Economic bonding, corporate governance and accrual-based earnings management.

(1)

Amsterdam Business School

Faculty of Economics and Business

Master Thesis Accountancy and Control

Track Accountancy

Economic bonding, corporate governance and

accrual-based earnings management

Student Kasper Groen

Student number 10625747

Course Master Thesis Accountancy & Control

Course code 6314M0044Y

Version Final

Date June 22, 2014

Supervisor Dr. Bo Qin

(2)

Abstract

This study examines whether different types of audit fees are associated with accrual-based earnings management. More specifically, whether different fee types contribute to the economic bonding between client and audit firm, which could in turn cause situations in which the auditor acquiesce to the pressure from a client to allow earnings management. Furthermore, the association of corporate governance with earnings management and the potential mitigating role of corporate governance on the relationship between economic bonding of an auditor on a client and the allowance of earnings management are examined. Evidence is found that audit fees, non-audit fees, non-audit fee ratio and total fees are all positively associated with accrual-based earnings management. These findings are in line with the economic bonding theory. Furthermore, corporate governance is negatively associated with magnitude of discretionary accruals. In contrast, no evidence is obtained for a moderating effect of corporate governance on the relationship between one of the fee variables and accrual-based earnings management. Substantially all results are robust to (1) potential problems that may arise from endogeneity issues, (2) an alternative measure for accrual-based earnings management, and (3) a different operationalization of the fee variables. Moreover, an additional analysis to control for the potential influence of the 2008 financial crisis on this research setting shows that this has no impact on the obtained findings.

Keywords: conflicts of interest, economic bonding, audit fees, non-audit fees, non-audit fee

ratio, total fees, auditor independence, (accrual-based) earnings management, corporate governance.

Data availability: All data incorporated in the examination of this research is available

(3)

1. Introduction

In the past decade, the European Union (EU) introduced new regulations and amendments related to the audit profession. Those are the result of the recent debate about the independence of an auditor. Initiator of this debate was the European Commission. They published a Green Paper concerning the audit profession on October 13th_{, 2010. EU} commissioner Michel Barnier, who is responsible for internal market and services, addresses the role of the statutory audit of financial statements and the wider environment in which the audit profession is active (European Commission, 2010). Moreover, he stated the question whether the audit profession can be improved to decrease financial risks in the future (European Commission, 2010).

The purpose of these changes is to increase the objectivity and independence of the auditor in relation to the client and restore the confidence in the audit profession. One of the central issues in recent debates was the view that auditors tend to lose their objectivity and independence since they face conflicts of interest as result of a longstanding client-auditor relation (Ewelt-Knauer et al., 2013) and an economic dependence on the client firm (Knapp, 1985; Moore et al., 2006). Conflicting situations between an auditor and client firm may arise in terms of the need for adjustments in financial statements, the inappropriate use of certain accounting practices that needs to be corrected, or certain modification with respect to the adequacy of disclosures (Knapp, 1985). It is argued that auditors allow high fee client firms more discretion over their financial reporting as compared to low fee client firms (Ashbaugh et al., 2003), since the economic bond between the client firm and the auditor increases and therefore the economic dependence on this specific client (DeAngelo, 1981).

Conflicts of interest arising from an economic bond, and therefore indirect impaired independence, are extensively argued as one of the causes of the recent scandals (Moore et al., 2006), which led to the ongoing debate about the independence of an auditor in relation to the client and the recently proposed regulations related to the audit profession (European Commission, 2010). Opposite to these restricting regulations to impair the client-auditor relationship stands the need for specific knowledge of the client to perform an efficient and effective audit (Bamber and Iyer, 2007). It is necessary that auditors deepen their understanding about the client and the processes within the specific organizations (Stanley and Dezoort, 2007). In other words, for an efficient and effective audit, and to guarantee a high quality audit, the auditor should increase the familiarity with the client.

(5)

As a result of this familiarity, the auditor is capable of effectively providing non-audit services, since there is already a deepened understanding of the clients businesses. This is also a threat to the independence of the auditor since providing non-audit services could result in an economic bond between the auditor and the client (Cahan et al., 2008). Furthermore, it is argued that providing non-audit services will strengthen the economic dependence between the auditor and the client firm, which may result in the occurrence that audit firms will avoid issuing negative opinions to safeguard the revenue stream from that specific client (Frankel et al., 2002; Moore et al., 2006). The nature of the economic dependence resulting from fees paid to the auditor is the main focus of this research with respect to conflicts of interest that may arise between an auditor and a client firm. This is build upon the central research question: “Do conflicts of interest arising from economic dependence on the client firm relate

to earnings management?”

Furthermore, in the aftermath of the financial crisis, it is frequently discussed how the role of the auditor should be strengthened to decrease financial risks in the future (European Commission, 2010). Implemented laws and regulations, such as the separation between audit and advisory, solely focus on the auditor side. Since the focus is mainly on the auditor side, the question raises which role is reserved for the less emphasized side, namely the client firm. The question remains whether firms have the ability to contribute in creating an environment in which financial risks and threats for misleading financial reporting could be reduced. The adoption of a strong corporate governance system will result in an additional monitoring mechanism mitigating the agency problems within a firm (Larcker and Richardson, 2004) and will help to ensure that a manager acts in the best interest of the shareholders. Corporate governance includes, in the broadest sense, the fundamental principles on which the firm is controlled and managed (Lin and Hwang, 2010). Therefore, it is relevant to examine whether corporate governance will help impair the ability for management to engage in earnings management. The second research question is therefore: “How is corporate governance

related to earnings management?”

In addition, another relevant aspect worth examining is the incorporating of corporate governance as a potential moderating factor on the relationship between the auditor’s economic dependence on the client and the allowance of earnings management. Aforementioned is relevant since in practice audit firms are economically depended on their clients and accounting policies leave room for earnings management. Corporate governance is seen as a mechanism that could contributes in ensuring auditor’s independence and hence could restrain the economic bond between the auditor and client firm (Gold et al., 2004).

(6)

More specifically, a well-functioning corporate governance system can help other governance bodies – like an audit committee – in mitigating the likelihood that an auditor will acquiesce to the pressure from a client to allow earnings management. Ultimately, corporate governance supports the overall monitoring of the financial reporting (Gold et al., 2004). This situation highlights the importance of a well-functioning governance system within the organization in order to mitigate the relationship between the economic bonding on a client and the allowance of earnings management. Therefore it is relevant to incorporate corporate governance as a potential moderating factor in this research. Hence, this study aims to obtain evidence for the influence of a good corporate governance structure in constraining the ability for management to engage in earnings management and strengthen the role of the auditor. This is addressed in the third research question: “Can corporate governance moderate the relationship between

economic dependence and earnings management?”

The results of this study provide evidence that audit fees, non-audit fees, non-audit fee ratio and total fees are all positively associated with accrual-based earnings management. These findings support the economic bonding theory, implying an increased dependence of the auditor on the client. Aforementioned is based upon an observable increased allowance of accrual-based earnings management as the magnitude of fees increases. Furthermore, this study finds a negative association between corporate governance and the absolute value of discretionary accruals. This implies support for the role of a well-functioning system in ensuring high quality financial reporting and audit processes through impairing the ability for management to engage in earnings management. Contrary to what is expected, no evidence is obtained for a moderating effect of corporate governance on the relationship between one of the fee variables and accrual-based earnings management. Substantially all results are robust to (1) potential problems that may arise from endogeneity issues, (2) an alternative measure for accrual-based earnings management, and (3) a different operationalization of the fee variables. Moreover, an additional analysis to control for the potential influence of the 2008 financial crisis on this research setting shows that this has no impact on the findings.

This research contributes to the academic literature in several ways. Firstly, this study highlights that a corporate governance structure is associated with accrual-based earnings management. Incorporating the influence of a corporate governance structure based on a governance index instead of separate governance elements is new and an addition to the earnings management literature. This study provides evidence that corporate governance is associated with lower discretionary accruals and therefore highlights the importance of examining a corporate governance structure for example for users of financial statements,

(7)

since higher assurance is expected through a more reliable audit report. Secondly, this study extends and deepens the earnings management literature by further examining the relationship between fees and accrual-based earnings management. Although multiple studies explore the relationship between one of the fee variables and earnings management, findings are not coherent which limit the ability to understand the separate associations with earnings management. The findings support the current debate on European level concerning the separation of advisory and audit services in achieving more auditor independence and ultimately in ensuring high-quality financial reporting. Thirdly, this study finds no evidence of a possible effect of the financial crisis on discretionary accruals and on the relationship between one of the fee variables and earnings management. This provides support that there is limited impact of the financial crisis on the magnitude of accrual-based earnings management.

The next section discusses related literature and presents the hypotheses development. Applied methodology is described in section 3. Section 4 elaborates on the findings obtained and section 5 will discuss the findings and concluding remarks.

2. Literature review and hypotheses development

This section elaborates more on the separate concepts underlying this research and provides insight in the interconnectedness of earnings management, conflicts which may arise from the economic bonding between an auditor and a client firm and the functioning of a corporate governance structure.

Section 2.1 elaborates more on the concept of earnings management, underlying incentives for a manager and briefly describes accrual-based earnings management. Section 2.2 discusses conflicts of interest between an auditor and client firm, which may arise from the economic bonding. Subsequently, the economic bond in terms of fees paid to the auditor will be discussed in relation to earnings management. Thereafter, section 2.3 elaborates more on the corporate governance structure and which role the presence of a well-functioning governance system can play in the relationship between fees and earnings management.

2.1 Earnings management

It is argued that the occurrence of earnings management may arise from the agency theory as a result of the separation between ownership and management of the organization. This results in control difficulties due to agency problems and information asymmetry (Beatty and Harris, 1998). Agency problems occur when the manager has different interests as compared to the owners and therefore will act to maximize his own personal gain at the

(8)

expense of the shareholders (Beatty and Harris, 1998). The control difficulty of information asymmetry arise from the fact that managers have more inside information about the organization than the shareholders, which could result in more difficulty in attracting external capital or lower share prices than would be the case if no information asymmetry would exists (Beatty and Harris, 1998).

Furthermore, accounting standards gives managers substantive discretion over their financial reporting because judgment is required in the application of these standards (Hail et al., 2010). An example of the aforementioned is accounting accruals, which requires subjective estimation and judgment based on managers inside information (Jackson and Pitman, 2001; Hail et al., 2010). So, in addition to the control difficulties of the agency problems and information asymmetry, the flexibility in the application of accounting standards – management discretion – gives the opportunity for earnings management (Hail et al., 2010).

Earnings management is defined in several different forms. Two common used definitions in academic research that are used are from Healy and Wahlen (1999): “Earnings

management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” (Healy and Wahlen, 1999: 368) and Scott (2009): “Earnings management is the choice by a manager of accounting policies, or actions affecting earnings, so as to achieve some specific reported earnings objective.” (Scott, 2009: 403). This research

builds upon these two definitions.

More specific, accrual-based earnings management is best explained as “the process

of taking deliberate steps within the constraints of generally accepted accounting principles to bring about a desired level of reported earnings” (Davidson et al., 1987 in Schipper 1989:

92). This flexibility in financial reporting allows managers to communicate preferable information about the company through the financial reports to present more accurate and more informative figures to third parties (Watts and Zimmermann, 1978). Contrary, the management discretion gives the opportunity to engage in earnings management (Hail et al., 2010) to reach certain goals or targets. The accounting literature outlines several incentives for earnings management, which will be discussed in the next paragraph.

(9)

2.1.1 Motives for earnings management

The academic literature mentions various motives for the application of earnings management by managers. The most common used motives are categorized in capital market motivations, contracting motivations and regulatory motivations (Healy and Wahlen, 1999; Jackson and Pitman, 2001). These three categories will be addressed in respective order.

Capital market motivations

The first category is capital market incentives. When managers are motivated by capital incentives to influence earnings, it is expected that stakeholders are provided with misleading information in order to manipulate the decision-making of investors and financial analysts. Investors and financial analysts use the provided information for investment decisions and in their valuation process. In this way, managers could have incentives to manage short-term stock price performance to attract or retain capital provided by investors (Healy and Wahlen, 1999). In addition, earnings management is used to meet expectations of the capital market, where managers expect that not meeting pre-set expectations will have a negative effect on the expectations of future performance of the company (Xie et al., 2003).

Contracting motivations

Secondly, contracting motivations could drive earnings management. Contracting incentives are based on the appearance that accounting data is needed to monitor and regulate contracts between the company, the manager and related stakeholders (Healy and Wahlen, 1999). These contracts are needed to align the different interests as a result of the separation between ownership and management (Scott, 2009). Two of the most common contracting incentives for management to manipulate earnings are management compensation and lending contracts (Healy and Wahlen, 1999).

Management compensation contracts most often include compensation directly through annual pay and bonuses and indirectly through promotions and job security (Xie et al., 2003). When these forms of compensation are based on measures of earnings performance against pre-set targets, it is expected that managers will have incentives to manipulate earnings in order to boost their compensation and improve job security (Holthausen et al., 1995; Healy and Wahlen, 1999; Xie et al., 2003).

Lending contracts close to violation could also provide incentives for the manager to influence earnings. These contracts are formed in order to ensure that the manager will not undertake actions to improve shareholders wealth at the expense of the firms’ debt holders

(10)

(Healy and Wahlen, 1999). The violation of lending contracts is costly for the firm because the firm could, for example, be subjected to more strict conditions for new debt-agreements.

In general, it is argued that management compensation and lending contracts are needed to overcome conflicting interests. Despite the general accepted view that contracts help in reducing these conflicts (Scott, 2009), it is argued that perfect contracts could not be established, which leaves room for earnings management. This originates from the occurrence of managers discretion over compliance with these contracts and the difficulty for the other party to see through this (Watts and Zimmerman, 1978; Healy and Wahlen, 1999).

Regulatory motivations

The third category of earnings management motivations is related to regulatory incentives. Regulatory incentives arise from the assumption that reported earnings will influence actions of regulating and governmental bodies (Jackson and Pitman, 2001). Managing earnings for regulatory reasons mostly focuses on minimizing the impact of regulations on the company and minimizing political scrutiny. One of the regulatory incentives to manipulate earnings is shifting tax over time in order to reduce the tax payable in a certain period for certain reasons. In the case that earnings are managed to minimize political scrutiny, the manager beliefs that lowering current earnings can lower political heat by shifting focus. In addition, political costs – associated with highly profitable firms – could be lowered (Watts and Zimmerman, 1978). Political costs could consist of, for example, higher taxes and more strict enforcement of laws by government of becoming in a range that requires more strict accounting standards (Cahan, 1992).

2.1.2 Accrual-based earnings management

Aforementioned, managers have incentives to engage in earnings management. One of the common used methods to determine accrual-based earnings management is the estimation of discretionary accruals (e.g. DeAngelo, 1986; Dechow et al., 1995; Jones, 1991; Kothari et al., 2005). Reported earnings comprise operating cash flows and total accruals, where total accruals incorporate both non-discretionary and discretionary accruals (Scott, 2009). The latter is most relevant since it is subjective to management assumptions, which allows discretion and therefore room for earnings management. Examples are the determination of provisions, devaluation of assets and one-time depreciations. Non-discretionary accruals are not subject (or to a lesser extent) to management subjectivity, for example depreciation costs (Scott, 2009). To determine the magnitude of earnings management it is most relevant to

(11)

estimate the discretionary accruals since this part represents management’s intention to manipulate earnings (Scott, 2009).

2.2 Conflicts of interest: auditor and client firm

The demand for the audit profession is originated from the agency theory – the separation between ownership and management – where the auditor needs to ensure reliability of the information provided by organizations (Limperg, 1932). In this way, the auditor takes the role as a fiduciary of society to ensure accurate (accounting) numbers (Limperg, 1932). Therefore, the society relies to a great extent on the opinion of the auditor by evaluating and assessing the performance of a company. As stated in the Code of Ethics of the International Federation of Accountants (IFAC), it is the responsibility of a professional auditor to act in the public interest and not “exclusively to satisfy the needs of an individual client or

employer” (IFAC, 2006: 1104).

Notwithstanding the fact that this is one of the fundamental requirements of the audit profession, in practice situations could occur in which the auditor face conflicting interest between their professional commitment and the tendency to act in their own self-interest (Moore et al., 2006). Conflicting interest, and therefore indirect impaired independence, is extensively argued as one of the causes of the recent scandals (Moore et al., 2006), which led to the ongoing debate about the independence of an auditor in relation to the client (European Commission, 2010). Conflicting situations between an auditor and client firm may arise in terms of the need for adjustments in financial statements, the inappropriate use of certain accounting practices that need to be corrected or certain modification with respect to the adequacy of disclosures (Knapp, 1985).

With respect to earnings management, auditors should impair this in order to ensure financial reports that present a true and fair view of the organization and increases trust in financial statement audits (Quick and Warming-Rasmussen, 2009). Contrary, for reasons as aforementioned, managers have incentives to engage in earnings management. It is argued that audit firms are confronted with conflicting interests because the firm that needs to receive a clean audit opinion is the same firm that hires and pays for the audit. Financial statement users and regulators argue that auditors allow high fee client firms more discretion over their financial reporting as compared to low fee client firms (Ashbaugh et al., 2003), since the economic bond between the client firm and the auditor increases and therefore the economic dependence on this specific client (DeAngelo, 1981). Moreover, audit firms are frequently involved in providing non-audit services to clients for which they are also conducting the

(12)

audit. It is argued that providing non-audit services will strengthen the economic dependence between the auditor and the client firm, which will result in situations in which audit firms will avoid issuing qualifying opinions to safeguard the revenue stream from that specific client (Frankel et al., 2002; Moore et al., 2006).

The nature of the economic dependence resulting from the fees paid to the auditor is the main focus of this research with respect to conflicts of interest that may arise between an auditor and a client firm. The next paragraph will elaborate more on the audit firm’s economic dependence on the basis of economic bonding theory.

2.2.1 Economic bonding theory: audit fees, non-audit fees, non-audit fee ratio and total fees

Conflict situations between the auditor and client may arise in terms of the need for adjustments in financial statements, the inappropriate use of certain accounting practices that needs to be corrected or certain modifications with respect to the adequacy of disclosures (Knapp, 1985). While discovering an error, the auditor needs to decide whether the client should to correct this or whether no further adjustments are needed. It is argued that auditor’s allowance of earnings management will increase as a result of the economic dependence on the client firm (Beeler and Hunton, 2001; Frankel et al., 2002). This economic dependence – defined in the literature as the ‘economic bonding theory’ (Simunic, 1984; Quick and Warming-Rasmussen, 2009) – will increase as the stream of revenues obtained from the client increases. The revenues may exist of both audit fees and non-audit fees. Audit fees are fees directly related to the financial audit and non-audit fees incorporate all other fees received for services not related to the audit as taxation, advisory, and management consultancy services.

The economic bond between the auditor and the client firm is previous examined in relation to earnings management. These studies most often incorporate only one of the different fee types, namely audit fees, non-audit fees, total fees or fee ratio. The latter is the proportion of non-audit fees as compared to audit fees. Prior accounting research found no consensus for the relationship between providing non-audit services and the allowance of earnings management (Lin and Hwang, 2010). Studies focusing solely on audit fees found both a positive (Lin et al., 2006; Antle et al., 2006) and negative association (Frankel et al., 2002) between the magnitude of audit fees and the earnings management. Other studies, focusing on non-audit fees only, also found different results for both a positive association between non-audit fees and earnings management (Frankel et al., 2002; Ferguson et al., 2004; Larcker and Richardson, 2004; Cahan et al., 2008) and a negative association between

(13)

non-audit fees and earnings management (Antle et al., 2006). This negative association could be explained by the knowledge spillover theory (Antle et al., 2006; Habib, 2012). This theory state that providing non-audit services contribute in the understanding of the clients business and that this knowledge will result in a better-performed audit with less accrual-based earnings management (Kinney et al., 2004; Antle et al., 2006; Habib, 2012). Studies focusing on total fees and fee ratio in relation to earnings management all find a positive relationship between total fees and fee ratio and auditor’s allowance of earnings management (Frankel et al., 2002; Larcker and Richardson, 2004; Cahan et al., 2008). These findings provide evidence that an auditor who becomes more economically depended on the client is more susceptible for pressure from management to allow earnings management.

Empirical studies conducting a meta-analysis in the field of fees paid to the auditor and earnings management, state that non-audit fee ratio and total fees are most relevant when examining the influence of fees on the allowance of earnings management (Lin and Hwang, 2010; Habib, 2012), but that the dependence on different types of fee (audit and non-audit) could have various effects on the allowance of earnings management. Therefore, all separate fees are incorporated since they all contribute to the revenue stream from the client to the audit firm and therefore potentially strengthen the economic bond between audit and client firm. Prior literature state that expected audit fees in an auditor-client relation create in essence an economic bond (DeAngelo, 1981; Simunic, 1984) and therefore a higher likelihood that an auditor will acquiesce to the pressure from a client to allow earnings management (e.g. Beeler and Hunton, 2001; Frankel et al., 2002).

Other prior research state that the magnitude of fees not necessarily harms auditor independence, but that the disclosure of fee information itself result in a tendency of auditors to issue a qualified report sooner when an error or breach is discovered and therefore less accrual-based earnings management is allowed (Craswell et al., 2002). So contrary to the economic bonding theory, it could be argued that audit fee represent audit effort and therefore the level of assurance. This means that the audit fee increases with the amount of work performed by the auditor or the quality delivered by the auditor (Hay et al., 2006). In this respect, a larger audit fee may indicate a higher level of assurance. It could therefore be expected that accrual-based earnings management is impaired with an increase in audit fees. These competing expectations are operationalized in the following hypothesis, which will be tested two-sided.

(14)

Although, this research builds on the economic bonding theory and therefore aimed in providing evidence for a positive relationship between non-audit fees and earnings management, the hypotheses related to non-audit fees are operationalized and tested two-sided. This is based on the aforementioned reasoning that a negative relationship could be explained by the knowledge spillover theory (Antle et al., 2006; Habib, 2012).

H1b: Non-audit fee is associated with accrual-based earnings management.

H1c: Non-audit fee ratio is associated with accrual-based earnings management.

Furthermore, for the reason that all separate fees together contribute to the economic bond between audit and client firm, it is relevant to examine the relationship between total fees and earnings management. As mentioned above, prior studies find a positive relationship between the magnitude of total fees and auditor’s allowance of earnings management (Frankel et al., 2002; Dee et al., 2002; Larcker and Richardson, 2004; Cahan et al., 2008). However, there are also competing findings supporting an opposite effect of the magnitude of audit and non-audit fees on accrual-based earnings management. This is operationalized through the following hypothesis, which will also be tested two-sided.

H1d: Total fee is associated with accrual-based earnings management.

2.3 Corporate governance

A system of corporate governance is originated from the separation between ownership and control and hence the accompanying agency problems (Lin and Hwang, 2010). Corporate governance includes in the broadest sense the fundamental principles on which the firm is controlled and managed (Lin and Hwang, 2010). One of the primary goals of a well-functioning corporate governance structure is to ensure and improve the quality of both the financial reporting and the audit process within a firm (Larcker and Richardson, 2004; Beasly et al., 2009; Pergola and Joseph, 2011; Zaman et al., 2011). Recent legislation changes held board members more responsible for controlling the management and in ensuring a properly functioning of the corporate governance system. It focuses on constraining the possibilities for a manager to engage in managing earnings (Farinha, 2003).

The academic literature comprises various definitions of corporate governance. These definitions result from the ability to approach the concept of corporate governance from different perspectives, namely an internal and external perspective. From an external

(15)

perspective, corporate governance is seen as a way “in which suppliers of finance to

corporations assure themselves of getting a return on their investment.” (Shleifer and Vishny,

1997: 737). From a more internal perspective, corporate governance is seen as “the system of

laws, rules and factors that control operations at a company.” (Gillan and Starks, 1998: 4).

This set of structures to create boundaries for operations includes the ones who are involved in corporate activities and suppliers of capitals (Gillan and Starks, 2003). Embroidering this definition, corporate governance relates to “a set of contracts that defines the relationships

among the three principal actors in the corporation” (Gomez and Korine, 2005: 739). The

three principal actors to which the definition refers to are the sovereign, which is the shareowner; the governed, which are all stakeholders; and the governing, which refer to the ones who manage and control the firm (Gomez and Korine, 2005). Based on the above definitions, this thesis defines corporate governance as a system that helps to ensure the quality of financial reporting and audit processes and therefore strengthens the role of the shareholders and moreover helps the audit firm impair the ability for management to engage in earnings management.

2.3.1 Corporate governance, economic bonding and earnings management

Corporate governance is often studied in relation to firm performance. Researchers state that companies which have a stronger corporate governance structure – based on governance indexes – relate to higher market return, growing firm value and a better operating performance (Gompers et al., 2003; Brown and Caylor, 2006; Krafft et al., 2009).

Another stream of prior research examines corporate governance in relation to earnings management. These studies are most often focused on some specific elements of a corporate governance structure such as the role of the board and the presence of an audit committee (Klein, 2002; Xie et al., 2002; Bedard et al., 2004). Studies focusing on the effect of an audit committee show that the presence of a financial and governance expert, the presence of a clear mandate for responsibilities of the committee and perceptible independence relate negatively to aggressive earnings management (Bedard et al., 2004) and show lower levels of discretionary accruals (Klein, 2002; Xie et al., 2002; Davidson et al., 2005). In addition, frequency of meetings is also associated with lower levels of discretionary accruals (Xie et al., 2002). Taken together, these studies conclude that elements of a corporate governance structure – such as board structure, policy, and presence of an audit committee – help in reducing the ability for managers to engage in earnings management. The above-expected association is operationalized in the following hypothesis.

(16)

H2: Corporate governance is negatively associated with accrual-based earnings

management.

Aforementioned, the incorporation of corporate governance in this research is relevant since in practice audit firms are economically depended on their client and accounting policies leave room for earnings management. Corporate governance is seen as a mechanism that could contribute in ensuring auditor independence and hence could restrain the economic bond between the auditor and client firm (Gold et al., 2004). This will likely have an effect on the relationship between the auditor’s economic dependence on the client and the allowance of earnings management. More specifically, a well-functioning corporate governance system can help other governance bodies – like an audit committee – in mitigating the likelihood that an auditor will acquiesce to the pressure from a client to allow earnings management (Gold et al., 2004). Ultimately, corporate governance supports the overall monitoring of the financial reporting (Gold et al., 2004).

This situation highlights the importance of a properly functioning governance system within the organization in order to mitigate the relationship between the economic bonding of an auditor on a client firm and the allowance of earnings management. Based on the aforementioned theory and findings of previous research, this study predicts that corporate governance will moderate the relationship between the auditor’s economic dependence on the client and the allowance of earnings management and therefore the following associated hypotheses will be tested:

H3a: Corporate governance moderates the association between audit fee and accrual-based

earnings management.

H3b: Corporate governance moderates the association between non-audit fee and

accrual-based earnings management.

H3c: Corporate governance moderates the association between non-audit fee ratio and

accrual-based earnings management.

H3d: Corporate governance moderates the association between total fee and accrual-based

earnings management.

2.4 Overview research and expectations

Figure I provides a conceptual model for this research. As previously stated, the association between the independent variables – economic bonding and corporate governance

(17)

– and the dependent variable – earnings management – is examined. Moreover, a moderating effect is studied as to whether corporate governance influences the association between economic bonding and earnings management.

The above-described variables are measured through the following proxy’s. The dependent variable – earnings management – is measured by estimating the discretionary accruals. Audit fees, non-audit fees, non-audit fee ratio and total fees are incorporated to express the independent variable economic bonding. The Governance Metrics International (GMI) index is used as proxy for the strength of corporate governance. This operationalization of the variables will be explained in more detail in the following section.

Furthermore, the constructed hypotheses are displayed including the related expectations based on the literature review and hypotheses development.

3. Methodology

This section elaborates more on which research design is selected to test the aforementioned hypotheses. Section 3.1 explains which sample is selected for testing. Thereafter, section 3.2 elaborates more on the dependent variable in this research, namely earnings management and the measure used in this research to estimate the discretionary accruals. Section 3.3 discusses the incorporated independent variables, namely audit, non-audit, non-audit fee ratio and total fees to express the economic bond between the auditor and client firm. Subsequently, section 3.4 elaborates more on the selected control variables and

Figure I

(18)

finally section 3.5 presents the empirical models that will be used in order to test the hypotheses.

3.1 Research samples

The selected research approach for this study is an archival study, using quantitative data to test the stated hypotheses. The data sample is comprised of companies, which are included in the S&P 1500 index of the United States during the years 2003-2012. The United States is studied since US companies are relatively homogenous with respect to audit fees as well as the importance for audit firms to generate revenues from providing non-audit services (Dee et al., 2002). This relative broad time span is chosen to compose a rich dataset in order to be able to build solid conclusion upon. Year 2003 is chosen as starting point, since this is the year after the implementation of the Sarbanes-Oxley act, a major change in the audit standards within the United States. A starting point before the year 2003 could create a biased sample.

For this research, two different research samples are (indispensable) created as a result of data unavailability for the corporate governance index after year 2006. From now on ‘Data sample I’ will refer to the dataset which cover the fiscal years 2003-2012 and will primary test the relationship between fees and earnings management. ‘Data sample II’ will refer to the second dataset which covers the fiscal years 2003-2006 and will test both the primary relationship between fees and earnings management and moreover, the moderating effect of corporate governance on this primary relationship. As stated above, the selected corporate governance index is not provided after fiscal year 2006. With respect to test results of hypothesis 1, data sample II is used to validate the results obtained from data sample I. Table I provides an overview of both testing samples.

Sample Period Observations Databases Hypothesis

I 2003-2012 33.606 AuditAnalytics, Compustat 1

II 2003-2006 4.903 RiskMetrics, AuditAnalytics, Compustat 1/2/3 Start of year 01/01 and year-end 31/12.

Table I

(19)

Furthermore, in line with similar studies, firms related to regulation, financial and government sectors (SIC codes 6000-6999) are excluded from the data observations since it is difficult to estimate the discretionary accruals as a result of special accounting practices within these sectors (Frankel et al., 2002). In the case that data is unavailable for the required variables, companies will be excluded from the sample.

The first data sample is comprised from data obtained from AuditAnalytics (data regarding audit fees, non-audit fees and total fees) and Compustat (remainder data). In addition, information is required for year 2002 since some variables need lagged information. The Central Index Key (CIK) numbers are used as common identification code for the two different databases. Firm-year combinations are derived on basis of CIK numbers and associated year. This combination will be used to merge both datasets.

For the second data sample, in order to test the association of corporate governance with accrual-based earnings management and the moderating effect of corporate governance on the primary relationship, it is necessary to retrieve data from three different databases. The corresponding databases of interest are: RiskMetrics for the corporate governance index, AuditAnalytics for data related to audit fees, non-audit fees, total fees and name of audit firm, and Compustat for all other required data needed for this research.

Since the corporate governance index is the main point of interest in the second data sample, the RiskMetrics database is the starting point in retrieving data for the years 2003-2006. In line with the data collection of the first data sample, the CIK numbers are used as common identification code for all three different databases. Based on the retrieved data and accompanying CIK numbers of the RiksMetrics database, the required data for corresponding firm and year is extracted from AuditAnalytics. The remaining data is, again based on CIK numbers, extracted from Compustat. In this case, additional information is also needed for the year 2002 to be able to derive lagged variables.

The different datasets are merged by using the statistical software program SPSS. After merging, the missing data fields were deleted and the different variables were computed by using Microsoft Excel. After performing all required calculations except the discretionary accrual calculation, the datasets were imported in the statistical software program STATA. The discretionary accrual calculations and all regressions were performed in STATA. Finally, to control for outliers, all variables were winsorized at the 1st_{and 99}th_{percentile. Table II} provides an overview of the sample derivation process and accompanying steps in creating both data samples.

(20)

3.2 Dependent variable: accrual based earnings management

This research will build on prior research in determining accrual-based earnings management through the cross-sectional modified Jones model. This model is one of the most used models to estimate discretionary accruals and is introduced by Jones (1991). This model estimates discretionary accruals as part of total accruals. Total accruals are the sum of discretionary and non-discretionary accruals. With this measure, it is identified if firms are using discretionary accruals to manipulate earnings and therefore engage in earnings management.

The model used in this study extends the modified Jones model used by previous research (Dechow et al., 1995; Frankel et al., 2002; Chung and Kallapur, 2003) by including a performance variable, namely previous year’s operating performance (Kothari et al., 2005; Kim et al., 2012). The model to determine discretionary accruals is as follows:

TACCit/ASSETSit-1 = β0 + β1 (1/ASSETSit-1) + β2 (∆REVit - ∆RECit)/ASSETSit-1+

β3 PPEit/ASSETSit-1+ β4 ROAit-1 + εit (1)

Where:

TACCit Total accruals calculated by income before extraordinary items minus cash flow from operations for each firm i in year t.

ASSETSit-1 Total assets of yeart-1 for each firm i in year t.

∆REVit Change in revenue from prior year for each firm i in year t.

∆RECit Change in accounts receivables from prior year for each firm i in year t.

PPEit Gross property, plant and equipment for each firm i in year t.

Description Observations Description Observations

Initial sample after merging (based on CIK numbers) 136.355 Initial sample after merging 11.634

Removal financial institutions (SIC 6000-6799) (51.724) Removal missing values (4.489)

Removal missing values (33.405) Raw data before calculations 7.145

Raw data before calculations 51.226 Removal lagged year (2002) (1.506) Removal lagged year (2002) (5.668) Removal missing values through non-consecutive years (167) Removal missing values through non-consecutive years (11.315) Removal financial institutions (SIC 6000-6799) (115) Data before DACC calculations (STATA) 34.243 Data before DACC calculations (STATA) 5.357 Removal cases less than 10 firms in an industry year group (637) Removal cases less than 10 firms in an industry year group (454) Total firm-year observations 33.606 Total firm-year observations 4.903 Observations presentto the number firm-year observations in total sample.

Data sample I Data sample II

Table II Sample derivation

(21)

ROAit-1 Return on assets measured by net income (NI) for yeart-1 divided by ASSETSit-1 for each firm i in year t.

εit Error term, which is the residual of the estimated model (1) and is interpreted as discretionary accruals for each firm i in year t (DACCit).

Total accruals (TACCit) are equated against non-discretionary variables. Expected level of accruals, based on industry and firm specific characteristics, are estimated for every industry group with at least ten firms in a specific year. To estimate earnings management more closely by using this cross-sectional accrual model, all cases where less than ten firms were available in an industry-year group were deleted. In this research, industry groups are based on the accompanying 2-digit Standard Industrial Classification (SIC) codes. Within STATA, a loop is created that runs the modified Jones model for the specific 2-digit SIC industry-year groups and the residual term of the regressions is equal to the industry-year adjusted discretionary accruals (DACCit).

The estimate of the residual is transformed to an absolute value, because earnings are managed upwards and downwards and therefore the combined effect of income-increasing and income-decreasing earnings management decisions needs to be measured (Becker et al., 1998; Frankel et al., 2002; Chung and Kallapur, 2003). The absolute value of discretionary accruals is hereafter referred as |DACCit|.

3.3 Independent variables

3.3.1 Economic bonding: audit fees, non-audit fees, non-audit fee ratio, total fees

As stated in the literature section, the economic bond between the auditor and client firm in relation to the magnitude of earnings management will be examined in multiple ways, namely audit fees, non-audit fees, non-audit fee ratio and total fees paid to the auditor. Non-audit fee ratio is the ratio of non-Non-audit fee to Non-audit fee. Total fees cover all fees paid to the audit firm, including audit fees and non-audit fees related to other activities performed by the auditor such as taxation, advisory, and management consultancy services. In order to eliminate bias in the sample and for better comparability between firms, fees are scales by total assets. This is based on the assumption that larger clients simply require more resources in terms of staff and labor hours of the audit firm (Hay et al., 2006).

(22)

3.3.2 Corporate governance

As aforementioned, this paper examines the association between corporate governance and earnings management and whether strong corporate governance within firms moderates the relationship between the economic bond between an auditor and a client firm and the magnitude of earnings management. Earlier conducted research on corporate governance in relation to firm performance, selects a governance index rating such as the Governance Metrics International (GMI) and RiskMetrics/International Shareholder Services (RM/ISS) index to indicate the quality of a firms’ governance system (Gompers et al., 2003; Brown and Caylor, 2006; Krafft et al., 2009). These indices reasonable assess weaknesses, strengths and firms’ characteristics of their governance system and incorporate this in one combined index (Krafft et al., 2009; Pergola and Joseph, 2011).

This study adopts the measure provided by the RiskMetrics database, which is introduced by Gompers, Ishii and Metrick (2003). This governance index also referred as the ‘G-index’ indicates the quality of the corporate governance. The G-index is rated on 24 separate points, where the allocation of one point indicates an observation of an unfavorable provision for a shareholder. So, a high corporate governance score represents a low quality of a corporate governance system since shareholder rights are lower represented and management power therefore increased and vice versa (Gompers et al., 2003). The points are assigned over five categories related to shareholders rights, namely delay (tactics for delaying hostile bidders), protection (management protection), voting (shareholder voting rights), other (diverse takeover defenses) and state (presence of state laws) (Gompers et al., 2003; Jiraporn et al., 2006). Appendix A provides a list of the individual governance provisions included in the construction of the G-index (Gompers et al., 2003).

3.4 Control variables

Previous research has shown that different variables influence the variables of interest in this research: economic bonding, earnings management and corporate governance. These variables will be included in the empirical model as control variables, in order to ensure that these have no effect on the results of this research. To control for the size of the client company, all continuous variables are scaled by total assets.

SIZE is included since the size of the firm may affect earnings management (Frankel

et al., 2002; Ashbaugh et al., 2003; Cahan et al., 2008; Kim et al., 2012) and corporate governance variables (Becker et al., 1998). It is argued that larger firms have potentially more stable accruals (Myers et al., 2003), but on the other hand larger firms are expected to lower

(23)

earnings in line with the political cost hypothesis (Watts and Zimmerman, 1986). SIZE is measured as the natural logarithm of total assets for each firm i in year t.

BIG4 is included since prior studies show that firms audited by Big 4 auditors are

associated with lower levels of earnings management than firms audited by non-Big 4 auditors (DeAngelo, 1981; Becker et al., 1998; Frankel et al., 2002, Ashbaugh et al., 2003; Cahan et al., 2008; Kim et al., 2012). BIG4 is an indicator variable, which equals 1 if firm is audited by a Big 4 auditor and 0 if not, for each firm i in year t.

LOSS is included as prior studies expect that firms in financial difficulties have more

incentives to engage in earnings management (Frankel et al., 2002; Cahan et al., 2008). LOSS is an indicator variable, which equals 1 if net income is negative and 0 when positive, for each firm i in year t.

LEV is included since prior studies have shown that leverage is associated with

earnings management. It is argued that there could be both a positive or negative association (Frankel et al., 2002; Ashbaugh et al., 2003; Kim et al., 2012). The variable leverage is measured as long-term debt scaled by total assets.

SCFO is included to control for non-discretionary accruals with respect to firm

performance (Dechow et al., 1995; Frankel et al., 2002; Chung and Kallapur, 2003; Cahan et al., 2008). It is argued that firms with extreme positive cash flows have more negative non-discretionary accruals and vice versa. Cash flow from operations is scaled by total assets for each firm i in year t.

3.5 Empirical models

In order to test the stated hypotheses, this research will make use of ordinary least squares multivariable regression models. Model (2) will be used to test hypotheses 1a, 1b, 1c and 1d. Therefore test the relationship between the different fee variables, namely scaled audit fees, scaled non-audit fees, scaled total fees and non-audit fee ratio and the absolute value of discretionary accruals.

|DACC|it = β0 + β1 FEE_VARIABLEit + β2 SIZEit + β3 BIG4it + β4 LOSSit +

β5 LEVit + β6 SCFOit + εit (2)

Hypothesis 2 is tested by model (3). Model (3) will measure the association between the corporate governance variable and the absolute value of discretionary accruals.

(24)

|DACC|it = β0 + β1 CORP_GOVit + β2 SIZEit + β3 BIG4it + β4 LOSSit +

β5 LEVit + β6 SCFOit + εit (3)

To measure the moderating effect of corporate governance, this research will create an interaction term of the variables of interest, namely SC_AUD_FEESit, SC_NONAUD_FEESit,

NAF_RATIOit and SC_TOT_FEESit, multiplied by variable CORP_GOVit to capture the

interaction effect. Hypothesis 3a, 3b, 3c and 3d will be tested by model (4).

|DACC|it = β0 + β1 FEE_VARIABLEit + β2 CORP_GOVit +

β3 FEE_VARIABLEit* CORP_GOVit + β4 SIZEit + β5 BIG4it +

β6 LOSSit + β7 LEVit + β8 SCFOit + εit (4)

Model (2), (3) and (4) variables explained:

|DACC|it Absolute value of discretionary accruals as estimated by model

(1).

FEE_VARIABLEit One of the different tested fee variables, namely

SC_AUD_FEESit, SC_NONAUD_FEESit,, NAF_RATIOit or

SC_TOT_FEESit.

SC_AUD_FEESit, Audit fees scaled by total assets for each firm i in year t.

SC_NONAUD_FEESit Non-audit fees scaled by total assets for each firm i in year t.

SC_TOT_FEESit Total of audit fees and non-audit fees scaled by total assets for

each firm i in year t.

NAF_RATIOit Non-audit fee ratio, which is the ratio of non-audit fees to audit

fees for each firm i in year t.

CORP_GOVit Corporate governance index score for each firm i in year t.

SIZEit Natural logarithm of total assets for each firm i in year t.

BIG4it BIG4 as a Big 4 auditor, otherwise 0, audits a dummy variable, which equals 1 if firm, for each firm i in year t.

LOSSit Loss is a dummy variable, which equals 1 if net income is negative and 0 if positive, for each firm i in year t.

LEVit Leverage, long-term debt scaled by total assets.

SCFOit Cash flow from operations is scaled by total assets for each firm

(25)

Above specified models are created to test the hypotheses of interest in this research. With respect to model (2) – designed to test hypotheses 1a, 1b, 1c and 1d – the following findings will support the stated hypotheses. If the coefficient of the variable SC_AUD_FEESit has a significant positive outcome, hypothesis 1a would be supported by the economic bonding theory. If there is a significant negative outcome, this will support the expectation that higher audit fee means higher effort and therefore higher quality with less allowance of accrual-based earnings management. For the variables SC_NONAUD_FEESit and

NAF_RATIOit – respectively hypothesis 1b and 1c – a significant positive coefficient would

support the economic bonding theory, implying that auditors tend to lose their objectivity and independence because they are subject to the dependency on the revenue stream obtaining from the client. A significant negative coefficient would mean evidence to support the knowledge spillover theory, which implies that providing non-audit services and audit services for the same client will result in a higher quality audit and more specifically constrain earnings management attempts. In line with above reasoning, hypothesis 1d will support the economic bonding theory if the coefficient of the variable SC_TOT_FEESit has a significant positive outcome. Contrary, when the coefficient is significant negative, this would imply support for the knowledge spillover theory and the expectation that higher fees obtained from the client will result in higher quality in terms of less accrual-based earnings management.

Moreover, with respect to model (3) – designed to test hypothesis 2 – hypothesis 2 is supported if the coefficient of the variable CORP_GOVit has a negative significant coefficient. This would imply that corporate governance could impair accrual-based earnings management.

Furthermore, with respect to model (4) – designed to test hypotheses 3a, 3b, 3c and 3d – the following findings will support the stated hypotheses. Irrespective of the obtained association between one of the fee variables and accrual-based earnings management, these hypotheses will be supported if the coefficient of the interaction variable FEE_VARIABLEit * CORP_GOVit is significant negative, since this implies that the governance variable (good governance) moderates the relationship between the examined fees and the absolute value of discretionary accruals (i.e. earnings management).

This research will observe a coefficient as significant when the associated p-value is smaller than .1. A higher level of significance is obtained when the associated p-value is below .05 or lower than .01. These significance levels will be marked respectively as (*), (**) and (***).

(26)

4. Empirical findings

This section will outline the empirical findings obtained through testing of the stated models (2), (3) and (4). These findings provide evidence on whether the stated hypotheses will be supported. Section 4.1 describes the descriptive statistics of both research samples, where after section 4.2 will discuss the correlation analysis. Section 4.3 elaborates on the results of the regression analyses. Finally, section 4.4 discusses several robustness and sensitivity analyses.

4.1 Descriptive statistics

Table III, IV and V provide descriptive statistics with respect to the variables of interest for both research samples. Table III is constructed to provide an overview of the distribution of audit, non-audit and total fees for both research samples. Notwithstanding that these variables are not directly used as input for one of the regression models, it is relevant to provide information about the (percentage) distribution of the fees and means per year and over the whole research period. Table III describes mean audit fees of $ 1,7 million dollar and $ 2,9 million dollar, respectively for research samples I and sample II. The mean non-audit fees amount to almost $ 0,5 million and $ 1,0 million for both samples I and II. These statistics conveyed to a percentage distribution show approximately 75% audit fees to 25% non-audit fees of total fees as seen over both periods. For both samples, non-audit fees as compared to audit fees declined after the implementation of SOX in the year 2002, but the audit fees, and therefore simultaneous the total fees increases rapidly over time. This trend of increasing audit costs is in line with the higher costs as result of the implemented SOX legislation and accompanied higher compliance costs for firms. So, increasing audit fees offsets the decrease in revenues obtained from providing non-audit fees as a result of the restricting regulations. To clarify the (percentage) distribution of audit, non-audit and total fees, a graphical overview is presented in Figure II.

(27)

In Table IV, the sample distribution is presented by the two-digit SIC code industry. For data sample I, the most heavily represented industry is SIC code 73 ‘Business Services’ with 13,54 per cent. Secondly, SIC code 28 ‘Chemicals and Allied Products’ with 9,91 per cent followed by SIC code 36 ‘Electronic and other Electric Equipment’. With respect to data sample II, the most heavily represented industry is SIC code 36 ‘Electronic and other Electric Equipment with 10,93 percent. Followed by SIC code 28 ‘Chemicals and Allied Products’ with 8,75 percent. Thirdly, SIC code 35 ‘Industrial Machinery and Computer Equipment’ with 8,10 percent.

Year Mean Percentage Mean Percentage Mean Percentage Mean Percentage Mean Percentage Mean Percentage 2003 783.048 58,64% 552.274 41,36% 1.335.322 100,00% 1.724.273 56,40% 1.332.694 43,60% 3.056.967 100,00% 2004 1.376.176 73,88% 486.556 26,12% 1.862.732 100,00% 2.851.237 73,15% 1.046.494 26,85% 3.897.731 100,00% 2005 1.619.396 79,95% 405.992 20,05% 2.025.388 100,00% 3.475.604 80,53% 840.475 19,47% 4.316.079 100,00% 2006 1.786.654 81,65% 401.426 18,35% 2.188.080 100,00% 3.623.571 81,54% 820.366 18,46% 4.443.937 100,00% 2007 1.869.299 80,20% 461.522 19,80% 2.330.821 100,00% 2008 1.961.449 81,29% 451.306 18,71% 2.412.755 100,00% 2009 1.875.502 81,19% 434.634 18,81% 2.310.136 100,00% 2010 1.911.339 79,43% 494.960 20,57% 2.406.299 100,00% 2011 2.012.639 79,15% 530.084 20,85% 2.542.723 100,00% 2012 2.128.036 78,27% 590.853 21,73% 2.718.889 100,00% Mean 1.732.354 77,37% 480.961 22,63% 2.213.315 100,00% 2.918.671 72,91% 1.010.007 27,09% 3.928.679 100,00% Fees are presented in USD.

Table III

Distribution audit, non-audit and total fees over time

Data sample I

Audit Fees Non-Audit Fees Total Fees

Data sample II

Audit Fees Non-Audit Fees Total Fees

Fees are presented in USD.

Figure II

(28)

Two-digit Percentage Percentage

Industry SIC N of sample N of sample

Metal Mining, Ores 10 157 0,47 0 0,00

Oil and Gas 13 1.393 4,15 192 3,92

Food, Beverage 20 748 2,23 150 3,06

Textile Mill Products 22 116 0,35 0 0,00

Apparel and Other Textile Products 23 319 0,95 63 1,28 Lumber and Wood Products 24 143 0,43 0 0,00 Furniture and Fixtures 25 201 0,60 21 0,43 Paper and Allied Products 26 248 0,74 71 1,45 Printing and Publishing 27 418 1,24 112 2,28 Chemicals and Allied Products 28 3.331 9,91 429 8,75

Petroleum 29 182 0,54 48 0,98

Rubber 30 440 1,31 55 1,12

Primary Metal Industries 33 480 1,43 107 2,18 Fabricated Metal Products 34 516 1,54 80 1,63 Industrial Machinery and Computer Equipment 35 1.968 5,86 397 8,10 Electronic and Other Electric Equipment 36 3.143 9,35 536 10,93 Transportation Equipment 37 842 2,51 144 2,94 Instruments and Related Products 38 2.788 8,30 353 7,20 Miscellaneous Manufacturing 39 319 0,95 53 1,08 Trucking and Warehousing 42 193 0,57 42 0,86

Water Transportation 44 175 0,52 0 0,00

Air Transportation 45 288 0,86 53 1,08

Communication 48 1.250 3,72 175 3,57

Electric, Gas, Sanitary Services 49 2.016 6,00 321 6,55 Wholesale-Durable Goods 50 769 2,29 125 2,55 Wholesale-Non-Durable Goods 51 534 1,59 63 1,28 General Merchandise Store 53 220 0,65 70 1,43 Auto Dealers, Gas Stations 55 188 0,56 31 0,63 Apparel and Accessory Stores 56 404 1,20 98 2,00

Eating and Drinking 58 582 1,73 100 2,04

Miscellaneous Retail 59 695 2,07 97 1,98

Business Services 73 4.550 13,54 0 0,00

Amusement and Recreation Services 79 298 0,89 0 0,00

Health Services 80 782 2,33 108 2,20

Educational Serives 82 218 0,65 0 0,00

Engineering and Management Services 87 781 2,32 88 1,79

Other 1.371 4,08 721 14,71

Total 33.606 100,00 4.903 100,00

Sample distribution based on the two-digit SIC code industry.

Table IV

(29)

Furthermore, descriptive statistics of the main variables of interest over total sample period are provided in Table V. The following summary statistics can be obtained with respect to the dependent and independent variables. Table V shows |DACC| with a mean of 0.0985 (sample II: 0.0432) and a standard deviation of 0.1470 (sample II: 0.0457). In addition, single year descriptive statistics show for year 2003 to 2007 no major deviations from the total sample mean, but in year 2008 an increase in mean |DACC| followed by a decrease from 2009 onwards (not tabled). Table V shows total sample means for

SC_AUD_FEES, SC_NONAUD_FEES, SC_TOT_FEES, and NAF_RATIO of 0.0055, 0.0012,

0.0069, and 0.3285 respectively (sample II: 0.0017, 0.0006, 0.0021, and 0.4454). The scaled fees compared against each other describe the same relationship as in Table III. CORP_GOV ranges from 4 to 15, with a mean and standard deviation of 9.1177 and 2.4522. Moreover, single year descriptive statistics (not tabled) show an increase of the mean CORP_GOV over time from 2003 to 2006.

With respect to the control variables, the following descriptive statistics can be obtained. A significantly large part of firms hire a big 4 audit firm. Table V show a mean of 0.7265 meaning 72,65% of sample hires a big 4 audit firm (sample II: 0.8823, 88,23%) and standard deviation of 0.4457 (sample II: 0.3222). The variable LOSS shows a mean of .3704 (sample II: 0.1940) and standard deviation of 0.4829 (sample II: 0.1940). This implies that 37,04% (sample II: 19,40%) of the sample firms show on average a loss as seen over total sample period.