Are female CEO’s more susceptible for Big-4 auditor monitoring on the usage of classification shifting?

Are female CEO’s more susceptible for Big-4 auditor monitoring on the usage of classification shifting?

Master Accountancy & Controlling 2019/2020 Student Name: Stevan Winkels

Student Number: S2370670 Supervisor: prof. dr. C.K. Hoi

Date: 21 June 2020 Word Count: 7966

Abstract

Female CEOs are become increasingly present within companies in the recent years, in part due to the implementation of quorums that countries are starting to enforce. Given the increasing number of female CEOs there is a need for research that analyses the differences between male

and female CEOs. This paper expands on previous literature regarding Big-4 auditors as part of the governance structure of companies in mitigating classification shifting and the research on

CEO gender differences. This research contributes to the existing literature in analysing the gender differences of the CEO regarding the monitoring of external auditors. Using financial data

from U.S. companies from 2003 to 2019 this research aims to determine if the mitigating effect of Big-4 auditors differ when the CEO is female compared to male CEOs. Based on the results there is evidence that female CEOs use less classification shifting when there is a Big-4 auditor

compared to male CEOs. The results provide useful insight in the differences of governance monitoring from the auditors for male and female CEOs. Shareholders and regulators who are aware of the gender differences can use these results and other research on this topic to set up

a more efficient and effective corporate governance structure.

1

Introduction:

Classification shifting has seen increased attention from researchers ever since McVay (2006) introduced this new type of earnings management. Classification shifting is a more subtle form of earnings management where the expertise of auditors lies regarding the accounting standards.

Due to more awareness and in some countries enforced quorums like Norway the number of female CEOs have increased over the recent years and researchers like Zalata et al. (2018) started to investigate more in depth the gender differences regarding classification shifting. This research aims to determine the influence of Big-4 auditors with female CEOs. Based on previous literature of Becker et al. (1998) and Francis et al. (1999) Big-4 auditors are more effective in mitigating earnings management, whereas Haw, Ho & Li (2011) found evidence of Big-4 auditor mitigation of classification shifting for East Asian companies. The finding of Zalata et al. (2018) indicate that there are indeed differences between male and female CEOs which this papers aims to further explore. This leads to the goal of this paper of analysing the mitigating effect of Big-4 auditors when the CEO is female compared to male CEOs. Depending on the CEO gender there could be different decisions needed to enable effective and efficient monitoring for

shareholders and regulators.

This paper uses financial data from U.S. companies from 2003 to 2019 to analyze the influence of Big-4 auditor presence and the gender differences of the CEO. This research expands the model of McVay (2006) to detect classification shifting. The results of this research indicates that Big-4 auditor presence mitigates classification shifting consistent with previous findings of Haw, Ho & Li (2011) from East Asian companies. There is also evidence based on the results that female CEOs use less classification shifting when there is a Big-4 auditor present supporting the literature of Zalata et al. (2018) who found that female CEOs used less classification shifting as well by expanding on the governance influences. This paper contributes to the existing literature of Big-4 auditor influence on earnings management and classification specifically and on the topic of CEO gender differences. CEO gender differences have been getting more attention due to the increasing availability of data, therefore this research gives future researchers a

foundation to investigate more differences due to gender.

Literature review:

Research regarding classification shifting is growing more abundant ever since McVay started

the field in 2006. Classification shifting is ‘’deliberate misclassification of items within the income

statement’’ (McVay, 2006). This means that the core expenses are shifted towards special items

that occur within the same year (Mcvay, 2006). Core expenses are the expenses associated

with the daily operations of the company and are therefore relatively stable (Lipe, 1986; Fairfield,

Sweeney & Yohn, 1996), whereas special items are unusual and infrequent and also highly

transitory (Lipe, 1986). Investors treat these costs accordingly (Lipe, 1986; Bradshaw and Sloan,

2002) because they know the characteristics of these special items. Previous research also

shows that special items generally tend to be excluded by managers (Lougee and Marquardt,

2004) and analysts (Philbrick and Ricks, 1991). Classification shifting differs from accrual

earnings management and real earning management because classification shifting doesn’t

change GAAP earnings (Mcvay, 2006). Classification shifting also doesn’t change past or future

earnings (Mcvay, 2006) whereas the other methods do. The reason classification shifting is a

problem, even though the profits in a given year doesn’t change, is that it tampers with the

2 informative function of the income statement. The users of financial statements value different components of the profit & loss statement differently (Lipe, 1986; Elliott and Hanna, 1996;

Francis, Hanna & Vincent, 1996; Davis, 2002; Bradshaw and Sloan, 2002). There is a distinction between core expenses and revenues and special/incidental expenses and revenues. The informative value of the core expenses is larger due to the more permanent/structural nature of these costs and revenues (Lipe, 1986; Fairfield et a, 1996). This indicates that core expenses are a better prediction of future performance of the company then the total expenses including the special items.

This research paper aims to add to the existing literature regarding the mitigating effect of Big-4 auditors on the usage of classification shifting and on the gender differences of the CEO. The first part will discuss the role of Big-4 auditors regarding earnings management and classification shifting specifically. The next part will discuss the relation between CEO gender and

classification shifting, combining it with the monitoring of Big-4 auditors.

Big-4 auditors:

The profession of the external auditor finds its value in fostering the trust of the public and encourages them to believe that the financial statements are true and fair (Sikka, 2009). The external auditor tries to reduce the agency problems due to the separation between ownership and the management of companies. Agency costs arise because of the separation of ownership from control since managers as agents will not always act in the best interest of the principal (Jensen and Meckling, 1976). As the agency costs due to the separation of ownership and control increase there is a demand for higher-quality information, either voluntarily undertaken by managers as a bonding mechanism or externally imposed as a monitoring mechanism by

stockholders and/or debtholders (Watts and Zimmerman, 1986). This is in line with other previous research that found that the principal(s) need to spend resources to monitor the agent to prevent actions that reduce firm value (Gomez-Meija & Wiseman, 1997; Tosi & Gomez-Meija, 1989). The audit service are demanded as monitoring devices because of the potential conflicts of interest between owners and managers as well as those among different classes of security holders (Watts, 1977; Watts and Zimmerman, 1981; Benston, 1980). The information asymmetry between a company’s manager and owners is also one of the factors that leads to a demand for auditing of the financial statements (Jensen and Meckling, 1976) to reduce the information asymmetry. The greater the extent of the agency conflicts, the higher the quality of auditing needed to make management credible to current and potential investors (DeFond, 1992). To increase the credibility of the financial statements for the users of these financial statements, the auditor performs a financial statement audit in which the auditors gather evidence and provide investors with ‘a high level of assurance that the financial statements follow generally accepted accounting principles’ (Whittington and Pany, 2001). This provides more faith in the financial statements for all the users of these financial statements about the company.

Earnings management:

Earnings management is one of the behaviours that can arise because of the separation of

ownership and control of companies. There are different definitions of earning management that

have been used in previous research. Schipper (1989) defined it as ‘purposeful intervention in

the external reporting process with the intent of obtaining private gains’. Another definition of

Healy & Wahlen (1999) state that ‘earnings management occurs when managers use judgement

3 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’. Whatever definition you use the essence is that there is purposeful adjusting of accounting numbers. Given the opportunity for selfish behaviour of the management the external auditor can be important to detect and correct the manipulation of the accounting numbers. Audit quality is important in this aspect to detect earnings management therefore Big-4 auditors should provide higher audit quality based on the previous research discussed above. Further research on the relation between earnings management and big-4 auditors support this assumption.

Constraining influence of Big-4 auditors:

Becker et al. (1998) researched the constraining effect of Big-6 Auditors (now Big-4) on

discretionary accruals. The underlying assumption of their research is that management wants to increase their earnings through the discretionary accruals. Based on their analysis on the accruals compared to Big-6 Auditor clients and the non Big-6 Auditor clients their research shows that the companies audited by non Big-6 auditors have more income increasing accruals.

These results do support the idea that Big-6 auditors have a higher mitigating effect. Recurring explanations of higher audit quality of Big-4 auditors are due to the increased wealth exposed in case of litigation (Dye, 1993) and the difference in reputational risk in case of litigation

(DeAngelo, 1981). The research of Eshleman and Guo (2014) builds further on the literature previously mentioned to determine if the Big-4 auditors indeed do provide higher audit quality.

Eshleman and Guo (2014) used a propensity-score matching procedure to determine if the Big-4 auditors do provide higher audit quality. They looked specifically at the chances of disclosing accounting restatements as a measure of audit quality. After controlling for the endogenous choice of auditor they find support that clients of Big-4 auditors are less likely to have accounting restatements compared to non Big-4 auditors. Although accounting restatements may not be completely reliable as a measurement it does indicate that Big-4 auditors provide higher audit quality. A meta-analysis of 48 studies performed by Lin and Hwang (2010) looked at the relation between earnings management and audit quality. The underlying assumption behind audit quality and earnings management was that higher audit quality is better able to constrain

opportunistic behaviour and reduce information risk that the financial statements contain material misstatements or omissions. Their research found a significant negative relation between the auditor size. These findings are in line with what Becker et al. (1998) and Francis et al. (1999) have found previously. Francis et al. (1999) also looked at the accruals similarly to Becker et al (1998) and this is especially important regarding earnings management. Their research focused on companies listed on the NASDAQ and focused on the discretionary accruals of the

companies. Even though they found that companies that had an Big- 6 auditor showed higher total amount of accruals the discretionary accruals were lower compared to companies that had a non Big-6 auditor. This supports the idea that Big-6 auditors are better able to mitigate

opportunistic behaviour by the management. There is plenty of research to support this relation:

Choi et al. (2018) concluded their research with the following findings ‘’Although we caution against reaching definitive conclusions at this early stage, our evidence lends some support to the intuition that mitigating real earnings management (REM) can be more effectively

accomplished by encouraging firms to hire higher-quality auditors (such as Big 4 auditors)’’ and

Houge et al. (2017) found in their research among companies in India that high audit quality

reduced earnings management. DeFond and Jimbalvo (1993) found that auditor - client conflicts

relating to income-increasing accounting practices are more likely to occur if the auditor is one of

4 the Big Eight (now Big-4), and conclude that these audit firms are more likely to resist

managerial pressure and maintain an independent opinion also enhancing the quality of the financial statements. This research specifically looks at classification shifting instead of the broader earnings management. The research of Kim et al (2003) provides insights that Big-4 auditors should be able to mitigate opportunistic income-increasing behaviour regarding the core earnings of the company. Kim et al. (2003) researched the differences of the effectiveness of Big-6 auditors (now Big-4) regarding income-increasing or income-decreasing behaviour of the management. They concluded that ‘Our results show that Big 6 auditors are more (less) effective than non–Big 6 auditors in the presence (absence) of reporting incentive conflicts between the two issuers of financial statements’ (Kim et al., 2003). After the research of McVay (2006) the field of classification shifting is getting more attention from researchers. McVay (2006) concluded in her research that even though classification shifting doesn’t use accruals it is still a form of earnings management and therefore previous research in the field of earnings management and the impact of audit quality can be used to build expectations and perform tests. Further research used this to investigate the relation between the Big-4 auditors and classification shifting

extending the previous literature on earnings management. There is interesting research of Desai & Nagar (2016) regarding Big-4 auditors and reporting on classification shifting. McVay (2006) indicates that it is difficult for auditors to detect misclassification of items. Considering this difficulty of detecting classification shifting Desai & Nagar (2016) wanted to research if auditors are unable to detect classification shifting or are unwilling to report on it. Desai & Nagar (2016) conducted their research through experiments with auditors who were working in the field with some experience. The results of the research of Desai & Nagar (2016) shows that the auditors were able to detect classification shifting but the reporting/acting on this was more dependent on the legal regime of the country. Although Desai & Nagar (2016) acknowledge that more research is needed regarding the impact of stronger legal regime it does show that auditors are able to detect classification shifting, but based on the study of Desai & Nagar (2016) ‘it can be inferred that the presence of weak legal institutions reduces the litigation risk faced by auditors which make them less likely to report misclassifications’. This finding is supported by previous research on the differences between Big-4 auditors and non Big-4 auditors on audit quality and earnings management mentioned above. Research of Haw, Ho & Li (2011) looked at classification shifting within East Asian economies to determine if there is also classification shifting inside these companies. They also looked at the impact governance mechanisms have on classification shifting. Haw, Ho & Li (2011) used financial data from eight East Asian economies for their analysis and found evidence that East Asian companies similar to American companies use classification shifting to increase the core earnings. They also found that Big-4 auditors were effective in mitigating the usage of classification shifting, but only in countries with strong legal institutions. Although they don’t analyse American companies their findings do support the idea that Big-4 auditors are effective in mitigating classification shifting when there are strong legal institutions present. Research of Joo & Chamberlain (2017) investigated the mitigating influence of different aspects of corporate governance on classification shifting and income shielding by the CEO. They used financial data from the U.S. for their research over the period 1995 to 2012.

This research covered both the pre SOX and post SOX environment regarding governance. The

results show that firms were less willing to shield CEO in the usage of classification shifting post

SOX regulation. And strong governance was associated with less classification shifting and

switching to a Big-4 auditor was part of the strong governance giving more support of the

mitigating effect of Big-4 auditors on classification shifting. Although Joo & Chamberlain (2017)

did not analyse the Big-4 auditor presence on its own it does provide some support and is

5 consistent with the findings of Haw, Ho & Li (2011) from the East Asian companies they researched.

This research aims to determine if this relation holds true in the case of classification shifting.

Although classification shifting isn’t income-increasing because the GAAP earnings aren’t changed (McVay, 2006), the core earnings are subjected to income-increasing behaviour. The research of Desai & Nagar (2016) indicated that auditors are able to identify classification shifting and with the increased litigation risks of Big-4 auditors and reputational damage based on previous research stated above the Big-4 auditors should be more effective in mitigating opportunistic classification shifting and the results from Haw, Ho & Li (2011) and Joo &

Chamberlain (2017) provide some evidence of the mitigating effect of Big-4 auditors on classification shifting.

This leads to the first hypothesis of this research:

H1: The presence of a Big-4 auditor has a stronger mitigating effect on the level of classification shifting compared to a non Big-4 auditor.

CEO gender:

Previous research on CEO gender found numerous differences between male and female CEOs that indicates that this variable is worth exploring in a new setting. The goal of this study is to determine if the presence of a big-4 auditor has a greater influence on mitigating the usage of classification shifting for female CEOs compared to male CEOs.

Researchers from the psychology field have been researching differences between men and women for a long time. Byrnes, Miller and Schafer (1999) have performed a large meta analysis of 150 studies to determine the differences regarding risk taking behaviour. In this meta analysis the researchers identified 16 different types of risk taking to allow for more in depth analysis of different types of risk taking. Byrnes, Miller and Schafer (1999) reviewed and coded research that directly compared men and women along various types of risk taking behaviour. Of the 16 different risk taking types men were more likely to engage in more risky behaviour in 14 out of the 16. The highest risk taking types for male out of the 14 were: Physical skills, Intellectual skills and risky experiments (Byrnes, Miller and Schafer, 1999). Intellectual skills and risky

experiments risk taking are 2 fields that are of interest when you translate this to the financial and corporate setting. Schubert et al. (2000) used an experimental setting to determine the differences in risk taking between men and women in a financial setting found in previous studies through surveys. They performed an experiment with undergraduates in Switzerland to determine if there were gender differences confronted with building investment portfolios and risk tolerance. The findings didn’t support the gender differences between men and women when there is enough information to base the decisions on. The gender difference observed only was prevalent whenever the information was ambiguous, which could be explained based on the risk perception and the perception of their own competence. This could lead to more

overconfident behaviour from the men which is supported by the findings of the research of

Huang and Kisgen (2013) who looked at the differences in a corporate setting. Powel and Anisc

(1997) also performed an experiment with undergraduates and postgraduates who are familiar

with financial decisions. The findings were interesting because in general women are more risk

averse than men, but there was also a difference in strategy between men and women. In

6 particular, Powel and Ansic (1997, page number 622) conclude that 1) ’’as females are less risk propensive, they tend to focus on strategies which avoid the worst situation to gain security.’’

and 2) ‘’As males are more risk propensive, they tend to focus on strategies which they think will achieve the best possible gains’’.

Ford and Richardson (1994) looked at ethical behaviour, Ford and Richardson (1994) performed a literature review to investigate ethical decision making. They reviewed 14 papers that

specifically looked at the variable gender and found that seven found that women acted more ethical and the remaining seven either found no significant results or found no difference between women and men. Although the results are somewhat mixed there is some support regarding increased ethical behaviour of women. On the other hand the research of Huang and Kisgen (2013) used financial data to investigate the differences in investment decisions the executives made and found that the male executives were more likely to engage in acquisitions and take on debt, both decisions that involve a higher degree of risk. Huang and Kisgen (2013) additionally specifically looked at the changes after a transition from a male to female executive compared to a male to male transition. The results of their research indicated that male

executives appear to be more overconfident which was measured as the number of projects undertaken by a CEO and the amount of higher risk projects undertaken such as acquisitions.

Female CEOs undertook a lower amount of projects then male CEOs. Barber and Odean (2001) looked at overconfidence in men regarding trading in the equity market. They looked at account data of 35.000 households from a large discount broker to investigate the investment decisions made by men and women. They found support that men are generally more overconfident based on the trading volume and the risk appetite of the trades, sometimes men took a trade when the expected gains were negative. Khan and Vieito (2013) looked at the differences in risk at the corporate level through the CEO gender. Their research looked at financial data of companies to determine the level of corporate risk. Khan and Vieito (2013) measured risk as the natural logarithm of the standard deviation of returns calculated over 60 months with Black and Scholes methodology. The results from their research show that companies that were led by female CEO’s were associated with less corporate risk. The research of Ho et al. (2015) used financial data to determine the differences in financial reporting between male and female CEO’s. Ho et al. (2015) linked ethical and risk aversion behaviour of women with more conservative financial reporting because women are more timely in reporting bad news. They found evidence that companies with a female CEO report earnings more conservatively especially for firms with a high risk of litigation. Palvia et al. (2014) also investigated risk aversion and accounting

conservatism of female CEO’s or chairwomen. They used archival data from the banking sector to investigate risk and conservatism. They looked in particular at the bank capital ratios and default risk and found that banks with female CEO’s or chairwoman are associated with more conservatism, lower default risk and higher levels of equity capital. These findings support the more risk averse behaviour assumption of females within a corporate setting. Research of Sun et al. (2017) looked at fraudulent reporting and the influence of CFO demographic

characteristics. They used financial data from Chinese firms and CFO characteristic data to investigate fraudulent reporting. From their findings female CFO’s are associated with lower fraudulent reporting and although the research doesn’t cover CEO’s it does support the previous research on the differences between men and women in corporate settings.

Zalata et al. (2018) also researched gender differences in CEO’s regarding risk taking. Their

study looked at classification shifting over a period ranging from 1992 to 2014 to determine if

there were any differences in the usage of classification shifting between male and female

7 CEO’s. Interestingly enough they found that before the implementation of the Sarbanes-Oxley (SOX) regulation there were no differences between the male and female CEO’s regarding the usage of classification shifting. After the implementation of SOX the results show that female CEO’s significantly reduced the usage of classification shifting, whereas this wasn’t the same for male CEO’s. This indicates that female CEO’s are more risk-averse compared to male CEO’s, although the researchers point out that this could also be the result of the enhanced regulatory environment after SOX. Unfortunately Zalata et al. (2018) didn’t analyse the influence of stronger regulatory environments, although the results do suggest that female CEO’s are more

susceptible to a stronger governance environment. The results of Zalata et al (2018) regarding risk-aversion of female CEO’s are consistent with previous research on gender differences.

These findings are in line with the results from the previous literature mentioned above and provide enough evidence that there are differences between male and female CEO’s.

Based on previous literature mentioned above this research and particularly the findings of Zalata et al. (2018). I hypothesize that there is a stronger mitigating effect of Big-4 auditors regarding classification shifting whenever the CEO is female.

This leads to the second hypothesis of this research:

H2: The mitigating effect of the Big-4 auditor on classification shifting is higher when the CEO is female.

Research method:

This study aims to analyse the influence of Big-4 auditors on classification shifting and the role of CEO gender in this. This study focuses on the deliberate misclassification of core expenses, as special items within the income statement (McVay, 2006). I follow McVay (2006) in developing a proxy for normal core earnings being a function of last year’s core earnings, current and last year’s accruals, asset turnover, change in sale and negative sales. This leads to model (1) to calculate the expected core earnings:

∆

1

The variables in the models are measured as follows:

CE Core earnings/Sales. Where core earnings

equals (Sales - cost of goods sold - selling, general and administrative expenses)

ATO Asset turnover, defined as Sales/average

net operating assets, where net operating assets is the difference between operating assets and operating liabilities. Operating assets = Total assets - cash and short-term investments. Operating liabilities = Total assets - Total debt - Book value of common and preferred equity - minority interest.

Accruals (Net income before extraordinary items -

8 cash from operations)/Sales all at time t, where cash from operations is calculated as cash flow from operating activities/(Net cash flow - extraordinary items and Discontinued operations).

∆Sales Change in sales = (Sales ‐ Sales /

Sales

∆Neg_Sales This refers to ∆Sales if it is negative, and

zero otherwise

UE_CEt Core earnings predicted using model 1 -

Core earnings in the financial statement.

%SI (-1 * Special items)/Sales. Where income

increasing special items are set to 0.

BIG4 Value is 1 when the auditor is one of the

following: KPMG, Deloitte, Ernst & Young or PricewaterhouseCoopers. Otherwise the value is 0.

N_BIG4 Value is 1 when the auditor isn’t one of the

4 mentioned above. Otherwise the value is 0.

F Value is 1 when the CEO gender is female.

Otherwise the value is 0.

M Value is 1 when the CEO gender is male.

Otherwise the value is 0.

Model (1) is estimated per industry and fiscal year and the predicted core earnings is then subtracted from the actual earnings to calculate the unexpected core earnings. Using the unexpected core earnings the following model is used consistent with the method of McVay (2006) to determine if classification shifting using the special items exist in the data set.

_ ∝ ∝ % 2

Based on the method of McVay (2006) model (2) is used to test for classification shifting inside the data set. The coefficient of ∝ % should be positive if there is classification shifting inside the data set. This would indicate that the special items as a percentage of sales would be dominant in explaining the unexpected core earnings instead of other factors.

_ ∝ ∝ % ∗ 4 ∝ % ∗ _ 4 3

9 To test the first hypothesis model (3) is used to determine if the level of classification shifting used by companies is influenced by the presence of a Big-4 auditor. This model has 2

interaction variables between the special items as a percentage of sales and whether the auditor is a Big-4 auditor (BIG4) or a non Big-4 auditor(N_BIG4). I expect that the coefficient of both variables to be positive because based on the literature there should still be classification shifting present even with a Big-4 auditor. The coefficient of %SI*BIG4 should be lower than the

coefficient than the coefficient of %SI*N_BIG4 due to the expected higher audit quality and mitigating effect of BIG-4 auditors. Consistent with the research of McVay (2006) I do not add extra control variables to model (3) due to the many variables used to predict the value of the expected core earnings in model (1).

_ ∝ ∝ % ∗ 4 ∗ ∝ % ∗ 4 ∗ ∝ % ∗ _ 4 4

To test the second hypothesis the following model (4) is used to determine if the presence of a BIG4 auditor is further influenced by the gender of the CEO. The variable %SI*BIG4*F is the special items as a percentage of sales when the auditor is a Big-4 auditor and the CEO is a female. The variable %SI*BIG4*M is the special items as a percentage of sales when the auditor is a Big-4 auditor and the CEO is a male. Based on the discussed literature the mitigating effect of a Big-4 auditor should be higher when the CEO is female compared to a male CEO, therefore I predict that the coefficient of %SI*BIG4*F should be lower than the coefficient of %SI*BIG4*M.

Both variables should have a positive coefficient as there should still be classification shifting present consistent with the expectation of model (3). Similar to model (3) I do not add extra control variables due to the many variables used to predict the value of the expected core earnings in model (1).

Data and Sample selection

Financial data is collected from the COMPUSTAT database for the years 2003-2019 from the Annual File. CEO gender is collected from the Compustat Executive Compensation - Annual compensation database. Each firm year observation is required to have the variables available to predict the core earnings according to model (1). Observations with sales less than 1 million and industries with less than 15 observations per fiscal year are excluded consistent with the research of McVay (2006). After calculating the core earnings all observations where the CEO gender information is missing are excluded from the sample. This results in a full sample of 16.510 observations spanning the period from 2003 to 2019.

Results:

Descriptive statistics:

Table 1 presents the results of the descriptive statistics of unexpected core earnings, Special

items, CEO gender, Big-4 Auditor and other variables used in the analysis. We report the mean,

standard deviation, minimum and maximum of each variable. The mean of the unexpected core

earnings is 0.3e-10 with a minimum of -0.88 and a maximum of 0.908. The mean of the special

items as a percentage of the sales is -0.002 with a minimum of -0.088 and a maximum of 0

because the research only takes into account income-decreasing special items. The mean of the

CEO gender is 0.03 which means that approximately 3% of CEO’s are female in our sample.

10 The mean of the Big-4 auditors is 0.886 which means that approximately 88% of the auditors are Big-4 auditors. The mean of the accruals is -0.089 with a minimum of -0.91 and a maximum of 0.305. The mean of the asset turnover ratio is 1.652 with a minimum of -1.349 and a maximum of 7.923. The mean of the ∆Sales is 0.094 with a minimum of -0.917 and a maximum of 34.127.

The mean of the ΔNeg Sales is -0.031 with a minimum of -0.917 and a maximum of 0. The mean of the core earnings is 0.181 with a minimum of -10.105 with a maximum of 0.887.

Table 1 Descriptive statistics

Variable Obs Mean Std. Dev. Min Max Unexpected Core

Earnings 16,510 -3.00e-11 0.072 -0.880 0.908 Special items as a

percentage of Sales 16,510 -0.002 0.108 -0.088 0 CEO Gender 16,510 0.030 0.171 0 1 Big-4 Auditor 16,510 0.886 0.317 0 1

Accruals 16,510 -0.089 0.155 -0.910 0.305 Asset turnover ratio 16,510 1.652 1.388 -1.349 7.923

ΔSales 16,510 0.094 0.415 -0.917 34.127 ΔNeg Sales 16,510 -0.031 0.076 -0.917 0

Core Earnings 16,510 0.181 0.245 -10.105 0.887

Table 2 presents the correlation matrix between the variables used in the analysis. There is a positive significant relation between the unexpected core earnings and special items as a percentage of sales, which suggests that companies might have misclassified core expenses as special items to increase core earnings. There is also a high correlation between core earnings and unexpected core earnings which is to be expected because for the calculation of the unexpected core earnings the variable core earnings is used. Consistent with the correlation matrix McVay (2006) presented, the previous year core earnings are highly correlated at 0.8631.

∆Neg_Sales is also correlated with ∆Sales at 0.6149 which is to be expected because it is a

smaller part of the same variable. For the rest of the variables there are no high correlations

between the variables used in the analysis to test the hypotheses.

11 Table 2 Correlation matrix

Variable Unexpected Core

Earnings

Special Items as a

% of Sales

CEO

Gender Auditor Asset turnover Ratio

Accruals ΔSales ΔNeg Sales Core

Earnings Core Earnings Unexpected

Core Earnings 1.0000

Special Items as a

% of Sales

0.0161** 1.0000

CEO Gender 0.0005 0.0084 1.0000 Auditor -0.0061 0.0398*** 0.0183** 1.0000 Asset turnover

Ratio -0.1025*** 0.0624*** 0.026*** 0.0517*** 1.0000

Accruals 0.0000 -0.0502*** 0.0176** -0.0113 0.2592*** 1.0000 ΔSales 0.0000 0.016** -0.0386*** -0.05*** 0.0375*** 0.1365*** 1.0000

ΔNeg Sales 0.0000 0.0611*** -0.01 0.0224*** 0.0866*** 0.2763*** 0.6149*** 1.0000 Core Earnings 0.4583*** -0.0185** -0.0193** -0.0053 -0.3979*** -0.0793*** 0.1998*** 0.2225*** 1.0000

Core Earnings -0.0000 -0.0117 -0.0166** -0.0004 -0.4143*** -0.1799*** 0.0657*** 0.1020*** 0.8631*** 1.0000

12 Estimating core earnings:

This research uses the same model (1) that McVay (2006) uses to calculate predicted core earnings based on the lagged core earnings, asset turnover ratio, lagged accruals, accruals, change in sales and the change in negative sales. The lagged core earnings tend to be very persistent (McVay, 2006) therefore this should be useful in predicting the core earnings. Asset turnover ratio has been shown to be negatively related to profit margin (McVay, 2006 and Nissim and Penman, 2001) and the core earnings McVay (2006) uses is closely related to profit margin.

McVay (2006) added the asset turnover ratio in the estimation because firms that have large income-decreasing special items are likely to be making changes to their operating strategy, possibly altering their mix and turnover. Prior year accruals are an explanatory variable for future performance based on previous research of Sloan (1996), therefore McVay (2006) added this variable to the estimation. Earnings performance attributable to the accrual components of earnings exhibits lower persistence than earnings performance attributable to the cash flow component of earnings (McVay, 2006). Research of DeAngelo et al. (1994) found that extreme performance is highly correlated with changes in accrual levels, therefore the current year accruals are added to the estimation. This research follows the estimation of McVay (2006) therefore focussing on special items and controlling for changes in accruals due to extreme performance and making stronger predictions of the core earnings. Core earnings are correlated to sales but this relation is not constant because the fixed costs per sales differ because they don’t increase and decrease as much. Therefore in the estimation∆Sales is added as an

explanatory variable by McVay (2006). The slope of sales may differ based on previous research

of Anderson et al. (2003) therefore the variable ∆NegSales is added to make the differentiation

between increase in sales and a decrease in sales.

13 Table 3 Estimation Core Earnings

VARIABLES Model

(1) Core Earnings 0.842***

(0.004)

Asset turnover ratio -0.007***

(0.000)

Accruals -0.152***

(0.004)

Accruals 0.128***

(0.004)

∆Sales 0.067***

(0.004)

∆Neg Sales 0.158***

(0.010)

Constant 0.040***

(0.002)

Observations 16,510 R-squared 0.9747 Number of GlobalCompanyKey 1,443

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

The results of the regression of model (1) to predict the Core earnings are presented in table 3.

The coefficient of the asset turnover ratio (-0.007) and prior year accruals (-0.152) are both negative consistent with the prediction and significant at the 1% level. Consistent with the research of Anderson et al. (2003) the coefficient of the ∆NegSales is larger (0.158) than the coefficient of the ∆Sales (0.067) which is also in line with the results from McVay (2006). This support the addition of ∆NegSales to the model due to the different influence of changes in sales. As predicted the core earnings of the previous year positively influences the core earnings with a coefficient of 0.842 and significant at the 1% level. The results from the regression are in line with the regression results of the research of McVay (2006). The results from the regression of model (1) are used to calculate predicted core earnings for the companies, with these

predicted values the unexpected core earnings are calculated by subtracting the predicted core

earnings from the actual presented core earnings. The unexpected core earnings are used in the

next models (2,3 & 4).

14 Main analysis:

To determine if there is classification shifting inside the data set to test my hypotheses model (2) is used. Based on the research of McVay (2006) the coefficient of %SI should be positive and that is in line with the results of the regression analysis. The results presented in table 4 model (2) shows that the coefficient of %SI is 0.243 and significant at the 1% level which confirms that the data set contains classification shifting. The data set can therefore be used to test for the influences of Big-4 auditors and CEO gender on classification shifting.

To test the first hypothesis that the mitigating effect of Big-4 auditors on classification shifting is higher compared to non Big-4 auditors model (3) is used and the results are presented in table 4. To confirm the first hypothesis the results should show that the coefficient of %SI*BIG4 should be lower than the coefficient of %SI*N_BIG4. The results show that the coefficient of %SI*BIG4 is 0.219 and significant at the 1% level and the coefficient of %SI*N_BIG4 is 0.346 and also significant at the 1% level. These coefficients of both variables are positive which is also in line with the prediction, because there should still be classification shifting despite the presence of an auditor to test the first hypothesis. The results suggest that the mitigating effect of a Big-4 auditor on classification shifting is indeed higher than the non Big-4 auditor, the coefficient of the

%SI*BIG4 is lower than the coefficient of %SI*N_BIG4 as predicted. To determine if the first hypothesis can be accepted a two sample t test is used to determine if the usage of

classification shifting is lower whenever the auditor is a Big-4 auditor. The results show that the companies with a non Big-4 auditors use less classification shifting (M=-0.0004, SD=0.00004) compared to companies with a Big-4 auditor (M=-0.0018, SD=0.00007), t(24369) = -17, P = <

.001. Based on these results there is not enough evidence that companies with Big-4 auditors use less classification shifting, therefore I am unable to accept the first hypothesis.

For testing the second hypothesis that the mitigating effect of the Big-4 auditor is higher when the CEO is female, model (4) is used and the results are presented in table 4. The results of model (3) discussed above confirm that the Big-4 auditor has a higher mitigating effect on classification shifting and therefore the second hypothesis can be tested. To confirm the second hypothesis the coefficient of %SI*BIG4*F should be higher than the coefficient of %SI*BIG4*M.

The results show that the coefficient of %SI*BIG4*F is 0.192 and it is not significant at the conventional levels. The coefficient of %SI*BIG4*M is 0.220 and significant at the 1% level.

These results suggest that the mitigating effect of Big-4 auditors is higher when the CEO is female because the coefficient of %SI*BIG4*F is lower compared to the coefficient of

%SI*BIG4*M. The coefficient of %SI*BIG4*F is not significant which suggests that in the firms lead by a female CEO with a Big-4 auditor the monitoring is strong to the point there is no indications of classification shifting being used. This would prove the assumptions based from the literature that the monitoring effect of Big-4 auditors is higher for female CEOs. The coefficient of %SI*N_BIG4 is consistent with the value of the analysis of model (3). To test the second hypothesis a two sample t test is used to determine if the usage of classification shifting is lower when the CEO is female in the presence of a Big-4 auditor. The results show that the companies with a female CEO use less classification shifting (M=-0.00005, SD=0.00001) compared to companies with a male CEO (M=-0.0017, SD=0.00007), t(17489) = 22.9, P = <

.001. Based on these results there is enough evidence to accept the hypothesis that female

CEOs use less classification shifting compared to male CEOs.

15 Table 4 Regression results

Model Model Model

VARIABLES (2) (3) (4)

%SI 0.243***

(0.057)

%SI * BIG4 0.219***

(0.063)

%SI* N_BIG4 0.346*** 0.346***

(0.131) (0.131)

%SI* BIG4 * F 0.192

(0.356)

%SI * BIG4 * M 0.220***

(0.064)

Constant 0.001 0.001 0.001 (0.001) (0.001) (0.001) Observations 16,510 16,510 16,510 R-squared 0.001 0.001 0.001 Number of GlobalCompanyKey 1,443 1,443 1,443 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Additional analysis:

Consistent with the research of McVay (2006) and Zalata et al (2018) I perform the same 3 analyses on a data sample only consisting of observations with income-decreasing special items. The data set consists of 2.384 observations after eliminating the observations without the income-decreasing special items. Table 5 shows the results of the 3 analyses (model 2, model 3

& model 4).

The first analysis (model 2) shows that in the data set the special items as a percentage of sales is still dominant because the coefficient of %SI is 0.392 and significant at the 1% level. The same analyses can be performed to test the influence of Big-4 auditors and CEO gender on this data set.

The results of model (3) in table 5 show that the difference in the coefficients between %SI*BIG4 and %SI*N_BIG4 is larger compared to the analysis on the full data set. The coefficient of

%SI*BIG4 is 0.295 and significant at the 1% level and the coefficient of %SI*N_BIG4 is 0.854

and also significant at the 1% level. Based on the results the case for accepting the first

hypothesis is strengthened because the gap between the coefficients of both variables has

increased and therefore suggest that there is indeed a difference between the mitigating effect of

16 Big-4 auditors and non Big-4 auditors on classification shifting. To determine if the results of the t test performed on the full sample differ from the sample with only observations with income decreasing special items the test is reperformed using this sample. The results show similar to the test on the full sample that the companies with a non Big-4 auditors use less classification shifting (M=-0.003, SD=0.0003) compared to companies with a Big-4 auditor (M=-0.013, SD=0.0004), t(4098) = -19.6, P = < .001. Based on these results there is not enough evidence that companies with Big-4 auditors use less classification shifting, therefore unable to accept the first hypothesis on the sample with only observations with income decreasing special items as well.

The results of model (4) are presented in table 5 and show a similar view of the influence of CEO gender. The coefficient of %SI*BIG4*F is 0.259 and not significant at the conventional levels just like the results presented in table 4. The coefficient of %SI*BIG4*M is 0.296 and significant at the 5% level. The results of the analysis on only the firms that have income decreasing special items are in line with the results derived from the full sample. This

strengthens the case that the monitoring effect of Big-4 auditors is higher for firms that have a

female CEO. Similar to the testing of the first hypothesis the t test is reperformed for the second

hypothesis as well to determine if the results are the same with the sample of only observations

with income decreasing special items. The results show that the companies with a female CEO

use less classification shifting (M=-0.0003, SD=0.00008) compared to companies with a male

CEO (M=-0.012, SD=0.0004), t(2782) = 26.9, P = < .001. Based on these results there is

enough evidence to accept the hypothesis that female CEOs use less classification shifting

compared to male CEOs. These results are consistent to the results based on the full sample

which supports the acceptance of the second hypothesis.

17 Table 5 Regression results (special items only)

Model Model Model

VARIABLES (2) (3) (4)

%SI 0.392***

(0.112)

%SI * BIG4 0.295***

(0.121)

%SI* N_BIG4 0.854*** 0.854***

(0.248) (0.248)

%SI* BIG4 * F 0.259

(0..587)

%SI * BIG4 * M 0.296**

(0.122)

Constant 0.006** 0.006** 0.006**

(0.002) (0.002) (0.002)

Observations 2,384 2,384 2,384

R-squared 0.009 0.012 0.012

Number of GlobalCompanyKey 1,017 1,017 1,017

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

18

Conclusion :

This research analysed the influence of Big-4 auditors and CEO gender on classification shifting.

Previous research found evidence of the influence of Big-4 auditors on classification shifting (Haw, Ho & Li, 2011 & Desai & Nagar, 2016) and CEO gender on classification shifting (Zalata et al., 2018). The goal of this research was to combine the two to determine if the gender of the CEO influenced the mitigating effect of Big-4 auditors. The results indicate that Big-4 auditor have a higher mitigating effect on classification shifting compared to non Big-4 auditors, although care is needed with interpreting these results because there isn’t enough evidence to state this with certainty. The results were more conclusive for the differences between male and female CEOs, because the results indicate that female CEOs didn’t use classification shifting whereas the male CEOs did use classification shifting. There is enough evidence to state that for the data used female CEOs use less classification shifting then male CEOs, although more research is needed to determine if this relation is robust with different data sets and different variables of governance influences. The results were consistent when reperformed on a smaller sample with only observations that have income-decreasing special items. Although there was evidence that female CEOs use less classification shifting in the presence of a Big-4 auditor, the expected relation of higher mitigating influence of Big-4 auditors wasn’t observed which makes it more difficult to state that female CEOs are more susceptible for auditor monitoring. Also the analysis uses financial data from U.S. companies which makes it hard to generalize the results to

different economies like Europe and Asia. The results are consistent with previous research and provide more evidence on the differences of CEO gender for future researchers to expand on.

This research contributes to the field of classification shifting by analysing the role of the Big-4 auditor on mitigating classification shifting. This interaction has been observed by Haw, Ho & Li (2011) in East Asian companies as well and now the same evidence can be provided for U.S.

companies after the implementation of the SOX regulation. The research also provides more results regarding the influence of CEO gender on classification shifting that Zalata et al. (2018) have analysed from an ethical perspective. The results provide more support that female CEOs behave differently than male CEOs, which future researchers can use and expand on. Future researchers could analyse different CEO characteristics to determine their role in the use of classification shifting and also more and different governance aspects. Future research could also analyse the influence of Big-4 auditors and CEO gender in different economies than the U.S. to determine if the results are a global phenomenon or if it is limited to certain regions due to other external factors like the strong regulatory environment. Future research could

investigate the gender differences with other C-level directors like the CFO as well as to

determine their influence on the usage of classification shifting.

19 References

Anderson, M., Banker, R., Janakiraman, S. (2003). Are selling, general and administrative costs

‘’sticky’’?. Journal of Accounting Research. Vol. 41, issue 1: 47-63.

Barber, B., Odean,T. (2001). Boys will be boys: gender, overconfidence and common stock investment. The Quarterly journal of Economics. Vol. 116, issue 1: 261-292.

Becker,m C., DeFond, M., Jimbalvo, J. and Subramanyam, K.R. (1998). The effect of audit quality on earnings management. Contemporary Accounting Research. Vol. 15: 1-24

Benston, G. (1980). The market for public accounting services:Demand, supply and regulation.

Accounting Journal. Forthcoming.

Bradshaw, M., Sloan, R. (2002). GAAP versus the Street: An empirical assessment of two alternative definitions of earnings. Journal of Accounting Research. Vol. 40, issue 1: 41-66 Byrnes, J.P., Miller, D.C., Schafer, W.D. (1999). Gender differences in risk taking: A meta- analysis. Psychological Bulletin. Vol. 125, issue 3: 367-383.

Choi, A. Choi, J., Sohn, B.C. (2018). The Joint Effect of Audit Quality and Legal Regimes on the Use of Real Earnings Management: International Evidence. Contemporary Accounting

Research. Vol. 35, issue 4: 2225-2257.

Davis, A. (2002). The value relevance of revenue for Internet firms: Does reporting grossed-up or barter revenue make a difference?. Journal of Accounting Research. Vol. 40, issue 2: 445- 477.

DeAngelo, L.E. (1981). Auditor size and audit quality. Journal of Accounting and Economics.

Vol. 3, issue 3: 183-199.

DeAngelo, H., DeAngelo, L., Skinner, D. (1994). Accounting choice in troubled companies.

Journal of accounting and economics. Vol. 17, issue 1-2: 113-143.

DeFond, M.L., (1992). The association between changes in client firm agency costs and auditor switching. A Journal of Practice and Theory. Vol. 11, issue 1: 16-31.

DeFond, M.L., Jimbalvo, J. (1993)., Factors related to auditor - client disagreements over income-increasing accounting methods. Contemporary Accounting Research. Vol. 9: 415-431.

Desai, N., Nagar, N. (2016). A research note: Are auditors unable to detect classification shifting or merely to report it? Evidence from India. Journal of Contemporary Accounting & Economics.

Vol. 12: 111-120.

Dye, D.A. (1993). Auditing standards, legal liability and auditor wealth. Journal of Political

Economy. Vol. 101, issue 5: 29-40.

20 Elliot, J., Hanna, J. (1996). Repeated accounting write-offs and the information content of earnings. Journal of Accounting Research. Vol. 34 (supplement): 135-155

Eshleman, J.D., Guo, P. (2014). Do Big 4 Auditors provide higher audit quality after controlling for endogenous choice of auditor? Auditing: A journal of Practice & Theory. Vol. 33, issue 4:

197-219.

Fairfield, P., Sweeney, R., Yohn, T., (1996). Accounting classification and the predictive content of earnings. The accounting Review. Vol. 71, issue 3: 337-355.

Ford, R.C., Richardson, W.D. (1994). Ethical Decision making: A review of the empirical literature. Journal of Business ethics. Vol. 13: 205-221.

Francis, J., Hanna, J., Vincent, L. (1996). Causes and effects of discretionary asset write-offs.

Journal of Accounting Research. Vol. 34 (supplement): 117-134.

Gomez-Meija, L. & Wiseman, R. (1997). Refraining executive compensation: An assessment and outlook. J. Manag. Vol. 23: 291-374.

Haw, I., Ho, S.S.M, Li, A.Y. (2011). Corporate Governance and Earnings management by classification shifting. Contemporary Accounting Research. Vol. 28, issue 2: 517-553.

Ho, S.S., Li, A.Y., Tam, K. & Zhang, F. (2015). CEO gender, ethical leadership and accounting conservatism. Journal of business ethics. Vol. 127, issue 2: 351-370.

Houge, M.N., Ahmed, K., Zijl, T. (2017). Audit Quality, Earnings Management, and Cost of Equity Capital: Evidence from India. International journal of Auditing. Vol. 21: 177-189.

Huang, J., Kisgen, D.J. (2013). Gender and corporate finance: are male executives

overconfident relative to female executives? Journal of Finance and Economics. Vol. 108: 822- 839.

Healy, P.M., Wahlen, J.M. (1999). A review of the earnings management literature and its implications for standard setting. Accounting Horizons. Vol. 13, issue 4: 365-383.

Jensen, M., & Meckling, W. (1976). Theory of the firm: managerial behaviour, agency costs and ownership structure. Journal of finance & Economics. Vol. 3, issue 4: 305-360.

Joo, J.H., Chamberlain, S.L. (2017). The effects of corporate governance on classification shifting and compensation shielding. Contemporary Accounting Research. Vol. 34, issue 4:

1779-1811

Khan, W.A., Vieito, J.P. (2013). Ceo gender and firm performance. Journal of Economics and Business. Vol. 67: 55-66.

Kim, J., Chung, R., Firth, M. (2003). Auditor conservatism, asymmetric monitoring and earnings

management. Contemporary Accounting Research. Vol. 20, issue 2: 323-359.

21 Lin, J.W., Hwang, M.I. (2010). Audit Quality, Corporate governance, and Earnings management:

A Meta-analysis. International journal of Auditing. Vol. 14: 57-77.

Lipe, R. (1986). The information contained in the components of earnings. Journal of Accounting Research. Vol. 24 (supplement): 37-64.

Lougee, B., Marquardt, C., (2004). Earnings informativeness and strategic disclosure: An empirical examination of ‘’pro forma’’ earnings. The Accounting Review. Vol. 79, issue 3: 769- 795.

McVay, S.E. (2006). Earnings management using classification shifting: an Examination of core earnings and special items. The accounting review. Vol. 81 Issue 3: 501-531.

Nissim, D., Penman, S., (2001). Ratio analysis and equity valuation: From research to practise.

Review of Accounting studies. Vol. 6, issue 1: 109-154.

Palvia, A., Vähämaa, E. & Vähämaa, S. (2014). Are female CEOs and chairwomen more

conservative and risk averse? Evidence from the banking industry during financial crisis. Journal of business ethics. Vol. 131, issue 3: 577-594.

Philbrick, D., Ricks, W., (1991). Using value line and I/B/E/S analyst forecasts in accounting research. Journal of Accounting Research. Vol. 29, issue 2: 397-417.

Powell, M., Ansic, D. (1997). Gender differences in risk behaviour in financial decision-making:

An experimental analysis. Journal of Economic Psychology. Vol. 18, issue 6: 605-628.

Schipper, K. (1989). Commentary on earnings management. Accounting Horizons. Vol. 3: 1-102.

Schubert, M., Brown, M., Brachinger, H. (2000). Gender specific attitudes towards risk and ambiguity: An experimental investigation, Center of Economic Research, Swiss Federal Institute of Technology Zurich Working Paper.

Sikka, P. (2009) Financial crisis and the silence of the auditors. Accounting, Organisations and Society. Vol. 34, issue 6: 868-873.

Sloan, R. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? Accounting Review. Vol 71, issue 3: 289-315.

Sun, J., Kent, P., Qi, B.& Wang, J. (2017). Chief financial officer demographic characteristics and fraudulent financial reporting in China. Auditing and finance. Vol. 59, issue 4: 2705-2734.

Tosi, H. & Gomez-Meija, L. (1989). The decoupling of CEO pay and performance: An agency theory perspective. Adm. Sci. Q. Vol 34: 169-189.

Watts, R. (1977). Corporate financial statements, a product of the market and political

processes. Australian Journal of Management. April, 53-75.

22 Watts, R., Zimmerman, J. (1981). The markets for independence and independent auditors.

Unpublished manuscript (University of Rochester, Rochester NY.

Watts, R., Zimmerman, J. (1986). Positive accounting theory (Prentice-hall, Englewood Cliffs, NJ).

Whittington, O.R., Pany, K. (2001). Principles of auditing and other assurance services, Irwin McGraw-Hill, pp 4-6.

Zalata, A, Tauringana,V. And Tingbani, I. (2018). Audit committee financial expertise, gender and earnings management: does gender of the financial expert matter?. International review of financial analysis. Vol. 55: 170-183.

Zalata, A., Ntim, C., Aboud, A. and Gyapong, E. (2018). Female CEOs and core earnings quality: New evidence on the ethics versus risk-aversion puzzle. Journal of business ethics. Vol.

160: 515-534.