more risk averse (Faccio et al., 2015) and are less involved in unethical behaviour (Bersoff, 1999). The first two hypothesis are accrual-based earnings management hypothesis. The real earnings management hypothesis are hypothesis 3 and 4.
11
H1: Female CEOs are more conservative in the use of accrual-based earnings management in the Netherlands compared to their male counterparts.
H2: Female CFOs are more conservative in the use of accrual-based earnings management in the Netherlands compared to their male counterparts.
H3: Female CEOs are more conservative in the use of real earnings management in the Netherlands compared to their male counterparts.
H4: Female CFOs are more conservative in the use of real earnings management in the Netherlands compared to their male counterparts.
4. RESEARCH DESIGN
This chapter will be organized as follows: first, the collection of data will be depicted in paragraph 4.1. The following two paragraphs after that discuss accrual-based earnings management and real earnings management respectively. In the last paragraph, 4.5, the control variables used are elaborated on.
4.1 Data
The sample consists of Dutch companies listed on the AEX, AMX and the AScX for the period 2010 to 2016. During this time period, no big legislative changes have occurred since then in the Netherlands. IFRS is obligatory since 2005 for companies that trade equity or debt securities5. Financial information was retrieved from Datastream and CEO and CFO gender data was collected from the BoardEx database. Companies are divided in nine industries based on the FTSE Industry Classification Benchmark. I follow Roychowdhury (2006) and exclude companies from that operate in regulated industries, banks and financial institutions (FTSE Industry Classification Benchmark 7: utilities and 8: financials). As for the remaining industries, no minimum number of observations were required. The Dutch dataset is small, therefore every observation is very valuable for the analysis of the data. This thesis follows Walker’s (2013) definition of earnings management and therefore focuses on accruals and real earnings management.
5 http://www.ifrs.org/Use-around-the-world/Documents/Jurisdiction-profiles/European-Union-IFRS-Profile.pdf accessed May 24, 2017
12
4.2 Accrual-based earnings management
Due to the parallels between this thesis and the papers of Peni and Vähämaa (2010), Arun et al. (2015), and Lakhal et al. (2015), the research design will have its similarities. First, the modified Jones model (Dechow et al., 1995) is used to calculate the discretionary accruals, since the latter is a tool for management to use discretion in reported earnings (Cohen & Zarowin, 2010). Dechow et al. (1995) conclude that the modified version of the Jones model works best for determining discretionary accruals. This will be done by the next cross-sectional regression, equation (1):
𝑁𝐷𝐴𝐶𝐶𝑖𝑡 = 𝛼1 1
𝑎𝑡𝑖,𝑡−1+ 𝛼2 ((∆𝑠𝑎𝑙𝑒𝑠𝑖,𝑡− ∆𝑟𝑒𝑐𝑡𝑖,𝑡)/𝑎𝑡𝑖,𝑡−1) + 𝛼3(𝑝𝑝𝑒𝑔𝑡𝑖,𝑡/𝑎𝑡𝑖,𝑡−1 ) (1)
Where NDACC = estimated non-discretionary accruals for firm 𝑖 in year 𝑡 ; 𝛼1, 𝛼2, 𝑎𝑛𝑑 𝛼3 = firm-specific parameters; ∆𝑠𝑎𝑙𝑒𝑠𝑡 = change in revenues in year 𝑡 for firm 𝑖; ∆𝑟𝑒𝑐𝑡𝑡= change in net receivables in year 𝑡 for firm 𝑖; 𝑝𝑝𝑒𝑔𝑡𝑡 = gross property plant and equipment for firm 𝑖 in year 𝑡 -1 scaled by total assets at 𝑡 -1.
Total accruals are calculated by use of equation (2), where 𝑎𝑐𝑡 = current assets, 𝑙𝑐𝑡 = current liabilities, 𝑐𝑎𝑠ℎ = cash and cash equivalents, 𝑑𝑙𝑐 = debt in current liabilities, 𝑑𝑝 = depreciation and amortization.
𝑇𝐴𝐶𝐶𝑖𝑡 = (∆𝑎𝑐𝑡𝑖,𝑡− ∆𝑙𝑐𝑡𝑖,𝑡− ∆𝑐𝑎𝑠ℎ𝑖,𝑡 + ∆ 𝑑𝑙𝑐𝑖,𝑡− 𝑑𝑝𝑖,𝑡 )/𝐴𝑖,𝑡−1 (2) Total accruals minus the non-discretionary accruals results in the discretionary accrual amount (ABN_Accr). The absolute value of discretionary accruals (ABS_DA) is calculated as well, because it can be used to manage earnings upwards as well as downwards (Dechow and Dichev, 2002).
4.3 Real earnings management
Where previous research of Peni and Vähämaa (2010), Arun et al. (2015), and Lakhal et al. (2015) only focuses on accrual-based earnings management this thesis focuses on real earnings management as well. Roychowdhury (2006) mentions three methods that can influence real activities: (1) manipulation of sales by alteration of credit terms, increasing discounts and/or quicker timing, (2) lower reported cost of goods sold (COGS) by increased production, and (3) decrease discretionary expenses. The real earnings management proxies will be calculated according to the method of the same author. With this method the abnormal cash flow from operations, production, and discretionary expenses are calculated to determine the height of the management in real activities.
13
The following equations are used to calculate the real earnings management proxies: 𝐶𝐹𝑂𝑡 𝑎𝑡𝑡−1 = 𝛼0+ 𝛼1 1 𝑎𝑡𝑡−1+ 𝛼2 𝑠𝑎𝑙𝑒𝑠𝑡 𝑎𝑡𝑡−1 + 𝛼3 ∆𝑠𝑎𝑙𝑒𝑠𝑡 𝑎𝑡𝑡−1 + 𝜀𝑡 (3) 𝑃𝑅𝑂𝐷𝑡 𝑎𝑡𝑡−1 = 𝛼0+ 𝛼1 1 𝑎𝑡𝑡−1+ 𝛼2 𝑠𝑎𝑙𝑒𝑠𝑡 𝑎𝑡𝑡−1 + 𝛼3 ∆𝑠𝑎𝑙𝑒𝑠𝑡 𝑎𝑡𝑡−1 + 𝛼4 ∆𝑠𝑎𝑙𝑒𝑠𝑡−1 𝑎𝑡𝑡−1 + 𝜀𝑡 (4) 𝐷𝐼𝑆𝐸𝑋𝑃𝑡 𝑎𝑡𝑡−1 = 𝛼0+ 𝛼1 1 𝑎𝑡𝑡−1+ 𝛼2 𝑠𝑎𝑙𝑒𝑠𝑡−1 𝑎𝑡 (5)
Abnormal cash flow from operations is consequently calculated by the realized CFO minus the calculated by equation (3). Equation (4) shows the calculation of the abnormal production. Discretionary expenses at the normal level are calculated, but to overcome too low residuals by heightened sales in year 𝑡, delayed sales are used as a function for discretionary expenses (equation 5). These three proxies are compared with the actual level. The difference is the abnormal level and results in abnormal cash flow from operations (ABN_CFO), abnormal production (ABN_PROD), and abnormal discretionary expenditures (ABN_DISEXP). ABN_PROD and ABN_DISEXP were used to calculate COMBINED_RAM in equation (6).
𝐶𝑂𝑀𝐵𝐼𝑁𝐸𝐷_𝑅𝐴𝑀 = 𝐴𝐵𝑁_𝑃𝑅𝑂𝐷 − 𝐴𝐵𝑁_𝐷𝐼𝑆𝐸𝑋𝑃 (6)
ABN_CFO is likely to decrease, ABN_PROD to increase and ABN_DISEXP will decrease when real earnings are managed to upward reported earnings (Roychowdhury, 2006). COMBINED_RAM is calculated following this reasoning, and will be higher when activities are managed.
4.4 Models
The control variables are briefly discussed here, and are more elaborated on in paragraph 4.5. CEOGender, CFOGender, and CEOCFOGender are the most important variables of this thesis. These are dummy variables for the gender of the CEO and CFO, where the dummy variable is 1 when female and 0 when male. Several control variables will be used to control for the possible influence on the amount of earnings management. The following control variables are used: LEV, the leverage of the firm is calculated by dividing long term debt with total assets; SIZE, natural logarithm of total assets of the firm; MB, the market to book ratio of the firm; SGROWTH, is the percentage of sales growth; LOSS, this dummy variable if 1 the firm did not make a profit and 0 if the firm did; ROA, return on assets of the firm. Section 5.4 is further dedicated to control variables. SMALLP (small profit) is a dummy variable: 1 if a firm has a small profit, and 0 if not. STOCKISSUANCE is a control variable since earnings are sometimes managed upward before stock offerings (DuCharme, Malatesta, & Sefchik, 2004).
14
INDUSTRY is used as a control variable to control for industry specific characteristics (MacKay & Phillips, 2005).
Next, the main regression models for the earnings management proxies of this thesis are depicted in equation (7) and (8) where 𝐸𝑀𝑡 is one of the following:𝐴𝐵𝑁_𝐶𝐹𝑂𝑡,𝐴𝐵𝑁_𝑃𝑅𝑂𝐷𝑡,𝐴𝐵𝑁_𝐷𝐼𝑆𝐸𝑋𝑃𝑡, 𝐶𝑂𝑀𝐵𝐼𝑁𝐸𝐷_𝑅𝐴𝑀𝑡, 𝐴𝐵𝑆_𝐴𝑐𝑐𝑟𝑡 or 𝐴𝐵𝑆_𝐷𝐴𝑡.These equations include the thesis’ most important variables (CEO, CFO, and CEOCFOGender) and the control variables.
𝐸𝑀𝑡 = 𝛼1+ 𝛼2𝐶𝐸𝑂𝐺𝑒𝑛𝑑𝑒𝑟𝑖,𝑡+ 𝛼3𝐶𝐹𝑂𝐺𝑒𝑛𝑑𝑒𝑟𝑖,𝑡+ 𝛼4𝐿𝐸𝑉𝑖,𝑡+ 𝛼5𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛼6𝑀𝐵𝑖,𝑡+ 𝛼7𝑆𝐺𝑅𝑂𝑊𝑇𝐻𝑖,𝑡+ 𝛼8𝐿𝑂𝑆𝑆𝑖,𝑡+ 𝛼9𝑅𝑂𝐴𝑖,𝑡+ 𝛼10𝑆𝑀𝐴𝐿𝐿𝑃 + 𝛼11𝑆𝑇𝑂𝐶𝐾𝐼𝑆𝑆𝑈𝐴𝑁𝐶𝐸 + ∑𝑛−1𝑘=1𝛼𝑘𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖𝑘+ 𝜀𝑖,𝑡 (7) 𝐸𝑀𝑡 = 𝛼1+ 𝛼2𝐶𝐸𝑂𝐶𝐹𝑂𝐺𝑒𝑛𝑑𝑒𝑟𝑖,𝑡+ 𝛼3𝐿𝐸𝑉𝑖,𝑡+ 𝛼4𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛼5𝑀𝐵𝑖,𝑡+ 𝛼6𝑆𝐺𝑅𝑂𝑊𝑇𝐻𝑖,𝑡+ 𝛼7𝐿𝑂𝑆𝑆𝑖,𝑡+ 𝛼8𝑅𝑂𝐴𝑖,𝑡+ 𝛼9𝑆𝑀𝐴𝐿𝐿𝑃 + 𝛼10𝑆𝑇𝑂𝐶𝐾𝐼𝑆𝑆𝑈𝐴𝑁𝐶𝐸 + ∑𝑛−1𝑘=1𝛼𝑘𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖𝑘+ 𝜀𝑖,𝑡 (8) 4.5 Control variables
Several proxies were used, which were briefly mentioned in paragraph 4.4. The proxy for debt covenant violation is LEV, since that earnings management is more likely to occur in leveraged firms (Chih et al., 2008; Elayan, Li, & Meyer, 2008). Debt covenants are contracting motive for earnings management (Walker, 2013). Contradicting evidence has been found on size of the company (SIZE). According to Richardson (2000) and Chih et al. (2008) bigger firms want to manage their earnings upward. However, Watts and Zimmerman (1990) have found evidence that larger companies want to do the opposite. According to them bigger firms want to manage their earnings downward. An example is an ‘earnings bath’, where companies manage their earnings downwards if it is unlikely that earnings reach the earnings threshold that will grant managers their bonus. An earnings bath will create a buffer for next year’s earnings, and a better prospect of receiving a bonus in that year. SIZE is calculated as a natural logarithm of the total assets of the firm. Firms with high sales growth and market to book ratio (MB) tend to be more indulged in earnings management practices (Chih et al., 2008). This to keep up current performance. Expected is that firms which make a loss, use earnings management to reduce earnings (Healy, 1985; DeAngelo, DeAngelo, & Skinner, 1994). The control variable LOSS is used to control for this effect. Return on assets (ROA) is used as an indicator for the performance of the organization. Industry is used as a control variable to control for industry specific characteristics. Industries 7: utilities and 8: financials in the FTSE
15
Industry Classification Benchmark are removed for this reason to prevent skewed results (Roychowdhury, 2006). A firm has a SMALLP (small profit) if the outcome of income before extraordinary items scaled by lagged total assets is between zero and 0.01 (Jeanjean and Stolowy, 2008). SMALLP is a dummy variable: one if small profit, and zero if not. Burgstahler and Dichev (1997) found that firms use earnings management to prevent them to write in the red and have a loss. STOCKISSUANCE is a control variable since earnings are sometimes managed upward before stock offerings (DuCharme et al., 2004).
5. RESULTS
5.1 Descriptive statistics
The total sample by industry is presented in Table 1. The industries were divided following the Industry Classification Benchmark (Equity) of FTSE Russel. The continuous variables were winsorized at 2.5 percent to control for outliers. The CEO and CFO data was merged to increase data availability. This is depicted in the last two columns. In totality there were 262 observations before the exclusion of industry 7: utilities and 8: financials. During the sample period, there were no female CEOs or CFOs active in the deleted industries, so no female executive data was lost when these industries were removed. The control variable SMALLP was not used, since no firm in the data had a small profit. Compared to other studies, this thesis has relatively few observations. However, this is explained by the fact that the Netherlands is a small country with a small exchange. In total, 7.3 percent of the CEOs and CFOs observations were from a female executive. Female CFOs accounted for 11 percent in the total CFO population and for CEOs this was 4 percent.
Table 1: Total sample by industry
FTSE INDUSTRY FEMALE CEO MALE CEO FEMALE CFO MALE CFO FEMALE CEO/CFO MALE CEO/CFO 1: Basic Materials 0 22 6 16 6 38 2: Industrials 5 70 7 68 12 138 3: Consumer Goods 0 46 2 44 2 90 4: Health Care 0 14 1 13 1 27 5: Consumer Services 5 47 0 52 5 99 6: Telecommunications 0 10 3 7 3 17 7: Utilities 0 5 0 5 0 10 8: Financials 0 5 0 5 0 10 9: Technology 0 33 7 26 7 59 Total 10 252 26 236 36 488
16
Table 2 presents the descriptive statistics for the sample after the exclusion of FTSE industry 7 and 8. As mentioned in the chapter 5, CEOGender, CFOGender, and CEOCFOGender are dummy variables: 1 if female and 0 if male. The real activities management proxies are the variables ABN_CFO, ABN_PROD, ABN_DISEXP, and COMBINED_RAM and are calculated according to the Roychowdhury (2006) model. The accruals-based earnings management proxy is calculated according to the modified Jones model (Dechow et al., 1995). This results in the variable ABN_Accr which are the discretionary accruals.
Table 2: descriptive statistics full sample
n=170 Quantiles
VARIABLE MEAN Q1 MEDIAN Q3
CEOGender 0.05 0.00 0.00 0.00 CFOGender 0.15 0.00 0.00 0.00 CEOCFOGender 0.19 0.00 0.00 0.00 ABN_CFO -0.01 -0.05 -0.01 0.04 ABN_PROD 0.02 -0.10 0.03 0.18 ABN_DISEXP -0.01 -0.16 -0.02 0.10 ABN_Accr -0.02 -0.11 -0.05 0.03 ABS_DA 0.14 0.04 0.09 0.16 COMBINED_RAM 0.02 -0.17 0.06 0.32 LEV 0.18 0.06 0.15 0.27 SIZE 14.71 13.22 14.70 16.25 MB 2.09 1.18 1.77 2.54 SGROTWH 0.01 -0.06 0.00 0.06 LOSS 0.16 0.00 0.00 0.00 ROA 0.06 0.03 0.07 0.10 INDUSTRY 4.07 2.00 3.00 5.00 STOCKISSUANCE 0.00 0.00 0.00 0.00
Regarding table 2, one can see that in the Netherlands real earnings management is used, since it has a low mean of 0.02, it is not used on a large scale. Accruals are a popular method to manage earnings as well. ABS_DA has a mean of 0.14. ABN_Accr mean is negative, so accruals are mostly used to manage earnings downward. The means of the other earnings management proxies are close to zero, which implies that Dutch top executives on average are not very willing to use these methods.
Tables 3, 4, and 5 are an extended overview of the descriptive statistics. The sample is divided in six groups and are depicted in these three tables. They contain the statistics of female and male CEOs, CFOs, and the merged CEO and CFO data in respectively table 3, 4, and 5. These univariate analysis already show the differences between the various earnings management and control variables.
17
ABN_Accr (p-value: 0.0342) and ABS_DA (p-value: 0.0424) are significant for the gender of the CEO. However, a remark has to be made: these two variables are linked to each other. ABS_DA is the absolute value of ABN_Accr, and the prior is used for this main analysis. ABN_Accr will be used for an additional check in paragraph 5.3. So, the use of accruals is both reflected in these variables. As for real earnings management, female CEOs use real earnings management methods in a different manner than men. Woman are more inclined to use ABN_CFO, while men prefer ABN_PROD and ABN_DISEXP. However, the combined measure is not significant for CEO gender (COMBINED_RAM, p-value: 0.2725). What is significant for CEO gender, is the use of accruals. A positive accrual mean indicates a more aggressive accrual use policy and vice versa (Arun et al., 2015). Female CEOs make more use of accruals (ABS_DA, p-value: 0.0424), and use accruals to manage earnings upwards (ABN_Accr, mean: 0.13) while men use accruals to lower earnings (ABN_Accr, mean: -0.02). For the control variables the market to book ratio (MB, p-value: 0.0443) seems to be significant for CEO gender. Female CEOs operate in firms with a higher market to book ratio then male CEOs and are thus more often ‘overvalued’. Next, CFOGender tends to have less influence on the earnings management practices deployed, since no earnings management proxy is significant. The control variable return on assets (ROA, p-value: 0.0176) is significant for CFO gender. Male get a better return on their investment compared to women.
For the merged data, ABN_CFO is the only earnings management proxy that is significant. With a p-value of 0.0550 it is significant at the 10 percent level. Which means that the overall preference of female CEOs and CFOs is to use cash flow from operations to manage earnings. Next to that, the control variables SIZE (p-value: 0.0642) and return on assets (ROA, p-value: 0.0511) are significant for the merged CEO and CFO data.
5.2 CEO and CFO gender and earnings management proxies
In this paragraph the main analysis will be conducted. The influence of the gender of CEOs and CFOs on earnings management proxies will be analysed and compared with the hypothesis. Regarding the accruals, the absolute value of accruals is used primarily, since earnings can be managed upwards and downwards (Dechow and Dichev, 2002). Paragraph 5.3 elaborates more extensively on ABN_Accr.
The results of the multivariate analysis between all variables in table 6 are partly in line with the results of the univariate analysis for the earnings management proxies. Accrual-based earnings management variables – ABN_Accr and ABS_DA – are still significant for the gender of the CEO, and are significant at the five percent and ten percent level respectively. Accruals
18
are positively correlated with female CEOs, and women tend to make more use of accruals than men. This is not supporting H1 since it is in the opposite direction than expected. This is also contradicting evidence with the result of Barua et al. (2010). They argue that females are not keen on using accruals and therefore accrual height is lower for women than men. But male CEOs seem to be more conservative in accrual use compared to females. There is no significant relation between female CFOs and accrual-based earnings management, which is not in line with the expectation of H2. Hypothesis 2 is rejected as well. In the multivariate analysis, COMBINED_RAM became significant for CEOGender at the 5 percent level. The sign is negative, which mean that female CEOs are more conservative in the use of real earnings management. This is in line with H3. An additional analysis is performed for this hypothesis in the next paragraph. For CFOGender, the real earnings management proxy is not significant, and therefore H4 is rejected. In the univariate analysis, CEOCFOGender was significant for ABN_CFO, however in the multivariate analysis, and is still significant at the 10 percent level in the multivariate analysis. Tables 7 and 8 contain the regression results of CEO and CFO gender, the merged CEO and CFO data, the accrual-based and real earnings management proxies, and the control variables. The results of this regression differ slightly from the univariate and the multivariate analysis. Abnormal cash flow from operations is not significant for CEOCFOGender were it was significant in the previous analysis, and COMBINED_RAM is no longer significant for CEOGender.
19
Table 3: descriptive statistics female and male CEO
test of difference between mean (median) Female CEO Male CEO n=8 Quantiles n=162
VARIABLE MEAN Q1 MEDIAN Q3 MEAN Q1 MEDIAN Q3 p-values
ABN_CFO -0.03 -0.08 -0.01 0.03 -0.01 -0.05 -0.01 0.04 0.3842 (1.000) ABN_PROD -0.04 -0.14 -0.12 0.16 0.02 -0.10 0.03 0.15 0.3444 (0.469) ABN_DISEXP 0.09 0.03 0.05 0.16 -0.01 -0.17 -0.03 0.10 0.1967 (0.004) ABN_Accr 0.13 -0.08 -0.04 0.36 -0.02 -0.12 -0.05 0.03 0.0342 (1.000) ABS_DA 0.25 0.04 0.08 0.49 0.13 0.04 0.09 0.16 0.0424 (1.000) COMBINED_RAM -0.14 -0.18 -0.15 0.02 0.02 -0.17 0.08 0.32 0.2725 (0.147) LEV 0.23 0.17 0.27 0.28 0.17 0.06 0.15 0.26 0.3073 (0.030) SIZE 15.57 15.29 15.76 5.90 14.67 13.11 14.46 6.28 0.1650 (0.030) MB 3.10 1.97 3.04 3.43 2.04 1.18 1.71 2.49 0.0443 (0.147) SGROTWH 0.02 -0.00 0.01 0.02 0.01 -0.06 -0.00 0.06 0.7405 (0.469) LOSS 0.13 0.00 0.00 0.00 0.17 0.00 0.00 0.00 0.7581 (0.756) ROA 0.07 0.04 0.08 0.09 0.06 0.02 0.07 0.10 0.7537 (0.469) INDUSTRY 3.50 2.00 3.50 5.00 4.10 2.00 3.00 5.00 0.5177 (0.838) SMALLP 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.7016 (0.700) STOCKISSUANCE -0.00 -0.00 -0.00 -0.00 0.00 0.00 0.00 0.00 0.4888 (0.469)
20
Table 4: descriptive statistics female and male CFO
test of difference between mean
(median)
Female CFO Male CFO
n=25 Quantiles n=145
VARIABLE MEAN Q1 MEDIAN Q3 MEAN Q1 MEDIAN Q3 p-values
ABN_CFO -0.03 -0.05 -0.02 0.01 -0.00 -0.05 -0.01 0.04 0.1057 (0.279) ABN_PROD 0.06 -0.06 0.07 0.22 0.01 -0.11 0.02 0.14 0.1873 (0.279) ABN_DISEXP -0.01 -0.22 -0.08 0.10 -0.00 -0.16 0.00 0.10 0.8568 (0.279) ABN_Accr -0.06 -0.11 -0.07 -0.02 -0.01 -0.11 -0.04 0.03 0.2466 (0.516) ABS_DA 0.10 0.04 0.08 0.13 0.14 0.06 0.09 0.17 0.2715 (0.829) COMBINED_RAM -0.06 -0.16 0.21 0.43 -0.01 -0.17 0.04 0.29 0.5286 (0.279) LEV 0.20 0.07 0.15 0.28 0.17 0.06 0.15 0.27 0.3708 (0.829) SIZE 15.12 13.42 15.76 16.52 14.64 13.11 14.59 15.98 0.2184 (0.279) MB 1.76 0.90 1.38 2.33 2.15 1.22 1.83 2.56 0.2266 (0.279) SGROTWH 0.01 -0.03 0.03 0.05 0.01 -0.06 -0.01 0.07 0.9061 (0.130) LOSS 0.20 0.00 0.00 0.00 0.16 0.00 0.00 0.00 0.6090 (0.606) ROA 0.03 0.01 0.03 0.06 0.07 0.03 0.08 0.11 0.0176 (0.001) INDUSTRY 4.08 2.00 2.00 6.00 4.07 2.00 3.00 5.00 0.9841 (0.482) SMALLP 0.04 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.3543 (0.351) STOCKISSUANCE 0.01 -0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.2996 (0.279)
21
Table 5: descriptive statistics merged data CEOs and CFOs
test of difference between mean
(median)
Female CEO/CFO Male CEO/CFO
n=33 Quantiles n=137
VARIABLE MEAN Q1 MEDIAN Q3 MEAN Q1 MEDIAN Q3 p-values
ABN_CFO -0.03 -0.06 -0.02 0.02 -0.00 -0.04 -0.01 0.04 0.0550 (0.332) ABN_PROD 0.04 -0.11 0.05 0.21 0.01 -0.10 0.02 0.14 0.5013 (0.561) ABN_DISEXP 0.01 -0.17 0.01 0.10 -0.01 -0.16 -0.02 0.10 0.5974 (0.561) ABN_Accr -0.01 -0.11 -0.05 -0.02 -0.02 -0.12 -0.04 0.03 0.9272 (0.561) ABS_DA 0.14 0.04 0.08 0.14 0.14 0.04 0.09 0.17 0.9335 (0.846) COMBINED_RAM 0.01 -0.17 0.08 0.30 0.02 -0.17 0.06 0.32 0.9820 (0.846) LEV 0.21 0.07 0.17 0.28 0.17 0.05 0.15 0.24 0.1771 (0.175) SIZE 15.23 14.33 15.76 6.35 14.59 13.07 14.39 6.16 0.0642 (0.033) MB 2.09 0.94 1.66 2.88 2.09 1.22 1.79 2.53 0.9921 (0.846) SGROTWH 0.01 -0.01 0.01 0.05 0.01 -0.06 -0.01 0.07 0.9492 (0.081) LOSS 0.18 0.00 0.00 0.00 0.16 0.00 0.00 0.00 0.7695 (0.768) ROA 0.04 0.01 0.04 0.07 0.07 0.03 0.08 0.11 0.0511 (0.012) INDUSTRY 3.94 2.00 2.00 6.00 4.10 2.00 3.00 5.00 0.7426 (0.604) SMALLP 0.03 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.5329 (0.530) STOCKISSUANCE 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.5773 (0.561)
23 Table 7: regression results
ABN_CFO ABN_PROD ABN_DISEXP COMBINED_RAM ABS_DA ABN_Accr
Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient
CEOGender -0.0321 -0.0272 0.0919 -0.1171 0.1194** 0.1613** CFOGender -0.0107 0.0346 -0.0086 0.0405 -0.0447 -0.0653 LEV 0.0236 -0.0038 -0.1861 0.1485 -0.0490 -0.3209*** SIZE -0.0080*** 0.0056 0.0073 0.0031 -0.0060 0.0184** MB 0.0139*** -0.0350*** 0.0126 -0.0461* 0.0046 0.0083 SGROWTH -0.0099 0.0401 0.0147 -0.0124 0.3544*** -0.2164** LOSS -0.0192 0.1066* -0.0770 0.2138* -0.0121 -0.1059* ROA 0.0216** 0.0836 -0.3030 0.5423 -0.1564 -0.3691 INDUSTRY 0.0076*** -0.0067 -0.0070 -0.0021 0.0025 -0.0067 STOCKISSUANCE -0.1242 -0.2385 -0.3992 -0.6144 0.5978 0.9511* constant 0.0395 0.0106 -0.0531 -0.0152 0.1156 -.2507* N 170 170 170 170 170 170 R² 0.3555 0.1547 0.0410 0.0561 0.2127 0.2273
*** Correlation is significant at the 1% level ** Correlation is significant at the 5% level * Correlation is significant at the 10% level
24
Table 8 regression results
ABN_CFO ABN_PROD ABN_DISEXP COMBINED_RAM ABS_DA ABN_Accr
Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient
CEOCFOGender -0.0160 0.01931 0.0162 0.0015 -0.0041 -0.0092 LEV 0.0233 -0.00048 -0.185 0.1461 -0.0465 -0.03174*** SIZE -0.0079*** 0.0056 0.0073 0.0031 -0.0007 0.0184** MB 0.0134*** -0.0362*** 0.0146 -0.0492** 0.0078 0.0128 SGROWTH -0.0107 0.0377 0.01866 -0.0186 0.3609*** -0.2076 LOSS -0.0200 0.1045* -0.0737 0.2086 -0.0067 -0.0984 ROA 0.2107* 0.0694 -0.2800 0.5062 -0.1188 -0.3162 INDUSTRY 0.0078*** -0.0064 -0.0075 -0.0012 0.0016 0.0055 STOCKISSUANCE -0.0942 -0.1521 -0.0735 -0.2322 0.6284 0.1074 constant 0.0404 0.0134 -0.0576 -0.0081 0.1083 -0.2609 N 170 170 170 170 170 170 R² 0.3526 0.1508 0.0339 0.0508 0.1795 0.1837 *** Correlation is significant at the 1% level
** Correlation is significant at the 5% level * Correlation is significant at the 10% level
25 5.3 Additional and robustness checks
Some additional and robustness checks were performed to test the results. First, the continuous variables were winsorized at 5 percent to control for outliers. This did not lead into differences in the univariate analysis. For the multivariate analysis, some differences emerged. For CEOGender both accrual proxies are no longer significant, but ABN_DISEXP became significant at the 10 percent level. For CEOCFOGender the ABN_CFO proxy is significant now at the 10 percent level, and this was not the case before. The regression results of CEOGender, CFOGender and CEOCFOGender lost some of its explanatory power and largely remained the same.
Second, in the main analysis, Dechow and Dichev (2002) were followed in the use of the absolute value of accruals. This because earnings can be managed upwards (positive value of accruals) and downwards (negative accrual value) (Arun et al., 2015). As an additional analysis, the dataset was split in these two groups to test if gender has an influence on upwards or downward managing of accruals. The expectation was that women are more conservative and thus manage earnings downward. The results of this analysis are depicted in table 9 and 10. A positive coefficient is expected between CEOGender, CFOGender, and CEOCFOGender and negative abnormal accruals. For negative ABN_Accr there is a positive coefficient CEOGender (0.0551), which is in line with the expectation, however it is not significant. For CFOGender and CEOCFOGender. The coefficients are in line with the expectation, but both are not significant either. For positive abnormal accruals a negative coefficient is expected. The analysis for positive ABN_Accr, CEOGender, and CFOGender yields the following results. The coefficients for CFOGender (-0.1283) and CEOCFOGender (0.1091) and negative ABN_Accr are in the expected direction but not significant. The most remarkable outcome is the coefficient between CEOGender and ABN_Accr. This is contrary to the expectation positive and significant at the 1% level. But with only 55 observations of positive accruals, outliers can have an impact.
26 Table 9: regression results
ABN_Accr ABN_Accr Negative Positive Coefficient Coefficient CEOGender 0.0551 0.6729*** CFOGender 0.0016 -0.1283 LEV -0.1109* -0.2716 SIZE 0.0077 0.0028 MB -0.0089 0.008 SGROWTH -0.3008*** 0.3514** LOSS -0.0548 -0.1546* ROA -0.1845 -1.0283* INDUSTRY 0.0032 0.0220** STOCKISSUANCE -0.7985 0.6569 constant -0.1756 0.1332 N 115 55 R² 0.2368 0.5722
*** Correlation is significant at the 1% level ** Correlation is significant at the 5% level * Correlation is significant at the 10% level
Table 10: regression results
ABN_Accr ABN_Accr Negative Positive Coefficient Coefficient CEOCFOGender 0.0136 0.1091 LEV -0.1080* -0.3016 SIZE 0.0074 0.0133 MB -0.0071 0.0031 SGROWTH -0.2948*** 0.2634 LOSS 0.0541 -0.1058 ROA -0.1851 -0.4263 INDUSTRY 0.0030 0.0180 STOCKISSUANCE -0.8999 0.2342 constant -0.1746** -0.0499 N 115 55 R² 0.2239 0.2773
*** Correlation is significant at the 1% level ** Correlation is significant at the 5% level
27
Third, an extra control variable was calculated: the Altman z-score. The z-score is an indicator for distressed firms. The lower the score the higher the indication for bankruptcy. A value of 1.8 or lower, means a company is in the bankruptcy danger zone, while a score above 3 indicates a very healthy company. The z-score is calculated by the next formula (Altman, 2000):
𝑧 − 𝑠𝑐𝑜𝑟𝑒 = 1.2𝑥1+ 1.4𝑥2+ 3.3𝑥3+ 0.6𝑥4+ 1.0𝑥5
Where 𝑥1= working capital/total assets, 𝑥2= retained earnings/total assets, 𝑥3= earnings before interest and taxes/total assets, 𝑥4= market value of equity/book value of total liabilities, and 𝑥5= sales/total assets.
The analysis is twofold: first the effect of the z-score on the earnings management proxies will be analyzed, and after that the distressed firms are subjected to a second analysis for gender influence. First, the total sample. The descriptive statistics show that the average z-score of female executives is significantly lower. With means of 1.51, 1.75, and 1.69 for female CEOs, CFOs and the merged data compared to means of 3.23, 3.39, and 3.50 respectively. One can conclude that female CEOs and CFOs operate their firms in dangerous z-score zones, while men operate in the more healthy zones. Z-scores are significant in the multivariate analysis for ABN_CFO (0.2621***) and ABN_PROD (-0.1580**). The coefficients in the regression analysis are significant for these two proxies as well. A company with a low z-score is very likely to manage earnings by cash flows from operations (coefficient: 0.0116***) and abnormal production (-0.0163**). Accrual earnings management is not influenced by this z-score in the regression (ABS_DA: -0.0084; ABN_Accr: -0.0025). The multivariate and the regression results match, and suggest a significant change in abnormal cash flow from operations and production when the z-score lowers.
Next, only the distressed firms were analysed to compare gender. There firms with a z-score lower than 2 were analysed to only include firms that are on the brink of financial distress. This results in 63 observations. Only CEOGender had a significant influence on earnings management proxies in the multivariate analysis. The proxies ABN_PROD (-0.2956**) ABN_DISEXP (0.3470***), ABN_Accr (0.2297*), and COMBINED_RAM (-0.3313***) are significant for CEOGender in distressed firms. This indicates that female CEOs working in a distressed firm, use earnings management significantly more than men to alter their earnings. In the regression, for both CEO and CFO gender abnormal discretionary expenditures and COMBINED_RAM are significant at the 1 percent level. Female CEOs and CFOs in distressed firms use earnings management more to alter their earnings. However, female CEOs are only
28
active in two sectors and these sectors specifically had difficulty climbing out of the recession after the crisis in 20086. Poor economic performance seems to enhance earnings management, and can be driven by all the three motives (contractual, capital market, & third-party motives) for earnings management mentioned in paragraph 3.2 (Walker, 2013).
Fourth, an extra analysis was performed to check the significance of CEOGender for accruals and COMBINED_RAM. Next to that, female CEOs only operate in two of the nine identified sectors. Therefore, the same analysis were performed, but only for these two industries: industrials and consumer services. The sample contains 91 observations. Female CEOs still use significantly (coefficients: ABN_Accr: 0.2904*** & ABS_DA: 0.2873***) more accruals than men. CEOGender was not significant for COMBINED_RAM. Hypothesis 3 has to be rejected based on this finding. On the contrary, CFOGender was at the 1 percent level with a coefficient of 0.3337. Which implies that female CFOs use real earnings management techniques less than their male counterparts in these sectors to manage earnings downwards.
The last additional analysis is the analysis of the top 25 percent high growth companies. These high growth firms are the top 25 percent firms with the highest sales growth. These firms have an incentive to keep up the growing sales. There are no significant relations between executive gender and high growth firms, but there is a significant relation between high growth firms and ABS_DA (0.2128***). This indicates that firms which are in the top 25 percent of growing firms, tend to use accruals significantly more, but the direction of these accruals is not significant.
6. DISCUSSION AND CONCLUSION
In this last section, the results are compared with the hypothesis and with previous research. Prior research mainly concludes that female CEOs and CFOs are more conservative in their earnings management practices (Peni and Vähämaa, 2010; Arun et al, 2015; Francis et al., 2015; Ho et al., 2015). Next to that, female gender characteristics indicated that females are more conservative (Francis et al., 2015), risk averse (Byrnes et al., 1999; Powell & Ansic, 1997; Croson and Gneezy, 2009), and are less involved in unethical behaviour (Bersoff, 1999). In line with these findings, the hypothesis were drafted.
6
29
One result of this thesis is that accruals are positively correlated with female CEOs, so women tend to make more use of accruals than men. This is contradicting evidence with the result of Ho et al. (2015) who found a significant relation between female CEOs and conservative accounting. But female CEOs in the dataset only operated in two of the nine defined sectors: industrials and consumer services. Based on the additional analysis in paragraph 5.3, a better conclusion is that female CEOs use significantly more accruals than male CEOs in these two sectors. For CFOs, females operate in all sectors but the excluded sectors; the CFO sample is spread more evenly. Male CFOs use accruals more, but no significant difference was found. This is in line with the result of Barua et al. (2010). They argue that female CFOs are not keen on using accruals and therefore accrual height is lower for women. A possible explanation for CEO dominance comes from Feng et al. (2011). Feng et al. (2011) argue that earnings management is initiated by the coerce of the CEO, despite the CFOs accountability for the financial statements. The CEO has the highest position in the firm, and therefore the main driver for earnings management according to that paper. The result of this thesis support the findings of Feng et al. (2011). The significant earnings management proxies for CEOs are consistent over the various analysis performed. However, the regression results show that the significant association between CEOCFOGender and accruals does not sustain. Implying that the function of CEO is not that dominant as expected. Arun et al. (2015) argue that the CFO has more influence, since the person on that position is accountable for the financial statements, and focus for that reason on that executive position. But this thesis did not find any evidence to support Arun et al. (2015). The CEOGender was significant at the 5 percent level for COMBINED_RAM. Female CEOs are more conservative in real earnings management, but this was only the case in the multivariate analysis. In the univariate and the regression analysis this was not the case.
Only hypothesis 3 can be accepted based on the main analysis. However, an additional analysis was required since the result was only significant in the multivariate analysis, and female CEOs only operated in two industries: industrials and consumer services. The additional analysis of COMBINED_RAM of these two industries showed that CEOGender is not significant for this proxy. Therefore, hypothesis 3 has to be rejected, since it was previously only significant for the total sample and for the inclusion of industries were no female CEOs operated in.
Hypothesis 2, and 4 are rejected since the relation is not significant. Hypothesis 1 is significant, but is not in the expected direction. The expectation was that female CEOs were more conservative in the use of accruals. However, results show that female Dutch CEOs are
30
more inclined to use them. This is remarkable, but the results may be not be generalizable to future female CEOs in the Netherlands. The dataset only included 10 female CEO observations compared to 252 observations for male CEOs. When female participation rises or when the desired gender ratio of the EU and the Dutch initiative is realised, results could be very different.
According to Francis et al. (2015) and Ho et al. (2015) women are more conservative when risk increases. Therefore an additional control variable was created: Altman’s z-score. A z-score below 1.8 indicates a distressed firm. This is used as a proxy for increased default risk. The results of this thesis are in line with these two papers. A low z-score is significantly associated with the use of abnormal cash flow from operations and abnormal production. For the analysis of distressed firms only, female CEOs make more use of earnings management compared to their male counterparts. However, this seems to be driven by poor economic factors. The industries were female CEOs were operating in, had more difficulty to climb out of the crisis of 20087. Even during the time of the used sample.
Lakhal et al. (2015) are the only authors that wrote about executive gender and earnings management with data from an insider country. As mentioned before, the focus was on the critical mass theory and female executives on the board with respect to earnings management. They conclude that more women in the board reduces earnings management and even more if the chair is female. Next, they conclude that there is a glass ceiling in France. This because the relations are not significant, but are in the expected direction. So Lakhal et al. (2015) argue that there are not enough women to make the relations significant. In this thesis this is for example the case for the relation of CEOGender and ABN_CFO in the multivariate regression (table 4). It has a p-value of 0.1057, and could be significant with a larger proportion of female CFOs. Following this reasoning, then there is a glass ceiling as well in the Netherlands. The CFO gender relations to both earnings management techniques and the relation of CEO gender on real earnings management, are in the expected direction and not significant. There are too little female CEOs and CFOs to make the relation significant.
This research examined the role of CEO and CFO gender on several earnings management proxies which were either accrual-based or real earnings management proxies in the Netherlands. A limited dataset was available, since the Netherlands is a small country and relatively few women work in a top management function. Female CEOs have a more
7
