CFO gender and accrual quality in the post-SOX period

(1)

Amsterdam Business School

CFO gender and accrual quality in the post-SOX period

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

Name: Sourjanie Elisa

Student number: 10571795 Thesis supervisor: Nan Jiang

(2)

2

Statement of Originality

This document is written by student Sourjanie Elisa 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.

(3)

3

Abstract

This research examine the whether the implementation of the Sarbanes-Oxley act (hereafter, SOX) has led a different impact on the accrual quality of female CFOs and male CFOs. SOX was passed into law after some major accounting scandals such as Enron and WorldCom. After the implementation of SOX, CEO and CFO have become more accountable for the financial reporting and disclosure of the firm. The existing literature provides evidence that firm report more conservative after the implementation of SOX. Prior research in psychology, sociology and accounting field provide evidence that women are more risk-averse and

conservative than men. Therefore, one might expect that due it to the high scrutiny, litigation risk and significantly increase in penalties after the implementation of SOX, that female executives will behave even more conservative than their male counterparts. The empirical results are based on an S&P 1500 firm’s sample over the period 1996 to 2006. The sample period was divided in two, 1996 to 2001 to examine the pre SOX period and 2002 to 2006 to examine the post SOX period. To measure accrual quality, which function as determinant for earning quality in this study, the residual from the Jones, Modified Jones, Dechow and Dichev (2002) and the extended version of Dechow and Dichev (2002) were used. The empirical result of this study show that accrual quality have significantly increase in the post SOX period, but fail to provide evidence whether accrual quality have increase more for female than male executive in the post SOX period. Overall, the findings of this research can help future researcher when investigating the relation between executive gender and accrual quality, while using a different approach.

(4)

4

Acknowledgement

I would like to express my sincere gratitude to my supervisor Nan Jiang who has actively instructed me through the whole master thesis process. She offered me useful comments and remarks at every stage of the thesis. Furthermore, I would like to thank my family because without their support I would have not been able to finish my thesis. At the same time, I would like to thank my friend Sui Ting Cheung for her encouragement and unconditional support during this whole period.

(5)

5

1. Introduction

1.1 Background

The purpose of this study is to provide empirical evidence regarding the relationship between chief financial officer (hereafter, CFO) gender and accrual quality in the post-SOX period. More specifically, this study attempt to assess whether the accrual quality for firms with female CFO have increased more than for firm with male CFO in the post-SOX period.

In 2002 the Sarbanes-Oxley Act (2002) has introduced requirements regarding the responsibilities of CEO and CFO to improve the quality of the financial reporting. The Sarbanes- Oxley act (hereafter, SOX) is a legislative response to the major corporate and accounting scandals such as Enron and WorldCom. The Act was passed in to law with the objective to make financial reporting more transparent and executives more accountable, and by this means restore investor confidence in the security market and enhance corporate governance (Collins et al., 2009). Section 302 of SOX makes CEO and CFO personally responsible for the financial information and disclosures of a company by requiring executives to personally certify the accuracy and completeness of the financial statement. Section 906 of SOX elevate the CEO and CFO responsibility toward financial statement by imposing criminal penalties for CEO and CFO who certify financial statement knowing that they not comply with requirements of the Security Exchange ACT of 1934 (Small et al. 2007).

Prior literature regarding the effect of SOX suggests that the quality of financial statement is enhanced in the post-SOX period (Li et al., 2008; Cohen, 2008). Lobo and Zhou (2006) investigate the change in managerial discretion over financial reporting following the introduction of SOX. Their findings suggest a reduction in discretionary accruals and that firms on average are more conservative in their financial reporting after the implementation of SOX. The research of Wang et al. (2011) do not find evidence of a lower signed of

discretionary accruals in the post SOX period. Although there is a body of research about the effect of SOX on financial reporting and earnings management the results contradicts each other. Therefore, this study will focus on the association between the implementation of SOX and accrual quality.

Existing research suggest that the quality of financial reporting is affected by the characteristics and incentives of the firm’s executives (e.g., Matsunaga and Yeung, 2008; Peni and Vähämaa, 2010; Srinidhi et al., 2011). A body of literature in the sociology, psychology,

(7)

7

and economics provide evidence that women are generally more risk averse than men ( Eckel and Grossman,2004; Croson and Gneezy,2009). This is in line with the research of Francis et al. (2014), which suggests that female CFOs are more risk averse than male CFOs. Francis et al. (2014) examine the impact of CFO gender on financial reporting decision-making in the context of accounting conservatism. They found a positive relation between female CFOs and accounting conservatism meaning that women are more risk–averse then men. Peni and Vähämaa (2010) examine the effect of CFO gender on financial reporting quality in the context of accruals. They found evidence that female CFOs are associated with income-decreasing discretionary accruals and adopt more conservative earnings management strategies. Srinidhi et al. (2011) found evidence that corporations with greater female participation on boards exhibit higher earnings quality. Barua et al. (2010) also examine the relationship between CFO gender and the quality of accruals and they found that firms with female CFO are likely to have lower absolute abnormal accruals. Unlike Mei et al. (2011) found limited evidence on the effect of CFO gender and discretionary accruals. Although there is some research about CFO gender differences and financial reporting quality in the context of accruals quality the results contradict each other. Hence, this provides a good foundation to further investigate the difference effect that the implementation of SOX might has led on the reporting earnings quality of female and male CFOs.

(8)

8

1.2 Research question

This study focuses on the association between the passage of the Sarbanes-Oxley act, chief financial officer gender and the quality of accruals. Accrual quality will be used in this study to determine earning quality. Furthermore, several studies use models of discretionary accruals to examine the manipulation of accruals (Healy and Wahlen, 1999). Healy and Wahlen (1999) state that a range of literature suggest that managerial intent affects the

incidence and magnitude of accrual estimation error. Dichow and Dichev (2002) illustrate the usefulness of accrual quality as a measure of earning persistence. They provide evidence that firms with low accrual quality have more accrual that is unrelated to cash flow, and have more noise and less persistence in their earnings. Prior research also document the benefit of the accrual process and provide evidence that earnings is a better measure of performance than cash flows are (e.g., Dechow, 1994; Lui et al., 2002). Furthermore, accrual quality has already been used by in other studies as determinant for earnings quality. Therefore, objective of this study resulted in the following research question:

Did accrual quality improved more for female CFOs than for male CFOs after the implementation of the Sarbanes-Oxley ACT of 2002?

Motivation for this study comes primary because to the best of my knowledge there is no research yet that examines the association between the passage of the Sarbanes-Oxley act, chief financial officer gender and the quality of accruals. Subsequently, a range of accounting research indicates that financial reporting quality is affected by the characteristics of the firm’s executives. Barua et al. (2010) argue that the passage of the Sarbanes-Oxley act might have led a different impact on female executives than on their male counterparts and

encourage other researches to study the differences. Beside, according to Grant Thorton’s International Business Report (2015, p.5) woman currently holds worldwide 22% of senior management positions. That is an increase from 19 % in 2004 to 22% today and this percentage never has been higher than 24% over the intervening period. Is this because women are more conservative and risk-averse or is it because women provide better earning quality, or because SOX had a bigger impact on female executives? I expect that SOX had a bigger impact on female executives than male executives because of their conservative and risk-averse nature.

This research will contribute to prior literature in several ways. Firstly, because this research will be the first research that provides evidence whether the reported earnings

(9)

9

quality significantly differ after the implementation of SOX for firms with female versus male executives. Subsequently, this research helps in evaluating the effectiveness of SOX. The purpose of SOX is to enhance financial reporting quality by making executives more accountable for intentional errors. This means that an improvement in accrual quality implies a reduction in intentional and unintentional estimation errors. Therefore the findings of this study can be useful for regulators when measuring the effectiveness of SOX on financial reporting practices. The result of this research might have also implication for recruiter in their hiring decisions.

To test the hypothesis in this study different approaches were used. First, the methods suggested in Kothari et al. (2005) were used to measure the absolute value of performance-matched abnormal total accrual and abnormal current accruals. Subsequently, the cross-sectional version of Dechow and Dichev (2002) and its extended version by Francis et al. (2005) and Jones et al. (2008) where used to measure accruals quality.

The results of this study indicate that accrual quality has increase after the

implementation of SOX. Moreover, the result suggests that the level of absolute abnormal accruals and accrual estimation errors was not significantly lower for female CFOs after the implementation of SOX.

The remainder of this paper is as follows: chapter 2 will discuss the relevant literature regarding earning and accrual quality, the influence of executives officer characteristics on accrual quality, the passage of the Sarbanes-Oxley act and the hypothesis development. Chapter 3 will address the research methodology of this study. Chapter 4 will discuss the result of this study. Finally, chapter 5 will provide a discussion, conclusion and the limitations of this study.

(10)

10

2. Literature Review and Hypothesis

In this chapter a literature review will be provided concerning, earning quality, accrual quality, gender differences and the passage of the Sarbanes-Oxley Act. Finally, the hypothesis development will be presented.

2.1 Earnings quality

Earnings numbers are not only important for investors to measure firm performance (Dechow, 1994) but also for third parties such as creditors and suppliers. The quality of earnings is an important aspect when third parties have to make decisions that are based on earnings figures (Dechow et al., 2010). The information in earnings is essential because it contained components that have different consequences on the decision usefulness of for example investors. Investor can herewith predict future dividends or decide whether or not to invest in a firm. Firm’s earnings numbers is an accrual accounting measure of the firm profit or loss from business activities and events during a quarter or annual period (Nichols and Whalen, 2004). This means that reported earnings is a summary measure of firm

performance.

Statement of Financial Accounting concept No.1 define earnings quality as follows: “Higher quality earnings provide more information about the future of a firm’s financial performance that are relevant to a specific decision made by a specific decision-maker”. Several researches used the term earnings quality with a different interpretation in mind. Dechow et al. (2010) define three characteristics of earnings quality. First, they state that earning quality is conditional on the decision relevance of the information provided. Subsequently, that the quality of reported earnings number depends on whether it provide helpful information, in understanding the firm’s financial performance. Thirdly, earnings quality is determined by the relevance of underlying financial performance to the decision, together with the ability of the accounting system to measure performance.

There are several proxies to measure earnings quality. Some proxies that can be used to give an indication of earnings quality are: smoothness of earnings, earnings persistence, and residuals from accrual model. Smoothness of earning is a good proxy but it is difficult to disentangle smoothness of reported earnings that reflect smoothness of the fundamental earnings process, accounting rules and intentional earnings manipulation (Dechow et al., 2010). Earnings persistence is a good measure as well but required a great amount of time in

(11)

11

order to make a reasonable evaluation of persistence as a measure of earnings quality. Due to time constraint the proxy earnings persistence in not used in this research. Residual of accrual models represent management discretion or estimation error, both reduce the decision

usefulness. This measure intent to isolate the managed error component of accruals and has become the accepted methodology in accounting research to capture discretion (Dechow et al., 2010). In these research residuals of accrual models is used as a proxy for earnings quality.

A common used indicator of earning quality in previous studies is accrual quality (Dechow et al., 2010). High quality accruals indicate high quality earnings. Therefore, accrual quality will be used in this research as determinant for earnings quality. Hence, the residuals from the Jones model and Dechow and Dichev (2002) model will be used to measure accrual quality, following the method of Barua et al. (2010).

2.2 Earnings and accruals

Earnings consist of cash and accruals. Accrual is the recognition of an expense or revenue for which an entity expects to expend cash or receives cash, in a future period. Their main

purpose is to provide a better insight of firm performance. Cash flow as an element of earnings, suffer from timing and matching problems. Accruals resolve the timeliness and matching problems of the cash flow item. Accruals capture the actuals economic performance of the firm when it occurs.

Dechow (1994) examines the circumstances under which accruals improve earnings to measure for firm performance. Dechow (1994) suggest that accruals are a better measure of firm performance than cash flow are, this is in line with the view of the Financial

Accounting Standard Board Statement of Financial Accounting Concepts No. 1, paragraph 44 states: “Information about enterprise earnings and its components measured by accrual accounting generally provides a better indication of enterprise performance than does information about current cash receipts and payments”.

Sloan (1996) examines two components of earning, accruals and cash flow. Sloan (1996) suggests that earning quality is affected by the size of accruals. Sloan provides evidence that accruals are less persistent that cash flows are, and that earnings with small amount of accruals will more likely persist. However, Dechow (1994) states that firms operating in volatile environments with large changes in their working capital, investment

(12)

12

and financing activities have severe timing and matching problems. The accrual process provide hereby rules on the timing of cash flow recognition in earnings in this way earnings will better reflect performance than cash flow.

Dechow (1994) also recognized that accruals introduced problems because management have discretion over the recognition of accruals. This is consistent with the stream of research that examines the opportunistic use of accruals to window-dress and mislead user of financial statements and they suggest that managerial intent affects the incidence and magnitude of accrual estimation errors (Dechow and Dichev, 2002).

Accruals are subject to estimation errors, which are split in intentional, and unintentional estimation errors. An intentional error is caused by manager’s intent to overstate earnings in order to get for example a higher bonus. Unintentional errors are caused due to lack of information or a predictive ability concerning for example future revenue or expenses.

In general, prior studies provide evidence that a high amount of accruals is associated with lower quality accruals and earnings (e.g., Dechow and Dichev, 2002; Sloan, 1996). Accruals are essential to evaluate the subjective part of earnings. Therefore, accruals are important for financial reporting. This implicates that determining accrual quality also determines earnings quality.

2.3 Accrual quality

(Francis et al., 2005) investigate the association between accruals quality and the costs of debt and equity, they suggest that lower accrual quality is associated with higher cost of capital. Accrual quality informs investors about the mapping of accounting earnings into cash flow. This implicate that when accrual quality is low, the mapping of accounting earning into cash flow is weakened and therefore information risk increased (Francis et al. 2005). Francis et al. (2005) state that accruals have two components; accruals that reflect economic

fundamentals (innate factors) and accruals that represent managerial choices (discretionary factors).

In contrast, the study of Dichow and Dichev (2002) do not discriminate between these different types of accruals and they based their research around total accrual quality. Dechow and Dichev (2002) examine whether accruals are useful and better in evaluating firm

performance. They suggest that the quality of accruals and earnings is decreasing in the magnitude of estimation error in accruals. This means that the magnitude of the estimation

(13)

13

error determines accrual quality. Furthermore, they state that the notion of estimation errors includes both intentional and unintentional error, and that the distinction is important because accrual and earnings quality is not only affected by management intent to manipulate

(unobservable actions) earnings. The result of their study indicate that accrual quality is related to observable firm characteristics such as for example firm size and volatility of operations. Dechow and Dichev (2002) finds strong evidence of a positive association between accrual quality and earning persistence.

Sloan (1996) suggests that earnings quality is influenced by the size of accruals and, provide evidence that the persistence of earnings depends on the relative magnitude of the cash and accruals fraction. This implicates that when earnings consist of a high level of accruals, it will be less persistent indicating a sign of lower earnings quality.

2.4 Earnings manipulation

Schipper (1989) defines earnings management as a purposeful intervention in the external financial reporting process, with the intent of obtaining private gain, as opposed to merely facilitating the neutral operation of the process. According to Scott (2011) earnings management is the choice by a manager of accounting policies, or real actions, affecting earnings to achieve some specific earnings objective. There are two forms of earnings management: accruals based and real earnings management. Real earnings management occurs when manager purposeful alter cash flow such as for example decreasing advertising expenditures or R&D. Accruals based earnings management occurs when manager make use of opportunistic accrual accounting choices to report expected earnings number. In this study we focus on accrual accounting.

There are several explanations for the fact that managers try to manage earning through accrual. A potential cause for the excessive level of accrual is managerial

opportunism. Positive accounting theory suggest that managers who are agents to principal, act to their self-interest and that they will more likely take advantage of less efficient market to maximize their own wealth. Graham et al. (2005) offer some insight why managers managed reported earnings. Firstly, because accounting earnings matters more to manager than cash flow. Because managers believe that earnings have more information content about firm value, therefore will more likely focus on earnings rather than on cash flow. Second, managers are interested in meeting and beating earnings benchmarks. In this way they can influence stock prices and their own welfare. Executives believe that hitting earnings

(14)

14

benchmarks builds credibility with the market and helps to maintain or increase their firm’s stock price. Jensen & Meckling (1976) also state that due to the two conflicting functions of financial reporting managers are incentives to manage earnings. Financial reporting serves to allocate resources and to monitor executives. This two functions conflict because, executives are the ones in charge to produce earnings report for investor. But the rewards executives get are also based on the same earnings report. This incentivizes manager to get involved in earnings manipulation. Since, executive are more able to affect accruals through discretion they will more likely make use accrual earnings manipulation.

2.4 Gender differences

2.4.1 Decision making

Female are playing an important role in business management and there is a significant increase in the number of females belonging to top executive positions. Women are evaluated in terms of their success in decision-making. Because of this several researches investigate the differences between female and male executives in the context of decision-making. Researcher focuses on the difference in decision-making between male and females, because this is the main responsibility of an executive. Decision-making is associated with risk taking, balancing of potential rewards, against possible negative consequences of a particular action (Johnson and Powell, 1994). The outcomes of decision-making are not predictable and this involves a greater risk. Decision-making is for this fact one of a firm’s most critical factor for success.

A body of research indicates that women tend to be less aggressive and more cautious in financial decision-making. Powel and Ansic (1994) examine the differences in the nature of decision taken by male and females. They provide evidence that females have lower risk preferences because of their greater desire for security. In contrast, male have a greater desire for return. Females tend to focus on strategies that avoid the worst situation to gain security. Unlike, male tend to focus on strategies, which they think will achieve a higher return despite the higher risk. For example females will choose the widest insurance cover to avoid loss when male will mostly choose the lowest cover for insurance. The findings of Powel and Ansic (1994) are in line with the findings of Byrnes et al. (1999). Byrnes et al. (1999) review 150 studies about gender differences in risk taking behavior of male and females; they state that females will take less risk even in situation that is not really harmful. As well, they

(15)

15

provide evidence that women on average are more cautious and less aggressive than men in decision context. The study of Mason and Mudrack (1996) also suggest that men are

interested in economic benefit and a successful career and will more likely break rules to achieve success, when women are more focus on harmonious relationship and are less likely to be unethical. Bernarci and Arnold (1997) also state that female executives have higher moral standards than their male counterparts.

Francis et al. (2009) investigate the effect of CFO gender on accounting conservatism. They find a positive relation between female CFOs and conservatism when firms face higher litigation risk, systematic risk or management turnover risk. They also find evidence that female CFOs apparently reduce dividend payout. Prior research also provides evidence that men are more overconfident compare to women (Huang and Kisgen 2013). Huang and Kisgen (2013) also find that female CFOs in contrast to their male counterparts issue less debt and are more likely to reduce leverage ratio. Nguyen et al. (2008) examined moral issues and gender difference in ethical judgment with the Reidenbach and Robin’s multidimensional ethics scale. They looked at the differences in ethical perceptions between men and women and they find that women are more ethical than men. In sum, several studies provide evidence that women compare to men are more risk averse, less confident and more ethical.

2.4.2 Executives gender and Accrual quality

Barua et al. (2010) study the association between CFO gender and the quality of accruals. They use the absolute value of abnormal accruals estimated by the Modified Jones Model and the accrual estimation errors from Dichow and Dichev (2002) to measure accrual quality. Barua et al. (2010) provide evidence that firms with female CFOs have lower level of absolute abnormal accruals and lower accrual estimation errors. This suggests that firm with female CFOs have higher accrual quality compared to firms with male CFOs. Peni and

Vähämaa (2010) also provide support for this finding. Their results indicate that female CFOs are associated with income-decreasing discretionary accruals. Gravious et al. (2012) also provide evidence that firms with female executives have lower level of abnormal accruals. Srinidhi et al. (2011) investigate the association between female participation on the board of directors and earnings quality. As a measure for earnings quality they used discretionary accruals computed as the absolute value of the estimation error in accruals after controlling for current, past and future cash flows, sales and long-term assets, and operating cycle and

(16)

16

volatility in sales. They provide evidence that firms with greater female participation on their boards exhibit higher earnings quality.

The study of Ye et al. (2011) examines the possible association between executives gender and earnings quality in China, the largest emerging market. The results suggest that there are no significant differences between the level of absolute discretionary accruals for firms with female executive and male executives.

2.5 The Sarbanes Oxley Act

Extreme forms of earnings management have led to several financial scandals around the world, which has raised serious concerns about quality of financial reporting and the integrity of top executives. This resulted in a decrease in the investor confidence level. As a response regulators have signed the Sarbanes-Oxley act into law. SOX was passed into law with the objective to protect investors by improving the accuracy and reliability of corporate

disclosures and to restore investors confidence in the integrity of firms financial reporting (Lobo and Zhou 2006).

Two of the most important section of SOX is section 302 and 906. In this section the SOX intend to hold CEOs and CFOs more accountable for financial statements. Section 302 requires the CEOs and CFOs of publicly traded firms to certify the fairness of the

presentation of the financial statements. Section906 imposes criminal sanctions for CEOs and CFOs that misstated financial statements (Collins et al. 2009). This can be $1 million fine and 10 years imprisonment, and can increase to 5$ million and 20 years imprisonment if the executives engage in wrongdoing.

Prior literature has provided diverse results about earnings quality of firms after the implementation of SOX. Cohen et al. (2008) investigate the prevalence of accrual-based and real earnings management in the period leading to the passage of SOX and in the period following the passage of SOX. They document a significant increase in absolute discretionary accruals before the passage of SOX, which indicate a decrease in accrual quality. They

provide evidence that absolute discretionary accruals decline significantly after the passage of SOX, which implicate an increase in accrual quality.

Bedard (2006) investigate the association between the different disclosures of SOX and earnings quality. He used unexpected and actual accruals as a measure for earnings quality and provides evidence that the level of absolute and unexpected accruals increase in the year internal control weakness are disclosed. Doyle et al. (2007) conduct a similar study,

(17)

17

and they provide evidence that the material weakness disclosure of SOX (705) is associated with poor estimated accruals that are not realized to cash flow. This suggests that weak internal controls are associated with lower accruals quality.

Another research that examines the change in managerial discretion over financial reporting following the passage of SOX is the research from Lobo and Zhou (2006). They compare discretionary accruals across the period before and after SOX. The results indicate that discretionary accruals are lower in the post SOX period indicating that firms are more conservative in the post SOX period.

A body of research has been investigating the implication of SOX, there is still a big debate about the effect of SOX on the quality of financial statements. Most of the literature about the association between SOX and earnings quality indicate that the quality of earnings have increase after the implementation of SOX.

2.6 Hypotheses development

After reviewing literature concerning the quality of earnings, executives gender differences and the passage of SOX. Three hypotheses have been developed.

The first hypothesis concerns the extent to which accrual quality has increased in the period after the passage of the Sarbanes-Oxley Act. As was mention before the study of Cohen et al. (2008) and Lobo and Zhou (2006) provide evidence that discretionary accruals are significantly lower in the period after the implementation of SOX implying higher accrual quality. Therefore, I expect that accrual quality have increased after the implementation of SOX. The first hypotheses is formulated as follow:

H1: Accrual quality have increased in the period after the passage of the Sarbanes-Oxley act

As already comment before, previous research provides evidence that women are more risk averse, less overconfident and more ethical than men. The study of Barua et al. (2010), Peni and Vähämaa (2010) and Gravious et al. (2012) all indicate that the level of discretionary is lower for firm with female executives than for firms with male executives. Hence, I expect that female executive are positively related to accrual quality. The second hypothesis is as follow:

(18)

18

H2: Firms with female CFO’s are positively related to accrual quality

Prior research provides evidence that the level of discretionary accruals have decrease in the period after the implementation of SOX indicating more conservative reporting (e.g., Cohen et al., 2008;Lobo and Zhou, 2006). After the implementation of SOX CEO and CFO has to personally certify the fairness and reliability of financial statement and disclosures. CEO and CFO can face serious penalties by certifying financial statement and disclosures that does not meet the requirements. In this study the focus is on CFOs, because, CFOs oversees the firm’s financial reporting process and therefore he/she has the most direct impact of all senior managers on the accounting related decisions of the firm (Francis et al., 2015).

After the implementation of SOX firms face higher litigation cost when they overstate earnings. In most of the cases when firms overstate earnings the litigation cost is higher than the cost of understating earnings. Beside, SOX has led a significant increase in penalties, high scrutiny and higher risk of relatively higher fines. Due to this fact executive are more

conservative in their reporting style. There is a range of literature that documents evidence that females are more risk-averse and conservative than their male counterparts (e.g., Francis 2009;Francis 2014). Based on this evidence I predict that female executive will do anything to avoid higher litigation cost for the firm, to avoid the risk of higher fines and imprisonment. With this expectation in mind the third hypothesis is as follow:

H3: Accrual quality has increased more for firms with female CFOs compared to firms with male CFO’s in the period after the passage of the Sarbanes-Oxley act

(19)

19

3 Methodology

This chapter will discuss the empirical method used to answer the research question. The regression models used to test the hypotheses. Subsequently, the selected sample will be described. Finally, the regression models are presented and explained.

3.1 Method

Several accrual models are available to measure the magnitude of accruals, earnings

management or accrual quality. In this study the focus is on accrual quality in the post SOX-period. Following the method of Barrua et al. (2010) four different models were used to measure accrual quality.

The first model used in order to measure accrual quality in this research is the

performance –matched Modified Jones Model, as applied in Kothari et al. (2005). A body of research regarding discretionary accruals is based on the Jones Model or the Modified Jones model because of the strong interpretation power. Kothari et al. (2005) point out that

discretionary accruals estimated using the Jones or the Modified-Jones model; adjusted for a performance-matched firm’s discretionary accrual, tend to be the best measures of

discretionary accruals. Residuals usually represent discretionary accruals and are used as proxy for accrual quality (unsigned residuals). The following equation is estimated cross-sectionally by each two-digit SIC industry and by year:

(1) Where:

TA =Total accruals

Ai,τ-1 =Total assets of firm I for year t-1

ΔREV iτ =Change in revenues for firm I from year t-1 to year t ΔREC iτ =Change in receivables for firm I from year t-1 to year t PPEiτ =Firm i’s year t gross property plant and equipment ɛ iτ =Error term

The residuals from the equation are a measure of abnormal total accruals (ATA). The measure intent to disengage the managed or error component of accruals. Residual from accrual models exhibit management discretion or estimation errors, both of which reduce decision usefulness (Dechow et al. 2010). The equation is estimated by industry-year with at

(20)

20

least 5 observations. The ATA it is adjusted for performance matching following the approach used in Francis et al. (2008). Performance-matched abnormal total accruals

(PMATAit) are estimated as the difference between ATAit and median ATA for the ROA to which firm i belong. Large values of PMATA in both directions reflect lower accrual quality (Barua et al.2010).

Prior studies argue that corporate executives have more flexibility in manipulating earnings when using current accruals instead of long-term accruals (e.g., Bradshaw et al.2001; Guenther 1994; Becker et al 1998). Taking this into account the second measure to proxy for accrual quality is abnormal current accrual. The same approach to measure abnormal total accruals will be used to estimate abnormal current accruals. PPEit will be deleted from the equation and the following equation will be estimated by two-digit SIC industry and by year:

(2) Where:

TCA iτ =Total current accruals in year t= (∆CA iτ - ∆CLiτ ) - (∆Cash iτ - ∆STDEBT iτ)

(∆CA iτ =Change in current assets for firm i from year t-1 to year t; ∆CLiτ =Change in current liabilities for firm i from year t-1 to year t;

∆Cash iτ =Change in cash and short-term investment for firm i from year t-1 to year t; ∆STDEBT iτ =Change in current portion of long-term liabilities for firm i from year t-1 to year t

ɛ iτ =Error term

The residuals in the equation are a measure of abnormal current accruals (ACA). After

adjusting ACAi by median ACA for the industry ROA group performance-matched abnormal current accrual PMACA are derived. The absolute value of performance-matched current accruals is used to measure accrual quality.

The third model used to measure accrual quality in this study is the accrual estimation error from Dechow and Dichev (2002). This model separate accruals based on their

association with cash flows by regressing working capital accruals on cash from operations in the current period, prior period, and future period Francis et al. (2008). The model was

(21)

21

estimated cross-sectionally by year and by two-digit SIC industry as in (Francis et al 2008;Jones et al. 2008).

(3)

OCFit=Operating Cash flow of firm i in year t, and all variables are scaled by average total assets. The unexplained part of the change in working capital accruals is an inverse measure of accrual quality. A greater unexplained portion implies lower quality Francis et al. (2008).

Finally, the specification of the Dichow and Dichev (2002) model as was suggested by McNichols (2002) and as implemented by Jones et al. (2008) is used as the fourth measure of accrual quality. In this model beside regressing working capital accruals on cash from operations, the change in revenues and property, plant and equipment will be also take into consideration.

(4)

The equation was estimated cross-sectionally by year and two-digit SIC industry. In this equation the absolute value of the residuals is the measure of accrual quality. All the variables in the equation are define in previous equation.

3.1.1 Regression models

In order to examine whether accrual quality increase after the implementation of SOX for firms with female CFOs, the four different proxy for accruals quality will be regressed on the variable SOXPERIOD and FEMALECFO together with six other variables representing firm characteristics. To test hypotheses 1 and 2 the following equation was estimated:

AQiτ = a0 + a1 SOXPERIODiτ + a2 FEMALECFOiτ + a3 SIZEiτ + a4 BMiτ + a5 SGROWTHiτ + a6 ROAiτ + a7 OCFiτ + a8 LEVERAGEiτ + a8 OCF2iτ (5)

(22)

22

AQiτ =accruals quality measures (ABS_PMATAit, ABS_PMACAit,ABS_DDit, ABS_MDDit)

SOXPERIOD = dummy: if pre-SOX period:0 and post_SOX period:1 FEMALE =dummy: if CFO is a woman 1 and if CFO is man :0 SIZEit =natural logarithm of total assets for firm i in year t; BMit =ratio of book value of equity to market value

SGROWTHit =change in revenue for firm i from year t-1 to year t divided by revenue in year t-1

ROAit =return on assets measured as income before extraordinary items for firm i in year t divided by average total assets in year t;

OCFit =operating cash flow for firm I for year t divided by total assets for year t-1

OCF2 = quadratic of operating cash flow for firm I for year t divided by total assets for year t-1

LEVERAGE =book value of long-term debt divided by book value of equity

I expect, a negative relationship between the variable SOXPERIOD and the dependent variable accrual quality. Because the dependent variables are all inverse measures of accrual quality. Second, because of the higher penalties CFO’s will face due to misreporting, I expect that they will act more conservative to avoid higher fines and imprisonment.

Prior research (e.g., Vähämaa 2010; Geigher and North 2006) suggests that female executives are more risk-averse and conservative than their male counterparts. Hence, I predict that the coefficient of female CFO will be negative, also because the dependent variables are inverse measure of accrual quality.

Firm size has been used as control variable because prior research suggests firm size is negatively related to discretionary accruals Peni and Vähämaa (2010). In general large firms have higher political cost this incentive them to use more conservative accounting. I predict a negative coefficient for SIZE.

BM and SGROWTH are used as control variables for firm growth. Prior studies find evidence of a positive relationship between growth and accruals and negative relationship for book-to-market ratio (e.g., Lui et al., 2014; Menon and Williams, 2004). Sales growth can inflate the market expectations of future cash flow and also change the market value of the firm (Ahmed and Diellman, 2007). Therefore I predict a positive coefficient for growth and a negative coefficient for BM.

(23)

23

For the firm performance I included ROA and OCF as variable. ROA, because poor

performance is associated with lower accrual quality (Dichow and Dichev, 2002). I predict, a negative coefficient for ROA. Lui eta al. (2014) and Subramayan (1996) both find a negative asociation between operating cash flow and abnormal accruals. Considering that

the dependent variables are absolute accrual measure OCF2 is also included as control variable, and I predict a positive coefficient.

LEVERAGE is included as variable for the financial condition of a firm. Troubled companies may have strong incentive to use income-decreasing accruals (DeAngelo, 1994). Previous research finds evidence that LEVERAGE is negatively related to abnormal accruals ( Lui et al., 2014;Peni and Vähämaa, 2010). Sweeney (1994) suggest that firms manage earnings upward even after the violation of debt covenants. Evidence are ambiguous about the relation between LEVERAGE and abnormal accruals, therefore no prediction can be made.

To investigate whether the quality of accruals increase more for firms with female CFOs compared to firms with male CFOs in the post SOX period, the variable FEM*SOX is included in the following equation:

AQiτ = a0 + a1 SOXPERIODiτ + a2 FEMALECFOiτ + a3 FEM*SOX + a4 SIZEiτ + a5 BMiτ + a6 SGROWTHiτ + a7 ROAiτ + a8 OCFiτ + a9 LEVERAGEiτ + a10 OCF2iτ (6)

Because preceding research show evidence that female executives are more risk –averse and conservative, I predict that abnormal accruals decrease more for female CFOs than male CFOs after the passage of SOX. As a result I expect a negative coefficient between accruals and FEMSOX.

(24)

24

3.2 Sample and data selection

Data regarding CFO’s gender will be collected primarily from Execucomp database as was collected by Francis et al. (2014). In cases where data information about CFO gender was not available, I collected them manually from the 10-K fillings of the firm year. Data to measure accrual quality is collected from Compustat Annual Fundamentals, which covers most of the S&P1500.The preference for this data was mainly because SOX has been

implemented for US-based firms initially. Financial firms and utility firms are excluded from the sample SIC codes between 6000-6900 and between 4900-4900.

The sample period is from 1996-2006, and is divided into periods. The first period is from 1996-2001 this is the period before SOX (pre-SOX period) and the second period is from 2002-2006 for the period after SOX (post-SOX period). Reason why this sample was selected is because finding a larger sample was expected to be problematic in a short time, since data about CFO gender was not always available in Execucomp. The initial sample consisted of 12.858 firm-year observations after dropping firm-year observations with SIC codes 6000-6900 and 4900-4999 (1095), the sample consisted of 11.768 firm-year

observations.

(25)

25

4 Results

In this chapter the results of this study will be presented and discuss. In the first paragraph the descriptive statistics will be elaborated and the main differences will be explained. The second paragraph will present the result of the regression models for H1 and H2. The third paragraph will report the regression results regarding the difference in accrual quality for female CFOs in the post SOX-period (H3).

4.1 Descriptive statistics

Table 1 present the descriptive statistics of the accrual quality measures and the control variables by dividing the sample based on the SOX-period, pre-SOX period =0 and post-SOX period =1. The result of the Ranksum t-test is also included to examine the difference

between mean (median) in pre and post-SOX period. All continuous variables (except log transformed variables) are winsorized at 1 percent and 99 percent because there is a possibility, that outliers might influence the results. The total amount of firm years observations for the pre-SOX sample consist of .6727 firm-year observations and 6.233 respectively for the post-SOX period after dropping observations without data to estimate accruals based on the Jones model.

The mean (median) absolute performance-matched abnormal total accruals and current accruals are 0.037 (0.026) and 0.035(0.041) respectively in the pre-SOX period, compared with 0.043 (0.039) and 0.041 (0.039) in the post-SOX period. Both abnormal accruals measures, ABS_PMATA and ABS_PMACA are slightly more positive in the post-SOX period. Indicating a higher level of abnormal current and total accrual after the

implementation of SOX. Ranksum t-test show that the magnitudes of both measures differ on a 1% level p=-5.931 and p=-7.033, respectively for mean test and p=0.000 and p=0.000 for median test indicating that abnormal accruals increased in the period after SOX. These results are not consistent with hypotheses one; however the results are preliminary, and inferences can be made only after controlling for other factors in the regression models.

The mean and median of the other control variables are also provided in table 1 and Ranksum t-test show that ROA, OCF and OCF2 do not significantly differ between the pre and post- SOX period. In contrast, SIZE is significantly higher in the post-SOX period suggesting a higher level of total assets in comparison with the pre-SOX period. This also might implicate firm size have become larger over the years. BM, GROWTH and

(26)

26

LEVERAGE are significantly lower in the post-SOX period indicating a lower level of growth in comparison with the pre-SOX period. This is consistent with the results of Wang et al. (2011).

Table 2 presents the summary statistics for the accrual quality measures and control variables, by dividing the sample based on CFO gender. The sample consisted 209 firm-year

observations with a female CFO (1.64%) and 12.751 (98.36%) firms –year observations with male CFOs. The mean (median) absolute performance-matched abnormal total accruals and current accruals are 0.056 (0.044) and 0.050(0.044) for firms with female CFOs, compared with 0.040 (0.032) and 0.037 (0.033) respectively for firms with male CFOs. Ranksum t-test shows that although firms with female CFOs have a higher mean, absolute performance-matched abnormal total accruals and current accruals do not differ significantly between female and male CFOs (p=-0.293 and p=-0.309). The result contradict the study of Barua et. (2010). Given the fact that Barua et al. (2010) provide a significant lower mean of absolute

TABLE 1 Descriptive Statistics H1

Pre-SOX Post-SOX

Variables Mean Std.dev Q1 Median Q3 Mean Std.dev Q1 Median Q3

Ranksum t-test ABS_PMATA 0.037 0.143 -0.031 0.026 0.102 0.043 0.121 -0.016 0.039 0.106 -5.931 (0.000***) ABS_PMACA 0.035 0.134 -0.028 0.026 0.093 0.041 0.109 -0.012 0.039 0.099 -7.033 (0.000***) SIZE 6.640 0.178 5.588 6.590 7.655 6.920 1.925 5.907 6.893 8.026 -11.659 (0.000***) BM 5.313 8.033 1.341 0.233 4.245 4.397 7.026 1.456 2.267 0.364 2.354 (0.019**) SGROWTH 0.187 0.417 -0.007 0.097 0.262 0.129 0.300 0.009 0.092 0.198 3.161 (0.002**) ROA 0.028 0.213 0.008 0.053 0.104 0.020 0.235 0.009 0.053 0.102 0.711 (-0.477) OCF 0.103 0.149 0.046 0.102 0.170 0.099 0.151 0.049 0.104 0.167 -0.484 -0.629 LEVERAGE 0.794 9.765 0.009 0.346 0.868 0.649 8.155 0.000 0.268 0.686 8.515 (0.019**) OCF2 0.033 0.062 0.004 0.013 0.032 0.032 0.068 0.004 0.013 0.031 -0.140 (-0.889) ***,**,** Represent a significant levelof respectively 1%,5% and 10%

(27)

27

performance-matched abnormal total accruals and current accruals for female CFOs than for male CFOs.

SIZE and LEVERAGE are significantly lower for firms with female CFOs then for firms with male CFOs. Barua et al. (2010) also provide a lower mean for SIZE and

LEVERAGE for firms with female CFOs. Hence, this implicate, that firms with female CFOs have lower level of total assets and debt compared with firms with male CFOs. The control variables ROA, OCF and OCF2 have a higher mean for female CFOs than for male CFOs, indicating that female CFOs generate a higher positive cash flow and a higher level of return. Despite, the higher mean this does not significantly differ from firms with male CFOs. The mean for BM and GROWTH is also higher for firms with female CFO but does not differ significantly from firms with male CFOs.

The correlation matrix on table 2A page 48 provides the Pearson correlations between the abnormal current and the other control variables in the model. The correlation matrix provides the strength of the relationship between independent variable, dependent variable and the control variables. The abnormal current accrual is the dependent variable instead of the abnormal total accrual in the correlation matrix because prior research suggests that executives have more flexibility in manipulating current accruals. Abnormal current accruals are positively related to SOXPERIOD, which contradict the expectation for H1. The control variables SGROWTH, ROA, OCF and OCF2 are also positive related to abnormal current accruals. The variable SIZE and OCF is negatively related to abnormal current accruals. The variable of interest FEMALECFO is not significantly related to abnormal current accruals. Furthermore, there is only one sign of high collinearity between OCF and ROA this might influence the reliability of the model.

(28)

28

TABLE 2 Descriptive Statistics H2

FEMALE MALE

Ranksum t-test ABS_PMATA 0.056 0.132 -0.026 0.044 0.123 0.040 0.131 -0.024 0.032 0.104 -1.051 (-0.293) ABS_PMACA 0.050 0.122 -0.029 0.044 0.118 0.037 0.129 -0.021 0.033 0.096 -1.016 (-0.309) SIZE 6.447 1.585 5.219 5.939 7.654 6.959 1.687 5.882 6.816 7.908 5.014 (0.000***) BM 4.946 7.967 1.367 2.461 4.619 5.011 1.249 0.015 2.444 4.095 -0.259 (-0.795) SGROWTH 0.156 0.328 0.003 0.073 0.186 0.152 0.367 0.005 0.095 0.218 1.326 (-0.185) ROA 0.056 0.173 0.024 0.082 0.135 0.042 0.217 0.014 0.057 0.108 -3.759 (0.000***) OCF 0.123 0.129 0.055 0.122 0.200 0.114 0.159 0.057 0.11 0.174 -1,670 (0.095*) LEVERAGE 0.215 7,637 0.000 0.032 0.382 0.638 1.525 0.006 0.304 0.737 5.941 (0.000***) OCF2 0.040 0.052 0.005 0.017 0.044 0.030 0.074 0.005 0.013 0.033 -1.613 -0.107 ***,**,** Represent a significant levelof respectively 1%,5% and 10%

(29)

29

4.2 Regression Results

4.2.1 Results H1 and H2

To further examine whether accrual quality have increased in the post-SOX period (H1) and if firms with female CFOs are positively related to accruals quality (H2) the variables SOXPERIOD and FEMALE are regressed in the following equation:

AQiτ = a0 + a1 SOXPERIODiτ + a2 FEMALECFOiτ + a3 SIZEiτ + a4 BMiτ + a5 SGROWTHiτ + a6 ROAiτ + a7 OCFiτ + a8 LEVERAGEiτ + a8 OCF2iτ + ɛ iτ (5)

The dependent variables in this equation are absolute performance-matched abnormal total accruals (ABS_PMATA) and (ABS_PMACA). SIZE, BM, GROWTH, ROA, OCF, LEVERAGE and OCF2 are included as control variables. The regression results for H1 and H2 are presented in table 4.The adjusted R2 of the estimated model are relatively high, varying between 31.4 and 42.1 percent. This is a relatively high-adjusted R2 in comparison with other study (e.g., Barua el al., 2010; Peni and Vähämaa, 2010). The coefficient of SIZE is significantly negative in both regressions, indicating that size is associated with lower level of abnormal accruals. This is in line with the result of Barua et al. (2010) and Francis et al. (2008). Consistent with prior studies GROWTH, ROA and OCF2 are positive and

significantly related to higher level abnormal accruals (e.g., Dichow and Dichev 2002; Barua et al., 2010).

The variable LEVERAGE is negatively related to the absolute measure of performance matched total accruals but this is not significant. In contrast, there is a negative significant coefficient in relation to the absolute measure of performance matched current accruals.

The first variable of interest SOXPERIOD is in both regression not significant, there is a negative coefficient for absolute value of performance matched total accrual and for absolute abnormal current accruals is the coefficient zero, indicating that accrual quality did not significantly increased in the post SOX period. This result contradicts the evidence from Lobo and Zhou (2006). The second variable of interest is FEMALECFO; in both regressions shows an insignificant coefficient. Which implicates that female CFOs are not significantly associated with higher level of accrual quality. This is in line with the research of Ye et al. (2010). However this can also be due to the lack of female CFOs in the examined sample.

(30)

30

To test whether there is no high correlation between the independents variable the Variance Inflation Factor (VIF) have been performed. The results are presented in table 5 on page 46.The highest VIF is 2.1 and the mean VIF is 1.31, indicating no sign of multi-collinearity.

4.2.2 Results H3

Furthermore, to examine whether accrual quality have increased more for firms with female CFOs than for firms with male CFOs in the post-SOX-period, accrual quality have been regressed on the variable FEMCFO*SOX, FEMALECFO and SOXPERIOD in equation 6 and together with the control variables from equation 5. AQ represent the two accrual quality measures; the absolute value of performance-matched abnormal total accruals and the

absolute value of performance-matched abnormal current accruals. The regression results of the following equation are presented in table 6:

AQiτ = a0 + a1 SOXPERIODiτ + a2 FEMALECFOiτ + a3 FEMCFO*SOX + a4 SIZEiτ + a5 BMiτ + a6 SGROWTHiτ + a7 ROAiτ + a8 OCFiτ + a9 LEVERAGEiτ + a10 OCF2iτ (6)

Table 4

Regression of Absolute Value of Performance-Matched Abnormal Accruals

ABS_PMATA ABS_PMACA

Variables Predicted Sign Coefficient p-value Coefficient p-value Intercept 0.091 0.000*** 0.142 0.000*** SOXPERIOD - -0.001 0.634 0.000 0.883 FEMALECFO - 0.004 0.661 0.000 0.936 SIZE - -0.001 0.038** -0.008 0.000*** BM - 0.000 0.987 0.000 0.686 SGROWTH + 0.041 0.000*** 0.053 0.000*** ROA - 0.602 0.000*** 0.619 0.000*** OCF - -0.636 0.000*** -0.673 0.000*** LEVERAGE ? -0.001 0.493 -0.001 0.038** OCF2 + 0.119 0.000*** 0.016 0.000*** Adjusted R2 31.4% 42.1% No.observations 11.477 11.477 Mean VIF 1.31 1.31

(31)

31

The result for H3 is similar to the results of H1 and H2. The adjusted R2 for model of H3 is the same as H1 and H2.The variable of interest FEMCFO*SOX is negatively related to the absolute value of performance-matched abnormal total accruals and the absolute value of performance-matched abnormal current accruals, this is conform expectation. But the p-value of both regressions indicates that these results are nog significantly negative. This indicates that the implantation of SOX did not have an incremental effect on the reported accrual quality of female CFO. A (VIF) have been performed (table 7,page 46) the mean VIF is 1.63 and the highest VIF is 2.79 indicating no sign of multi-collinearity.

Table 6

Regression of Absolute Value of Performance-Matched Abnormal Accruals

ABS_PMATA ABS_PMACA

Variables Predicted Sign Coefficient p-value Coefficient p-value Intercept 0.091 0.000*** 0.142 0.000*** SOXPERIOD - -0.001 0.743 0.000 0.822 FEMALECFO - 0.017 0.221 0.006 0.573 FEMCFOSOX - -0.020 0.230 -0.009 0.518 SIZE - -0.001 0.039** -0.008 0.000*** BM - 0.000 0.987 0.000 0.686 SGROWTH + 0.041 0.000*** 0.053 0.000*** ROA - 0.603 0.000*** 0.619 0.000*** OCF - -0.637 0.000*** -0.673 0.000*** LEVERAGE ? 0.000 0.497 -0.001 0.039** OCF2 + 0.120 0.000*** 0.016 0.419 Adjusted R2 31.4% 42,2% No.observations 11.477 11.477 Mean VIF 1.63 1.63

(32)

32

4.3.3 Additional Analysis

To control whether the sign and significance of the variables in table 1, 2, 4 and 6 are generally similar for accruals quality measures based on the model of Dichow and Dichev (2002). The descriptive statistics of the two accrual quality measure based on DD (2002) model is presented in table 8 and 9.

The results for H1 are provided in table 8. The total amount of firm years observations for the pre-SOX period 37.330 and 24.623 respectively for the post-SOX period The mean (median) of the accrual estimation errors according the DD (2002) model are 0.003 (0.015) and

0.003(0.079) respectively in the pre SOX period, compared with 0.002(0.085) and

0.001(0.083) in the post SOX period. This suggests a decrease in accrual estimation error in the period after the implementation of SOX. Except, the Ranksum t-test shows that the accrual estimation error in the pre and post SOX do not differ significantly. The mean

(median) changes for the variables SIZE, BM, SGROWTH, ROA, OCF, LEVERAGE, OCF2 is in comparison with table 1 the same. Table 8A on page 49 provide the correlation matrix with the accrual estimation error based on the extended DD (2002) model as dependent variable. Accrual estimation errors are negatively related to SOXPERIOD, this in line with the predicted expectations for H1.

(33)

33

***,**,** Represent a significant level of respectively 1%,5% and 10%

Table 9 provides the descriptive statistics for female and male CFO based on the DD (2002) model. The sample consisted of 11.088 firm year observations which of 10.924 have a male CFO and 164 have female CFO (1.47%). The mean (median) for the accrual estimation errors ABS_DD and ABS_MDD are smaller for firms with male CFOs -0.003(0.002) and

0.007(0.000) than for firms with female CFO 0.004 (-0.006) and 0.006 (0.000). The mean for the variable SIZE is significantly lower for firms with female CFO compared to firms with male CFOs. The mean of the variables in table 9 are similar to those in table 2. Except, the mean (median) of the variable SGROWTH is 0.142 (0.169) for firms with female CFO and 0.144 (0.212) respectively for firms with male CFO, indicating a lower level but not

significant level of growth for firms with female CFOs. TABLE 8

Descriptive Statistics H1

Pre-SOX Post-SOX

Ranksum t-test ABS_DD 0.003 0.224 -0.041 0.015 0.08 0.002 0.002 -0.049 0.014 0.085 0.265 -0.791 ABS_MDD 0.003 0.215 -0.045 0.013 0.079 0.001 0.001 -0.052 0.011 0.083 1494 -0.135 SIZE 6.787 1.563 5.697 6.633 7.721 7.139 7.139 6.087 6.985 8.079 -13.218 (0.000***) BM 5.708 8.515 1.497 2.542 4.585 4.316 4.316 1.571 2.352 3.704 5996 (0.000***) SGROWTH 0.169 0.347 -0.007 0.093 0.245 0.124 0.124 0.013 0.091 0.19 1.655 (0.098*) ROA 0.050 0.152 0.012 0.059 0.113 0.0478 0.048 0.015 0.056 0.105 1.397 0.162 OCF 0.118 0.124 0.057 0.112 0.179 0.115 0.115 0.059 0.109 0.169 (-1.472) -0.141 LEVERAGE 0.734 1.053 0.009 0.319 0.179 0.514 0.514 0.002 0.276 0.674 4906 (0.000***) OCF2 0.029 0.042 0.005 0.014 0.035 0.025 0.025 0.004 0.013 0.031 3.703 (0.000***)

(34)

34

***,**,** Represent a significant level of respectively 1%,5% and 10%

A separate regression is estimated where the dependent variables AQ represent the two accrual quality measures based on accruals estimation errors of the Dichow and Dichev 2002 model (ABS_DD and ABS_MDD). The regression result for H1 and H2 are presented in table 10. The adjusted R2 for the estimated model are relatively high again, varying between 36.5 and 45.8. The VIF is also conducted to measure the extent of multi-collinearity. The mean VIF is 1.54 indicating no sign of multi-collinearity. Further, result about the VIF is provided in table 10A page 47.

The results in table 10 are quite similar to the results in table 4, except for the variable SIZE and the main variable SOXPERIOD. The variables SGROWTH and OCF2 are

significant and in the predicted directions and this align with the study of Barua et al. (2010). The variable SIZE is not significant related to accrual quality. SOXPERIOD has a significant negative coefficient on a 5% level in both regressions, which means lower accrual estimation error in the post SOX period. Implicating higher accrual quality in the post-SOX period. This is in line with the majority of literature indicating an improvement in earnings quality after the implementation of SOX (e.g., Cohen et al.2008; Lobo and Zhou 2006). The result for H3 based on the DD (2002) model is presented in table 11. Additional information about the

TABLE 9

Descriptive Statistics H2

FEMALE MALE

Ranksum t-test ABS_DD 0.004 0.076 -0.038 -0.006 0.046 -0.003 0.111 -0.042 0.002 0.043 0.036 0.971 ABS_MDD 0.006 0.087 -0.043 0.000 0.042 -0.007 0.111 -0.045 0.000 0.039 -0.521 0.602 SIZE 0.646 0.166 5.219 5.939 7.654 6.964 1.569 5.879 6.812 7.905 4.998 (0.000***) BM 5.024 1.239 1.367 2.461 4.619 5.035 7.975 1.538 2.451 4.098 -0.234 0.815 SGROWTH 0.142 0.316 -0.004 0.066 0.169 0.144 0.293 0.006 0.092 0.212 1.554 0.120 ROA 0.066 0.153 0.024 0.082 0.135 0.046 0.140 0.014 0.057 0.107 -3.766 (0.000***) OCF 0.129 0.129 0.051 0.121 0.200 0.116 0.116 0.058 0.11 0.174 -1.493 0.135 LEVERAGE 0.209 1.501 0.000 0.032 0.382 0.635 7.706 0.005 0.302 0.735 5.916 (0.000***) OCF2 0.033 0.043 0.005 0.017 0.045 0.027 0.039 0.005 0.013 0.032 -1.198 0.231

(35)

35

mean VIF is provided in table 11A page 47.The results in table 11 are also similar to table 6 with the same exceptions as in table 8 for the variables SIZE and SOXPERIOD.

TABLE 10

Regression of Absolute Value of Accruals Estimation Errors

ABS_DD ABS_MDD

Variables Predicted Sign Coefficient p-value Coefficient p-value Intercept 0.064 0.000*** 0.056 0.000*** SOXPERIOD - -0.005 0.002** -0.005 0.003** FEMALECFO - 0.003 0.659 0.010 0.203 SIZE - -0.001 0.300 0.000 0.975 BM - 0.000 0.966 0.000 0.413 SGROWTH + 0.034 0.000*** -0.001 0.710 ROA - 0.642 0.000*** 0.565 0.000*** OCF - -0.855 0.000*** -0.767 0.000*** LEVERAGE ? 0.000 0.461 0.000 0.581 OCF2 + 0.196 0.000*** 0.145 0.000*** Adjusted R2 45.8% 36.5% No.observations 9.249 9.249 Mean VIF 1.54 1.54

(36)

36

.

TABLE 11

Regression of Absolute Value of Accruals Estimation Errors

ABS_DD ABS_MDD

Variables Predicted Sign Coefficient p-value Coefficient p-value Intercept 0.064 0.000*** 0.056 0.000*** SOXPERIOD - -0.005 0.002** -0.006 0.003** FEMALECFO - -0.001 0.942 0.003 0.773 FEMCFOSOX - 0.007 0.621 0.011 0.450 SIZE - -0.001 0.297 0.000 0.968 BM - 0.000 0.966 0.000 0.413 SGROWTH + 0.034 0.000*** -0.001 0.715 ROA - 0.642 0.000*** 0.565 0.000*** OCF - -0.855 0.000*** -0.767 0.000*** LEVERAGE ? 0.000 0.465 0.000 0.588 OCF2 + 0.196 0.000*** 0.145 0.000*** Adjusted R2 45.8% 36.5% No.observations 9.249 9.249 Mean VIF 1.73 1.73

(37)

37

5. Conclusion

In the previous section, the descriptive statistics and the regression results were presented. In this section the results will be evaluated and discuss. After that the results are evaluated, a conclusion will be formulated. Furthermore, the limitation of this study will be discussed and avenues for future research will be provided.

5.1 Discussion

In order to answer the first hypothesis and the second hypothesis the variables SOXPERIOD and FEMALCFO were regressed together with other six control variables for firm

characteristics. For the first hypothesis: Accrual quality has increased in the period after the implementation of SOX, a positive relation was expected. The result based on the Modified Jones model show a negative coefficient indicating an increase in accrual quality, this was however not significant. The result based on the Jones Model show a coefficient of 0.000 and a p-value 0.936, which is also indicate an insignificant relation. This contradict the study of Lobo and Zhou (2006) and Cohen et al. (2008) who find a significant negative association between the level of accruals and SOX period. An explanation for this can be that that the sample used in this study might be too small to produce significant results. In contrast, when the first hypothesis was tested based on the DD (2002) and the extended version of the DD (2002) model the results show a significant negative coefficient. Indicating that the

estimation error in post SOX period was relatively low, implying an increase in accrual quality. This is in line with the predicted expectation and also in line with the research of Lobo and Zhou (2006) and Cohen et al. (2008). An explanation for this can be the fact that the sample based on the Jones model was smaller than the sample based on the DD model due it to insufficient information to calculate the level of accruals. Therefore, H1 is supported.

The result of whether female CFOs are positively related to accrual quality based on the Jones model shows, an insignificant positive coefficient. This suggests that the level of accrual is higher for female CFOs but not significant. The result based on the DD (model) show also a positive coefficient indication a higher level of accrual estimation error for female CFOs but not really significant. The result based on both accrual quality measures contradicts previous studies that indicate that female executives are related to higher level of accrual quality (e.g., Barua et al., 2010; Peni and Vähämaa, 2010). A possible explanation for

(38)

38

the contradicting results is the relatively low number of female CFOs in the examined sample. A possible solution could have been using CEO gender instead of CFO gender but female CEO number was lower than CFO. In this research the focus was mainly on female CFO because as mention before it is the main responsibly of a CFO to watch over the

financial reporting process and make related decisions (Feng et al., 2011). The sample of Peni and Vähämaa (2010) has a lot more firm-year observations than this study. This can also function as an explanation for why the result in this study might be insignificant. Another solution to increase the explanatory power of this research might have been increasing the sample but it would have been unfeasible in the amount of time for this study because data regarding CFO gender was hand-collected.

To test whether accrual quality have increase more for female in the post-SOX period the variable FEMCFO* SOX was included in the model together with the other control variables. The two accrual quality measure based on the Jones model bot indicate an insignificant negative coefficient. Implicating that the level of accrual is lower for female CFOs but not significant lower than their male counterparts. The accrual quality measures based on the DD (2002) model both show an insignificant positive coefficient, contradicting the predicted expectations. As was mention before that a possible explanation for the

insignificant result might be because of the small sample. Another explanation can be that male CFOs might have become more conservative than women after the implementation of SOX, but this has to be further investigated. Summarizing, the results of the hypothesis together accrual quality did not improved more for female CFOs than for male CFOs after the implementation of the Sarbanes-Oxley ACT of 2002.

5.2. Conclusion

The objective of this research was to examine whether the Passage of the Sarbanes-Oxley act have led a different impact on female CFOs than male CFOs in the context of accrual quality. The Passage of the Sarbanes-Oxley act was passed into law after the major accounting

scandals such as Enron and WorldCom. The main objective of this law was to restore investor confidence in the market. After the implementation of SOX CFO are more accountable for the financial reporting and disclosure of the firm. Several researches have provided evidence that firms are more conservative in their reporting after the

CFO gender and accrual quality in the post-SOX period