Amsterdam Business School
CFO Gender and Accrual-based Earnings Management
After the Passage of the Sarbanes-Oxley Act
Name:
Carlijn Zijleman
Student number:
10867899
Date of final version:
21-06-2015
Word count:
14.436
Thesis supervisor:
Dr. W.H.P. Janssen
MSc Accountancy & Control, variant Control
Statement of Originality
This document is written by student Carlijn Zijleman who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This research has been the first that provides empirical evidence that the passage of the Sarbanes-Oxley has had a different impact on female CFOs compared to male CFOs, in the context of accrual-based earnings management. Prior research provides evidence that female CFOs adopt more conservative earnings management strategies: discretionary accruals are negatively related to female CFOs. Other research shows that after the passage of SOX, corporate executives adopt more conservative earnings management strategies: discretionary accruals declined after the passage of SOX. Considering this evidence, I predicted that female CFOs adopted even more conservative earnings management strategies after the passage of SOX.
The performance-matched Jones model has been used to estimate discretionary accruals, the proxy for earnings management. Subsequently, discretionary accruals have been regressed on FEMALE, SOXPERIOD and FEM*SOX and eight control variables, to investigate whether there is an incremental effect for female CFOs of the passage of SOX on discretionary accruals. I expected that discretionary accruals would have declined more for female CFOs compared to male CFOs after the passage of SOX, because they would have been more deterred to engage in earnings management compared to male CFOs. However, in contrast with this expectation, I find that discretionary accruals increased more for female CFOs compared to male CFOs after the passage of the Sarbanes-Oxley Act. The findings indicate that female CFOs adopt less conservative earnings management strategies in the post-SOX period compared to the pre-SOX period.
Table of Content
1. Introduction 1 1.1 Background 1 1.2 Research Question 3 1.3 Results 3 1.4 Contribution 4 1.5 Structure 52. Literature Review & Hypothesis Development 6
2.1 Earnings Management 6
2.1.1 Definition of Earnings Management 6
2.1.2 Reasons for Earnings Management 7
2.1.3 Earnings Management Mechanisms 8
2.1.3.1 Accrual-based Earnings Management 8
2.1.4 Constraints to reduce Earnings Management 9
2.2 CFO Influence on Accrual-based Earnings Management 10
2.3 Gender Differences 11
2.3.1 Gender Differences in General 11
2.3.2 Gender Differences in Economic Decision-making 12
2.3.3 Gender Differences in Financial Reporting 13
2.4 Passage of the Sarbanes-Oxley Act 15
2.5 Hypothesis Development 18
3. Methodology 20 3.1 Data and sample Description 20
3.2 Method 21
3.2.1 Performance-matched Jones model 22 3.2.2 Regression Models 24 4. Results 28 4.1 Descriptive Statistics 28 4.1.1 CFO Gender 28 4.1.2 SOX period 30 4.2 Regression Results 33 5. Summary and Conclusion 37
5.2 Conclusion 40
5.3 Limitations 41
References 43
1
1
Introduction
In the last decade the number of female executives is increasing. Especially a rise in female CFOs is noticeable. The rise of female executives resulted in research into gender differences in economic decision making and financial reporting. Prior research finds that female CFOs adopt different financial reporting strategies. In 2002, after major corporate and accounting scandals, the Sarbanes-Oxley Act was signed into law to enhance the quality of financial reporting. Evidence has been found that corporate executives changed their financial reporting strategies after the passage of SOX. These findings provided an interesting basis for me to investigate whether the passage of SOX has had a different impact on female CFOs compared to male CFOs.
1.1
Background
Prior research finds evidence that women may act differently than men do. The findings of research from diverse areas suggest that women tend to be less aggressive and more cautious in economic decision making (e.g. Byrnes et al. 1999; Huang and Kisgen 2013), tend to be more risk-averse in general and in the context of investment choices (e.g. Riley and Chow 1992; Byrnes et al. 1999;), tend to be less (over)confident (e.g. Bengtssona et al. 2005; Niederle and Vesterlund 2007) and are more likely to be in compliance with rules and regulations (Baldry 1987) compared to men. These differences between women and men provided an interesting basis for research in the accounting environment.
Prior research finds evidence that female CFOs adopt different earnings management strategies compared to male CFOs. Earnings management occurs when corporate executives intervene in the external financial reporting process with the purpose to obtain some private gain or use judgement in financial reporting to either mislead stakeholders about the real financial performance or influence contractual outcomes (Schipper 1989; Healy and Wahlen 1999). Earnings management can be used to manage earnings up or down.
Firms with female CFOs are associated with income-decreasing discretionary accruals, indicating more conservative earnings management strategies (Peni and Vähämaa 2010). Francis et al. (2014) also find strong support that female CFOs are more risk-averse and adopt more conservative financial reporting policies. This evidence indicate that female CFOs are more cautious in recognizing gains compared to losses. Liu et al. (2014) show that female CFO
2
have significantly lower discretionary accruals, lower total accruals, lower abnormal production costs and higher abnormal discretionary expenditures. Their empirical evidence also suggests that female CFOs act more conservatively compared to male CFOs. Other research show that companies with female CFOs have a lower level of absolute abnormal accruals and accrual estimation errors, indicating a higher level of earnings quality (Barua et al. 2010).
In July 2002 the Sarbanes-Oxley Act (SOX), a U.S. federal law, was signed into law. SOX was a reaction to a number of major corporate and accounting scandals. These major scandals decreased the investor’s confidence in the security market. The Act was passed to enhance corporate governance and professional responsibility and improve the quality of financial reporting in order to restore investor confidence (Jain and Rezaee 2006). SOX has been considered as the most far-reaching Securities legislation since the Securities Acts of 1933 and 1934 (Romano 2004).
Section 302 of SOX requires CEOs and CFOs of publicly traded companies to certify that the financial statements and disclosures are reliable and fairly present the operations and financial condition of the company (Gordon et al. 2006). After the passage of SOX, CEOs and CFOs became personally accountable for the reliability of financial reporting. They face significant penalties for a knowingly false certification, e.g. imprisonment, fines and barred by the SEC from serving as CEO/CFO (Rezaee 2005).
Prior research finds evidence that suggests that the level of accrual-based earnings management declined significantly after the passage of the Sarbanes-Oxley Act in 2002 (e.g. Cohen et al. 2008; Li et al. 2008). Cohen et al. (2008) show a significant increase in accrual-based earnings management just before the passage of SOX and a significant decrease after the passage of SOX. Li et al. (2008) find evidence that suggests that accrual-based earnings management is constrained by the passage of SOX and that the quality of financial statement information is enhanced in the post-SOX period. However, Ghosh et al. (2010) do not find evidence suggesting that the level of earnings management declined after the passage of SOX. Other research shows significantly lower signed discretionary accruals after the passage of SOX (Lobo and Zhou 2006; Zang 2012). This evidence indicates that corporate executives use more conservative earnings management strategies after the passage of SOX. In contrast, although Wang et al. (2011) do find evidence for a decrease in the level of absolute discretionary accruals, they do not find evidence for lower signed discretionary accruals in the post-SOX period.
3
1.2
Research Question
Was the impact of the passage of the Sarbanes-Oxley Act the same for male and female CFOs? Or were female CFOs more affected by the passage of SOX, due to their more risk averse and conservative nature? I will investigate whether discretionary accruals declined more for female CFOs just after the passage of SOX compared to male CFOs. This objective resulted in the following research question:
Did the passage of the Sarbanes-Oxley Act have a bigger impact on female CFOs compared to male CFOs, in the context of accrual-based earnings management?
I predict that the passage of SOX has had a bigger impact on female CFOs, due to their more risk-averse and conservative nature. CEOs and CFOs face imprisonment up to twenty years and fines up to one million dollars for knowingly false certifications. I expect that these significantly increased penalties deterred female CFOs more to engage in earnings management compared to male CFOs.
I choose to investigate differences in CFO gender instead of CEO gender. One argument for this is that prior research shows that earnings management is more affected by CFOs compared to CEOs. Jiang et al. (2008) investigate the influence of CEO and CFO equity incentives and find that earnings management is more increasing in CFO equity incentives than in CEO equity incentives. This evidence may indicate that earnings management is more affected by CFOs compared to CEOs.
Another reason why I choose to investigate CFOs is that the number of female CFOs is higher than the number of female CEOs in the S&P500. At the end of 2006, ExecuComp reported more than three times as many female CFOs than female CEOs for S&P500 companies (source: WRDS database). The number of female CFOs at the S&P500 Index companies increases also faster than the number of female CEOs (Frier and Hymowitz 2013). One of the limitations of prior research in this area in the U.S. was that the number of female executives was very low.
1.3
Results
The performance-matched Jones model, as applied in Kothari et al. (2005), has been used to estimate the discretionary component of total accruals. This model decomposes total accruals into nondiscretionary and discretionary accruals. Discretionary accruals can be used as a proxy
4
for earnings management. I choose to investigate the signed version of discretionary accruals because the direction of earnings management is also relevant in this research. The sample consists of all companies of the S&P500 Index for which data were available in Compustat. The sample period extends from 1998 to 2005 and is divided into two periods: the pre-SOX period (1998-2001) and the post-SOX period (2002-2005).
I find that discretionary accruals are significantly lower for female CFOs compared to male CFOs, even after controlling for other factors. This evidence is consistent with prior research that shows more conservative earnings management strategies for female CFOs (e.g. Peni and Vähämaa 2010; Francis et al. 2014; Liu et al. 2014) and my first hypothesis that female CFOs and discretionary accruals are negatively related. I also find that discretionary accruals are not significantly lower after the passage of the Sarbanes-Oxley Act. This evidence is in contrast with prior research (Lobo and Zhou 2006; Zang 2012) and my second hypothesis that the passage of SOX and discretionary accruals are negatively related.
Finally, I find that for female CFOs there is a significant incremental effect of the passage of SOX on discretionary accruals. However, in contrast with my third hypothesis, the incremental effect of the passage of SOX for female CFOs is positively related to discretionary accruals. Discretionary accruals increased more for female CFOs after the passage of the Sarbanes-Oxley act compared to male CFOs. I predicted that discretionary accruals would have
declined more for female CFOS after the passage of SOX compared to male CFOs. My finding
indicates that female CFOs became less conservative after the passage of SOX.
1.4
Contribution
Prior research investigates the effect of CFO gender and the impact of the passage of Sarbanes-Oxley Act on earnings management separately. To the best of my knowledge, I am not aware of any empirical research that investigates whether the passage of SOX has had a different impact on female CFOs compared to male CFOs. This research will be the first that investigates whether the possible change in discretionary accruals after the passage of SOX differs significantly between female and male CFOs.
From an academic point of view, this research contributes to the existing literature that indicates a relation between CFO gender and earnings management by providing evidence regarding the relation between the two. Although some research has been undertaken in this area, the literature is limited. This research also contributes to the existing literature that
5
indicates that the passage of SOX has had an impact on earnings management. There is only little research that investigates the impact of the passage of SOX on signed discretionary accruals. Most prior research uses the absolute value of discretionary accruals to measure the impact of the Sarbanes-Oxley Act. In this research signed discretionary accruals has been used as a proxy for accrual-based earnings management.
As mentioned before, I did not find any empirical work that investigates whether the passage of the Sarbanes-Oxley Act has had a different impact on female CFOs compared to male CFOs in the context of accrual-based earnings management. No research has been undertaken to investigate whether there is an incremental effect for female CFOs on discretionary accruals after the passage of SOX. This research is the first empirical work that provides evidence regarding the incremental effect of the passage of SOX for female CFOs on discretionary accruals.
From a societal point of view, this research is relevant as well. The number of female CFOs has increased significantly in the last decade. In 2012 the number of female CFOs increased by 35% (Frier and Hymowitz 2013). Although the number of female CFOs is increasing, the world of CFOs is still dominated by man. Some countries imposed (or threatened to impose) percentage quota’s for female board directors. This research could contribute to the discussion about female CFOs (or more in general, corporate executives) in society. The research will also contribute to the debate of the effectiveness of SOX. Several articles suggests that the passage of SOX did not necessarily caused the differences between the pre- and post-SOX period (e.g. Coates IV 2007; Ball 2009).
1.5
Structure
The structure of this thesis is as follows. In the next chapter, I will provide a literature review geared towards the development of the hypotheses. Important issues will be explained and discussed in order to develop the hypotheses. In the third chapter, the research method applied in this research will be explained. The results of empirical research will be presented and explained in the fourth chapter. In the final chapter I will present a summary, I will discuss my results in the conclusion and I will explain the limitations of this research.
6
2
Literature Review and Hypothesis Development
In the literature review, I will introduce the most important issues in this thesis geared to the development of the hypotheses. The most important issues are earnings management, CFO influence on accrual-based earnings management, differences between women and men in different settings and the passage of the Sarbanes-Oxley Act. After all the necessary information has been provided I will develop the hypotheses.
2.1
Earnings Management
In this section of the literature review, the concept earnings management will be introduced. First, I will provide a definition of earnings management and explain reasons to engage in earnings management. Subsequently, I will describe the two different forms of earnings management, i.e. accrual-based and real earnings management. Finally, I will discuss the constraints to reduce earnings management.
2.1.1 Definition of Earnings Management
One of the first definitions of earnings management was provided by Schipper (1989). She described earnings management as disclosure management in the sense of a purposeful intervention in the external financial reporting process, with the extent of obtaining some private gain, as opposed to merely facilitating the neutral operation of the process (Schipper 1989). According to Healy and Wahlen (1999), earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company, or to influence contractual outcomes that depend on reported accounting numbers.
In their definition, Healy and Wahlen (1999) state that managers alter financial reports to either mislead stakeholders about firm performance or to influence contractual outcomes that depend on accounting information. Schipper (1989) states that earnings management is a purposeful intervention, opposed to facilitating the neutral operation of the process. Both definition reflect that earnings management impacts the reliability and transparency of financial reporting. Earnings become a less reliable measure of firm performance. Several studies use earnings management as a proxy for earnings quality (e.g. Krishnan and Parsons 2008; Ye et
7
al. 2010; Srinidhi et al. 2011). A high level of earnings management indicates a lower level of earnings quality.
2.1.2 Reasons for Earnings Management
Healy and Wahlen (1999) provide several reasons to engage in earnings management in their study. One reason addresses the capital market. Most investors and financial analysts use accounting information to value the stock price of a firm. Influencing the (short-term) stock price can be an incentive for managers to manipulate earnings. Capital market motives are likely to be high in periods prior to management buy-outs and equity offers for example. If a firm reveals high earnings in a period prior to such events, it is likely this will be reflected in the stock price, which will lead to a higher sales price.
Another reason for earnings management is associated with contracts (Dechow 1994). In most firms, accounting information is used to monitor and regulate contracts between the firm and its stakeholders (Healy and Wahlen 1999). It is common that corporate executives of a firm do not have the same interests as its shareholders. An explanation is that all individuals are driven by desires to maximize their own wealth (Deegan 2011). Delegation of decision-making authority to corporate executives could lead to non-goal congruence problems. Through the use of compensation contracts, e.g. bonuses based on income, shareholders try to align the interests of the corporate executives with the company’s interest (Eisenhardt 1985). However, these compensation contracts could also be an incentive to defer income when a target would not be met or when the bonus is constrained, which is not in the interest of the company. Another example of contracts are lending contracts. DeFond and Jambalvo (1994) find evidence that firms have higher discretionary accruals in the year prior to and in the year of debt covenant violations.
Regulation can be a reason to engage in earnings management as well. Some industries face regulatory monitoring that is explicitly tied to accounting data, e.g. minimum capital requirements or a maximum return on assets. Other regulatory reasons are anti-trust or other adverse political consequences, which can be an incentive to manage income downwards to appear less profitable. This is also applicable to companies which are seeking for subsidy or protection (Healy and Wahlen 1999).
8
2.1.3 Earnings Management Mechanisms
Earnings consist of two components: cash flows and accruals. Both components can be used for earnings management. Real earnings management, the management of cash flows, refers to the purposeful altering of reported earnings in a particular direction by changing the timing or structuring of an operating, investing or financing decision (Badertscher 2009). Examples of real earnings management are cutting R&D costs and the decision to forgo a project. An advantage of this method for corporate executives is that it is less easier to detect than accrual-based earnings management, because it will not be scrutinized by auditors or regulators. However, real earnings management is considered to be more costly from a cash flow perspective, because of its adverse impact on optimal business operations and its potential to destroy long-term firm value (Badertscher 2009).
2.1.3.1 Accrual-based Earnings Management
In this thesis, the focus will be on accrual-based earnings management. Accrual-based earnings management occurs when managers adjust revenue and/or expense accruals to manage financial reports (Badertscher 2009).
Accrual accounting is an accounting method that is used when firms are in continuous operation and is primarily used to overcome problems with measuring firm performance (Dechow 1994). When companies are in continuous operation, it is difficult to measure firm performance accurately. Cash flows are not necessarily useful information for stakeholders due to timing and matching problems. The economic transactions are not recognized until the moment that cash transactions occur. Managers use accruals to recognize economic transactions by matching revenues to expenses at the time in which the transaction occurs, regardless of the payment is made or received (matching principle). Accruals are also used to recognize revenue when services are performed and it is reasonably certain that the payment will be received (revenue recognition principle). Accruals are expected to mitigate timing and matching
problems associated with cash flows by taking future cash flows into account, resulting in a
more accurate picture of a firm’s current financial condition.
Unfortunately, accruals can also be used by managers to opportunistically manipulate earnings. This is called accrual-based earnings management. Accrual-based earnings management is popular by managers because it is less likely to destroy long-term firm value, because it has no effect on cash flows. Accrual-based earnings management can be done within
9
the boundaries of GAAP, the general accepted accounting principles. Because GAAP allows judgments in financial reporting, detection costs will be relatively low. A disadvantage of this mechanism is that the ability is restricted, because of the reversing nature of accruals.
The net incentive (benefits minus costs) of real earnings management and accrual-based earnings management will determine the choice of the mechanism. Potential benefits are enhanced welfare for managers and advantages arising from hiding the real economic performance of the firm. Costs can be a decline in firm value resulting from not making optimal decisions, a decline in a firm’s stock price after discovery of accrual manipulation and reputation loss (Badertscher 2009). Because CFOs are more able to effect accruals, I will focus on accrual-based earnings management in this research. Accrual-based earnings management is also easier to detect than real earnings management.
2.1.4 Constraints to reduce Earnings Management
The internal governance structure should constrain or limit the ability of earnings management. With an appropriate internal governance structure, management should have less discretionary to manage earnings (Dechow et al. 1995). Auditing should constrain earnings management as well, because reputation will be damaged and firm value will be reduced if misreporting is detected (Becker et al. 1998). Further, accounting standards are expected to constrain the ability of earnings management. However, tighter accounting standards can cause a substitution effect between accrual-based and real earnings management. If accounting standards become stricter, a decrease in accrual-based and an increase in real earnings management is expected (Ewert and Wagenhofer 2005).
In 2002, the Sarbanes-Oxley Act (SOX) was signed into law. SOX was a reaction to the major accountings scandals in 2001-2002. The purpose of SOX is to enhance corporate governance and professional responsibility and improve the quality of financial reporting, in order to restore investor confidence (Jain and Rezaee 2006). Earnings management should be constrained by SOX. In section 2.4, I will go into further detail about SOX.
10
2.2
CFO Influence on Earnings Management
In this section, I will present a review of prior research that investigates the influence that CFOs (and CEOs) have on accrual-based earnings management. Does a CFO have significant influence on accrual-based earnings management? How is this influence related to the influence of the CEO?
Prior research shows that CEO equity incentives are positively related to accrual-based earnings management (e.g Baker et al. 2003; Cheng and Warfield 2005; Bergstresser and Philippon 2006). Jiang et al. (2010) investigate whether CFO equity incentives are independently related to accrual-based earnings management. One thought is that a CFO merely acts as a CEO agent (Graham and Harvey 2001). If a CFO only acts in the interest of the CEO, it is expected that CFO equity incentives would not have an incremental effect on accrual-based earnings management beyond the equity incentives of the CEO (Jiang et al. 2010). On the other hand, the primary responsibility of CFOs is preparing financial reports. CFOs have more knowledge about financial reporting than CEOs. Therefore CFOs know exactly what discretionary accruals are favourable to them, before the moment the accrual decision has to be made. For this reason it is easier for CFOs to influence their bonus than it is for CEOs (Beaudoin et al. 2013). Given that financial reporting is the primary responsibility of CFOs, Jiang et al. (2010) expected that CFOs can affect earnings management activities independently.
Jiang et al. (2010) find evidence that CFO equity incentives are positively related to accrual-based earnings management and that accrual-based earnings management is more increasing in CFO equity incentives than in CEO equity incentives. This evidence suggests that CFO equity incentives have a more significant role in accrual-based earnings management compared to CEO equity incentives. CFO wield independent influence on accrual-based earnings management, even after controlling for the effect of CEO equity incentives.
Geiger and North (2006) show that accrual-based earnings management decrease significantly after the appointment of a new CFO. This effect is not observable after the appointment of a new CEO. Peni and Vähämaa (2010) find similar evidence. They find a significant relation between CFO gender and accrual-based earnings management. This relation is not noticeable for CEOs. This evidence also suggests that CFOs have independent influence on accrual-based earnings management.
11
2.3
Gender Differences
In this section, I will present a review of prior research that investigates differences between women and men in general, in economic decision making and in financial reporting.
2.3.1 Gender Differences in General
Prior research shows that women are less competitive, less (over)confident, and more risk averse on average than men are. Croson and Gneezy (2009) reviewed the literature on gender differences in risk preferences, social preferences and competitive preferences. They document significant differences between women and men in all of the three categories. Most studies indicate that women are more risk averse and more sensitive than men and that men prefer competitive situations more than women do.
Risk taking involves the implementation of options that could lead to different outcomes. The kind of options is determined by the goals and values that a person has. We talk about risk taking, if at least one of the possible outcomes is undesirable or even dangerous (Furby and Beyth-Marom 1992). This definition indicates that a wide range of actions can be qualified as risk taking. Examples are smoking, buying a lottery ticket, online shopping and driving too hard.
Byrnes et al. (1999) provide evidence suggesting that women are less aggressive at a general level and are less likely to take risk than men. Their results show that men take more risk, even in situations where it is obvious that it is not a good idea to take the risk. Women, in contrast, seemed to take less risk, even in fairly harmless situations or in situations when risk taking is without doubt a good idea. However, the gap in risk taking between women and men seems to get smaller over time. Other research provide evidence that indicates that women appear to be slightly more risk averse than men, but this may be more a function of age, income and wealth (Riley and Chow 1992).
Another difference between women and men is the level of confidence. Men tend to be more overconfidence than women are (Bengtssona et al. 2005). According this research, the difference in confidence can be attributed to constraints, preferences and technology. Niederle and Vesterlund (2007) find that men are more overconfident about their relative performance compared to women, both in a competitive and a non-competitive setting. Beliefs of relative performance can be an indicator for preferences for performing in a competition. Because men are more confident about their performance, they may be more inclined to compete.
12
Beyer and Bowden (1997) examine the ability self-perceptions of women and men with similar abilities. They find that if the domain is more male oriented, the ability self-perceptions of women is inaccurately low and their response bias more conservative compared to men. This indicates that women have the tendency to underestimate their own ability, revealing lower confidence than men. Wankel (2008) and Montouchet (2013) state that women are more inclined to attribute their performance to luck than to their skills. In contrast, men often attribute their performance to their skills. This is also an indication of lower self-perceptions for women.
2.3.2 Gender Differences in Economic Decision Making
In the previous section, a review of gender differences in general has been present. Women tend to be less aggressive, less (over)confident and more risk averse compared to men. What are the implications of these differences between women and men for economic decision making?
Odean (1998) finds evidence that overconfident investors, who think they can predict the value of a security more accurately than they actually can, trade more than rational investors. By trading more, expected utility is reduced. Because prior research show that men are more overconfident than women (Bengtssona et al. 2005), it is suggested that male investors trade more than female investors. This may result in lower expected utility for male investors. Barber and Odean (2001) find evidence that male investors trade 45 percent more than female investors do. Net return of male investors is reduced by 2.65 percentage points due to trading, compared to 1.72 percentage point for female investors.
Barsky et al. (1997) also find evidence that women tend to be less risk-seeking in economic-decision making compared to men. They show that men more often choose the most risk-tolerant option, indicating a more risk tolerant level. Sunden and Surette (1998) find that women are less likely to invest their retirement funds in risky assets, such as stocks. They are more inclined to invest in ‘mostly bonds’, indicating a more risk-averse strategy for their retirement funds. Huang and Kisgen (2013) show that firms with female executives are less likely to engage in acquisitions and issue less debt compared to firms with male executives. This evidence indicates a more cautious strategy regarding acquisition and debt issues for female executives.
Powell and Ansic (1997) explain the difference in decision-making between women and men through the motivational theory (Schneider and Lopes 1986). Women would have lower risk preference, because of a greater desire for security. Men, in contrast, have a greater desire
13
for return. A greater desire for return may result in higher risk preference. Women attempt to avoid the worst situation, by taking less risk. Men attempt to achieve the highest possible gains. In order to do so, more risk has to be taken. So women tend to act more conservative than men do.
Schubert et al. (1999) show that differences in risk taking between women and men do not exist in controlled economic conditions. Differences only arise in abstract gambles, in which men are more sensible to gains and women to losses. This evidence suggest that an abstract gambling setting may not be an adequate setting to analyse risk attitudes towards financial decision making. Klenke (2003) finds evidence that differences in decision making outcomes are not per se determined by gender differences, but rather determined by the structure of power, the organizational politics, conflict management and trust that matters. This results in an indirect relation between gender differences and economic decision making.
2.3.3 Gender Differences in Financial Reporting
The differences between women and men described in the previous sections provided an interesting basis for research in the accounting environment. In combination with the increasing number of female executives, especially CFOs, the literature about female executives and financial reporting increased significantly in the last decade (e.g. Peni and Vähämaa 2010; Barua et al. 2010; Srinidhi et al. 2011; Francis et al. 2014).
Prior research shows that firms with female CFOs act differently in the context of accrual-based earnings management. Peni and Vähämaa (2010) find that firms with female CFOs are associated with income-decreasing discretionary accruals. This evidence indicates that firms with female CFOs pursue more conservative earnings management. Basu (1997) define conservatism as a tendency to require a stronger degree of verification to recognize good news as gains than to recognize bad news as losses. Female CFOs need more security before they recognize a gain.
Francis et al. (2014) also find strong evidence that female CFOs adopt more risk-averse and more conservative financial reporting policies than male CFOs. This effect is clearly noticeable when a male CFO is succeeded by a female CFO: the degree of accounting conservatism increases significantly after the turnover. Liu et al. (2014) examine CFO turnovers as well and find similar evidence. They find that the sign of the mean of discretionary accruals changed from positive to negative when a male CFO was succeeded by a female CFO. This
14
evidence indicates that earnings are managed upwards by male CFOs and downwards by female CFOs. Vähämaa (2014) finds evidence that discretionary accruals decreased when a female CFO succeeded a man and that discretionary accruals increased when the new CFO was a man (independent of the gender of the former CFO). This evidence also suggests that female CFOs use more cautious earnings management strategies.
Other research shows that firms with female CFOs tend to have lower absolute discretionary accruals and lower absolute accrual estimation errors. This effect is also noticeable after controlling for factors that are associated with accruals (Barua et al. 2010). This evidence suggests that firms with female CFOs have higher accrual quality compared to firms with male CFOs. Gavious et al. (2012) find that firms with women on the board of directors, audit committee and in the function of CEO or CFO have a lower level of accrual-based earnings management. Firms with a higher female representation seem to have higher earnings quality, although the level of external monitoring is lower. Srinidhi et al. (2011) find that firms with a higher ratio of women on their board of directors exhibit higher earnings quality. This evidence suggests that firms can improve their earnings quality by increasing the ratio of women on the board of directors and in executive functions.
Ye et al. (2011) investigate the relation between female executives and earnings quality in the context of an emerging market. They find evidence that there are no significant differences in the level of absolute discretionary accruals for listed Chinese firms between female and male CFOs. Arun et al. (2015) neither find a relation between female CFOs and the level of earnings management.
15
2.4
Passage of the Sarbanes-Oxley Act
In this section I will present general information about the Sarbanes-Oxley Act (SOX). Why did SOX pass? What are the consequences of SOX? I will also provide a summary of the findings of prior research in earnings management after the passage of SOX.
Before the passage of SOX, the number accounting restatements and the level of earnings management were rising and liquidity and investor confidence were declining (Coates IV 2007). In July 2002, the Sarbanes-Oxley Act was signed into law. SOX is a federal law of the United States that set new or enhanced standards for all U.S. public company boards, management and public accounting firms. The Act was a reaction to a number of major corporate and accounting scandals, including Enron and WorldCom. These major scandals decreased the public’s confidence in the security market. The Act was passed to enhance corporate governance, audit effectiveness and accounting quality in order to restore public’s confidence (Jain and Rezaee 2006). SOX has been considered as the most far-reaching securities legislation since the Securities Acts of 1933 and 1934 (Ball 2009). It does not only impose additional disclosure requirements, but also proposed substantive corporate governance mandates. This practice is unprecedented in the history of federal securities legislation (Romano 2004).
Section 302 of SOX requires that CEOs and CFOs of publicly traded companies certify that the financial statements and disclosures are reliable and fairly represent the operations and financial condition of the company. CEOs and CFOs became personally responsible for the reporting. When they certify that the financial statements and disclosure are reliable and fairly represent the operations and financial condition of the company when in fact this is not true, CEOs and CFOs can be punished with significant penalties. They could face imprisonment up to twenty years and fines up to one million dollars, or even being barred by the SEC from serving as a CEO or CFO (Jain and Rezaee 2006).
The benefits of the passage of SOX are hard to isolate. The chances are good that manage behaviour would have changed anyway as a reaction to the scandals in the early 2000s, even when SOX had not been enacted (Coates IV 2007). Ball (2009) states that it is not right to compare the pre-SOX period to the post-SOX period without taking into account the changes that would have occurred in market mechanisms. Many of the observed changes would have occurred anyway, due to market forces.
16
However, prior research find evidence showing that earnings management has decreased and investor confidence has increased after the passage of SOX (Jain and Rezaee 2006). Ahmed et al. (2006) find that before the passage of SOX, a firm’s discretionary accruals were positively related with the importance of that firm to audit firms as a client. They find that this relation has vanished after the passage of SOX. This evidence indicates a higher level of audit quality after the passage of SOX.
The costs of SOX are hard to measure as well. However, it is for sure that they are substantive, especially for small firms. Direct costs of SOX include PCAOB fees, firms’ compliance costs, and increased audit fees (Coates IV 2007). However, increased audit fees do not necessarily have been incurred due to SOX. Before the passage of SOX, audit fees were already rising due to the riskier environment and less competition (Asthana et al. 2009). Indirect costs include manager time and more risk aversion and are even harder to estimate (Coates IV 2007). Unfortunately it is not possible to undertake a cost-benefit analysis of the passage of SOX, because benefits could not yet be expressed in dollars (and as mentioned above, the consequences of SOX are hard to isolate).
Ball (2009) discusses the effectiveness of SOX, whether increased penalties are an effective regulatory move to prevent accounting fraud in the United States. Before the passage of SOX, the CEO and CFO were required to sign and attest the financial statements and face heavy penalties as well. However, this did not prevent accounting frauds. He states that the passage of SOX, which is likely to be just ‘more of the same’, seems not to impact the ones that commit accounting frauds, but rather cause even more conservative behaviour for prudent managers. This does not necessary imply that the passage of SOX is not effective and has been an overreaction to the accounting scandals (Ball 2009).
Prior research investigate the effect of the Sarbanes-Oxley Act on earnings management. The results of Cohen et al. (2008) show that the level of accrual-based earnings management declined significantly after the passage of the Sarbanes-Oxley Act in 2002. Just before the passage of the Act, a significant increase in absolute discretionary accruals could be observed. After the passage of SOX absolute discretionary accruals declined significantly. In contrast, the level of real earnings management declined significantly before the passage of SOX and increased significantly after the passage of SOX. This evidence shows a trade-off between accrual-based and real earnings management after the passage of SOX.
17
Li et al. (2008) show that accrual-based earnings management is constrained by the passage of SOX. According to Li et al. (2008), the quality of financial statement information is enhanced in the period after the implementation of SOX. Wang et al. (2011) find evidence for a significant lower level of absolute abnormal accruals after the passage of SOX. However, they do not find evidence that the passage of SOX caused lower signed abnormal accruals.
Lobo and Zhou (2006) were one of the first who find evidence for significant lower signed discretionary accruals after the passage of SOX. They also show that firms recognize losses more quickly compared to gains after the Act had been passed. These findings indicate more conservative earnings management strategies after the passage of SOX. Zang (2012) also find evidence for significant lower discretionary accruals after the passage of SOX. Consistent with Cohen et al. (2008), she also shows a significant increase in real earnings management. Zang (2012) states that accrual-based earnings management is constrained after the passage of SOX due to higher scrutiny. For this reason, managers changed their earnings management strategies.
However, Ghosh et al. (2010) find no evidence suggesting that the overall level of earnings management declined after the passage of SOX. They do find evidence that suggest that the relation between accrual-based earnings management and several board and audit committee characteristics is weaker in the period after the implementation of the Sarbanes-Oxley Act. Wang et al. (2011) do not find a significant relation between signed discretionary accruals and the passage of SOX.
18
2.5
Hypotheses Development
As mentioned in the previous literature review, prior research shows that female CFOs are more risk averse and tend to following more conservative earnings management strategies (Peni and Vähämaa 2010; Francis et al. 2014; Liu et al. 2014) and have a higher level of accrual quality (Barua et al. 2010). Conservative financial reporting strategies are characterized by lower discretionary accruals. Consistent with these findings, I predict that there will be a negative relation between firms with female CFOs and discretionary accruals. Therefore, I developed the following hypothesis about CFO gender and accrual-based earnings management:
H1: There is a negative relation between firms with female CFOs and discretionary accruals.
The previous literature review also shows that discretionary accruals declined after the passage of the Sarbanes-Oxley Act (Lobo and Zhou 2006; Zang 2012). Lobo and Zhou (2006) find evidence that discretionary accruals are significantly lower after the passage of SOX and that firms recognize losses more quickly compared to gains. This indicates more conservative earnings management strategies after the passage of SOX. Consistent with the finding of Lobo and Zhou (2006) and Zang (2012), I predict a significant decline in discretionary accruals after the passage of SOX. Therefore, I developed the following hypothesis about the passage of SOX and accrual-based earnings management:
H2: Discretionary accruals declined after the passage of the Sarbanes-Oxley Act.
Considering prior research that shows evidence that the passage of the Sarbanes-Oxley Act has had an impact on discretionary accruals (Lobo and Zhou 2006; Zang 2012) and evidence that suggests more risk-averse and conservative financial reporting decisions for female CFOs (Peni and Vähämaa 2010; Francis et al. 2014; Liu et al. 2014), I predict that the passage of SOX has had a bigger impact on female CFOs compared to male CFOs. I expect female CFOs to have lower discretionary accruals compared to male CFOs.
After the passage of SOX, CEOs and CFOs have to personally certify that the financial statements and disclosures are reliable and fairly present the operations and financial condition of the company. They could face significantly increased penalties for knowingly false certifications (Coates IV 2007; Ball 2009). Due to this significantly increased penalties and higher scrutiny, the risk of significant penalties increased after the passage of SOX. Because of the more risk-averse and conservative nature of women, I expect that this higher risk deterred female CFOs more to engage in earnings management compared to male CFOs. For this reason,
19
I predict that female CFOs have adopted even more conservative financial reporting strategies after the passage of SOX, resulting in a higher decrease in discretionary accruals for female CFOs after the passage of SOX compared to male CFOs. This resulted in the following hypothesis:
H3: Discretionary accruals declined more for female CFOs after the passage of the Sarbanes-Oxley Act compared to male CFOs.
In the next chapters, I will address the methodology to answer the research question. The model used to estimate the earnings management proxy will be explained, just as the regression models I will use to examine if the just mentioned hypotheses are true. The chosen sample selection and period will be presented and explained as well.
20
3
Methodology
In this chapter, the research method that has been used to answer the research question will be introduced. The selected sample and sample period will be described. Subsequently, the model that has been used to estimate discretionary accruals per firm-year observation will be presented. Finally, the regression models that has been used to regress discretionary accruals on the variables of interest FEMALE, SOXPERIOD and FEM*SOX and the control variables will be presented and explained.
3.1
Data & Sample Description
Data for this research are collected from the Wharton Research Data Services (WRDS) database. Data to estimate the earnings management proxy discretionary accruals and data to calculate the different control variables are acquired from the Compustat Annual Fundamentals database. Unfortunately, CFO gender data before 2006 could not be obtained from ExecuComp
immediately1. Therefore CFO gender data are manually obtained from the 10K filings of the
firm-year observations. In this manner, a CFO turnover could easily be identified as well. The sample period extends from 1998 to 2005 and is divided into two time periods: the period before the passage of the Sarbanes-Oxley Act in 2002 (pre-SOX period) and the period after the passage of the Sarbanes-Oxley Act (post-SOX period). The pre-SOX period extends from 1998 to 2001, the post-SOX period extends from 2002 to 2005. Because this research attempts to measure whether discretionary accruals has declined after the passage of SOX, the period after the passage should not be too long to observe a possible decline.
The sample consist of all firms listed in the Standard & Poor 500 Index (S&P 500). The S&P 500 is widely regarded as the best single gauge of large-cap U.S. equities. Financial institutions (SIC code 60-69) are excluded from the sample, because this type of companies do not have the opportunity to influence discretionary accruals. Electric, Gas & Sanitary Services companies (SIC 49) are excluded as well. One condition for estimating discretionary accruals per industry per year is that the companies have to be relatively similar in accounting. For this type of companies, this is not the case. Finally, firm-year observations with negative equity and incomplete firm-year observations are excluded as well.
21
The original dataset consisted of 3,383 firm-year observations. After deleting firm-year observations with two-digit SIC code 49 and 60-69 (825), firm-year observations with negative equity (59) and incomplete firm-year observations (177), the sample consist of 2,322 firm-year observations.
3.2
Method
Detecting earnings management is quite difficult, since managers use the grey area in accounting standards to manage the earnings. Accrual models are the most common method to measure earnings management (Dechow 1994). Several models are available to measure earnings management. Commonly used models are the Healy-model, the DeAngelo-model, the Jones-model, the Modified Jones-model and the Industry-model. All these models use discretionary accruals as a proxy to measure earnings management. In this research the performance-matched Jones Model is used, as applied in Kothari et al. (2005).
The Jones model was developed by Jones in 1991 (Jones et al. 2008). As all the above mentioned models do, the Jones model decomposes total accrual into a nondiscretionary and a discretionary component. Accounting standard-setting bodies require to make accounting adjustments to cash flows. These accounting adjustments are called nondiscretionary accruals (Healy 1985). Discretionary accruals are adjustments to cash flows selected by corporate executives and can be used as a proxy for earnings management.
Kothari et al. (2005) find that the performance-matched Jones Model and the performance-matched Modified Jones model are the most appropriate models to measure
earnings management. For this reason lagged ROA (ROAi,τ-1) is added to the original Jones Model
to control for performance. The performance-matched Jones Model has been chosen instead of the modified version, because the performance-matched modified Jones Model may lead to underestimated non-discretionary and overestimated discretionary accruals for firms with substantial growth. Firms with substantial growth usually have growth in receivables. Growth in receivables is not necessarily an indicator of earnings management (Kothari et al. 2005).
The original time-series version of the Jones model has been heavily criticized, because it requires long series of data (at least 10 years). Subsequently, it assumes that independent variables remain stable over time. The cross-sectional version of the Jones Model attempts to solve these problems (DeFond and Jiambalvo 1994). Consistent with DeFond and Jiambalvo (1994), regression coefficients are calculated for each two-digit Standard Industry Code (SIC)
22
sample per year with at least six firm-year observations. These estimated regression coefficients and firm-specific data are used to estimate the nondiscretionary accruals for the firm-year observations. Compared to the time-series version, the cross-sectional version generates a much larger sample. A larger sample increases both the reliability and efficiency of the results (Peasnell et al. 2000).
Some investigations have examined the absolute value of discretionary accruals (e.g. Ashbaugh et al. 2003; Meek et al. 2007; Barua et al. 2010). In this research, signed discretionary accruals will be examined. According to DeFond and Jiambalvo (1994), Becker et al. (1998), Geiger and North (2006) and Vähämaa (2014), this is the most appropriate measure for earnings management. The absolute value of discretionary accruals only reveals the extent of accrual-based earnings management. In contrast, signed discretionary accruals also reveal the direction of accrual-based earnings management (Vähämaa 2014). Because prior research shows that female CFOs use more conservative financial reporting strategies (Peni and Vähämaa 2010; Francis et al. 2014; Liu et al. 2014), the direction of earnings management is important in this research. Because firms are rarely sued for income-decreasing accruals (Geiger and North 2006), signed discretionary accruals are assumed to be also an appropriate measure to examine the effect of the passage of Sarbanes-Oxley Act on accrual-based earnings management.
3.2.1 Performance-Matched Jones Model
In the performance-matched Jones model, total accruals are decomposed into discretionary and nondiscretionary accruals. To estimate the nondiscretionary part, regression coefficients for each two-digit Standard Industry Code (SIC) sample per year are estimated. Only if a two-digit SIC sample per year did not contain the required six firm-year observations, the one-digit SIC sample per year is used to estimate the regression coefficients. One-digit SIC samples that did not contain the required six firm-year observations, are excluded from the sample. This resulted in my final sample size of 2,265 firm-year observations.
To estimate nondiscretionary and discretionary accruals, the following equation (1) is used. To reduce heteroscedasticity, all variables have been scaled by total assets at year τ–1.
23
Where, for fiscal year τ and firm i:
TAiτ = total accruals
Ai,τ-1 = total assets at year τ-1 (lagged total assets) (Compustat item #6)
ΔREViτ = revenues in year τ less revenues in year τ–1, scaled by
lagged total assets
(Compustat item #12) PPEiτ = gross property, plant and equipment, scaled by lagged total
assets
(Compustat item #7)
ROA i,τ-1 = income before extraordinary items year τ-1 divided by total
assets year τ-1
(Compustat item #123) α1 α2 α3 α4 = industry-specific parameters.
υτ = measurement error in year τ
Total accruals (TA) used in this research are the change in non-cash current assets minus the change in current liabilities excluding the current portion of long-term debt, minus depreciation and amortization. Total accruals are calculated in this way because prior research argues that corporate executives have more flexibility in earnings management using current accruals instead of long-term accruals (e.g. Guenther 1994; Becker et. al 1998; Bradshaw et al. 2001; Ashbaugh et al. 2003). This makes current accruals the most appropriate measure for accrual-based earnings management. TA are calculated using the following equation (2):
TAiτ = (∆CAiτ - ∆Cashiτ) - (∆CLiτ - ∆DcLiτ) - Deptiτ (2)
Where, for fiscal year τ and firm i:
∆CAiτ = current assets in year τ less current assets in year τ–1,
scaled by lagged total assets
(Compustat item #4)
∆Cashiτ = cash and short-term investments in year τ less cash and
short-term investments in year τ–1, scaled by lagged total assets
(Compustat item #1)
∆CLiτ = current liabilities in year τ less current liabilities in year τ–1, scaled by lagged total assets
(Compustat item #5)
∆DcLiτ = debt included in current liabilities in year τ less debt included in current liabilities in year τ–1, scaled by lagged total assets
(Compustat item #34)
Deptiτ = depreciation and amortization expense in year τ (Compustat item #14)
The nondiscretionary part of total accruals are estimated by equation (3):
24
Discretionary accruals are the difference between total accruals and nondiscretionary accruals. Discretionary accruals will be used as a proxy for accrual-based earnings management. The residuals of equation (1) are an estimation of discretionary accruals, i.e. the
measurement error term υτ. The measurement error term υτ is the difference between total and
nondiscretionary accruals and could also be calculated by equation (4):
DAiτ = TAiτ – NDAiτ (4)
3.2.2 Regression Models
After the discretionary accruals have been estimated, regression analyses can be performed to test the three hypotheses. For the first and second hypothesis, discretionary accruals will be regressed on the variables FEMALE and SOXPERIOD and eight control variables, to examine whether female CFOs and the passage of the Sarbanes-Oxley Act are negatively related to discretionary accruals. The following equation will be used (5):
DAiτ = β0 + β1 FEMALEiτ + β2 SOXPERIODiτ + β3 LNASSiτ + β4 MBiτ + β5 SGROWTHiτ +
β6 OCFiτ + β7 LEVERAGEiτ + β8 FINANCEiτ + β9 TURNOVERiτ + β10 LOSSiτ (5)
Where, for fiscal year τ and firm i:
FEMALE = dummy: if CFO is a woman: 1; if CFO is man: 0 SOXPERIOD = dummy: pre-SOX period: 0; post-SOX period: 1
LNASS = natural logarithm of total assets (Compustat item #6) MB = market value of equity divided by book value of equity at
year τ-1
(Compustat item #25 * item #199) /
(Compustat item #60) SGROWTH = change in revenue from year τ-1 to year t divided by
revenue in year τ-1
(Compustat item #12) OCF = cash flows from operations scaled by lagged assets (Compustat item #308) LEVERAGE = book value of long-term debt divided by book value of
equity
(Compustat item #9) / (Compustat item #60) FINANCE = dummy: ∆share outstanding >10% or ∆Long-term Debt
> 20%: 1; if not: 0
TURNOVER = dummy: CFO turnover: 1; if not: 0
LOSS = dummy: net income last year < 0: 1; if not: 0
I predict the variable FEMALE to have a negative relation with the dependent variable discretionary accruals, due to their more risk-averse and conservative nature. This means that I
25
expect that firm-year observations with a female CFO have lower discretionary accruals, indicating a more conservative level of discretionary accruals.
I predict a negative relation between the variable SOXPERIOD and the dependent variable discretionary accruals as well. Due to significantly increased penalties, I expect that CFOs will act more conservatively to reduce the chance to get punished with fines or even prison sentences. In other words, I expect that firm-year observations in the post-SOX period have lower discretionary accruals compared to the pre-SOX period.
. Prior research shows that different factors affect discretionary accruals (e.g. Becker et al. 1998; Geiger and North 2006; Meek et al. 2007). Eight variables are selected and included in the regression model to control for factors that affect discretionary accruals: LNASS, MB, SGROWTH, OCF, LEVERAGE, FINANCE, TURNOVER and LOSS.
LNASS is added as a control variable for firm size. According to the Political Cost theory, larger firms may have greater incentives to manage earnings to reduce political attention (Watts and Zimmerman 1990). Prior research shows that firm size is negatively associated with discretionary accruals (e.g. Peni and Vähämaa 2010; Becker et al. 1998, Klein 2002; Vähämaa 2014). However, larger firms may have less possibilities for earnings management because of higher government regulation and scrutiny. Another possibility is that firms with a high LNASS value have a male CFO more often. This may lead to higher discretionary accruals, due to the less conservative nature of men. Because the relation between firm size and discretionary accruals seems ambiguous, no prediction can be made about the relation between LNASS and discretionary accruals.
MB and SGROWTH are control factors for firm growth. Firms with high growth are less transparent and have more opportunities for earnings management (Meek et al. 2007). Geiger and North (2006) show that firm growth is positively associated with discretionary accruals. Liu et al. (2014) and Vähämaa (2014) find also evidence for a positive relation between both the variables MB and SGROWTH and discretionary accruals. For this reason, I expect MB and SGROWTH to have a positive relation with discretionary accruals.
OCF is used to control for periods with extreme positive cash flows. According to Dechow et al. (1995), discretionary (nondiscretionary) accruals will be overestimated (underestimated) in the case of extreme positive cash flows. Prior research find evidence for a negative relation between discretionary accruals and OCF (e.g. Liu et al. 2014; Peasnell et al.
26
2005). Consistent with this evidence, I expect OCF to have a negative relation with discretionary accruals.
LEVERAGE is included as a control variable for debt covenants. According to the debt covenants hypothesis, firms financed with debt may have incentives to manage earnings upwards to avoid sanctions (Becker et al. 1998). Prior research find evidence that LEVERAGE is positively related to discretionary accruals (DeFond and Jiambalvo 1994; Guenther 1994; Chang and Sun 2009; Chen and Huang 2013). However, firms in trouble usually have a higher debt-to-equity ratio. These kind of firms may have incentives to reduce earnings, due to contractual renegotiations. Peni and Vähämaa (2010) and Liu et al. (2014) find evidence that LEVERAGE is negatively associated with discretionary accruals. Due to this ambiguous evidence, no prediction can be made about the relation between LEVERAGE and discretionary accruals.
The dummy variable FINANCE controls for the effect of significant changes in the capital structure of the firm. Significant changes in the capital structure can affect accounting accruals and there may be limitations of accruals management imposed by the balance sheet structure (Barton and Simko 2002). Geiger and North (2006) find that FINANCE and discretionary accruals are positively related. Consistent with this evidence, I predict a positive relation between FINANCE and discretionary accruals.
TURNOVER is added as a control variable for the appointment of a new CFO. Geiger and North (2006) and Vähämaa (2014) find evidence for significant lower discretionary accruals following the appointment of a new CFO. Consistent with this evidence, I expect a negative relation between TURNOVER and discretionary accruals.
LOSS is added as a control variable for a negative income in the past year. Firms with a negative income might increase reported earnings to avoid losses (Burgstahler and Dichev 1997). However, Peni and Vähämaa (2010) and Vähämaa (2014) find evidence that LOSS and signed discretionary accruals are negatively related. A reason for this evidence could be that firms that reported a loss in the past year experience higher scrutiny. Consistent with this evidence, I predict a negative relation between LOSS and discretionary accruals.
Initially I also wanted to add a control variable for CFO equity incentives. Prior research shows that CFO equity incentives are significantly related to discretionary accruals (Jiang et al. 2010). Unfortunately, it was not possible to obtain data to calculate equity incentives for CFOs
27
from ExecuComp straightaway2. If I wanted to add a control variable for CFO equity incentives,
I should have searched the 10k filings for all firm-year observations manually. This was not conceivable in the short-time to wright this thesis.
For the third hypothesis, the variable FEMALE*SOX will be added to examine the incremental effect of the passage of SOX for female CFOs on discretionary accruals. To examine if discretionary accruals decline more for female CFOs after the passage of SOX compared to male CFOs, the following model will be used (6):
DAiτ = β0 + β1 FEMALEiτ + β2 SOXPERIODiτ + β3 FEM*SOXiτ + β4 LNASSiτ + β5 MBiτ
+ β6 SGROWTHiτ + β7 OCFiτ + β8 LEVERAGEiτ + β9 FINANCEiτ + β10 TURNOVERiτ +
β11 LOSSiτ (6)
I predict the variable FEM*SOX to be negatively related to discretionary accruals. Due to the more risk-averse and conservative nature of women, I expect that the passage of SOX have had a bigger impact on female CFOs compared to male CFOs. This means that I expect that after the passage of SOX discretionary accruals decrease more for female CFOs compared to male CFOs.
28
4
Results
In this chapter the research results will be presented and explained. I performed several regressions to examine whether my hypotheses are true. First, I will present a description of the statistics and explain the main differences. Subsequently, I will present and explain the results of the regression models.
4.1
Descriptive Statistics
Before regression analyses are performed, the continuous variables have been winsorized. Descriptive statistics and regression coefficients can be highly influenced by extreme observations, also known as ‘outliers’. With winsorization, the observations in the smallest percentile are set equal to the smallest value of the second percentile. Observations in the largest
percentile are set equal to the largest value of the 99th percentile. Winsorization reduces the
influence of extreme observations.
The correlation matrix (Table 4, page 48) presents the correlations between the different variables used in the regression models. Discretionary accruals are significantly related to FEMALE, SGROWTH, OCF, FINANCE and LOSS. They are negatively related to FEMALE, OCF and LOSS, and positively related to SGROWTH and FINANCE. This is consistent with my predictions. Furthermore, there is no sign of high correlation between two or more independent variables.
4.1.1 CFO gender
In the total sample size, 179 firm-year observations have a female CFO. This is 7.90% of the total sample size (2,265 firm-year observations). The table on the next page (Table 1) presents
the mean, the 25th percentile, the median and the 75th percentile of total accruals (TA),
nondiscretionary accruals (NDA), discretionary accruals (DA) and the control variables, based on CFO gender. A t-test (Ranksum test) has been performed to examine whether the mean (median) of the different variables differs significantly between the two groups. The p-values can be found in the last column.
